China to Take the Lead in 4D Millimeter-Wave Radar Following LiDAR?
4D millimeter-wave radar, an emerging sensor component in level-x autonomous driving (AD) system, is getting more and more attractions given its robustness in extreme environments and capabilities to generate rich point cloud while being only a fragment of LiDAR cost.
Back in 2019, I identified a long list of LiDAR players in China, among which Hesai and Robosense have grown to be the leaders in the global automotive LiDAR segment. Would the development of 4D millimeter wave radar follow the footstep of LiDAR, with Chinese companies to take the lead?
This article is part 1 to review the difference between 4D millimeter wave radar and other sensors in the AD sensor suite, the various technical paths in 4D millimeter wave radar innovations and a list of Chinese players. Part 2 will discuss several trends witnessed on the China market and the challenges 4D millimeter radar is facing.
4D millimeter-wave (mmWave) radar emerged as an advanced version of the traditional radar, also called 3D mmWave radar, with the capability of measuring the elevation of targets as the 4th dimension besides range, azimuth and velocity.
Additionally, 4D mmWave radar comes with improved detection range, ranging accuracy, velocity accuracy, and a pixel-level angular resolution. Hence it generates high-density point cloud data that can be used to identify and track objects in the surrounding environment. It is often referred as imaging radar or 4D imaging radar.
4D mmWave Radar vs. Other Sensors
Camera, mmWave radar and LiDAR are the most common sensors in a level-x autonomous driving system, each having its own strengths and limitations, as illustrated in exhibit 1.
Cameras can capture rich color and detailed information while suffers in low visibility conditions. LiDARs are excellent in ranging and resolution but the price tag is an obstacle to massive deployment. Radar systems see the longest history of application in the automotive industry by being all-weather reliable.
As 4D mmWave radar addresses the weak performance in angular resolution with 3D mmWave radar, it can be used to detect and track objects around the vehicle to potentially handle complex perception tasks.
4D mmWave Radar Innovations
The innovative solutions on 4D mmWave radar aim to improve the vertical resolution as well as other performance indicators in order to meet the requirements in ADAS/AD applications. Achieving higher resolutions requires a wider aperture, which is related to the number of transmitting and receiving channels and the antenna array design.
Additionally, it is looking to improve efficiency, directivity stability and impedance bandwidth while maintaining competitive manufacturing costs.
The typical components of 4D mmWave radar include a transceiver module, which is usually discrete MMIC and antenna PCB, and a signal-processing module. The following diagram (exhibit 2) shows 4D mmWave radar’s typical structure and the various technical paths at each level. A few of the common technologies are discussed in details.
Multi-Chip Cascading
The most straightforward method to improve angular resolution is to cascade several standard mmWave radar chips. For example, TI’s AWR1843 comes with 3 transmitters (TX) and 4 receivers (RX) on a single transceiver IC. By cascading four of the AWR1843 in a multi-input multi-output (MIMO) configuration, a radar system would have 192 virtual channels from a combined 12 TX/16 RX arrangement to sample both horizontal and vertical space ahead in much more details.
Single-Chip Integrated with Multi-Channels
As the name suggests, an alternative to multi-chip cascading is to simply integrate more TX/RX channels on a single chip. For instance, the Arbe patented chipset for 4D imaging radar scales to 48 TX/48 RX producing 2,304 virtual channels. Uhnder’s Radar-on-Chip solution features a 12 TX/16 RX arrangement generating 192 virtual channels.
This approach can potentially reduce the complexity of the hardware design and optimize the overall size of a radar sensor and its power consumption but the disturbance between antennas remains a challenge.
Virtual Aperture Imaging (VAI)
Virtual Aperture Imaging (VAI) technology aims to enhance the angular resolution by virtually expanding the aperture of antennas with algorithms. In contrast to the previous two hardware-centric approaches, VAI addresses purely on the software level, cost-effectively improving radar performance with conventional MMICs. For example, Oculii, which is now part of Ambarella, launched a reference product called EAGLE using two of AWR1843 for 768 virtual channels by applying its proprietary AI software.
In a long-range radar design, VAI usually needs to be paired with multi-chip cascading to reduce noise. There is also a suspected trade-off between the angular resolution and the confidence level of each point.
Antenna Design
Antenna design is crucial for 4D mmWave radar to achieve wide bandwidth, low side lobe levels, high gain and good radiation efficiency. Microstrip antennas have been commonly adopted for low cost and ease of manufacturing.
In recent years, the use of traditional waveguide antenna is being evaluated due to its lower transmission losses and higher power capacity. There are two major types for automotive applications namely substrate-integrated waveguide (SIW) slot antenna and 3D waveguide antenna.
With TX/RX channels scale in multi-chip cascaded structure, the 3D waveguide technology is an ideal option, where the antenna array could be distributed in the additional 3rd dimension. It allows for a much more compact design of the radar sensor as opposed to using the microstrip antenna.
List of 4D mmWave Radar Companies in China
Generally speaking, there are two types of local companies engaged in the development of 4D mmWave radar on the China market (exhibit 3). Automotive tier 1 suppliers in the ADAS domain usually look to build up hardware strength by developing key sensors in-house, which include 3D and naturally 4D mmWave radars.
The growing installation of mmWave radar on passenger cars has created a battlefield for startups. They entered the field since 2013 to create comparable radar products to global giants like Autoliv, Bosch, Continental and Aptiv that are still dominating the China market. Companies like WHST, Nova and Muniu have gained prominence with their technical expertise, qualifications, and experience in mass production. As 4D radar technology continues to mature, they certainly have the potential to shift the market landscape with advantages in cost-efficiency, flexibility, and localized delivery.
Stay tuned for part 2.
Sources:
4D Millimeter-Wave Radar Industry Chain Report by Smartautoclub and Aibang (2023.10)
4D Millimeter-Wave Radar Market and Technology Trend Analysis by Jiu Zhang Zhi Jia
