For millimeter-wave radar, 4D imaging millimeter-wave radar to be exact, 2021 is destined to be a very unusual year. In the tide of “new four modernizations” in the automotive industry, the blessings of various OEMs are pushing 4D imaging technology into the air, and its large-scale commercial use is gradually approaching.
Millimeter-wave radar is not unfamiliar to people, but 4D imaging is not. Among the many sensors used in autonomous driving, what are the advantages of its technology? What role can you play? Who is leading technology development? How will it be commercialized?
At present, automotive radars are used in ACC and AEB, but they are not enough to provide the best performance, nor can they meet the functions required for L2+/L3/L4 level autonomous driving applications. Arbe imaging technology provides the insight needed for autonomous driving systems to make 4D imaging radar the backbone of the sensor suite. Arbe’s unique and game-changing radar technology enables 4D imaging capabilities in autonomous driving systems, complementing the lack of vision sensors in challenging harsh light and weather conditions. In addition, 4D imaging radar can provide depth information and instant radial velocity data with almost zero latency.
1. The market pattern of millimeter wave radar is subtle
Over the years, Infineon and NXP have almost monopolized the millimeter-wave radar chipset market, and STMicroelectronics has been catching up. In early 2020, Infineon announced its entry into the automotive 4D high-definition point cloud imaging radar market, but no corresponding chips have been released.
As early as the end of 2016, Texas Instruments (TI) launched a highly integrated 77GHz millimeter-wave radar sensor based on CMOS technology, but it did not change the pattern of the millimeter-wave radar chipset market. As a result, Texas Instruments threw out the concept of 4D imaging millimeter-wave radar at the end of 2018.
In fact, Arbe, an Israeli startup founded in 2015, has been working hard on a new 4D imaging radar for automotive applications.
In January 2019, Arbe released a beta version of the automotive-grade 4D imaging radar product using patented chipset technology, Phoenix.
In September 2019, Arbe launched its first new high-density radar antenna with the highest number of channels, field of view and resolution on the market, capable of detecting pedestrians and separating them from sidewalks with unprecedented excellence, making them ideal for ADAS and autonomous vehicles. Driving adds a new level of safety.
In October 2020, Arbe launched the first 2K high-resolution imaging radar development platform, bringing disruptive changes to customers’ imaging radar system design.
Arbe’s solutions are small, lightweight and energy efficient, affordable and fully customizable to meet all levels of vehicle autonomous driving needs.
2. What is the difference between traditional radar and 4D imaging radar?
Radar systems estimate the range and velocity of a target by transmitting a signal and comparing the reflected signal from an object in the environment with the characteristics of the transmitted signal. A single transmitter, single receiver radar system is called a single input single output (SISO). If the angle of an object (azimuth, elevation, or both) is to be estimated, either mechanical or Electronic scanning must be used, so the angle estimation of fast-moving objects is not accurate.
A radar with multiple transmit and receive antennas is called a multiple-input multiple-output (MIMO) system, and includes multiple transmit antennas and multiple receive antennas.
This creates a virtual array that provides higher resolution. Radar performance is related to the number of Tx and Rx channels used in the radar, while using a 2×3 array would result in 6 virtual channels, Arbe provides 48Tx channels and 48Rx channels, resulting in a virtual array of 2304 virtual channels.
3. The sensor ruler is long and short
As vehicles become more electrified and electronic, the increase in ADAS functionality requires the use of more sensors, increasing the complexity, cost and weight of vehicles while improving safety technology. Each sensor has its own advantages and disadvantages, so the industry generally adopts the fusion method to complement each other.
Cameras and other optical solutions effectively detect targets, measure distances, provide precise imaging and track multiple targets. But it has a limited field of view, is powerless in inclement weather, bright light and shaded areas, and has privacy concerns.
Ordinary radar works well in bad weather, detects presence, direction, distance and speed, and is a robust and scalable solution. However, its small antenna array results in low data output resolution and cannot generate rich images; its narrow field of view is mainly focused on one axis, and its angular resolution is limited, making it impossible to distinguish adjacent targets.
Lidar (LiDAR) can detect objects in fine detail, but is also not suitable for severe weather and is expensive.
It can be seen that the current sensors are not mature enough to support future autonomous driving. However, the breakthrough technology of 4D imaging radar can just make up for the shortcomings of existing sensors while having all the advantages of existing sensors, which can detect objects with ultra-high resolution in any environment and conditions, and achieve the required level of security.
4. Why is the resolution of 4D imaging radar higher?
The inability to distinguish between threats and false alarms is the leading cause of accidents in autonomous vehicles. Due to the limitations of low spatial resolution, the inability to separate targets by direction of arrival while maintaining low false alarms has relegated radar to a supportive role in automotive sensors. Arbe’s 4D imaging radar is changing that awkward situation.
