What can Arbe’s radar sensing technology bring to autonomous driving?

The autonomous driving track is like a martial arts world. Different research directions are like several major sects. Each of them is practicing a basic martial arts, hoping to get the best at all levels. If you can obtain a true copy of the special skills and techniques, you will undoubtedly be able to become a being who can see all the small mountains at the "Ling Jueding" moment.

In fact, many consumers are still dismissive of self-driving cars so far, because self-driving accidents have been common in the headlines in recent years. For OEMs, in order to gain the trust of consumers, they need to integrate the right sensor technology from the beginning to prevent unnecessary collisions for everyone - not just luxury models.

In an environment with multiple people or objects, Radar Perception is the only radar technology that can sense their presence and dynamically track them to determine the relative relationship between them. Perception radar is a powerful tool that can provide cutting-edge radar technology at a price that every vehicle can afford, ensuring safety and winning consumer trust, thereby realizing the autonomous driving revolution.

On January 6, 2023, Arbe released 360° surround radar sensing technology, using AI-based integrated analysis of the vehicle environment for the first time. This solution leverages data from the Arbe Sensing Radar chipset to enable safe vehicle operation in all weather and lighting conditions. Together with the two previous products, Lynx Surround Imaging Radar and Phoenix Sensing Radar, this is like providing the industry with a martial arts secret book. Those who can learn from it will have the opportunity to stand out from their peers.

Sensor capabilities are a major shortcoming of autonomous driving

At this stage, autonomous driving technology may not be ready for the mass market. In order to develop and launch corresponding advanced functions, OEMs need to use appropriate sensors to collect large amounts of data in more real-life scenarios, thereby achieving higher safety.

不过，每个传感器类型都有至少一个显著的缺点。

Radar resolution is not high enough: Ordinary radar lacks the resolution required to support safe driving in daily use situations, while long-range radar is too expensive to install at the four corners of the vehicle.

LiDAR has low range and high cost: LiDAR cannot be used around vehicles in severe weather conditions.

The camera sensor has insufficient transmission distance and lacks depth perception: the short range and narrow field of view are insufficient to meet the requirements of L2+ and higher autonomy, and often fail in poor weather and lighting conditions.

Optical sensors (camera sensors and lidar) have installation limitations: they require a transparent surface, which greatly limits the ability to place them at the corners of the car, and dust attached to the surface of the optical sensor can also affect its functionality.

In terms of applications, existing angle radars can no longer meet the needs of ADAS (Advanced Driving Assistance Systems ) and autonomous driving, and there are more than one challenges.

First, due to limited distance, insufficient computing power and low resolution, angle radar cannot provide imaging and provide the required redundancy for optical sensors. Therefore, it is difficult to achieve sensor fusion perception and is only suitable for basic driving assistance functions such as side blind spots and intersections. Traffic alerts and dynamic object tracking.

Second, although 360° sensing has previously used angle radar to implement ADAS functions, so far, angle radar has not been able to be used for more challenging use cases, such as long-distance approaching vehicle safety alarms, stationary object recognition, distance detection between two vehicles, and Assess the shape and size of objects. In addition, high-resolution long-range radar is expensive and cannot be deployed at multiple locations on the vehicle.

To this end, the industry urgently needs to find a reliable and affordable alternative with redundancy and data diversity to meet the requirements of automotive grade safety and launch autonomous functions. Arbe imaging radar is 100 times more precise than other radars on the market, can work normally in any weather and lighting conditions, and has the advantage of long-distance operation, providing a powerful complement to optical sensors.

Features of Arbe Surround Radar Imaging Technology

Arbe's disruptive and unique radar technology brings high quality and safety, enabling 4D imaging capabilities to be used in autonomous driving systems to make up for the shortcomings of optical sensors in challenging harsh light and weather conditions; and with almost zero latency Depth information and instant radial velocity data are provided at all times.

Prior to the release of 360° surround radar sensing technology earlier this year, Arbe had designed two radar development platforms based on its imaging radar chipset: the Phoenix sensing radar that won the CES 2022 Innovation Award in January 2019 and the Phoenix sensing radar released in July 2022. Surround imaging radar Lynx, the former uses 48x48 radio frequency channels, and the latter uses 24x12 radio frequency channels.

