Arbe develops the world’s most advanced 4D imaging radar for ADAS and autonomous vehicles. Its radar delivers dense environmental imaging with the highest channel count in the industry, enabling unmatched resolution and accuracy. The system provides reliable object detection in any weather or lighting conditions and integrates with perception algorithms and AI systems.
Arbe’s technology powers:
Arbe’s radar chipset is the core technology behind high-resolution 4D imaging radar systems developed by tier 1s and OEMs, enabling advanced perception for ADAS and autonomous vehicles. It combines a processor, transmitter, and receiver into a compact, automotive-grade module.
The chipset delivers:
It operates in the 76–81 GHz automotive radar band, is AEC-Q100 and ASIL-B ready, and functions from –40°C to 125°C. Mass market ready, it can be installed behind the vehicle’s fascia.
The Phoenix Radar by Arbe is a high definition radar made for perception, delivering ultra-high-resolution data to power advanced driver-assistance systems (ADAS) and full autonomous driving. It is built on a 48×48 channel array, producing 2,304 virtual channels for 100× more detail than traditional radar systems.
Phoenix is designed for perception. It enables object tracking, free space mapping, and advanced cruise control, emergency braking and autonomous steering while reducing false alarms. It supports L2+ to L5 autonomy and is used in vehicles ranging from passenger cars to robotaxis and heavy machinery.
What makes Phoenix different:
In short, Phoenix sets itself apart with exceptional resolution, environmental robustness, and practical integration for scalable autonomy.
The main customers for Arbe’s radar technology are automotive OEMs and radar Tier 1 suppliers.
Arbe’s radar systems are used in passenger, commercial, and industrial vehicles, as well as other advanced safety applications. The Phoenix Perception Radar supports applications from L2+ ADAS to full autonomy (Level 3 and beyond), including OEMs pursuing hands-off and eyes-off capabilities at full highway speeds, where long-range, all-weather perception is essential for meeting L3 performance requirements. The radar installs discreetly behind the bumper or fascia of compatible vehicle types, including:
Passenger vehicles Trucks and commercial vehicles Robotaxis, delivery robots, and heavy machinery. This versatility makes Arbe’s radar suitable for both consumer and industrial use cases, including safety-critical and autonomous driving environments.
Arbe is headquartered in Tel Aviv, Israel, with offices in China, Germany, and the United States.
Arbe is a publicly traded company listed on NASDAQ (NASDAQ: ARBE) and on the Tel Aviv stock exchange (TASE: ARBE)
Arbe’s radar technology is available for commercial deployment and is designed for mass-market automotive use, enabling broad adoption across the global vehicle industry. Integrated by Tier 1 suppliers into radar modules for OEMs, Arbe’s radar powers next-generation radar systems for passenger vehicles, trucks, and commercial vehicles, as well as robotaxis and autonomous platforms. Its advanced high-resolution sensing capabilities make it a key enabler of ADAS applications ranging from Level 2+ driver assistance to full Level 5 autonomy. Built for safety-critical automotive applications, Arbe radar delivers the performance, reliability, and precision required to support both current and future generations of intelligent, automated, and safe mobility solutions.
To partner with Arbe or become a customer, you can reach out directly to their team to discuss radar integration, support, or collaboration opportunities. Arbe works with leading OEMs, Tier 1 suppliers, and autonomous vehicle developers to deploy 4D imaging radar for scalable autonomy from L2+ to full self-driving. To start a conversation: Email: [email protected] Arbe welcomes inquiries related to partnerships, technical integration, and commercial deployments.
Off-highway autonomy is harder than road autonomy because it operates in unstructured environments with no defined lanes, inconsistent terrain, and unpredictable obstacles. These environments include dust, mud, vibration, and total darkness, which degrade or disable traditional sensors. As a result, perception is less reliable, increasing the risk of machine stoppage and reduced operational uptime.
Sensors fail in off-highway environments because dust, glare, rain, darkness, and airborne particles interfere with optical sensing technologies like cameras and LiDAR. These conditions reduce visibility and distort sensor input, leading to perception errors or complete signal loss. When sensors fail, autonomous machines stop operating, directly impacting productivity and return on investment.
Ultra-HD radar improves perception in off-highway autonomy because it generates dense 4D point clouds that represent objects and terrain with high resolution. This allows autonomous systems to map uneven ground, detect obstacles, and understand free space in real time. Ultra-HD radar enables consistent perception even in conditions where other sensors fail.
Ultra-HD radar differs from traditional radar because it provides high-resolution imaging and object separation instead of basic detection. Traditional radar identifies the presence of objects, while ultra-HD radar distinguishes between multiple objects, their positions, and their motion. This higher fidelity enables better decision-making in complex and crowded environments.
Ultra-HD radar detects objects in dust, rain, and darkness because it operates in the 76–81 GHz radio frequency range, which penetrates airborne particles and low-visibility conditions. Unlike optical sensors, radar signals are not dependent on light, allowing consistent performance in all weather and lighting conditions .
Ultra-HD radar improves object detection and classification because it combines fine Doppler resolution with a high channel count to separate objects in dense scenes. This enables the system to distinguish between stationary and moving objects and identify small hazards near larger reflective surfaces. The result is more accurate perception in complex environments.
Ultra-HD radar supports sensor fusion systems because it provides a reliable second source of truth for perception data. It complements cameras and GNSS/IMU systems by maintaining performance when vision or satellite signals are degraded. This improves overall system robustness and enables more accurate environment understanding.
Sensor fusion is necessary for autonomous systems because no single sensor can provide complete and reliable perception in all conditions. Cameras offer high-resolution visual detail, while radar provides robustness and motion detection in low visibility. Combining multiple sensors creates a more accurate and resilient perception system.
Ultra-HD radar performs well in GPS-denied environments because it enables self-localization using real-time environmental data instead of relying on satellite signals. This allows autonomous systems to maintain accurate positioning in underground, remote, or obstructed environments where GPS is unavailable.
Reliability is critical for off-highway autonomy because machine uptime directly determines productivity and profitability. When perception systems fail, machines stop operating, causing delays and financial loss. Reliable sensing ensures continuous operation and consistent output.
Ultra-HD radar is more reliable than cameras and LiDAR because it is not affected by lighting conditions, weather, or surface contamination. Its solid-state design eliminates moving parts, reducing failure points and maintenance requirements. This allows radar to operate consistently in harsh industrial environments.
Ultra-HD radar improves safety around heavy machinery because it provides high-resolution near-field perception and accurate object separation. This enables detection of workers and obstacles close to the machine, reducing the risk of collisions. Improved perception supports safer human-machine interaction in active work zones.
Success for off-highway autonomous systems is defined by continuous operation, high uptime, and reliable performance in all conditions. Autonomous systems must function consistently in dust, mud, darkness, and harsh environments without frequent interruptions. Performance in real-world conditions, not controlled demos, determines system value.
Ultra-HD radar plays a critical role in scaling autonomous operations because it provides consistent perception across all environments and conditions. This reliability allows autonomous systems to move from pilot programs to full deployment. Ultra-HD radar enables autonomy to become a dependable, revenue-generating asset rather than an experimental technology.
The core bottleneck is grounding and perception rather than processing power or “thinking.” No matter how advanced the AI model is, the system can only act on what it can reliably perceive; if sensor inputs are incomplete, delayed, or noisy, all downstream layers face failure.
Observations: Raw sensor inputs capturing the environment and system state.
World Model: Predictive layer evaluating how the environment will evolve.
Control Policy: Decision layer generating actions based on state estimates.
Feedback Loop: Cycle where actions alter the environment to produce new observations.
Cameras are highly vulnerable to degraded visibility conditions, including poor lighting, rain, fog, glare, and darkness. Additionally, camera sensors cannot directly measure critical physical variables like velocity or distance.
Conventional radar systems lack sufficient resolution and dynamic range. They typically operate with a 5-degree angular resolution, which is far too coarse to support precise obstacle detection and classification.
Virtual Channels: 2,304 channels.
Resolution: 0.7° azimuth resolution.
Field of View: 120° × 30°.
Detection Range: Extends up to 300 meters.
The system delivers direct Doppler velocity measurements with every detection. This allows the AI to track motion instantaneously instead of inferring it across multiple video frames, enabling faster estimation of agent intent.
The hardware contains zero moving parts, requires no routine cleaning or specialized field maintenance, and delivers near-zero false alarms to eliminate phantom detections and unnecessary system interventions.
Can detect micro-movements and identify objects through certain materials.
Captures vital signs, such as heart rate, via subtle chest movements in short-range mode.
Adapts instantly between indoor and outdoor environments without requiring lens or hardware changes.
The radar is fully integrated into the NVIDIA DRIVE Hyperion platform, providing an all-weather perception foundation across automotive, industrial robotics, defense, marine, and off-road sectors.
Unlike paved roads with clear lanes and signs, off-highway environments—like mines, construction sites, and farms—are “unstructured.” They feature unpredictable terrain, extreme vibration, and self-generated hazards like massive dust clouds or mud, which quickly overwhelm standard sensors.
Cameras and LiDAR are easily “blinded” by airborne particles or grime on the lens. 4D Imaging Radar uses radio waves (76-81 GHz) that physically penetrate dust, heavy rain, and fog. Because it is solid-state with no moving parts, it is also immune to the constant jarring and vibration of heavy machinery.
In a mine or field, the ground isn’t flat. Ultra-HD radar provides a dense 3D point cloud that maps uneven terrain, steep slopes, and vegetation in real time. This allows the machine to distinguish between a “puddle” (driveable) and a “ditch” (hazard), preventing unnecessary and costly emergency braking.
In crowded work zones, “good enough” sensors see a single mass of pixels. Arbe’s radar uses fine Doppler resolution and a high channel count to distinguish a stationary post from a nearby walking worker. This “object separation” is critical for safe human-machine interaction in high-risk zones.
In heavy industry, “uptime” is profit. If a sensor fails due to a bit of glare or dust, the machine stops, and ROI disappears. Because radar works 24/7 in all conditions and requires zero cleaning (unlike camera lenses), it ensures the machine stays productive, directly impacting the bottom line.
Yes. Mines and deep forests are often “GPS-denied” environments. Ultra-HD radar provides robust data for Radar Odometry and Self-Localization. By “anchoring” the perception stack, the machine can estimate its position and navigate safely even when satellite signals are blocked or vision is zero.
As of March 2026, Arbe has pivoted its strategic focus to emphasize industrial and off-road markets. Recognizing that these sectors have shorter adoption cycles than consumer automotive, Arbe officially introduced its dedicated Off-Highway 4D Imaging Radar solution at the 6th Autonomous Off-Highway Machinery Technology Summit.
Arbe’s 2026 industrial solution is designed for Physical AI—AI systems that interact with the physical world. In agriculture, it supports “precision workflows” and extended operating hours; in mining, it enables 24/7 autonomous hauling by providing AI-ready, dense detections that “humanize” machine movement.
Yes. To accelerate ROI, Arbe now provides a dedicated Sensor Integration Toolbox with ROS 2 support. This allows industrial OEMs to retrofit existing fleets or build new autonomous platforms with minimal engineering effort, moving from “pilot project” to “site asset” faster than ever before.
By integrating with NVIDIA AI Computing(specifically Hyperion 10 ecosystems in 2026), the radar data is transformed into a highly detailed AI-based Occupancy Grid. This allows a massive 400-ton mining truck to “understand” its surroundings with the same level of nuance as a high-end passenger car, but with the ruggedness required for the pit.
Level 3 autonomy enables “eyes-off” driving where vehicles handle steering, acceleration, and braking on highways or in traffic. Drivers can engage in secondary activities like messaging or watching videos but must be ready to take control when prompted.
Conventional radar can only detect small objects like pedestrians up to 80 meters, providing just 2-3 seconds reaction time at highway speeds. Current sensing technology lacks the range and resolution needed for safe highway-speed autonomy, forcing most OEMs to cap systems at 60 km/h.
Level 3 systems should operate safely at 130 km/h (80 mph), the normal highway speed on the German Autobahn. This represents true highway autonomy compared to current 60 km/h traffic jam assistance systems.
Safe autonomous operation at 130 km/h requires detecting and classifying obstacles at 300 meters. Since vehicles cover 36 meters per second at this speed, 300-meter range provides adequate time for gradual braking or smooth lane changes without disrupting traffic. HD imaging radar is the only automotive sensor able to provide such detailed long range perception.
Conventional automotive radar detects small objects only up to 80 meters and cannot separate pedestrians or cargo from guardrails due to limited resolution and low dynamic range. This provides insufficient reaction time and object classification capability for safe highway-speed operation. HD imaging radar, such as Arbe’s, offers high resolution at long ranges and enables L3 with full highway speeds.
High-resolution radar detects from short range to over 300 meters in all weather and lighting conditions. Unlike cameras and LiDAR that depend on visibility, radar penetrates rain, fog, darkness, and snow while measuring range and velocity precisely. HD imaging radar, such as Arbe’s, enables L3 with full highway speeds.
Arbe’s ultra-HD radar features 2,304 channels (48 transmit x 48 receive) exceeding the 300-meter detection requirement. This architecture delivers exceptional distance and velocity measurement accuracy, fine horizontal and vertical resolution, and high dynamic range to distinguish closely-spaced objects and detect partially obscured targets at highway speeds.
Radar provides all-weather reliability that cameras and LiDAR cannot match. While sensor fusion combines multiple technologies, radar serves as the safety-critical backbone because it maintains consistent performance when cameras fail in heavy rain, fog, darkness, blinding sunlight, and snow.
Superior object classification through high resolution and dynamic range minimizes unnecessary takeover requests, keeping systems in autonomous mode longer and building driver trust. Arbe’s High-resolution radar for example can distinguish a tire beside a guardrail or detect a child partially obscured by a bus, reducing false alarms that erode confidence.
Radar ambiguities occur when a sensor views one physical object but reports multiple targets with similar power levels scattered across its field of view. The radar cannot determine which target is real, creating phantom signals that make the sensor fundamentally unreliable for autonomous driving applications.
When you can’t trust whether radar shows a real obstacle or phantom, the fundamental safety premise collapses. Azimuth and elevation ambiguities create phantom targets for all objects, making it impossible to distinguish actual road hazards from background objects like guardrails, tunnel walls, or parked cars.
Radar ambiguities stem from the Nyquist sampling theorem: radars with sparsely placed antenna channels cannot definitively determine the direction of incoming signals. Like a camera filming a spinning propeller at too low a frame rate, insufficient spatial sampling creates phantom signals that cannot be distinguished from real targets.
Doppler ambiguities affect velocity measurements and can be resolved by taking multiple measurements, though this adds 100-150 milliseconds of latency. Azimuth and elevation ambiguities affect angular direction, create phantom targets for all objects, and are unmanageable for autonomous driving applications.
Conventional automotive radars typically use 12 transmit x 16 receive channel configurations, providing sparse spatial sampling. This limited antenna channel density creates highly ambiguous detection patterns insufficient for reliably detecting and classifying stationary objects in autonomous driving scenarios.