Real 4D radar images
Arbe radar produces a true 4D radar image with ultra-high resolution in range, azimuth, elevation and Doppler dimensions. This is achieved through highly reliable target detection, low sidelobe level (SLL) and low false alarm rate. It has almost no ghost objects, eliminating false positive and false negative scenarios.
In addition to the much higher accuracy of center and velocity, high-resolution imaging provides more detailed information about the object being tracked, such as the object’s orientation and boundaries, making it more meaningful to fuse it with other sensors, such as cameras.
Unmatched Physical Resolution
Arbe’s high physical resolution is 2-10 times higher than competing solutions, supporting more than 100,000 detections per frame, with the highest point cloud density on the market. This stems from an array of 2,304 virtual channels created by its digital beamforming 48 transmit and 48 receive antennas. Instead of using unreliable synthetic or statistical resolution enhancements such as super-resolution (SR), Arbe utilizes a wide aperture array that provides a physical 3dB beamwidth of 1.25° azimuth and 1.5° elevation at low SLL for improved dynamic range reliability and safety. This allows the Arbe chipset to remain effective in low signal-to-noise ratio (SNR) and multi-target scenarios. High physical resolution and high dynamic range provide the ability to separate various objects, such as a motorcycle next to a truck, a car stuck under a bridge, and a pedestrian standing next to a fence or a vehicle with a flat tire.
The system can also track moving objects, map the environment and stationary obstacles, and generate free-space maps for both path planning and precise positioning.
The expert explained that from a physical point of view, time is the fourth dimension because the time element is obtained from Doppler. Imaging radar actually creates an array that dramatically increases the density of measurements. Whereas traditional 2D radars are rough, producing only one point per object, imaging radar can provide many points, resulting in vertical resolution that provides a better understanding of what exactly the object being tracked is.
In other words, the time factor has always been key to radar functionality. The fourth element of a 4D imaging sensor is “lateral resolution”. 4D imaging radar can identify not only horizontal planes but also vertical planes, for example, a car can decide whether to pass “under” or “over” an object.
For example, a car is traveling at 80 km/h on a highway, and a motorcycle (small object with low reflectivity) is coming from behind at 120 km/h. Unlike cameras and lidars, 4D radar can spot motorcycles that are initially far apart and recognize that the two objects are moving at different speeds; it can also show whether an object is moving towards itself or far away.
5. How does 4D imaging become the protagonist of autonomous driving?
Arbe’s world’s first 2K ultra-high-resolution 4D imaging radar platform removes previous resolution limitations and reinvents radar. It can assess distance, altitude, depth and speed with high resolution, providing a wide field of view while achieving a low false alarm rate at long range. With the advancement of this technology, it is expected that the radar will be upgraded from an auxiliary accessory to the core of safe autonomous driving from L2 to L5.
To sum up, 4D imaging radar has the following advantages:
Real-time obstacle detection: In all weather and lighting conditions, a wide field of view provides a highly detailed image of the environment, discovering all kinds of obstacles in real time, including smaller objects on the side of the road, such as people or bicycles, even if they are blocked by a tree or truck, etc. Large objects occlude and also determine if and in which direction they are moving, providing real-time situational data and alerts to the vehicle.
Long-distance detection: Achieve the longest-distance detection of all sensors, most likely to be the first device to detect hazards. It can then direct the camera and LiDAR to the area of interest, greatly improving safety performance.
Path Planning: Provides true path planning as it creates a detailed image of the road at a range of over 300 meters and captures size, position and velocity data of objects around the car.
Object height separation: Recognizes whether an object facing directly in front of the car (such as a bridge) is stationary, has to stop, or is safe to drive over.
Reduced processing and server requirements: Since only the camera and LiDAR are aimed at the region of interest, utilizing high-quality radar post-processing will solve the main problem of current prototypes – power consumption.
Dramatically reduce production costs: Even above L3, there is no need for more than one LiDAR unit per vehicle, or LiDAR may not be required at all, helping manufacturers reduce costs. The production cost of a self-driving sensor kit should be less than $1,000, and some vehicles tested today use components and systems that cost 100 times that price.
In this way, the ADAS system can trust the radar readings, enable quick responses, and prevent unnecessary stops. Therefore, 4D imaging radar provides the basis for navigation, path planning and obstacle avoidance, and can support the sensing requirements of L4 and L5 autonomous vehicles in terms of safety and accuracy.
6. Technological breakthrough of Arbe radar development platform
Technological breakthroughs of the Arbe radar development platform include: high-resolution separation of objects by azimuth and elevation; detection of object orientation and boundaries; elimination of false alarms; ultra-high physical resolution; reduced mutual interference; no Doppler blurring . These breakthroughs are embodied in the following six aspects:
The first is the use of state-of-the-art RF chipsets. The proprietary mmWave automotive radar RFIC chipset developed by Arbe consists of a 24-output channel transmitter chip and a 12-input channel receiver chip. Built on the new 22nm FD-SOI CMOS process (22FDX), the AEC-Q100 qualified RF chipset supports TD-MIMO with best-in-class performance in channel isolation, noise figure and transmit power. Utilizing the latest RF processing technology, Arbe achieves the most advanced RF performance at the lowest cost per channel on the market.
