At the end of January in Santa Monica, California, a Waymo autonomous vehicle, part of Alphabet, was involved in an incident when a small child unexpectedly ran onto the road. The LiDAR sensors installed on the vehicle detected the child emerging from behind a parked SUV, but the collision could not be avoided. Despite hard braking, the vehicle's speed decreased from 17 to 6 miles per hour, and it struck the child, throwing him onto the sidewalk. According to reports, the boy sustained only minor injuries; however, this does not lessen the anxiety of parents living in cities where autonomous vehicles operate.
Although in this case Waymo managed to detect the child at the moment of his appearance, the question arises: what if the vehicle could "see" him beforehand? This is the challenge researchers at the University of Pennsylvania are trying to solve, having developed a sensing system that uses radio waves to detect objects hidden behind walls.
In experiments, the HoloRadar system was installed on mobile robots, which successfully identified the locations of people behind barriers. Although the work is still in its early stages, the team hopes that such a sensing system can help robots better navigate their surroundings. This could significantly enhance the safety of autonomous vehicles by preventing potential collisions.
“It is extremely important for robots and autonomous transport to see beyond just what is in front of them,” noted the senior author of the study, Professor Minmin Zhao. He added that this is an important step towards creating a more complete perception of the environment for machines.
Most modern autonomous vehicles, with some exceptions like Tesla, use a combination of cameras, radars, and LiDAR systems. LiDAR uses laser pulses that spread in all directions to create a detailed 3D map of the surrounding space. However, this technology has its limitations: large buildings or other obstacles can block the lasers, making it impossible to detect objects that are around the corner.
To overcome these limitations, Zhao and his colleagues decided to explore the use of radio waves, which many companies had previously dismissed. Researchers believed that radio signals have lower resolution compared to visible light, making object recognition more challenging. However, the team from the University of Pennsylvania concluded that this limitation could turn into an advantage in the context of LiDAR scanning around corners.
The radio waves emitted by the robot can pass through small irregularities in walls, and only a portion of the signal is reflected back. This phenomenon allows walls to behave like mirrors, reflecting radio waves along predictable paths.
HoloRadar is capable of detecting obstacles around corners, which helps identify people at intersections. “It’s similar to how drivers use mirrors at blind intersections,” explained co-author of the study, graduate student Haowen Lai. He added that with radio waves, the environment becomes somewhat filled with mirrors, which does not require changes in the environment itself.
To process the information obtained, the team developed a two-component artificial intelligence model. It first filters out unnecessary data and makes assumptions about what is hidden around the corner, then creates a 3D model, allowing visualization of objects behind walls. After that, the system was tested on a mobile robot, which successfully navigated three corners and detected hidden objects.
How can radio waves enhance the safety of autonomous vehicles?
Previous attempts to solve the problem of visibility around corners in robotics have had limited success. For example, in 2019, researchers from the Massachusetts Institute of Technology used cameras to detect shadows that could indicate approaching vehicles. Just recently, engineers from Stony Brook University applied single-photon LiDAR to detect light reflected around corners, inspired by how people use convex mirrors.
However, the main issue with these approaches is that they rely on light waves, making them less effective in poor visibility or darkness. HoloRadar addresses this problem by using radio waves, but it is not a universal solution. Researchers emphasize that HoloRadar should complement existing sensing systems rather than replace them.
Autonomous vehicles are becoming increasingly common on the streets of cities around the world. In the United States, accidents have already been reported, resulting in at least a dozen deaths, and this number is likely to grow as the number of such vehicles increases. The ability to "see" through walls could offer a significant improvement in safety for these autonomous vehicles.