Dense spatial sampling from high antenna channel counts provides the data needed to definitively determine signal direction of arrival. More channels mean finer spatial resolution, directly solving the Nyquist sampling limitation that creates ambiguities in sparse antenna arrays, eliminating phantom targets at their source.
Arbe’s radar uses 48 transmitting and 48 receiving channels, creating 2,304 virtual channels. This high-channel count provides the dense spatial sampling needed to eliminate ambiguities, enabling thousands of virtual channels and tens of thousands of detections per frame for reliable object detection.
The Nyquist sampling theorem states that sparse sampling cannot definitively determine signal characteristics. In radar, insufficient antenna channel density means the system cannot determine the true direction of incoming signals, creating phantom targets similar to how low frame rates make spinning objects appear to move incorrectly.
AI-based ambiguity resolution attempts to distinguish real targets from phantoms through pattern learning, but this approach remains unproven for safety-critical applications. Autonomous driving requires deterministic sensor reliability, not probabilistic filtering that could fail to identify actual hazards or create false alarms.
Arbe’s custom chipset architecture processes 30 gigabytes of raw RF data per second, compressing it into a 100-megabyte point cloud output. This purpose-built processing enables thousands of virtual channels and tens of thousands of detections per frame without the latency that would compromise safety-critical applications.
Frame rate represents the number of times radar transmits, receives, and processes RF signals within a single second. Each frame provides a complete environmental scan, with higher frame rates delivering more frequent updates that enable faster detection of changes and quicker responses to emerging threats.
The automotive industry has identified 20 frames per second as a fundamental requirement for autonomous driving systems. This frame rate provides environmental updates every 50 milliseconds, reducing perception gaps and enabling systems to build higher confidence levels for object detection and tracking in safety-critical scenarios.
At 130 km/h, vehicles travel approximately 36 meters per second. Traditional radar systems operating at 10-12 FPS create dangerous blind spots where critical changes like lane changes or appearing obstacles might occur between scans, making high frame rates essential for highway-speed autonomy.
The primary challenge is processing enormous data volumes at high speed. Each radar frame generates gigabytes of raw data that must be processed, filtered, and converted into actionable information within milliseconds. High frame rates also increase power consumption and heat generation, requiring advanced thermal management solutions.
Arbe’s proprietary processor architecture, built as custom intellectual property, delivers the processing power to handle 30 gigabytes of raw RF data per second, compressing it into a 1-gigabyte point cloud output. Innovative heat dissipation solutions and power-efficient processing enable sustained 20 FPS operation without excessive cooling requirements.
Legacy radar systems designed for basic functions like automatic cruise control and emergency braking can technically reach 20 FPS but lack the computational capacity to handle the data volumes and complexity required for autonomous operation. Without adequate processing power, faster transmission speeds become meaningless for complex driving tasks.
Urban environments present scenarios where objects appear and disappear within milliseconds: children stepping into crosswalks, cyclists emerging from behind parked cars, or vehicle doors opening unexpectedly. More frequent environmental updates enable immediate detection and response, building higher confidence in object detection and tracking.
Operating at high frame rates increases power consumption and heat generation. More frequent RF transmissions require more electrical power, while intensive real-time processing generates additional heat. Both the sensor and vehicle’s central processing unit must handle these demands without compromising reliability or requiring excessive cooling infrastructure.
High frame rate radar aligns temporally with cameras and other sensors, reducing perception latency and improving real-time object tracking and classification. Traditional radar at lower frame rates creates timing mismatches with high-frame-rate cameras, while 20 FPS provides detailed motion cues and dynamic scene understanding that improves fused sensor stack reliability.
Basic imaging radar (12×16 channels) operates around 16 FPS with limited processing capability. Legacy radar for driver assistance reaches 20 FPS but solves simpler problems without autonomous driving processing power. Arbe delivers 20 FPS with 10x richer imaging, superior processing capability, and the thermal management needed for sustained operation.
Perception radar is ultra-high-resolution 4D imaging radar technology specifically designed for autonomous driving. Unlike conventional radar that lacks resolution to separate closely positioned objects, perception radar detects and separates objects from short range to over 300 meters while maintaining performance in all weather and lighting conditions.
At 120 km/h (75 mph), vehicles cover more than 30 meters every second. This leaves minimal time for detection, decision-making, and safe maneuvering, making extended perception range and high-resolution object separation critical for autonomous systems operating at highway speeds.
Emergency braking requires nearly 100 meters, but comfortable, human-like braking begins closer to 300 meters from hazards. Extended range enables gradual braking instead of sudden stops, early lane changes to avoid obstacles, and gives other drivers time to react, preventing uncomfortable or dangerous emergency maneuvers.
Cameras offer detail but struggle in rain, fog, or darkness with limited range. LiDAR provides accurate 3D mapping but is limited by weather and range. Conventional radar works in all weather at long ranges but lacks resolution to separate closely positioned obstacles like a tire from an adjacent guardrail.
Systems must detect lost cargo at 150-200+ meters, recognize pedestrians at 200-300 meters, and identify traffic jams at 300 meters. Sensors that detect hazards only when emergency maneuvers are necessary risk being too late to prevent collisions or causing uncomfortable, frightening sudden stops.
Perception radar maintains ultra-high resolution consistently across weather transitions including rain, snow, fog, and darkness. Unlike cameras and LiDAR that may deactivate driver assistance features when needed most, radar’s immunity to environmental variability ensures safety features operate with full confidence and continuity in compromised visibility.
Tunnels present inconsistent lighting that causes optical systems to struggle, while traditional radar systems generate excessive reflections leading to false alarms and degraded performance. High-resolution perception radar maintains reliability in tunnels, filtering out false positives and sustaining automatic emergency braking and adaptive cruise control without interruption.
Free space mapping identifies drivable space by continuously assessing surroundings. Radar’s ability to model the environment enables vehicles to determine safe paths and plan maneuvers without interruption, which is essential for safe highway navigation and path planning at high speeds.
Lane changes at highway speeds require accurate orientation data and real-time tracking of neighboring vehicles. High-resolution radar provides precise object-level data including turn rate, even for partially occluded vehicles, enabling confident maneuvering during merges and overtakes without relying solely on optical sensors.
Perception radar repositions radar from backup sensor to essential perception tool by providing high-resolution environmental mapping at high speeds and long ranges regardless of external conditions. Its immunity to weather variability makes it the foundation for reliable highway-grade perception, not just a supplement to optical sensors.
LiDAR generates detailed 3D maps of a vehicle’s surroundings using light waves, enabling precise object classification across diverse scenarios. It excels at differentiating pedestrians, bicycles, and vehicles with high spatial resolution, and provides a wide field of view valuable for urban navigation, route planning, and collision avoidance.
LiDAR’s effectiveness diminishes dramatically in rain, fog, snow, smoke, and dust, as light waves scatter and absorb in adverse conditions. Range is limited to 150-250 meters, insufficient for highway scenarios requiring 300+ meters. LiDAR also lacks Doppler capability, cannot directly measure object speeds, and carries high production costs limiting mass-market adoption.
LiDAR performance suffers in harsh weather conditions including rain, fog, and snow, precisely when reliable perception is most critical. Its weather sensitivity compromises reliability in regions with unpredictable seasonal conditions, making it unsuitable as the sole perception sensor for autonomous systems that must operate safely in all conditions.
4D perception radar uses millimeter-wave signals rather than light waves, functioning as an independent data source that operates in all weather and lighting conditions. Unlike LiDAR, it directly measures object speed via Doppler capability, provides real-time free-space mapping, and detects low-reflectivity objects like pedestrians at night or road debris.
Arbe’s perception radar provides detection from short range to over 300 meters, exceeding the minimum highway coverage requirement. This surpasses LiDAR’s 150-250 meter limitation, enabling high-speed hazard detection and advance warning at distances required for safe autonomous highway operation.
Arbe’s perception radar maintains reliable detection and tracking in rain, fog, snow, darkness, and other conditions where cameras and LiDAR are compromised. Millimeter-wave signals are unaffected by weather conditions that scatter or absorb light waves, ensuring consistent safety-critical perception when optical sensors cannot provide advance warning of hazards.
Arbe’s perception radar is significantly more cost-effective and scalable than LiDAR, supporting broad adoption across automotive applications. The system integrates discreetly behind the vehicle’s bumper without structural alterations, allowing automakers to incorporate advanced sensing into mass-market vehicles without substantially increasing costs or compromising vehicle design.
Arbe’s perception radar provides real-time free-space mapping, identifying clear pathways and open areas surrounding vehicles. This delivers comprehensive environmental understanding for ADAS and autonomous vehicles, enabling effective path planning and safe navigation even in conditions where optical sensors are compromised.
Perception radar functions as an independent data source that enables a more comprehensive perception framework when paired with cameras. It supplies real-time object distance, speed, and direction data critical for safe navigation, providing the all-weather backbone that maintains system reliability when optical sensors are compromised by weather or lighting.
High-resolution radar provides cost efficiency, all-weather robustness, long-range detection from short range to over 300 meters, and direct speed measurement that LiDAR cannot match. Arbe’s 4D radar transcends traditional radar constraints, offering automakers a highly efficient alternative that accelerates widespread adoption of advanced driving technology across mass-market vehicles.
Five major trends are transforming the automotive industry: AI-powered autonomy with the automotive AI software market projected to reach $200 billion by 2032, regulatory advancements mandating enhanced emergency braking systems, software-defined vehicles enabling OTA updates, explosive radar market growth from $6.6 billion to $33.6 billion by 2030, and strategic sensor fusion reaching $3.3 billion by 2030.
AI-powered features including real-time object recognition and path prediction are revolutionizing vehicle safety and enabling self-driving capabilities. Advanced sensors provide rich, high-resolution data as vital input for AI algorithms, enabling more informed decisions especially in challenging weather and lighting conditions where other sensors struggle.
NHTSA’s 2029 vehicle safety standard mandates automatic emergency braking systems, including pedestrian AEB, as default features in all passenger cars and light trucks. Euro NCAP has set Vision Zero targets by 2030. Meeting these stringent standards requires sensors like Arbe’s 4D imaging radar with long-range, high-resolution, and all-weather capabilities.
Software-defined vehicles enable over-the-air updates that continuously improve performance and safety. Imagine that over 90% of traffic deaths could potentially be prevented through autonomous driving functions enabled by SDVs. Shorter software cycles allow rapid functionality upgrades, turning vehicles into smart platforms that benefit from latest perception algorithms.
The radar market is projected to explode from $6.6 billion to $33.6 billion by 2030, reflecting the technology’s critical importance. Automakers increasingly rely on high-resolution, all-weather radar sensors to improve safety and efficiency as demand for robust, reliable sensing solutions grows for ADAS and autonomous driving applications.
Sensor fusion strategically combines optimal sensor types rather than simply adding more sensors. The sensor fusion market is predicted to reach $3.3 billion by 2030. Multi-modal perception systems integrating radar, cameras, and other sensors create more comprehensive and robust environmental understanding essential for higher autonomy levels.
Perception radar generates rich, high-resolution data serving as vital input for AI algorithms. Arbe’s 4D imaging radar provides accurate information that enables AI to make informed decisions in challenging weather and lighting conditions at long ranges and high speeds, where cameras and LiDAR struggle, creating the synergy essential for true autonomous driving.
Meeting NHTSA 2029 AEB mandates and Euro NCAP Vision Zero targets requires sensors with long-range detection from short range to over 300 meters, high resolution for accurate object detection and tracking, and all-weather performance to ensure effective automatic emergency braking and pedestrian detection even in adverse conditions.
Software-defined radar architectures allow continuous improvement through over-the-air updates, enabling vehicles to benefit from latest advancements in radar processing and perception algorithms without hardware changes. This adaptability ensures radar systems evolve with emerging autonomous driving requirements, maximizing the safety potential of software-defined vehicles.
Radar provides reliable data even in challenging environments where cameras and LiDAR are compromised, making it an ideal complement for multi-modal sensing. Arbe’s high-resolution perception radar seamlessly integrates with other sensors, becoming the cornerstone of intelligent vehicle sensing that creates comprehensive perception systems essential for achieving higher autonomy and safety levels.
Studies show AVs demonstrate higher accident rates in dawn/dusk conditions compared to their overall safety performance. IIHS nighttime tests found over half of pedestrian AEB systems failed in nighttime conditions. High-resolution radar, such as Arbe’s, provides reliable perception in low-light environments where optical sensors struggle, addressing these critical perception gaps.
A comprehensive analysis of 2,100 AV accidents and 35,133 human-driven vehicle accidents found that AVs generally outperform humans in accident avoidance, precision control, and decision-making. However, AVs showed higher propensity for accidents specifically during dawn/dusk conditions and turning maneuvers, revealing areas requiring improved perception capabilities.
NHTSA’s updated standards mandate automatic emergency braking as default in all passenger cars and light trucks by September 2029. Vehicles must discern pedestrians and vulnerable road users in varying lighting from daylight to darkness. Arbe’s high-resolution radar maintains reliable performance regardless of ambient light, meeting requirements that optical sensors alone struggle to achieve.
Turning maneuvers require detailed analysis of the vehicle’s own localization and other object orientation across long ranges and wide fields of view. High-resolution radar with 2,304 virtual channels (48 transmit x 48 receive) resolves environmental sensing challenges during turns, providing the spatial resolution needed for safe navigation through complex maneuvers.
Arbe’s radar features 48 transmitting and 48 receiving antennas forming 2,304 virtual channels, vastly outpacing even the latest 16×16 imaging radars. This massive channel count delivers the level of detail necessary to significantly enhance perception algorithms, providing diverse sensing data including depth, relative velocity, object orientation, and long-range detection.
High-resolution radar delivers diverse sensing data including depth, relative velocity, object orientation, and long-range detection from short range to over 300 meters—capabilities that optical sensors alone struggle to match. This comprehensive data enables reliable perception in conditions where cameras and LiDAR fall short, particularly during dawn/dusk and adverse weather.
High-resolution radar provides free-space mapping over long ranges and wide fields of view, vital for detecting pedestrians, vehicles, and obstacles well ahead or during complex maneuvers like intersection turns. High spatial resolution in all dimensions, regardless of lighting or weather conditions, ensures real-time, reliable identification of drivable space.
Multipath suppression eliminates false detections caused by radar signals reflecting off multiple surfaces before returning to the sensor. Advanced radar integrating AI and perception achieves unparalleled image reliability through multipath suppression and false alarm elimination, translating into safer path planning through clustered point cloud detections for joint object tracking.
High-resolution radar complements camera-based systems by providing superior spatial resolution, excelling at long-range detection, measuring distance and relative velocity, and connecting data across multiple frames. Together these capabilities offer holistic understanding of the driving environment, making radar an indispensable component for real-time mapping and safe autonomous navigation.
Software-defined radar architecture allows ongoing updates and refinements without vehicle redesign as technology advances. Arbe’s perception radar exemplifies this adaptability, enabling continuous improvements to perception algorithms and sensing capabilities through over-the-air updates, ensuring AV safety continues to improve in real time as the industry evolves.
