In the perception system of autonomous driving, millimeter-wave radar stands out for its all-weather operational capability. It can reliably measure the distance and speed of obstacles, even in challenging weather conditions such as rain, snow, fog, and smog. However, a topic of interest often arises in discussions about millimeter-wave radar: the assertion that it cannot detect pedestrians. Is this claim truly valid?

Why do pedestrians seem to elude radar detection?

To ascertain whether millimeter-wave radar can detect pedestrians, it's crucial to understand the concept of Radar Cross-Section (RCS). RCS is a measure of an object's ability to reflect electromagnetic waves. Generally, the larger an object, the better its material conductivity, and the smoother its surface, the greater its RCS. On the road, vehicles, with their substantial size and metal-clad bodies, are highly visible to millimeter-wave radar, akin to searchlights in the dark.

Image source: Internet

In contrast, the human body is primarily composed of water, which strongly absorbs millimeter waves, resulting in weak reflected signals. Additionally, pedestrians are significantly smaller in volume compared to vehicles, and soft materials like clothing further disperse radar beams. Consequently, pedestrians appear as faint, flickering gray dots on radar screens.

These inherent physical differences mean that traditional millimeter-wave radar encounters a low signal-to-noise ratio when detecting pedestrian signals, making them easily obscured by background clutter.

Moreover, the wavelength of millimeter-wave radar typically falls within the millimeter range. While it can penetrate rain and fog, its ability to accurately depict object shapes is relatively limited. For early three-dimensional millimeter-wave radar, pedestrians, utility poles, or metal guardrails on the roadside may produce very similar signal points. Due to the lack of height information and fine resolution, the system struggles to distinguish a pedestrian crossing the road based solely on a few isolated reflection points.

Is it 'invisible' or simply 'overlooked'?

If weak signals represent a physical limitation, then algorithmic filtering introduces another human-induced factor. Traditional millimeter-wave radar faces a significant trade-off when detecting static objects. To prevent frequent false braking caused by detecting manhole covers, road signs, or bridge piers, filtering mechanisms are implemented to directly eliminate static targets with near-zero relative velocity or unremarkable echo characteristics.

Pedestrians find themselves in this very predicament. When stationary or slowly crossing the road, their radial velocity component on the radar is minimal. If the radar's performance is inadequate to differentiate between road clutter and subtle pedestrian movements, the system may disregard these pedestrian signals as invalid interference to ensure smooth driving and minimize false alarms. This explains why some older vehicle models' millimeter-wave radars, while capable of detecting obstacles ahead, often choose to ignore them.

However, advancements in signal processing technology have led to the application of micro-Doppler techniques. When pedestrians walk, the swinging of their arms and legs generates subtle frequency variations distinct from those of their torsos. By capturing these dynamic characteristics, radar can determine, even with weak signals, that a moving target with life-like features is present ahead through algorithms.

How does 4D imaging technology enable radar to discern contours?

In recent years, the advent of 4D imaging radar has revolutionized radar's limitations in pedestrian perception. The term '4D' refers to the addition of a 'height' dimension to the existing range, velocity, and horizontal angle measurements. While traditional radar perceives the world like a thin, black-and-white photograph, 4D imaging radar provides dense point clouds akin to those of LiDAR.

By increasing the number of transmitting and receiving antennas, the angular resolution of 4D imaging radar has undergone a qualitative leap. It no longer perceives just a few isolated points but can outline the rough contours of targets. More importantly, with height information, the system can clearly distinguish between pedestrians on the ground, overhead height limit bars, and roadside guardrails, significantly reducing missed detections.

Currently, the explosive growth in radar point cloud density has enabled some 4D digital imaging radars to achieve point cloud densities exceeding 100,000 points per second. This level of resolution allows the radar to identify target types with the precision of a camera. This enhanced capability means that even in complete darkness or when cameras are blinded by strong light, radar can still independently and accurately detect vulnerable road users.

Does multi-sensor fusion propel the development of autonomous driving?

While millimeter-wave radar is rapidly evolving, no single sensor is all-powerful in the realm of autonomous driving. Pedestrian perception relies on the deep integration of multiple sensors. Cameras excel at classification and recognition, capable of discerning pedestrians' clothing and orientation; LiDAR excels at high-precision 3D reconstruction; and millimeter-wave radar is responsible for all-weather speed measurement and redundancy verification.

In current mass-production solutions, 4D millimeter-wave radar complements visual systems. When the visual system experiences perception fluctuations due to sudden lighting changes or rain obstruction, the distance and velocity data provided by millimeter-wave radar serve as the ultimate safety defense. This fusion solution not only enhances pedestrian perception accuracy but also significantly reduces safety hazards caused by system misjudgments.

With the reduction in 4D radar costs and the increase in installation volumes, this technology not only bridges the gap in range measurement of pure vision solutions but also compensates for the shortcomings of ordinary radar in pedestrian detection. It can be said that the era when millimeter-wave radar could not perceive pedestrians has passed, replaced by a clearer, smarter, and always-vigilant electronic eye.

-- END --