Cameras have proved unreliable under difficult lighting situations – we remember the infamous fatal accident earlier this year when a vehicle in autopilot mode crashed into a crossing truck. “A lidar sensor probably would have identified the obstacle correctly and thus the system could avoid the crash”, comments Werner Brockherde who oversees Fraunhofer IMS’ CMOS Image Sensor department. According to the researcher, lidar in combination with other components sets out the conditions for future self-learning systems that can brake and steer automatically. In automated vehicles, lidar therefore will have at least a complementing function besides radar and camera and help the systems to properly identify the surroundings of the vehicle.
In traditional lidar systems, a single laser beam is deflected to a rotating mirror. Through the rotation of this mirror, the system can capture data from the surroundings in a 360-degree angle around the sensor. Such sensors however are clumsy and, due to its high degree of mechanic parts, failure-prone. Brockherde and his colleagues at Fraunhofer IMS therefore use highly sensitive lidar sensors that do not require any a rotating mirror, and, actually no moving parts at all. The “Flash Lidar” developed by IMS transmits a single laser flash to acquire all the reflexes from the surroundings. The scientists utilize so-called Single-Photon Avalanche Diodes (SPADSs) also developed at the Duisburg, Germany based Fraunhofer institute. “We do not illuminate just a point but a rectangular measurement field”, explains Brockherde.