Tesla pursues a vision-only AI system using cameras and training data, while Waymo and Baidu employ sensor-heavy approaches combining radar, LiDAR, and cameras. This fundamental difference raises questions about the balance between scalability, technological advancement, and safety for commercial autonomous vehicle deployment.
Cameras struggle with poor visibility in low light, rain, fog, and snow, have significantly limited range compared to radar for detecting distant objects, and face challenges accurately determining distance and depth for unfamiliar objects. These limitations compromise system robustness in challenging real-world conditions.
Radar, LiDAR, and cameras each provide different information that complements one another, enhancing the system’s ability to detect objects, assess distances, and navigate various weather conditions. This sensor suite approach provides more reliable information helping AI make smarter, safer, swifter decisions.
AI processes vast amounts of information from sensors to make sophisticated decisions in demanding environments. Even advanced AI cannot overcome limitations of faulty or inadequate sensor information. High-definition radar, LiDAR, and cameras provide detailed, up-to-date environmental pictures that allow AI to make safer and more informed decisions.
Radar offers all-weather detection in precipitation and low visibility, longer range for spotting distant obstacles crucial for highway speeds, real-time depth perception without calculation latency, operation unaffected by lighting conditions, and Doppler capability for direct velocity measurement essential for predicting trajectories and potential collisions.
Radar’s Doppler effect measures relative velocity of objects directly, providing essential information for predicting trajectories and potential collisions of vehicles, pedestrians, and obstacles. Unlike cameras that rely on image analysis to estimate velocity, Doppler offers direct and accurate measurement enhancing timely, safe decision-making.
Expensive sensing platforms work for small fleets but mass-market deployment requires affordable sensors. The lack of cost-effective high-performance technologies drives companies to use fewer sensors, potentially compromising safety. Overall costs determine whether autonomous technology can compete commercially in real-world services.
Arbe’s perception radar delivers ultra-high-resolution imaging capable of detecting objects, pedestrians, and road obstacles in challenging conditions at an affordable price. It detects small objects like lost cargo at long ranges, provides free space mapping for AI systems, and offers the optimal performance-cost solution for mass-market autonomous deployment.
Free space mapping allows AI systems to determine where vehicles can safely move, providing detailed real-time environmental maps. Arbe’s perception radar represents the first radar-based technology offering this capability, enabling autonomous systems to identify safe navigation paths even in challenging weather and lighting conditions.
The future hinges on balancing cutting-edge AI with advanced sensor suites rather than choosing between them. AI provides intelligence while sensors offer the eyes and ears for safe navigation. Combining high-performance radar like Arbe’s with cameras and AI creates autonomous vehicles that are capable, reliable, and commercially viable.
More than 46,000 people died in preventable traffic crashes in 2023, with mileage death rates increasing 22% over pre-pandemic 2019. The automotive industry cannot afford to wait for full autonomy. Technology can significantly reduce accident risk and mitigate harm while human driving remains required, paving the way for secure, efficient transportation.
Perception radar offers deeper, more comprehensive environmental understanding compared to traditional radar systems. It provides unique benefits in challenging driving scenarios through high resolution, high sensitivity, and exceptional dynamic range, delivering reliable data that complements vision sensors in their weak spots including long range, all lighting conditions, and challenging weather.
Lost cargo like small objects blocking highway lanes poses critical safety concerns at high speeds. Cameras have limitations in long-range detection. Arbe’s perception radar detects small objects at distances over 150 meters even near larger reflective objects like vehicles or guardrails, enabling informed decisions about whether objects can be driven over or must be avoided.
Early detection of lost cargo allows vehicles to operate safely at speeds up to 130 km/h compared to current systems limited to 60 km/h maximum. This high-resolution radar capability enables highway autonomous driving at high speeds by providing sufficient advance warning for safe maneuvering at true highway velocities.
Perception radar provides reliable detections in all weather and lighting conditions with zero latency, empowering AEB systems to make informed decisions and execute necessary braking maneuvers. High-resolution, long-range capabilities enable early hazard identification, while exceptional reliability prevents both collisions and false alarms like phantom braking.
Autonomous Emergency Steering demands rapid decision-making when immediate braking is impossible in high-speed emergencies. Arbe’s perception radar provides critical long-range, high-resolution data to detect oncoming threats and assess lane availability, enabling vehicles to safely execute evasive maneuvers by detecting obstacles like motorcycles at significant distances with precise position and velocity information.
Detecting pedestrians at long distances in challenging conditions like darkness or adverse weather is critical for accident prevention. Arbe’s perception radar addresses this through exceptional sensitivity and dynamic range, identifying pedestrians at greater distances than traditional radar systems, empowering vehicles to execute safe path planning maneuvers that safeguard both pedestrians and occupants.
Tunnels and bridges pose unique challenges due to varying, rapidly changing lighting conditions and limited visibility. Traditional cameras struggle in these environments, creating safety risks. Perception radar overcomes light-sensitivity limitations when entering or exiting tunnels and provides exceptional resolution for navigating under bridges, reliably detecting obstructing vehicles or motorcycles from a distance.
High-resolution point cloud data supports accurate assessment of available driving space, allowing vehicles to safely navigate through underpasses without unnecessary braking or swerving. Combining high sensitivity and dynamic range, perception radar effectively overcomes camera limitations in varying tunnel lighting while reliably detecting and classifying objects that may obstruct roadways.
Beyond highway and urban driving use cases, perception radar improves or enables parking assistance, blind spot detection, and cross traffic alerts. While crucial for creating a future where self-driving vehicles navigate with confidence and precision, perception radar’s contribution to ADAS and L2+ features is available today to enhance current vehicle safety.
Radar measures speed and depth instantly and directly through the Doppler effect, whereas cameras must calculate these values, leading to processing delays and potential inaccuracies. This capability provides the essential long-range, all-weather, and all-speed perception required to transition driving responsibility from humans to technology safely.
Traditional radars suffer from low azimuth resolution and a lack of elevation separation, resulting in fewer than 5 detections per vehicle. These systems struggle to detect stationary objects and are prone to “phantom braking” because they cannot reliably distinguish between background clutter and actual obstacles on the road.
While Basic Imaging Radar (<300 channels) improves upon traditional systems, it still lacks the resolution for sophisticated object classification and free space mapping. Perception Radar (>1500 channels) provides ultra-high resolution with up to 10,000 detections per frame, enabling the detailed environmental mapping necessary for L2+ and higher levels of autonomy.
Due to limited processing resources and low resolution, traditional radars cannot differentiate between a stationary object (like a stalled car) and background clutter, often leading the system to ignore the object entirely. In contrast, Perception Radar reliably detects stationary obstacles at ranges of 150 meters and beyond.
Phantom braking occurs when a radar’s frequent false alarms cause a vehicle to brake unnecessarily for misidentified objects. Perception Radar virtually eliminates these false alarms through ultra-high resolution and advanced signal processing, ensuring the vehicle only reacts to genuine threats.
By utilizing optimal pulse repetition frequency and sophisticated antenna design, Perception Radars eliminate Doppler and range ambiguities. This allows the system to operate reliably in complex, “congested” environments with multiple targets and interference sources without misinterpreting data.
Traditional and basic imaging radars generally cannot; they detect an “object” without knowing its type. Perception Radar generates a detailed point cloud that allows AI-based algorithms to classify objects specifically as cars, trucks, pedestrians, bicycles, or motorcycles, even at long ranges.
Free Space Mapping identifies all drivable areas—essentially mapping the “empty” space where there are no obstacles. This is a critical requirement for safe path planning in hands-free and eyes-off driving, and it requires the ultra-high-resolution data that only Perception Radar provides.
Traditional radars provide limited, “noisy” data that is computationally expensive to fuse with cameras or LiDAR. Perception Radar produces high-resolution data that is readily compatible with other sensors, creating a more comprehensive and seamless picture of the environment for the vehicle’s AI.
Unlike older radar generations, Perception Radar is built for seamless integration with software-defined architectures. This allows carmakers to push over-the-air updates that unlock new functionalities, fix bugs, and ensure the vehicle meets evolving safety regulations like the NHTSA 2029 AEB standards.
Software-Defined Architecture is a design approach where a vehicle’s features and functions are primarily enabled through software, allowing it to evolve over time via updates. Unlike traditional designs that are fixed at the point of manufacture, SDA-compliant vehicles can receive new functionalities, performance improvements, and safety updates throughout their entire lifespan.
If a vehicle’s hardware is designed only for today’s basic requirements, it cannot support the advanced software of tomorrow, regardless of how many updates it receives. For example, a radar with a limited antenna array (like 16×16 Tx/Rx) is physically incapable of supporting L2++ or L3 autonomy; to upgrade the software, one would have to replace the entire sensor, which is commercially and technically impractical.
Arbe provides a hardware foundation with 48 transmitter and 48 receiver channels, supporting 2,304 virtual channels. This massive “headroom” ensures that as ADAS and autonomous software become more sophisticated, the radar hardware already installed on the vehicle has the resolution and processing capacity to handle those advanced algorithms without needing a physical replacement.
The Radar Application Framework allows Tier 1 suppliers and OEMs to customize radar performance for specific driving scenarios via a user-definable module. This flexibility enables manufacturers to optimize the radar for different modes—such as high-range highway sensing or dense stop-and-go urban traffic—and implement future features as consumer needs evolve.
Instead of relying on expensive, power-hungry general processors, Arbe uses a proprietary, optimized System-on-Chip (SoC) specifically built for perception radar. By implementing essential processing tasks directly into the hardware while keeping them software-configurable, Arbe delivers high-performance SDR that is both cost-effective and energy-efficient for mass-market deployment.
Yes. Because the architecture includes ample processing power and memory capacity, it can accommodate future algorithms that haven’t even been developed yet. The system is designed to hold multiple software versions during an update, which minimizes downtime and ensures the radar stays compatible with the evolving landscape of Software-Defined Vehicles (SDVs).
Industry leaders, including experts from Mercedes-Benz, have emphasized that the performance demands for reliable ADAS require more than 1,000 channels. Arbe’s architecture exceeds this benchmark, processing over 2,000 channels in real-time to generate tens of thousands of simultaneous detections, providing the high-resolution “image” necessary for safe autonomy.
SDR allows OEMs to offer new capabilities and safety improvements as “after-sales” services. Because the hardware is future-proof, manufacturers can sell performance upgrades or new autonomous features as paid software updates to vehicles already on the road, turning the sensor suite into a long-term revenue source.
The automotive-grade SoC integrates a Radar Processing Unit (RPU) with dual-core DSPs, microcontrollers, and a dedicated “lock-step” safety processor. This configuration allows the system to convert massive amounts of raw data into actionable insights in real-time while maintaining the low power consumption required for electric and high-efficiency vehicles.
By allowing vehicles to adhere to new regulations and safety standards (like the 2029 NHTSA AEB requirements) through software updates, SDR ensures that older vehicles on the road can benefit from the latest life-saving innovations. This continuous improvement is vital for reaching the goal of zero traffic fatalities.
The National Highway Traffic Safety Administration (NHTSA) has officially mandated that all passenger cars and light trucks must include AEB—including pedestrian AEB—as a standard feature by September 2029. This regulation requires systems to function effectively during both daylight and low-light conditions to significantly reduce rear-end and pedestrian collisions.
Euro NCAP’s Vision 2030 shifts the focus from traditional crash protection to proactive crash avoidance. The new rating criteria assess safe driving, crash avoidance, crash protection, and post-crash safety, with a specific emphasis on nighttime testing to ensure safety technologies are effective regardless of lighting conditions.
The mandates are stringent: vehicles must be capable of avoiding collisions at speeds up to 62 mph and detecting pedestrians in both daylight and darkness. Furthermore, systems must engage brakes for leading vehicles at speeds up to 90 mph and respond to pedestrians at speeds up to 45 mph.
While cameras are useful, they struggle with glare, low light, harsh weather, and long-range detection. Cameras also cannot measure speed directly; they must calculate it over multiple frames, which introduces latency that can compromise safety during high-speed maneuvers or sudden pedestrian crossings.
Perception Radar functions flawlessly in total darkness and adverse weather, providing surround-view awareness that cameras often lose in low-light. It can also detect “occluded” objects, such as a pedestrian hidden between parked cars who is about to step into the road, which is critical for meeting NCAP’s nighttime testing requirements.
Traditional radars and even many “advanced” radars lack the resolution and dynamic range provided by a 2,000+ channel count. Without this high resolution, a radar may fail to identify small objects—like a child or a bicycle—when they are near highly reflective objects like large commercial trucks.
Phantom braking is often caused by traditional radars misinterpreting data or being unable to distinguish background clutter from real threats. Perception Radar’s ultra-high resolution and advanced signal processing provide an accurate environmental map, virtually eliminating the false alarms that lead to unnecessary braking.
Sensor fusion combines the strengths of cameras (visual classification) with the strengths of Perception Radar (all-weather reliability, long-range detection, and instant speed measurement). This multi-sense approach is essential for achieving a 5-star safety rating and ensuring the vehicle remains compliant as regulations evolve toward 2030.
By providing the high-definition data required for proactive crash avoidance, Arbe’s perception radar serves as the foundational technology for Vision Zero—the industry goal of zero road fatalities. It empowers OEMs to meet and exceed strict global mandates, moving society closer to a future without traffic deaths.
Channel count directly determines a radar’s resolution and its ability to perceive the three-dimensional world. Low channel counts result in “blurry” data, whereas a massive channel array provides the high-definition “eyes” necessary for a vehicle to distinguish between closely spaced objects, detect small obstacles at long ranges, and accurately perceive elevation—all essential for safe hands-free and eyes-off operation.
While some manufacturers claim L3 capability with 192 or 256 channels, these systems often reveal reliability gaps that force OEMs to impose speed limitations. In complex environments, radars with fewer than 300 channels struggle to discern static objects or lost cargo, making them inadequate for the robust safety required for true eyes-off driving at highway speeds.
The spatial Nyquist criterion is a technical benchmark for signal sampling; meeting it requires a dense array (at least 32×32 channels). Radars that fail to meet this criterion suffer from high “sidelobes”—erroneous reflections that create multiple ghost images for a single object—which lead to the false alarms and “phantom braking” that plague lower-resolution systems.
Most current “imaging” radars are built by moderately expanding legacy 3×4 chips to reach 192 or 256 channels. Arbe’s architecture was designed from the ground up to support a massive 48×48 array, delivering 2,304 virtual channels—roughly ten times the capacity of systems touted as “advanced” by competitors.
Not necessarily. While using legacy chips to reach 2,000+ channels would be prohibitively expensive and hot, Arbe’s proprietary chipset achieves this massive array at a comparable cost, size, and power consumption to much lower-resolution systems. This is made possible by RF chips and a dedicated processor optimized specifically for high-density perception.
A massive channel array generates a point cloud with 100 times the density of traditional radars. This rich data stream provides the detailed environmental “image” needed for sophisticated AI algorithms to classify objects and map free space, rather than relying on mathematical assumptions to “fill in the gaps” left by low-resolution sensors.
