The sensors in autonomous driving are the eyes and ears of the vehicles.
Without the sensors, the autonomous vehicles cannot perceive their environment and especially the lidar plays a major role. The lidar allows a 3D image through a point cloud, which significantly improves perception. The rotating lidar primarily records the distance to objects and can identify objects. The lidar itself emits light and measures the time it takes for the light beam to return reflected.
Researchers from the University of Florida, the University of Michigan and the University of Electro-Communications in Japan have now revealed a dramatic vulnerability in this sensor: a laser attack. Carefully timed laser beams aimed at an approaching lidar system can virtually blind the lidar.
The laser attack creates a blind spot in front of the vehicle. Because the incoming laser beams cause the lidar to assume that these are reflections. The researchers stood about five meters away. With better equipment, the process could also be carried out from a greater distance. The laser must be optimally aligned with the approaching lidar.
In this way, no people or objects can be seen in the blind spot. The car therefore assumes that the road is clear and the car can continue on its way. You can also fool the lidar into thinking that there is another car in front of it. The researchers not only discovered this, they also developed an update for the lidar to eliminate this vulnerability. That’s what she’s looking for software for telltale signatures of the false reflections.