(1)Phoenix sensing radar

Phoenix is ​​the industry's first automotive chip using advanced 22nm radio frequency CMOS technology, with a frame frequency of approximately 30Hz, achieving multiple technological breakthroughs in radar performance. It has 100 times finer imaging and double radar sensitivity. It uses advanced algorithms and unique channel separation technology to effectively reduce false alarms. Phoenix's high dynamic range capture can differentiate between targets of different sizes and provide clear resolution based on dynamic and static targets.

Arbe's Phoenix perception radar uses 48X48 radio frequency channels, which can distinguish, track and identify hundreds of targets in azimuth and elevation angles, and generate high-definition and detailed environment images with a wide field of view and larger identification range. In addition, it can also distinguish according to altitude Differentiate goals. Currently, Phoenix is ​​suitable for various dynamic application scenarios and static application scenarios.

Currently, the radars on the market are generally 3x4 angle radars, which are composed of 12 channels; the most advanced 12x16 long-range radars are composed of 192 channels. Phoenix uses 48 transmit channels and 48 receive channels, with 2304 virtual channels. Phoenix's resolution is 2-10 times that of competing products, and it has the highest point cloud density on the market.

What Phoenix generates is a true 4D radar image with ultra-high resolution in the range, azimuth, elevation and Doppler dimensions. This is achieved through highly reliable target detection, low side lobe level (SLL) and low false alarm rate. False positive and false negative scenarios are eliminated since there are almost no virtual objects.

Phoenix's high resolution and high dynamic range provide the ability to distinguish a variety of objects, such as a motorcycle next to a truck, a car stuck under a bridge, and a pedestrian standing next to a fence.

In addition to high-resolution imaging with much higher accuracy of center and velocity, Phoenix also provides more detailed information about the object being tracked, such as the object's orientation and boundaries, making its fusion with other sensors (such as cameras) more meaningful.

The system can also track moving objects, map the environment and stationary obstacles, and create complete free-space mapping for both path planning and precise positioning.

Arbe's patented chip technology has brought further innovation to autonomous driving components, giving 4D radar imaging technology greater development prospects and commercial value in all levels of autonomous vehicles. Phoenix's patented chip can upgrade automotive radars , which were originally just auxiliary sensors , into core components of driver assistance and autonomous driving sensor suites.

(2) Lynx surround imaging radar

At present, the industry has used angle radar for 360° sensing, including blind spot detection, vehicle warning, dynamic object tracking and other ADAS functions. However, due to limited range, insufficient computing power and low resolution, corner radar is not yet able to provide solutions for more challenging use cases, such as long-distance approaching vehicle alarm, identifying stationary objects, detecting the distance between two vehicles, and distinguishing objects shape and size, etc. Since angular radar does not provide imaging, it cannot provide the required redundancy for optical sensors and therefore cannot achieve sensor fusion and perception.

In order to meet the urgent market demand for long-range surround sensing sensors, Arbe launched its surround imaging radar Lynx, which fills the gap in surround sensors and provides a solution that can cope with harsh lighting and weather conditions and has long-range detection capabilities. , and these are the limiting factors for camera and lidar performance.

Lynx imaging radar is the industry's first cost-effective sensing solution with 360° coverage around the vehicle, providing the redundancy and data diversity critical to vehicle safety and autonomous driving capabilities, such as Doppler, long-range , depth perception, etc., able to achieve free space mapping. It provides effective sensing in all environmental conditions and can collect speed, distance and other data that optical sensors lack.

Lynx is a solution that integrates long-range, high-resolution, 360° sensing and high cost-effectiveness. It uses 24×12 RF channels. Its performance is better than the current 3×4 angle radar in the industry and even better than advanced The 12×16 front-facing radar makes it a surround radar that can both enhance perception and facilitate sensor fusion.

Lynx surround imaging radar is an important complement to Phoenix perception radar and is suitable for the four corners and rear of L2+ and higher level autonomous vehicles. Each vehicle is equipped with 4-6 surround imaging radars, integrated into one or two perception radars, to achieve true safety and autonomy. Lynx can also be used as a corner radar and front radar on lower-priced models, installed at the corners and rear of the vehicle.