You cannot simply scale up a legacy architecture once it hits a certain limit; reaching 1,000+ channels requires a fundamental shift in chipset technology. OEMs that “aim low” now risk being handicapped in the future, as they will have to redesign their entire sensor architecture to launch advanced features that require more data.
Leading experts, including Dr. Jürgen Dickmann of Mercedes-Benz, have stated that safe ADAS requires a minimum of 1,024 channels (a 32×32 array). Arbe’s solution exceeds this industry-recognized threshold today, providing 2,304 channels to ensure maximum safety and redundancy.
By integrating a massive channel array early, OEMs can collect high-quality, high-resolution data from real-world scenarios immediately. This robust data foundation allows manufacturers to roll out new autonomous features and safety improvements to vehicles already on the road via software updates alone.
Smaller array radars are not “obsolete”; they are “complementary.” While they aren’t meant to power front-facing L3 autonomy, they are ideal for corner and rear installations to provide a cost-effective 360-degree view, or as front-facing solutions for basic driver-assist features below L2+.
Dr. Jürgen Dickmann, Head of Radar and Radar-Perception at Mercedes-Benz Group, has publicly stated that advanced driver assistance systems (ADAS) require an array of no fewer than 32×32 channels. Furthermore, the antenna must be dense to avoid spatial ambiguity, a standard that underscores the necessity of high-channel-count technology for the future of automotive safety.
A March 2024 report from the Insurance Institute for Highway Safety (IIHS) revealed worrying results: out of 14 partial driving automation systems tested, only one earned an “acceptable” rating. Eleven systems were rated as “poor,” highlighting a critical reliability gap in the hands-off and eyes-off features currently hitting the roadways.
While radars with 12×16 or 16×16 configurations (192-256 channels) improve basic features like emergency braking, they still demonstrate significant reliability gaps. These limited arrays struggle with dynamic range and sidelobe reduction, often forcing manufacturers to impose speed limitations on hands-free features because the sensors cannot reliably detect small objects in complex environments.
To avoid inherent spatial ambiguity—where the radar misinterprets the location or existence of an object—the antenna array must be dense and meet specific physical criteria (λ/2 spacing). Massive channel arrays fulfill these requirements, ensuring that the data processed by the vehicle is accurate and free from the “ghost reflections” that plague traditional sensors.
Sensor fusion requires a robust, high-density point cloud to “match” the detail provided by cameras. A massive channel array radar provides 10s to 100s of detections per object, creating a data-rich environment that allows the vehicle’s AI to cross-reference sensor inputs with extreme precision, achieving the safety levels demanded by IIHS ratings.
Yes, but only with high-resolution perception radar. Arbe’s technology pinpoint smaller road users like pedestrians and cyclists—even in total darkness or adverse weather—and can accurately identify small obstacles and object boundaries from long distances where traditional radars would see only “blurry” interference.
Arbe’s proprietary chipset includes a 24-channel Tx chip and a 12-channel Rx chip that together create a 2,304-channel array. This is the largest channel array in the automotive industry, specifically designed to process massive amounts of data—up to 3Tb/s—to support L2+ through L5 autonomy.
Free Space Mapping is the ability of a sensor to identify the “drivable” area around a vehicle. Arbe’s perception radar offers the industry’s only radar-based free space solution that works at long range and with a wide field of view, regardless of lighting or weather conditions that would typically blind a camera.
By moving away from legacy “off-the-shelf” components and engineering a purpose-built radar chipset from the ground up, Arbe delivers a hundredfold increase in point cloud density while maintaining a competitive cost, physical size, and power consumption profile comparable to much lower-performing “4D” radars.
Recent automotive RFQs (Requests for Quotation) show that manufacturers are mandating higher channel counts for radar rather than relying on LiDAR to solve safety problems. Radar’s ability to work in all weather conditions, combined with the high resolution of a massive array, makes it the most reliable foundation for the next generation of safety.
A software-defined vehicle transitions from being a traditional machine to a sophisticated computing platform on wheels. Success in this area requires a “mindset change” where automakers equip cars with high-performance hardware and flexible, open architectures that allow for continuous software updates and the addition of features that have not even been conceived yet.
The modern automotive landscape is too complex for any single company to master alone. By partnering with nimble tech startups, traditional automakers can bridge the resource gap, combining their manufacturing prowess with innovative, cutting-edge ideas to accelerate the development of next-generation driving solutions.
Modern drivers are moving away from “horsepower-centric” metrics in favor of holistic experiences focused on safety, convenience, and enjoyment. Features like automated parking, hands-free driving, and seamless digital connectivity—which transform the car into a connected extension of a driver’s digital life—are now the primary drivers of consumer choice.
Building trust requires transparency and education. Automakers must go beyond simply installing technology; they must effectively communicate exactly how sensors and AI systems perceive the environment to enhance safety. This education is critical for overcoming consumer hesitation and fostering the widespread adoption of autonomous features.
China has emerged as a frontrunner in the development of next-generation transportation technologies. To remain competitive, global automakers must prioritize the rapid development of ADAS (Advanced Driver Assistance Systems) at Levels 2+ and higher, ensuring they launch vehicles with cutting-edge autonomous features in the immediate coming years.
Future-proofing involves installing advanced hardware today that has the “headroom” to support the software of tomorrow. By investing in high-performance infrastructure now, automakers ensure that their vehicles remain relevant and competitive throughout their entire lifecycle via over-the-air updates.
Arbe’s perception radar is designed with an open application framework, meaning it is built to readily accommodate the unique software solutions of different OEMs and Tier 1 suppliers. This flexibility allows manufacturers to integrate “best of breed” solutions and evolve their technology without being locked into a rigid, closed system.
Yes. Arbe’s high-resolution 4D imaging radar provides superior perception capabilities, allowing it to detect and classify objects with unmatched accuracy. This is particularly vital in challenging environments where other sensors might fail, providing the data-rich foundation needed for advanced ADAS and AD.
The industry is at a crossroads similar to the invention of the automobile a century ago. 2024 marks the point where the transition to software-defined mobility becomes a commercial necessity, requiring agility, collaboration, and a relentless focus on innovation to survive the shift.
McKinsey highlights that advanced technologies have “tremendous potential” to provide new levels of safety. Arbe fulfills this by providing the foundational sensing hardware (Perception Radar) that enables the “Vision Zero” mission—using technology to proactively prevent accidents and transform how people travel safely.
Originally developed for aircraft during WWII, radar first entered the automotive mass market in 1998 with the Mercedes-Benz S-Class to enable adaptive cruise control. Today, it has transformed from a simple distance-measuring tool into “Perception Radar,” a high-resolution sensor capable of human-like environmental understanding through AI.
Radar is the only sensor that directly measures both the distance and the relative velocity (speed) of an object instantly. While cameras must calculate these figures over multiple frames—leading to dangerous processing latency—radar provides near-zero latency data, allowing a vehicle to decide immediately whether to accelerate or brake.
In autonomous mode, simple alerts aren’t enough; the vehicle must make life-critical decisions. Traditional radars have low resolution, often seeing the world as “blurry.” To achieve true safety, radar resolution must be high enough to detect challenging targets like a child crossing the street, a stray tire on the road, or a pedestrian in heavy snow.
In radar technology, spatial resolution is directly proportional to the “channel count” (virtual antennas). While typical radars use roughly 192 channels, Arbe’s technology utilizes 2,300 channels. This ten-fold increase in hardware capacity translates directly into a perception-level image that rivals the detail of other high-end sensors.
No. Because cameras and radar have “orthogonal” strengths and weaknesses, a vision-only approach cannot solve the challenge of autonomy. True safety requires a sensor fusion approach where radar provides the all-weather, long-range redundancy that cameras lack, creating a “sweet spot” of overlapping sensor data.
Arbe leverages AI for “multipath suppression,” a process that identifies and removes erroneous signals (ghost reflections) that often cause traditional radars to trigger “phantom braking.” This AI-driven filtering ensures that the vehicle only reacts to real, coherent objects.
AI-powered clustering groups thousands of individual point cloud detections into single, coherent objects (like a specific car or a group of pedestrians). By analyzing an object’s speed and direction in real-time, the AI can predict where that object will be in a fraction of a second, allowing for proactive path planning.
Yes. Arbe’s Perception Radar can perform “SLAM” (Simultaneous Localization and Mapping). Using AI, the radar continuously maps the vehicle’s surroundings and localizes the car within that map, allowing it to estimate its own velocity and position even in areas with zero connectivity.
Free Space Mapping is the ability of the radar to identify “drivable” versus “occupied” space. AI algorithms process high-resolution radar data to create a detailed map of the road, telling the vehicle’s decision-making system exactly where it can safely move to avoid obstacles.
Arbe’s system shares data intelligently between multiple radars placed at the front, rear, and corners. AI enables “seamless overlap,” meaning an object of interest is smoothly tracked as it moves from the field of view of one radar to the next, validated by two distinct perception algorithms for maximum redundancy.
Sensor fusion is the process of integrating data from multiple sensor types—such as cameras, radar, and LiDAR—to run a unified perception algorithm. It is essential because no single sensor is perfect; by merging data, the system overcomes the individual limitations of each technology, providing the data redundancy and complementary strengths required for safe L2+ and higher autonomy.
+1
Cameras are susceptible to light saturation and white balance issues, such as the sudden glare when exiting a tunnel or the blinding effect of oncoming high beams at night. They also suffer from “partial observability,” where they may fail to recognize an object’s true shape or depth if it is partially obscured by a guardrail or other obstacle.
Cameras do not see depth natively and must calculate it using complex algorithms or stereoscopic mounting, which introduces latency. Radar, however, uses radio frequency technology to measure distance and relative velocity directly and instantly via the Doppler effect, providing immediate and accurate depth perception with near-zero latency.
Yes. Because radar RF signals can “wrap around” objects, imaging radar can maintain a complete map of the field of view even when a line of sight is partially blocked. This allows the system to detect and track objects that a camera might miss due to visual obstructions.
Unlike traditional radars that provide sparse data, imaging radar provides a rich, high-resolution point cloud similar to a camera’s detail. This allows it to not only detect objects but also confirm or correct camera detections and provide critical trajectory insights, such as “time to collision.”
High resolution in the elevation dimension allows Arbe’s perception radar to accurately identify road inclines, bridges, and overpasses. This helps the vehicle decide if it can safely drive under an object (like a high bridge) or over it (like a steep incline), a task that is notoriously difficult for cameras and low-resolution radars.
Arbe’s radar offers exceptional angular resolution and frame registration, reaching a level of detail that makes it possible to seamlessly merge its data with camera inputs. This dense data input enables reliable free space mapping and object boundary detection that lower-resolution radars cannot support.
Yes. While cameras are the traditional leaders in classification, the detailed data from Arbe’s perception radar, coupled with AI algorithms, allows the system to classify objects (pedestrians, cars, trucks) based on radar data alone. This provides a vital failsafe if the camera becomes unavailable due to weather or lighting.
In scenarios like a dark tunnel or heavy fog, where cameras struggle and traditional radars often ignore stationary targets, Arbe’s perception radar can reliably detect stopped obstacles, such as an emergency vehicle parked on the shoulder.
The goal is to achieve the “Vision Zero” mission of zero road fatalities. By leveraging the visual precision of cameras and the all-weather, long-range reliability of Arbe’s perception radar, automakers can ensure that the vehicle has a comprehensive and redundant understanding of its environment in every possible scenario.
The U.S. Department of Transportation (DOT) and NHTSA have established a Federal Motor Vehicle Safety Standard mandating that all passenger cars and light trucks be equipped with Automatic Emergency Braking (AEB) and Pedestrian AEB. This landmark ruling is projected to save at least 360 lives and prevent 24,000 injuries annually by requiring vehicles to autonomously intervene when a collision is imminent.
To comply with the mandate, automakers need sensors that perform across four critical areas: long-range hazard identification (up to high highway speeds), stationary object detection, the ability to identify small objects/pedestrians, and the intelligence to differentiate between actual obstacles and non-threats like bridges or manhole covers.
Arbe’s perception radar operates at an industry-leading range of 350 meters. This extended reach allows the AEB system to detect vehicles, cyclists, and pedestrians much earlier than traditional sensors, providing the necessary “time-to-react” for the system to warn the driver or engage the brakes safely at higher speeds.
Standard radars often suffer from low dynamic range, meaning a highly reflective object (like a large commercial truck) can “blind” the sensor to smaller, nearby objects. Arbe’s high-channel-count radar provides the high dynamic range necessary to distinguish a pedestrian or cyclist even when they are positioned right next to a massive metal vehicle.
A major cause of consumer distrust in AEB is unnecessary braking for non-hazards. Arbe’s high resolution in elevation allows the radar to accurately discern that a bridge is “above” the driving path and a manhole cover is “below” it. This precise mapping supports better “drive vs. brake” decisions, eliminating false alarms.
Yes. While most radars on the market rely on movement (the Doppler effect) and struggle to see stationary targets, Arbe’s Perception Radar can detect static objects and precisely identify their boundaries. This allows the vehicle to recognize a stalled car or lost cargo as a hazard and stop or steer accordingly.
Arbe’s massive channel array provides 100 times more detail than traditional radars. With 1° of separation, the radar can move beyond seeing a “blurry” reflection to identifying the specific shape of smaller objects like tires, scooters, and pedestrians, which is essential for the “comprehension” required by the DOT.
Yes, if powered by perception radar. Unlike cameras, which can be blinded by weather or low light, Arbe’s radar is designed to maintain reliable object detection and collision avoidance performance in rain, thick fog, and total darkness, ensuring the AEB system is always “on.”
Arbe uses real-time signal processing algorithms to analyze massive amounts of data instantly. This near-zero latency is critical for AEB, as it enables the vehicle to identify and respond to a potential collision risk in fractions of a second—a speed that is physically impossible for systems that rely on multi-frame camera calculations.
The DOT mandate requires a rare combination of long range, high resolution, all-weather reliability, and stationary object detection. While other sensors might hit one or two of these requirements, Arbe’s perception radar is the only technology that fulfills the entire checklist in a single, production-ready package.
Front-facing radar is responsible for real-time ego-motion calculation (the vehicle’s own speed and turn rate), tracking and separating surrounding objects, and mapping stationary obstacles. Crucially, it provides high-resolution Free Space Mapping, which identifies drivable areas in both horizontal and vertical dimensions—a non-negotiable prerequisite for autonomous path planning.
While cameras provide visual coverage, they cannot directly measure speed and depth, and their functionality is severely degraded by darkness, glare, and adverse weather. Arbe’s 360° radar-based perception works in all environmental conditions, transforming “comfort” features like lane change assist into high-stakes safety features by providing reliable, all-weather data that cameras cannot match.
In an emergency steering scenario, the vehicle must instantaneously decide where to move. This requires a 360° view to estimate the distance, speed, and orientation of all surrounding objects—including vehicles approaching from behind. Radar-based Free Space Mapping tells the car exactly where its safe options are, whether that is the next lane or the shoulder.
Object sharing occurs when multiple radars with overlapping fields of view track the same object simultaneously. This redundancy allows the system to validate an object’s location through different perception algorithms, which reduces false alarms and provides higher resilience in “occlusion” scenarios where one sensor’s view might be blocked.