The second is radar processing technology. Arbe’s unique self-developed baseband processor (Everest) integrates a patented Radar Processing Unit (RPU) architecture and embedded proprietary radar signal processing algorithms to process and convert large amounts of raw materials in real-time while maintaining low silicon power consumption data. The RPU is capable of processing up to 48 Rx channels and 48 Tx channels in real-time, generating 30 full 4D images per second for an equivalent processing throughput of 3Tb/sec.
The third is mutual interference suppression. More and more vehicles are using radar sensors, some with as many as 8 sensors, most emitting in the same frequency band. As a result, with increasing radar loadings, there is a growing risk of radar interference with each other, especially in dense urban environments, and close-range (reverse and co-directional) interference. When radar jamming occurs, detection missed or false positives can again lead to accidents. Arbe’s patented FMCW2.0 system innovation effectively avoids and mitigates interference from other FMCW radar transmitters with minimal or no performance degradation.
The fourth is safety. The Everest processor includes a dedicated ASIL-D island that oversees the safe operation of the system. Arbe radars are designed with safety in mind, with a special emphasis on reducing false alarm rates. Arbe is ISO 26262 compliant and enables continuous built-in radar system self-checks.
The fifth is post-processing and SLAM algorithm. The proprietary post-processing software stack developed by Arbe includes a radar-based SLAM solution optimized for enhanced FMCW TD-MIMO imaging radar. The SLAM algorithm enables real-time clustering, tracking, self-localization, false target filtering, and radar/radar-camera-based target classification.
The sixth is the enhanced perception algorithm. As the foundation of advanced perception capabilities, Arbe’s 4D imaging radar platform includes the ability to infer vehicle speed and in-lane positioning accurately in real time. Radar data post-processing helps track and classify objects in the vehicle’s entire field of view, determine their orientation and motion vectors, and provide precise and accurate free-space mapping to distinguish drivable and non-drivable environments in any weather or lighting conditions .
7. Speed up the commercial process
Today, radar plays a vital role in safety systems such as Adaptive Cruise Control (ACC), Blind Spot Detection (BSD), and Automatic Emergency Braking (AEB). But current radar technologies on the market have to choose between medium resolution for a limited field of view and low resolution for a wide field of view.
To achieve L4 and L5 vehicles, OEMs must adopt next-generation sensing technology that senses the environment with 4D high resolution over a 100-degree wide-angle range by using high-resolution imaging radar at about 1 degree azimuth and 2 degrees elevation. .
Another important issue is the ability to filter false alarms. To provide the best sensitivity, radars typically use the lowest detection threshold, so some noise is reported, post-processing and tracking are required to filter out random noise, and calibration schemes can only achieve very low sidelobe levels.
To this end, with the advent of high-resolution imaging radars, many radar vendors are eager to make the radar the only high-speed sensor capable of operating in harsh weather and light conditions.
To accelerate commercialization, the Arbe 2K high-resolution imaging radar development platform provides Tier 1s, OEMs and newcomers the ability to revolutionize their imaging radar systems and enhance their perception algorithms. The development platform provides the following components:
A complete Arbe imaging radar chipset with RF transmitter and receiver chips, and a patented imaging radar processor;
Radar antenna with the densest channel array in the industry, available in a form factor that meets current OEM size specifications;
A software layer that abstracts hardware access and scheduling;
Reference designs to guide the development of Tier 1 and OEM systems.
Optical sensors are of course also required in the sensor suite for autonomous driving, and 4D imaging radar enables the vehicle to achieve the required safety performance because it solves the following problems:
Highest reliability in all weather conditions, including fog, heavy rain, dark nights and air pollution;
According to the requirements of the automotive industry, detect obstacles up to 300 meters away;
measure the Doppler (radial velocity) of each frame;
In-vehicle security solutions that prevent tampering or unauthorized access;
Compact design for easy integration in the front grille of the vehicle;
Support dynamic calibration;
The ability to move beyond the PoC (Proof-of-Certification) stage into mass production.
In December 2019, Arbe received investment from institutions such as Beijing Automobile Group and Hyundai Group; it is currently working with 25 Tier 1 and OEMs from the United States, Europe, China and Japan to develop new products based on the Arbe imaging radar development platform. Generation Radar System. Radars based on Arbe technology are expected to be available in production vehicles in 2022.
8. The industry standard is just around the corner
It is precisely because of the outstanding advantages of 4D imaging radar that the entire industry is using 4D imaging radar as an indispensable and important element in the autonomous driving sensor suite, equipping autonomous vehicles with more sensitive ears and eyes, thus forming a safer automotive market. . 4D imaging radar is also expected to become the industry standard for radar, with profound implications for autonomous vehicles.