Architecture to Detect, Track, and Classify Objects using LiDAR Measurements in Highway Scenarios

Self-driving cars require a holistic perception of their environment. To achieve this requirement, a plethora of sensor technologies exists e.g. RGB-camera, ultra-sonic and radar. Those sensor technologies have different range, as well as resolution and behave differently with varying weather conditions. Another technology is Light Detection and Ranging (LiDAR), which enables precise distance measurements. In combination with RGB-cameras, ultra-sonic, and radar, LiDAR closes the gap to enable the holistic perception of the environment. Due to limited experience with LiDAR sensors, there is a lack of understanding how to detect, track, and classify objects (e.g. cars, guardrails) using LiDAR data. In this paper, we propose an architecture to detect, track, and classify objects based on LiDAR measurements in highway scenarios.We evaluate our architecture using preliminary sensor data obtained from a setup including six Ibeo Lux sensors and additional a roof mounted Velodyne HDL-64E.