A single, independent sensor must track an object over several frames to achieve a high level of confidence. By contrast, multiple radars working together can draw accurate conclusions about classification, trajectory, and velocity much faster. This collective intelligence allows the system to detect threats and initiate safety responses in a fraction of the time.
Multipath refers to “ghost” reflections caused by radar signals bouncing off metal surfaces or other targets. While a single radar might take several frames to realize a reflection isn’t behaving like a real car, a 360° system compares the data across multiple radars to recognize unusual patterns instantly, allowing the system to reject false objects and maintain data integrity.
Cognition Mode allows the radar to adjust its own parameters in real-time based on the specific driving task. For example, if the car is preparing for a left-hand turn, the side radars automatically reconfigure to prioritize long-range detection, ensuring the system can spot distant oncoming traffic to determine a safe turning window.
A large metal truck can sometimes “blind” a radar to smaller nearby objects, like a pedestrian, due to its strong signal amplitude. In response, Arbe’s radar can automatically adjust its mode, range, or sensitivity threshold to prevent saturation, ensuring the system can still separate and track the smaller target next to the larger one.
Ultra-high resolution adds unique depth and velocity information that optical sensors cannot provide. It is the only technology detailed enough to offer reliable redundancy for cameras in conditions like heavy fog or blinding light, providing the “sensing diversity” required to move from basic driver assistance to full autonomous path planning.
By perfecting 360° perception and object-sharing capabilities, Arbe elevates automotive technology from “nice-to-have” comfort solutions to life-saving safety infrastructure. This complete and coherent environmental comprehension is the breakthrough necessary to eliminate road fatalities and achieve full autonomy.
NVIDIA DRIVE Sim is an end-to-end simulation platform that allows Arbe to test and validate its perception radars in a physically based, multi-sensor virtual environment. By integrating its Phoenix and Lynx radar models into the platform, Arbe can accelerate its time to market, improve productivity, and ensure its sensors are more thoroughly tested than would be possible through real-world driving alone.
Arbe’s capabilities—such as SLAM (Simultaneous Localization and Mapping) and super-resolution—require massive amounts of data for AI training. Industry standards suggest data must be generated 180 times faster than a sensor’s real-time acquisition rate. DRIVE Sim provides the necessary data volume and quality to meet these “mind-boggling” training demands.
Testing scenarios like accidents, red-light runners, or a child bolting into the street is difficult and often unethical in the real world. DRIVE Sim allows Arbe to simulate these rare and hazardous “corner cases” safely and repeatedly, ensuring the radar is evaluated against the highest safety standards from every possible angle.
DRIVE Sim leverages NVIDIA’s core technologies, including NVIDIA RTX (for realistic rendering), NVIDIA Omniverse, and advanced AI. This combination creates a cloud-based environment capable of generating highly realistic synthetic radar data that closely mimics the performance of actual hardware.
Arbe’s sensor models are built directly into DRIVE Sim, allowing customers to train and validate their own perception algorithms even before physical hardware is available. This reduces the need for thousands of expensive test drives and simplifies the integration process for vehicle manufacturers.
Because DRIVE Sim includes synchronized LiDAR and camera sensors in every virtual scenario, Arbe can train and format its radar data to align perfectly with these other sensor types. This ensures that the final sensor fusion—combining radar, camera, and LiDAR data—is as seamless and accurate as possible.
While cameras and LiDAR are important, radar is the only sensor that remains fully functional in all weather and lighting conditions, including heavy rain, fog, snow, and dust. By adding ultra-high resolution to this inherent reliability, Arbe repositions radar as the primary infrastructure for autonomous safety.
The blog highlights Arbe’s Phoenix and Lynx imaging radars. These models provide the perception-level data that is integrated into the NVIDIA DRIVE Sim environment for large-scale, multi-sensor simulation.
Yes. DRIVE Sim allows developers to introduce “randomization” into scenes, such as unexpected weather changes or varied lighting conditions. This helps Arbe discover and solve for “corner cases”—rare events that might otherwise be missed in standard testing but are critical for L2+ and L3 safety.
The partnership aims to “supercharge” the development of autonomous technology. By providing a risk-free environment for training, development, and validation, the two companies are working to accelerate the deployment of safe autonomous vehicles and advance the future of transportation.
The Insurance Institute for Highway Safety (IIHS) found that more than half of vehicles tested earned a “basic” score or no credit at all when detecting pedestrians at night. Most current systems struggle to identify pedestrians until they are illuminated by the vehicle’s headlights, which is often too late to avoid a collision.
Headlights have a limited range that, depending on vehicle speed, is often shorter than the safe stopping distance. Furthermore, the light can be insufficient or create a “washout” effect, making it difficult for optic sensors (cameras) to distinguish vulnerable road users from the background.
The “sensor gap” occurs when the shortcomings of both the radar and the camera overlap. In the IIHS tests, the cameras were blinded by darkness, and the traditional radars lacked the resolution to classify pedestrians correctly, leading to a total failure to protect road users at night.
Unlike cameras, Perception Radar is unaffected by lighting conditions. It operates with equal precision in bright daylight, pitch-black night, and adverse weather like snow or rain. It provides highly detailed Free Space Mapping at long ranges, allowing the vehicle to detect pedestrians far ahead or on the side of the road long before they enter the headlight beam.
False positives occur when a sensor “sees” an object that isn’t there, leading to phantom braking. False negatives occur when a vehicle detects a real person but mistakenly filters them out as a “ghost” or noise. Both errors are common in radars with sparse antenna arrays that lack the density to produce unambiguous data.
Arbe’s Perception Radar relies on a dense, unambiguous array that captures a high-resolution image in a single frame. This minimizes ghost targets and ensures low-latency detection, which is critical for the split-second decisions required for Automatic Emergency Braking (AEB).
A vehicle must distinguish between objects that look similar from a distance, such as a road sign, a tree, or a person. Without high resolution in the elevation (vertical) dimension, a car might mistake a tree for a person and stop needlessly, potentially causing a rear-end accident. Elevation data allows for correct classification and avoids “senseless collisions.”
No. Arbe is focused on making cutting-edge radar technology affordable for the mass market. The goal is to ensure that true safety is available for every vehicle class, not just elite models, which is essential for earning widespread consumer trust in autonomous technology.
High-profile accidents involving autonomous features have made consumers distrustful. To earn that trust back, the industry must integrate sensors that prevent these accidents consistently. Perception Radar is identified as the only sensor capable of achieving this level of safety at a mass-market price point.
The findings prove that current “camera-heavy” or “low-res radar” suites are insufficient for 24/7 safety. The industry must move toward Perception Radar to close the sensor gap, ensuring that vehicles can perceive and understand their environment in any condition, at any time.
Lynx is a Surround Imaging Radar designed for corner and back installation. Unlike traditional corner radars, which have low resolution and are limited to basic tasks like blind-spot alerts, Lynx provides high-resolution imaging, long-range performance, and a wide field of view. This allows it to be included in sensor fusion and perception, helping the vehicle identify stationary objects and map drivable “free space.”
While Phoenix serves as the “gold standard” primary front/rear sensor, Lynx is its ultimate partner. When installed together, they can be fully synced for unified perception and interference avoidance. This creates a harmonious 360-degree safety cocoon around the vehicle, ensuring that multiple radars work together to understand the driving environment coherently.
Every other sensor has a significant drawback:
Lynx packs massive performance into a small form factor (77x67x38 mm):
Merging into fast-moving highway traffic requires sensing high-speed vehicles coming from the back and sides. Lynx provides the long-range detection and “Doppler” (instant speed) data needed to track these vehicles accurately, allowing the car to make safe lane changes and merges that cameras or basic radars would struggle to manage.
Yes. Navigating intersections involves detecting lateral (side-to-side) traffic and vulnerable road users like cyclists. Lynx’s wide 140° field of view and high resolution allow it to detect exact object boundaries and create a reliable “Free Space Map,” making unprotected turns significantly safer.
Lynx is optimized for both. While it is a powerhouse for L2+ and higher autonomy, it can also be used as a standalone front-facing solution for basic ADAS applications. Its 12×24 RF channels represent a significant upgrade over existing ADAS sensors, providing a higher level of safety for any vehicle class.
Arbe designed the Lynx processor to be open and flexible. OEMs can add their own current and future algorithms—such as trackers or super-resolution—directly onto the radar processor without needing additional hardware. This allows the vehicle’s perception to evolve through software updates throughout its lifecycle.
When Phoenix and Lynx are installed together, they utilize unified perception and interference avoidance technology. This ensures that the various radar units on the vehicle do not jam each other, maintaining a clean and accurate data stream even when the environment is “noisy” with other radar signals.
Lynx proves that high-performance imaging doesn’t have to be limited to the front of the car or to luxury price points. By delivering perception-level data in a small, affordable package, Lynx enables 360-degree safety for the mass market, accelerating the transition to a world with zero road fatalities.
According to industry expert Alexander Hitzinger, mobility in 2030 will be safer (due to widespread ADAS), cleaner (as electrification goes mainstream), and more convenient (with expanded options like micromobility and highway autopilot).
CEO Kobi Marenko predicts that delivery robots will become a common sight, dramatically reducing delivery times and increasing accessibility. Long-haul trucking will also become increasingly autonomous, improving the safety and efficiency of transporting goods to our homes.
Level 2+ is expected to reach approximately 70% market penetration by 2030. The emphasis will shift from “comfort features” to “safety features,” driven by both consumer demand and stricter safety regulations.
Probably not. The panelists noted that full autonomy in complex urban areas is a “long tail” problem with too many edge cases. While commercial pilots will exist, widespread mainstream adoption of urban Robotaxis is not expected by 2030.
While cameras and LiDAR play roles, high-resolution perception radar is essential for reaching the safety levels required for mass-market autonomous features. It is the only sensor that provides the necessary redundancy to work in all weather and lighting conditions at a scale that is cost-effective for mainstream vehicles.
As drivers spend less time “holding the wheel,” the car will become an extension of the entertainment and advertising ecosystem. Expect larger screens, video streaming, gaming, and potentially even commercials integrated into the cabin experience.
New technologies typically penetrate the market first through “premium” products where buyers are less cost-sensitive. As volume increases, costs decrease, eventually allowing the technology to become a standard feature in mainstream, affordable vehicles.
The panelists expect “car-as-a-service” to take significant market share, especially in crowded cities where parking is difficult. Consumers may own a basic vehicle for daily use but “hire” a Level 3 or Level 4 autonomous car for weekend trips to enjoy a hands-off driving experience.
The automotive industry operates on long cycles. Technology selected today is for model years 2025 or 2026. It then takes several more years to ramp up from high-end models to the mainstream. Therefore, the radars and chips being chosen right now are the ones that will define the market in 2030.
The Chinese market moves significantly faster than the West. Processes that take five years in Western markets typically take only three years in China, meaning Arbe expects to see revenues and technology adoption from China much earlier than from other regions.
Arbe made significant progress with Tier-1 partners and major OEMs. Notable achievements included a joint pilot with Hyundai Mobis showcased at the EcoMotion event and the engagement of five new customers across Level 2+ and Level 4 mobility sectors.
Tier-1 manufacturers are developing radar systems based on Arbe’s chipset. In Q1 2022, these partners submitted five RFPs and RFQs to major automakers, with projected volumes ranging from 400,000 to 1 million units per year, signaling strong future demand.
Arbe strengthened its board and management with three heavyweights from the automotive and tech sectors:
Arbe introduced Free Space Mapping to its perception stack. This was a world-first for automotive radar, providing the high-accuracy environmental mapping necessary for autonomous navigation and redundancy.
The final production configuration of the chipset increased detection range to 800 meters (approx. half a mile). This makes it particularly valuable for the trucking industry and delivery robots, which require long-range sensing for safe operation at scale.
Arbe received two major accolades:
Operating expenses rose to $11.1 million (up from $4.5 million in Q1 2021). This was primarily due to increased labor costs in R&D as the company moved toward production, as well as general overhead costs associated with being a public company.
The company held $87.3 million in cash and cash equivalents. Total debt was reported at $5.1 million, which Arbe expected to pay off by July 2022.
Arbe reiterated its guidance for 2022 (revenue of $7M–$11M) and confirmed it remains on track to reach its $312 million revenue goal for 2025.
Arbe has delivered the first combination of ultra-high radar native resolution and design flexibility in a mass-production configuration. This chipset provides 12 to 20 times more channels than best-in-class radars currently at retail, resulting in the highest radar image quality on the market.
The chipset consists of a transmitter chip with 24 channels and a receiver chip with 12 channels. When integrated with Arbe’s processor, it creates a 48×48 channel array (2,304 virtual MIMO elements), providing the high-dynamic range needed to eliminate angular ambiguity and false alarms.
Using single-channel transmission, the radar can detect vehicles at over 300 meters. However, by utilizing beamforming(simultaneous transmission from multiple channels), the detection range increases to a staggering 800 meters for vehicles and 350 meters for vulnerable road users (VRUs).
In traditional 12×16 radars, using channels for beamforming drastically reduces the MIMO array size, which kills resolution and increases “sidelobes” (interference). Because Arbe has a massive 48×48 array, it can use beamforming with minimal impact on resolution, maintaining a clear image while extending range.
Arbe successfully reduced the power consumption per chip by 50% for both the transmitter and receiver. This makes it the most efficient solution in the industry in terms of “power consumption per channel,” a critical factor for electric vehicles (EVs) and complex sensor suites.
The new receiver chip features on-chip temperature auto-calibration. This allows the radar to maintain perfect calibration and performance stability across the entire automotive temperature range, from -40°C to 150°C (-40°F to 302°F).
The receiver has a noise figure of 11 dB, the best in the industry. A lower noise figure means higher sensitivity, which is what allows the radar to “see” very small, low-reflection objects like a flat tire on a dark highway or a distant motorcyclist.
The receiver is designed to meet the highest automotive standards, including ASIL-B, AEC-Q100, and is both Automotive Grade 1 and 2 ready, ensuring it meets the rigorous reliability requirements of global OEMs.
The high resolution and beamforming capabilities allow for precise environmental mapping at close ranges. This enables delivery robots to navigate sidewalks safely, identifying small obstacles and pedestrians with the precision needed for urban operation.
The chipset enables the advanced perception required for unprotected left turns, T-junctions, and highway merging. By providing forward and backward “Free Space Mapping,” the radar tells the car exactly where it is safe to move in these high-risk scenarios.
Humans cannot look forward and backward simultaneously, and even with mirrors, the brain struggles to process high-speed information from multiple directions. Autonomous sensors, specifically Imaging Radar, overcome this “human shortcoming” by providing a continuous, unified understanding of the entire environment without gaps or “guesses.”
Adding a rear-facing radar isn’t just about pointing a sensor backward. It requires doubling the workload for tracking, classification, and free-space mapping, and adjusting algorithms to distinguish between forward and backward motion. The goal is a “unified presentation” where the whole is greater than the sum of its parts.
If you are stopped in traffic, rear-facing radar can detect if a trailing vehicle is gaining too quickly or is too close. This data allows the perception system to anticipate a potential rear-end collision and inform the vehicle’s next moves to avert disaster.
Free Space Mapping identifies drivable vs. non-drivable areas. Rear-facing radar is critical for this when merging, exiting highways, or changing lanes. It creates a “superpower” nearing clairvoyance—anticipating how the environment will shift as vehicles pass and prepare to cross back into your lane.
In heavy traffic or at stoplights, front-facing radar may be occluded by other vehicles. Rear-facing radar provides additional stationary reference points from the environment behind the car. This diverse data ensures that the estimate of the vehicle’s own speed (ego velocity) remains accurate and doesn’t “diverge” from reality.
Yes. Imaging Radar’s ultra-high resolution allows it to identify obstructions and occlusions that other sensors might miss. By processing rear data alongside front data in a single, elaborate calculation, the system achieves a much higher level of accuracy in mapping the surroundings.
Through Vehicle-to-Vehicle (V2V) communication, a vehicle with rear-facing radar can share its data with the car behind it. Since Arbe’s radar has a 300m+ range, two connected vehicles can effectively “stack” their ranges, allowing the trailing car to “see” hazards nearly a kilometer away.
This involves building an accumulating map of all stationary environmental detections. By sharing localization information and high-resolution radar reflections, vehicles can create an extended free-space map, which makes autonomous path planning significantly more efficient.
To achieve zero traffic fatalities, a vehicle must perceive hazards from every direction. Overlapping and complementary detections from the front and back of the vehicle create a failsafe “cocoon” of data, making Perception Imaging Radar indispensable for true safety and autonomy.
No. While it is vital for highway merging, it is equally important for low-speed maneuvers like reversing and navigating dense urban environments where pedestrians or cyclists may approach from behind or from blind spots.
Optical sensors rely on the visible light spectrum, much like the human eye. In winter, light waves are easily absorbed or scattered by rain, fog, and snow. This can reduce a camera’s effective range from 150 meters down to just 20 meters, rendering it nearly useless in heavy precipitation.
Unlike light waves, the radio waves used by radar sensors penetrate environmental obstacles like fog, heavy rain, smoke, and dust. This allows radar to maintain its detection range and sensitivity even when a driver (or a camera) can no longer see the road.
On sunny winter days, glistening snow creates thousands of tiny specks of reflected light. Optical sensors may mistakenly identify these reflections as physical objects (targets), causing the vehicle to brake suddenly. This “phantom braking” increases the risk of being rear-ended by trailing vehicles.
No. Radar is immune to sun glare, which frequently saturates or “blinds” cameras. Additionally, while mud or ice on a sensor can disable a camera, radar-based sensors are significantly less sensitive to surface blockages and can often maintain consistent operation even when the sensor housing is dirty.
While traditional radar is stable in bad weather, it can still suffer from false alarms, such as reflections from snow-covered ground. Arbe’s Perception Imaging Radar uses ultra-high resolution to filter out these irrelevant signals, accurately sorting real hazards from environmental noise.
Most radars only see in 3D (range, azimuth, and speed). Arbe’s 4D Imaging Radar adds a vertical (elevation) dimension. This allows the car to distinguish between a pothole hidden by snow and a bridge obscured by fog, determining exactly which objects the vehicle must avoid.
When optical sensors fail, Arbe’s radar provides the necessary redundancy to maintain autonomous features. It can calculate Ego-Motion (the car’s own speed/turn rate), track surrounding objects, map stationary obstacles, and perform Free Space Mapping—all in total darkness or heavy snow.
Off-road environments are often filled with mud, stone dust, and water spray. These conditions are lethal to cameras but manageable for 4D Imaging Radar. This ensures that heavy machinery can operate safely around workers even in the most “blind” industrial environments.
Yes. The goal of perception radar is to be safer than a human driver could ever be. By seeing through obstacles that disable human vision and providing 360-degree awareness without the risk of “snow blindness,” it strives to eliminate winter-related accidents entirely.
It combines the inherent weather-resistance of radio waves with the high-definition detail of an optical sensor. This unique combination ensures that the vehicle’s “eyes” never fail, regardless of how extreme the winter weather becomes.
The event, titled “The Road Ahead for Imaging Radar,” was designed to educate investors and analysts on the rapidly expanding imaging radar market. It featured insights from industry experts on how 4D Imaging Radar is revolutionizing automotive safety and transforming perception for autonomous driving.
The event featured a high-profile roster of experts from across the automotive and technology ecosystem, including representatives from General Motors (GM), Ghost, Qamcom, and the global research firm Frost & Sullivan, alongside Arbe’s executive leadership.
Industry experts highlighted that 4D Imaging Radar is not just an incremental improvement but a foundational technology that allows
Tier-1s and automotive OEMs to achieve a level of perception previously impossible with traditional radar. It is viewed as a key driver for the mass adoption of autonomous applications.
As independent industry analysts, Frost & Sullivan provided a macro view of the fast-growing imaging radar market, identifying it as a critical component in the transition from basic driver assistance (ADAS) to higher levels of vehicle autonomy (L2+ and beyond).
The event detailed how Tier-1 suppliers use Arbe’s ultra-high-resolution data to build more robust safety applications. This includes better detection of stationary objects, superior performance in all weather conditions, and the high-definition mapping required for emergency braking and collision avoidance.
CEO Kobi Marenko provided a strategic update on Arbe’s business activities and long-term projections. His presentation emphasized Arbe’s role as the market leader in the imaging radar space and outlined the company’s roadmap for scaling production and securing design wins with global manufacturers.
Traditional radar “senses” objects, but Arbe’s 4D Imaging Radar “perceives” the environment. This means it provides the rich, detailed data—including elevation and velocity—needed for the vehicle to actually understand its surroundings, rather than just detecting reflections.
No. While it supports full autonomy, the event emphasized the immediate application in enhancing current vehicle safety. By making 4D Imaging Radar affordable and scalable, Arbe is enabling advanced safety features across a broad spectrum of passenger and commercial vehicles.
The discussion focused on how 4D Imaging Radar fills the gaps left by cameras and LiDAR, particularly in challenging lighting and weather. It provides the necessary redundancy to ensure that autonomous systems remain safe 100% of the time, not just in optimal conditions.
Arbe has made the full recording of the event and the transcript of CEO Kobi Marenko’s presentation available to the public, providing transparency into the company’s strategic vision and financial outlook.
Arbe officially announced its status as a public company on October 8, 2021. Shortly after, on October 12, 2021, CEO Kobi Marenko rang the closing bell at the Nasdaq MarketSite in New York City.
Arbe is listed on the Nasdaq under the ticker symbol $ARBE.
CEO Kobi Marenko was joined by his co-founders Noam Arkind (CTO) and Oz Fixman (COO), along with Arbe’s team, partners, and investors. He also shared a moving tribute to late co-founder Amos Baron, noting that his memory remains a blessing and a driving force for the company.
Since its beginnings in a small subleased office in Tel Aviv, Arbe’s slogan has been “We make miracles.” Marenko credits the team’s passion and hard work for turning the “miracle” of camera-like radar into a commercial reality.
As Marenko highlighted in his remarks, Arbe has created a radar that generates a “camera-like picture” of the environment but maintains its performance in any weather or lighting condition—a feat traditional optical sensors cannot achieve.
The reach extends far beyond passenger cars. Arbe is engaged with autonomous trucking companies, robo-taxidevelopers, and industrial sectors like construction and agriculture. They are also eyeing futuristic applications like delivery robots and smart city infrastructure.
Following the IPO, Marenko outlined three key focus areas:
Marenko emphasized that while the Nasdaq listing is a seminal milestone, it actually marks the beginning of a new phase of innovation. The public capital and visibility are tools to “race full speed ahead” toward the goal of full vehicle autonomy.
By becoming a public company, Arbe has secured the resources necessary to scale its perception radar globally. This scale is required to make high-definition safety technology standard in every vehicle, which is the only way to achieve Vision Zero (zero traffic fatalities).
Marenko described the experience as a “rush” and an “honor,” viewing it as a celebration of the collective brilliance and tireless commitment of the entire Arbe family and their support systems.
Humans are inherently flawed drivers, and while we often accept human error, there is a near-zero tolerance for mistakes made by machines—especially after billions of dollars have been invested in safety. To gain public trust, autonomous vehicles must achieve “superhuman” performance, effectively eliminating the possibility of accidents.
Like the superhero Spider-Man’s intuitive sense of danger, 4D Imaging Radar provides a vehicle with the ability to “see” hazards that are not yet visible to the human eye or standard cameras. It can detect danger lurking around corners or hidden by other objects, providing a proactive safety net.
Human drivers don’t just use their eyes; they use a combination of senses and intuition. Relying only on optical sensors (cameras) is risky because light-based systems fail in darkness, heavy rain, or fog. A true level of safety requires a “multi-sensory” approach that includes radar.
Unlike standard radar, 4D Imaging Radar can:
They create a “superhuman” sensory duo.Cameras provide high visual precision and color/text recognition (like reading signs), while 4D Imaging Radar provides depth, velocity, and all-weather reliability. Together, they give the vehicle’s AI the “perfect information” needed to make life-saving decisions.
The industry operates on the principle that a robot (the car) must not injure a human or allow a human to come to harm through inaction. Because the expectations for safety are so high, the sensor suite must be infallible, moving far beyond the capabilities of a typical human driver.
While it cannot literally see through solid brick walls, it can detect reflections and use its ultra-high dynamic range to spot objects partially obscured by obstacles—effectively giving the car the ability to “foresee” scenarios that would blind a human driver.
No. These “super senses” are just as critical for urban driving. They help a car navigate complex intersections, spot pedestrians stepping out from behind buses, and manage the unpredictable “edge cases” of city traffic.
By outperforming human senses in every category—range, reaction time, 360-degree awareness, and all-weather visibility—superhuman sensors make the goal of zero road fatalities a technical possibility.
By merging the strengths of different technologies, the vehicle creates a redundant and diverse “sensory envelope.” If one sensor is compromised (e.g.,a camera blinded by the sun), the others fill the gap, ensuring the car never drives “blind.”
Seeing is simply detecting a point of data. Understanding, achieved through Object Tracking, is the ability to identify targets and create a continuous description of their movement. Tracking tells the vehicle not just that something is there, but where it came from, where it is going, and at what speed.
+1
Free space mapping identifies drivable vs. non-drivable areas (occupancy grids), but it doesn’t account for object dynamics. To drive safely, a car needs to know if a “filled” space is a stationary guardrail or a speeding cyclist about to cross its path. Tracking provides the “velocity vector” missing from a simple map.
Cameras are the industry standard for classification (knowing a “car” is a “car”), but they struggle with:
Traditional radars struggle to distinguish height, often confusing a bridge (safe to drive under) with a stalled car (must stop). Arbe’s 4D radar tracks in elevation, allowing it to “see” that a train on an overpass is 20 feet above the road, preventing unnecessary and dangerous emergency braking (phantom braking).
Because Arbe has an order of magnitude better spatial resolution than traditional radar, it can attribute “tight” bounding boxes to objects. This means the radar doesn’t just see a “blob”; it sees the length, width, and orientation of a vehicle, which is critical for understanding if a car is merging or braking.
Yes. By using frame-to-frame memory and velocity data, the tracking algorithm can “predict” the path of a vehicle even if it passes behind a sign or another truck, minimizing the time it takes to “re-find” and actively track it once it reappears.
Human drivers naturally track surroundings to anticipate hazards. By automating this with 100x more detail than legacy systems, Arbe’s radar allows the vehicle to make smoother, more natural decisions—like slowing down gradually for a merging car rather than jerky, last-minute adjustments.
As of January 2026, Arbe has partnered with NVIDIA to integrate this ultra-HD radar data (20,000+ detections per frame) into the NVIDIA DRIVE platform. This combination allows for “Hands-Off, Eyes-Off” highway driving by using the radar’s tracking data to power AI-based occupancy grids that mimic human intuition at speeds up to 130 km/h.
A standard FMCW radar is designed for basic tasks like Adaptive Cruise Control (ACC) and Emergency Braking (AEB). It “senses” presence but cannot “visualize” the scene. Imaging Radar evolves this into a high-resolution sensor capable of separating, localizing, and classifying multiple objects (pedestrians, bikes, tire debris) using dense point clouds.
Many vendors use software algorithms (synthetic/statistical resolution) to “guess” better performance from limited hardware. While these work in simple labs, they fail in real-world “corner cases” like reflective environments or high-traffic scenes. Arbe relies on High Physical Resolution—actual hardware channel count—to ensure reliability that doesn’t rely on assumptions.
In radar, resolution is a product of the number of transmitting (Tx) and receiving (Rx) channels. Arbe’s chipset uses 48 of each, creating 2,304 virtual channels. This is an order of magnitude higher than competitors (who often max out at 192 virtual channels), providing the spatial density needed for true 2K imaging.
To avoid “ghost” detections called grating lobes, antenna elements must be spaced at half the wavelength ($\lambda/2$).
Stationary objects (like a stalled car under a bridge) are the “Achilles’ heel” of traditional radar. Because Arbe has high resolution in the elevation dimension, it can distinguish between a bridge 5 meters above the road and a hazard directly on the pavement, preventing unnecessary emergency braking.
More channels mean a massive explosion of data. Processing 2,304 channels in real-time requires a throughput of over 3 Terabits per second (Tbps). Arbe solved this by developing a dedicated, patented processor (Everest) designed specifically to handle this data volume with low power consumption.
Yes. Due to its 1° azimuth resolution, Arbe’s radar can provide object orientation and boundaries. It doesn’t just see a single “blob” for a group of vehicles; it can “slice” the scene to see a motorcycle riding parallel to a large truck or a pedestrian standing next to a guardrail.
By eliminating both false positives (braking for nothing) and false negatives (missing a real hazard), 4D imaging radar provides the redundancy needed for L3 and L4 autonomy. It serves as the “eyes” of the car when cameras are blinded by sun glare, fog, or darkness.
As of late 2024 and 2025, Arbe’s technology has been integrated into Tier-1 systems (like HiRain’s LRR610) and selected for L4 vehicle programs. The chipset is designed to be cost-effective and compact enough for standard automotive integration.
As showcased at CES 2026, Arbe has now integrated its 2K radar with the NVIDIA DRIVE platform. This allows the radar’s 20,000+ detections per frame to feed directly into AI-based occupancy grids, enabling “hands-off, eyes-off” driving at true highway speeds (up to 130 km/h).
Free Space Mapping is the process of distinguishing drivable areas from non-drivable obstacles. It creates an “occupancy grid” that serves as the foundation for a vehicle’s navigation, path planning, and obstacle avoidance. Without a reliable estimate of empty space, an autonomous vehicle cannot safely determine its trajectory.
While cameras are excellent at identifying textures and boundaries, they are “contrast-based” and lack native depth perception. This makes it difficult for them to accurately measure the distance to a frontal target—especially in low-light, glare, or heavy weather—which is a critical requirement for safe path planning.
Radar provides highly accurate, direct distance measurements and is completely unaffected by lighting or weather conditions (rain, fog, or darkness). By pairing the camera’s visual detail with the radar’s reliable depth sensing, the vehicle gains a redundant, fail-safe understanding of the road.
LiDAR offers high resolution (0.1°) but struggles in adverse weather and carries a significantly higher cost. Furthermore, driving decisions require wide safety margins where 0.1° vs 1° resolution has a negligible impact. Therefore, Imaging Radar is seen as the primary reliable partner to the camera, with LiDAR acting as a mid-range backup.
Traditional radars have a low resolution (5° or worse), which causes objects to appear much larger than they are, blurring their boundaries. They also lack elevation data and often discard stationary objects (like guardrails or parked cars) as “clutter,” making it impossible to generate a reliable occupancy grid.
Arbe’s 4D Imaging Radar excels at detecting stationary objects. It can accurately map road geometry, curvature, and fixed hazards like concrete walls or vehicles parked in-lane. It treats the stationary environment as vital data rather than noise to be filtered out.
Traditional radars suffer from high false-alarm rates, leading to “phantom” braking. Arbe’s high channel count (2,304 virtual channels) provides superior spatial separation. This allows advanced post-processing to reduce false alarms to nearly zero, ensuring the vehicle only reacts to real threats.
The 2D occupancy map doesn’t just use the current frame; it includes a memory component from previous frames. This increases the confidence level of the map, “washing out” transient noise or multipath reflections and maintaining awareness of areas that might be briefly occluded (hidden).
Even though the final map is 2D (a bird’s eye view), the system needs the third dimension (elevation) to build it accurately. Without elevation data, a radar cannot distinguish between a road sign (drivable underneath) and a construction barrier (must avoid). Elevation resolution is what makes the 2D map trustworthy.
As of January 2026, Arbe has integrated its 2K ultra-HD radar with NVIDIA DRIVE. This combination uses AI-based Occupancy Grids to process over 20,000 detections per frame, enabling “eyes-off” driving at highway speeds (up to 130 km/h) by providing a highly detailed, real-time map of the vehicle’s surroundings.
Early autonomous vehicles were bulky and relied on three distinct sensors with conflicting pros and cons:
Cameras: High resolution but fail in poor weather/lighting.
Standard Radar: Robust in bad weather but suffer from very low resolution.
LiDAR: High detail and accuracy but struggle in environmental conditions like fog or heavy rain.
The challenge was finding a single technology that combined high resolution with all-weather reliability.
Arbe’s 4D Imaging Radar was built from scratch to combine the best traits of all three sensors. It offers superior resolution and object detection in all environmental conditions, providing the “unmatched image quality” necessary for the higher levels of autonomy (L3, L4, and L5).
It is the first to provide high sensitivity, high resolution, and full spatial sensing—including elevation. This prevents the “doppler ambiguity” and lack of vertical awareness that cause traditional radars to miss obstacles like a truck under a bridge or a tire in the road.
Yes. High resolution allows the radar to reliably detect pedestrians and bicyclists, even in dense urban environments or at night. This precise detection is critical for resolving the safety gaps found in current-generation ADAS and autopilot systems.
By providing exact boundaries of obstacles and distinguishing between a motorcycle and a nearby truck, the radar provides trustworthy data. This allows perception algorithms to make confident decisions, preventing the vehicle from stopping needlessly (phantom braking) while ensuring a fast response time to real threats.
Ego-velocity is the vehicle’s own speed and turn rate. Because radar measures Doppler (velocity) directly, it can calculate the vehicle’s motion with extreme precision, even in scenarios where GPS or wheel-speed sensors might be less reliable.
As of the latest 2026 specifications, the platform supports over 100,000 detections per frame. This creates a point cloud density that is two orders of magnitude higher than traditional radar, rivaling the detail of LiDAR while maintaining radar’s all-weather strengths.
The channel count (48 Tx and 48 Rx) is the “engine” behind the resolution. It allows the radar to “see” the world in 2K ultra-high definition, enabling the vehicle to accurately identify small hazards—like a tire or a traffic cone—at distances up to 350 meters.
Yes. In January 2026, Arbe announced a partnership with NVIDIA to integrate this ultra-HD radar with NVIDIA’s AI computing. This combination enables “eyes-off” driving at highway speeds (up to 130 km/h) by providing the reliable, human-like perception required for consumer trust.
By providing the point cloud density of LiDAR at a cost-effective price point, it serves as the backbone of the sensor suite. It offers the failsafe redundancy needed to navigate complex urban and highway environments without the visibility limitations that plague other sensors.
The long-haul trucking industry has struggled for over 15 years to recruit new drivers. Younger generations are often unwilling to spend months away from home. Automation offers a solution by making the job less stressful and refocusing the driver’s role on “first and last-mile” logistics rather than exhausting highway shifts.
L4 autonomy allows trucks to drive themselves “exit-to-exit” on highways. Local drivers or teleoperators only take over for the complex urban navigation between highway exits and distribution centers. This model significantly reduces labor costs, improves fuel efficiency, and alleviates driver fatigue.
Trucks are larger, heavier, and require much longer distances to brake safely. While a car might need to see 150 meters ahead, a heavy truck traveling at highway speeds needs reliable detection at 300 meters or more to execute smooth, safe stops or lane changes.
While early autonomous trucks were covered in bulky equipment, Arbe’s 4D Imaging Radar provides “LiDAR-like” detail in a compact, automotive-grade chipset. It offers the high resolution needed for autonomy without the aesthetic or maintenance drawbacks of exposed, spinning sensors.
Traditional radars often trigger “phantom braking” because they can’t tell the difference between a manhole cover and a stalled car. Arbe’s radar uses ultra-high resolution (1° Azimuth, 2° Elevation) to determine exact boundaries, ensuring the truck only reacts to real obstacles like lost cargo or pedestrians.
Ego-velocity is the truck’s ability to sense its own speed and turn rate independently. Because Arbe’s radar directly measures Doppler (velocity), it provides a critical fail-safe for in-lane localization and path planning, even if other sensors like GPS are compromised.
As of the latest 2026 highway-safety standards, Arbe’s radar delivers a raw point cloud of over 20,000 detections per frame (and up to 100,000 in specialized configurations). This density allows the truck’s AI to identify small, high-risk objects—like a fallen tire—at long range.
Yes. Because 4D radar creates its own high-definition map of free space in real-time, it allows trucks to operate “highway autopilot” features even in geographies that haven’t been meticulously pre-mapped or in zones with shifting road geometry.
As of January 2026, Arbe has active partnerships with major Tier-1 suppliers and OEMs in the United States, Europe, China, and Japan. This includes a high-profile collaboration with a prominent European truck manufacturer to integrate these chipsets into their next-generation vehicle platforms.
By integrating with the NVIDIA DRIVE platform, Arbe’s radar data is processed by powerful AI computing. This enables “eyes-off” highway driving at full speeds (up to 130 km/h), providing the human-like predictability and safety required for large-scale commercial deployment.
Object orientation refers to the exact heading and direction of an object in a vehicle’s environment. It is the difference between knowing a truck is in the next lane and knowing it is currently angling toward your lane to merge. Orientation is the key to safe lane changes, intersection navigation, and predicting the flow of a dynamic road scene.
Most sensors, particularly cameras, require “multi-frame analysis” to estimate orientation. This means they must observe an object over several frames (time) to deduce where it is headed, creating dangerous latency. 4D Imaging Radar provides this data in a single frame by analyzing the Doppler distribution across the object.
Standard radars produce a “sparse” point cloud with only a few dots per car. Arbe’s ultra-high-resolution radar detects hundreds of points on a single vehicle. This density allows the system to see the “Doppler gradient”—the minute differences in speed across different parts of the car—to determine its exact orientation.
By comparing the Doppler (velocity) on the left side of a car versus the right:
Micro-Doppler refers to the secondary Doppler shifts from moving sub-parts, such as spinning wheels or a pedestrian’s swinging limbs. These tiny signals can reveal an “intent” to move—like wheels beginning to turn for a lane change—before the entire body of the vehicle actually shifts, giving the autonomous system a crucial head start.
When a radar has high enough spatial resolution, it can detect the “minimal bounding box” or “L-shape” of a vehicle (the side and the rear). Seeing this physical shape serves as a high-confidence confirmation of the heading direction derived from the Doppler data.
Stereo cameras require massive computational power and expensive dual-camera setups to estimate depth and orientation. Radar orientation is “model-based” and straightforward, utilizing native physics (Doppler) to get the result with far less processing and zero latency.
Yes. Arbe’s dedicated processor is designed to analyze the full scene simultaneously. It provides orientation, speed, and positioning for hundreds of environmental objects at the same time, ensuring the vehicle has a 360-degree understanding of every hazard.
As showcased at CES 2026, Arbe’s orientation data is now integrated with NVIDIA AI Computing. This allows the “intent detection” from Micro-Doppler to feed directly into advanced perception stacks, enabling the “human-like flow” required for hands-off, eyes-off highway driving at speeds up to 130 km/h.
In a split-second highway scenario, waiting even 100 milliseconds for a camera to “process” a lane change can be the difference between a safe stop and a collision. By providing orientation at zero latency, 4D Imaging Radar gives the vehicle’s “brain” the fastest possible information to make life-saving decisions.
Ego-velocity is the measurement of a vehicle’s own speed and direction. In autonomous driving, this is the “ground truth” needed to understand the world. Without precise ego-velocity, a car cannot tell if an object is moving toward it because the object is fast or because the car is moving—making safe navigation impossible.
SLAM is the process of building a map of an unknown environment while simultaneously keeping track of the car’s location within that map.
Optical sensors (Camera/LiDAR) are “static” per frame. To find speed, they must compare Frame A to Frame B and calculate the difference over time. This creates latency. Radar, however, uses the Doppler Effect to measure the velocity of every single point instantly in a single frame.
By identifying stationary objects in the environment (like the road, signs, or walls), the radar can measure its relative speed against them. Because it sees thousands of these points simultaneously, it can calculate its own velocity with extreme precision every 1/20th of a second, without needing to wait for the next frame.
To calculate true forward speed ($v$), the radar uses the geometric relationship:
The Arbe Everest chipset processes data from 2,304 virtual channels to ensure that even in “sparse” environments (like a smooth tunnel), it finds enough stationary points to maintain a perfect ego-velocity lock. This prevents the “drift” that causes other autonomous systems to become disoriented.
Yes. This is called Radar Odometry. Because the radar can calculate its own velocity and direction independently of any external signal, a car equipped with Arbe technology can navigate through a long tunnel or underground parking structure with high precision, purely by “feeling” its motion relative to the walls.
As of January 2026, Arbe’s high-precision ego-velocity data is fed directly into the NVIDIA DRIVE AGX Orinplatform. This allows the AI to “anchor” its visual perceptions (from cameras) to the rock-solid physical velocity data from the radar, resulting in smoother, more human-like braking and acceleration.
Unlike GPS, which struggles near 0 km/h, 4D Imaging Radar remains accurate even at crawling speeds. It provides continuous, jitter-free velocity data, which is essential for “Traffic Jam Assist” features to move the car smoothly without jerky “start-stop” movements.
The Development Platform (or “A Sample” in automotive terms) is a production-ready system that allows Tier-1 suppliers and OEMs to integrate Arbe’s 2K ultra-high-resolution radar into their vehicle’s “brain.” It moves the technology from a lab theory to a real-world tool that developers use to build the next generation of autonomous perception algorithms.
The difference is an order of magnitude:
It directly addresses the “blind spots” of traditional radar that have led to high-profile ADAS accidents. Because of its resolution, it can:
The platform includes:
Yes. It features a software layer that abstracts hardware access, allowing developers to focus on high-level perception tasks like ego-velocity inference, in-lane localization, and free space mapping without having to manage the raw radar signals manually.
The platform is designed to support over 100,000 detections per frame. In real-world highway scenarios, it typically delivers a raw point cloud of over 20,000 detections, providing the “point cloud density” needed for AI-based post-processing and deep learning.
Yes. As of January 2026, Arbe has fully integrated this platform with NVIDIA accelerated computing. This allows the radar to provide “eyes-off” capabilities at highway speeds (up to 130 km/h) by delivering the reliable, human-like flow and predictability required for true L3 autonomy.
Arbe is collaborating with over 20 Tier-1 and OEM customers across the United States, Europe, China, Korea, and Japan. A notable milestone for 2026 is the selection of this platform by a major Chinese state-owned automaker for their Level 4 autonomous vehicle program, with mass production starting in December 2026.
Yes. Through advanced Free Space Mapping, the platform uses its high-resolution data to create a real-time occupancy grid. This allows the car to “see” exactly where the road is clear, even in thick fog, heavy rain, or total darkness where cameras would be blinded.
By delivering LiDAR-like performance (0.7° Azimuth, 1.2° Elevation) at a radar price point, Arbe has repositioned radar from a “support” sensor to the primary backbone. It provides the essential redundancy that ensures the vehicle is never “blind,” regardless of the environmental conditions.
Traditional radar forced a compromise: you could have a wide field of view with poor detail, or high detail in a very narrow slice. Arbe’s technology breaks this trade-off by providing ultra-high resolution (1° Azimuth) across a wide 100° field of view, delivering the “optical-like” sensitivity of a camera with the “all-weather” reliability of radar.
Arbe’s chipset generates a high-resolution 4D image with an internal data rate equivalent to more than 2,000 Gbps (and processing throughput up to 3 Tbps). This allows the system to process massive amounts of environmental data in real-time, far exceeding the “sparse” point clouds of legacy radar.
Yes, and it often exceeds them. While market-leading optical sensors (cameras) typically reach around 100 meters in challenging conditions, Arbe’s 4D Imaging Radar reaches up to 350 meters. This long-range capability is essential for safe braking and lane-changing at high highway speeds.
Legacy radars often trigger “phantom braking” because they can’t distinguish between a metallic manhole cover and a stalled vehicle. Arbe uses a low detection threshold to catch everything, then employs advanced post-processing, tracking, and calibration to filter out noise. This ensures the car only reacts to genuine hazards.
Arbe’s radar is roughly 10 times more sensitive than traditional automotive radars. This allows it to detect “weak” reflectors—like a pedestrian in dark clothing or a bicycle—even when they are standing next to “strong” reflectors like a metal fence or a large truck.
Because 4D Imaging Radar is so detailed and reliable, it can perform tasks that previously required multiple overlapping sensors. By serving as the “backbone,” it allows OEMs to simplify the sensor suite, reducing the total number of cameras or expensive LiDAR units needed.
In January 2026, Arbe announced a major integration with NVIDIA DRIVE. This combined Arbe’s ultra-HD radar data with NVIDIA’s AI computing to create a platform for “eyes-off” driving. For the first time, this allowed vehicles to handle complex highway scenarios at speeds up to 130 km/h with human-like predictability.
As of December 2025, a major China-based state-owned automaker selected Arbe’s chipset (via Tier-1 partner HiRain) for its Level 4 autonomous vehicle program. With a Start of Production (SOP) in December 2026, this marks one of the world’s first large-scale deployments of ultra-HD radar in a fully autonomous fleet.
The “Cool Vendor” title is awarded to companies with innovative, impactful, and “cool” technologies that have the potential to disrupt their respective industries. For Arbe, this recognition validates their mission to redefine road safety and sets them apart as a key player in the autonomous vehicle (AV) ecosystem.
Developing a solution for autonomous vehicles is not just about a single sensor; it’s about how that sensor integrates into a massive, complex system of software, compute, and other hardware. Gartner’s report emphasizes that vendors must be ready to both compete and partner within this high-stakes environment to succeed.
Arbe’s 4D imaging radar directly tackles three of the biggest hurdles in the industry:
By providing high resolution in both azimuth (horizontal) and elevation (vertical), the radar can tell the difference between a car on the road and a bridge overhead. This spatial separation allows the vehicle’s AI to make much more accurate and safer decisions across all levels of autonomy (L2–L5).
The 2020 Gartner recognition proved to be a reliable early indicator of Arbe’s trajectory. By January 2026, Arbe has moved from a “Cool Vendor” to an industry standard-setter, evidenced by its major partnership with NVIDIA to power “eyes-off” highway driving on the NVIDIA DRIVE platform.
As of February 2026, Arbe’s 2,304-channel array remains the industry benchmark. It provides up to 100 times more detail than traditional radar systems, allowing for a point cloud so dense that it can distinguish between a child and a nearby fire hydrant at distances exceeding 300 meters.
Yes. Following the path paved by this early industry recognition, Arbe’s Tier-1 partner, HiRain Technologies, recently secured a major contract with a China-based state-owned automaker. The LRR610 radar, powered by Arbe’s chipset, is scheduled for Start of Production (SOP) in December 2026 for a Level 4 autonomous vehicle program.
By providing reliable sensing in conditions where cameras and LiDAR often fail (heavy rain, fog, and total darkness), Arbe’s radar acts as the ultimate safety backbone. It ensures that the vehicle’s perception system is never “blind,” which is the foundational requirement for zero-fatality goals.
Building on the momentum of the Gartner award, Arbe was recently named the “Sensor Technology Solution of the Year” in the 2025 AutoTech Breakthrough Awards. This confirms their continued leadership in the perception radar space five years after being named a “Cool Vendor.”
The “Geek out” mentioned in the original blog post refers to the technical breakthrough of building a dedicated radar processor from the ground up. In 2026, that “geeky” engineering foundation is what allows Arbe’s radar to run AI-based occupancy grids and human-like path planning that other radars simply cannot handle.
Launched as the first automotive-grade (AEC-Q100) dedicated imaging radar processor, it was built specifically to handle the massive data flow of 4D imaging. Unlike general-purpose chips, it is optimized to process raw data from 2,304 virtual channels (48 Rx $\times$ 48 Tx) while staying within the strict power and heat constraints of a vehicle.
The patented processor manages a staggering 30 Gbps of raw radar data (with an equivalent internal throughput of up to 3 Tbps). It converts this data into a dense point cloud of over 100,000 detections per frame at 30 frames per second—all with near-zero latency.
The chip isn’t just a data “cruncher”; it is a perception engine. It integrates:
As of January 2026, the latest generation (Everest) serves as the “brain” for systems like the HiRain LRR610. It delivers a raw point cloud of 20,000+ points per frame, allowing vehicles to distinguish between small hazards—like a lost tire or a child near a guardrail—at ranges exceeding 300 meters.
A major milestone at CES 2026 was the announcement that Arbe’s processor is now fully integrated with NVIDIA accelerated computing. The dense radar data is processed on the NVIDIA DRIVE AGX Orin platform, enabling “eyes-off” driving at highway speeds (up to 130 km/h) by feeding AI algorithms with high-fidelity 4D imagery.
Designed with a “safety-first” architecture, the processor is ISO 26262 compliant. It carries an ASIL B qualification, but its architecture allows it to be integrated into systems targeting ASIL D—the highest level of automotive safety integrity required for fully autonomous (L4/L5) driving.
Yes. One of its most “revolutionary” features is its ability to mitigate mutual radar interference. As more cars on the road use radar, their signals can jam each other; Arbe’s processor includes proprietary algorithms to filter out these “noisy” signals, ensuring the car only “sees” its own environment.
By integrating complex algorithms (like SLAM and tracking) directly into a single, low-power System-on-Chip (SoC), Arbe eliminates the need for expensive, power-hungry external CPUs or FPGAs. This makes high-resolution 4D radar affordable for mass-market vehicles, not just luxury robo-taxis.
As of early 2026, the processor is the core of production programs for:
Level 4 vehicles require “exceptional resolution and reliability” in complex urban traffic. The Arbe processor is the only dedicated chip capable of generating the density of data needed for AI to make human-like decisions in rain, fog, or darkness without the high cost of multiple LiDAR units.
Traditional radar measures three dimensions: Range(distance), Doppler (velocity), and Azimuth (horizontal angle). 4D Imaging Radar adds the critical fourth dimension:Elevation (vertical angle). This allows the vehicle to “see” the world in 3D space with speed data, preventing it from confusing a bridge overhead with a stalled car in its path.
While cameras have millions of pixels, they fail in bad weather. LiDAR has high precision but is expensive and struggles in fog. 4D Imaging Radar bridges this gap. Arbe’s technology provides 1° Azimuth and 1.7° Elevation resolution. As of January 2026, Arbe’s newest systems deliver a raw point cloud of 20,000+ detections per frame, providing a “LiDAR-like” image that works in all weather conditions.
Standard radars often see a “blob” when a motorcycle is riding next to a large truck. High-resolution 4D radar can physically separate these two objects. This allows the car to track the motorcycle’s independent path, providing the “situational awareness” required to avoid the types of accidents that have plagued early autonomous prototypes.
Human drivers and cameras typically struggle to identify small hazards beyond 100 meters at night or in rain. Arbe’s radar reaches up to 350 meters (over 1,100 feet). At highway speeds of 130 km/h (80 mph), this gives the vehicle’s computer significantly more time to react, brake, or steer around a hazard like a lost tire or a stopped vehicle.
Traditional radars are often too “sensitive” to metallic noise, causing the car to brake for manhole covers or soda cans. Arbe uses the industry’s lowest detection threshold combined with advanced AI-based occupancy grids. This allows the system to filter out random noise while keeping a “lock” on real threats, ensuring the radar data is finally trustworthy enough for the car to act on independently.
In many scenarios, yes. For Level 3 “eyes-off” highway driving, 4D radar provides the high-definition mapping and redundancy needed to move forward without a LiDAR unit. This is a massive cost-saver, as it helps automakers reach the goal of a sub-$1,000 sensor suite for mass-market vehicles.
The industry has moved past the Proof-of-Concept (PoC) phase. In December 2025, a major China-based state-owned automaker selected Arbe’s chipset for its Level 4 autonomous vehicle program, with Start of Production (SOP) scheduled for December 2026. This marks the shift from “testing” to “deployment” at a massive scale.
Announced at CES 2026, Arbe’s radar is now fully integrated with NVIDIA accelerated computing (specifically the NVIDIA DRIVE AGX Orin platform). This synergy allows the radar’s dense point cloud to be processed by AI in real-time, enabling “human-like” flow and “eyes-off” capabilities on the highway.
No. Because of its high inertia and long braking distances, autonomous trucking is one of the biggest adopters. The 350-meter range is even more critical for a 40-ton truck. Additionally, the technology is being used in “SMART city” initiatives in Sweden and China to monitor intersections and improve pedestrian safety.
Cameras can be blinded by the sun or rain, and LiDAR can be blocked by fog. Radar is the only sensor that performs 100% of the time, day or night, in any weather. By providing high resolution, it no longer just “supports” the other sensors—it becomes the primary, unshakeable foundation that the entire autonomous system relies on.
As Kobi Marenko explains, cameras and LiDAR are limited by environment (darkness, snow, fog), and legacy radar has historically been limited by poor resolution. Arbe’s 4D imaging radar is the “missing link” because it combines the all-weather reliability of radio waves with the high-resolution “vision” previously only possible with optical sensors.
By removing the resolution bottleneck, the radar no longer just “assists” the camera; it becomes the primary source of truth. It provides 4D data—distance, height, depth, and speed—functioning 100% of the time, regardless of whether it is high noon or a midnight blizzard.
Arbe is the first to leverage this advanced manufacturing process for radar. It allows for a breakthrough in performance (100x more detail) while simultaneously achieving the lowest power consumption and lowest cost per channel in the industry, making high-level autonomy affordable for mass-market vehicles.
While most radars use 12 virtual channels, Arbe’s Phoenix platform supports 2,304 virtual channels. This provides a physical angular resolution of 1° in azimuth and 2° in elevation, allowing the radar to see “pixels” of range and velocity at 30 frames per second.
Legacy 2D radars cannot tell if a stationary object is a car in the lane or a bridge overhead. Arbe’s high-resolution elevation sensing allows the vehicle to “see” the vertical gap, ensuring it brakes for a stalled truck while smoothly driving under an overpass or into a parking garage.
Low-resolution radars rely almost entirely on Doppler (movement) to see things. This is why some AVs have famously hit parked fire trucks. Arbe’s 4D resolution tracks all objects simultaneously regardless of speed, ensuring that stationary hazards are never “filtered out” as background noise.
False alarms (phantom objects) happen when radar noise is mistaken for a car. Arbe achieves near-zero phantom objects through superior channel separation and a patented processor that filters noise in real-time. This eliminates the “false positives” that irritate drivers and the “false negatives” that cause accidents.
As of January 2026, Kobi Marenko announced a major milestone at CES: the integration of Arbe’s raw 4D data with NVIDIA accelerated computing. This allows the radar’s dense point cloud (20,000+ points per frame) to power AI-based occupancy grids on the NVIDIA DRIVE platform for true “eyes-off” driving.
Yes. In December 2025, Arbe confirmed its first major Level 4 production deal with a China-based state-owned automaker. Using the LRR610 radar (powered by Arbe’s chipset),the first fleet of thousands of vehicles is scheduled for Start of Production in December 2026.
By detecting vulnerable road users (pedestrians and cyclists) even when they are partially concealed, and by providing a 350-meter safety buffer at highway speeds, Arbe’s technology is designed to eliminate the primary causes of ADAS-related accidents and drive toward a reality of zero road fatalities.
“Vision Zero” is an international safety initiative, championed by the European New Car Assessment Program (NCAP), aiming for zero road fatalities or serious injuries. To earn a 5-star safety rating in 2026, vehicles must demonstrate advanced protection not just for passengers, but for Vulnerable Road Users (VRUs) like pedestrians and cyclists.
Legacy sensors have significant “blind spots”:
It is the only sensor that provides a “continuous vision” of the environment 100% of the time. With a 100° wide field of view and a 350-meter range, it captures the environment in 4D (range, velocity, azimuth, and elevation). This ensures there are no gaps or blind spots, even in the harshest weather conditions.
A critical requirement for Vision Zero is the ability to distinguish a pedestrian from nearby objects. Arbe’s radar can compute the physical distance between a person and a metal guardrail or a parked car. This precision allows the vehicle to make instantaneous, life-saving decisions—like emergency braking or evasive steering—that lower-resolution radars simply cannot.
As of January 2026, Arbe’s ultra-HD radar is integrated with the NVIDIA DRIVE platform. This combination allows the vehicle to process a raw point cloud of over 20,000 detections per frame. This data powers “AI-based Occupancy Grids,” giving the car a human-like understanding of drivable space and potential hazards at highway speeds.
Yes. Unlike 2D radars that “flatten” the world, Arbe’s 4D radar senses elevation. This allows it to detect a small child, a tall cyclist, or a low-profile road hazard separately from overhead signs or bridges. This “vertical awareness” is fundamental to eliminating accidents in complex urban environments.
One of the core barriers to Vision Zero has been the high cost of autonomous sensors. By leveraging a 22nm RF CMOS process, Arbe has reduced the cost of high-resolution sensing to a fraction of the price of LiDAR. This allows automakers to implement “Super Sensors” in entry-level and mid-range cars, not just luxury models.
The “Race to Zero” has moved into production. In December 2025, a major Chinese state-owned automaker selected Arbe’s chipset for its Level 4 autonomous program. With a Start of Production (SOP) in December 2026, this marks the first time ultra-HD 4D radar will be deployed at a scale of thousands of vehicles to meet rigorous safety standards.
For heavy trucks, the 350-meter range is a game changer. It provides the long lead time necessary for a 40-ton vehicle to brake safely. In 2026, Arbe’s radar is increasingly used as the “front-facing” safety backbone for autonomous freight, where the stakes for “zero accidents” are highest.
Because it is the only sensor that works in pitch darkness and through physical obstructions like fog or dust, it acts as the “trigger” for the entire sensor suite. It can identify a potential hazard hundreds of meters away and “cue” the cameras or LiDAR to focus on that specific area, significantly boosting the system’s overall reliability.
While traditional radars use a handful of channels, the Phoenix antenna is the densest array in the industry. It packs 48 transmitting (Tx) and 48 receiving (Rx) channels into a compact 14×11 cm form factor. This enables 2,304 virtual channels, providing a level of detail that is 100 times higher than current market solutions.
Because of its ultra-high resolution, the antenna can “separate” a person from a nearby sidewalk or a stationary guardrail. Most radars see a single “blob,” but the Phoenix antenna can identify a pedestrian even if they are standing perfectly still or are partially obscured by a parked car.
Elevation sensing allows the radar to perceive height.This means a vehicle can drive with confidence under a bridge or into a tunnel while still being able to detect a low-profile hazard, like a lost tire or a small child, in its path. Traditional “2D” radars often fail here, causing “phantom braking.”
Despite its massive channel count, the antenna is designed to match the size of standard radar units used today. Its sleek, compact design allows it to be hidden behind a car’s grill or headlight, maintaining the vehicle’s aesthetics while providing “super-sensor” capabilities.
A high number of physical channels creates a significantly higher Signal-to-Noise Ratio (SNR). This allows the radar to filter out random “ghost” reflections (false positives) that typically cause autonomous systems to brake unexpectedly for no reason.
As of January 2026, the raw data from this antenna array is now fully integrated with NVIDIA accelerated computing. At CES 2026, Arbe demonstrated that this antenna provides the “detection density” (20,000+ points per frame) required for NVIDIA’s AI perception stacks to enable true eyes-off highway driving at speeds up to 130 km/h.
Arbe’s antenna technology is being integrated by global Tier-1s, including:
A major milestone was reached in December 2025: a China-based state-owned automaker selected the HiRain LRR610 (powered by Arbe) for its Level 4 autonomous vehicle program. The Start of Production (SOP) is scheduled for December 2026, with thousands of vehicles expected on the road by 2027.
By providing a wide 140° FoV in certain configurations and a nearly 0% false alarm rate, the antenna acts as the ultimate “first line of defense.” It can detect a cyclist emerging from a side street long before they enter the vehicle’s direct path, making it the foundational sensor for achieving zero-fatality safety ratings (NCAP).
