Object Detection and State Estimation of Autonomous Vehicles with Multi-Sensor Information Fusion

An accurate object detection and state estimation is the keystone for realizing motion decision of autonomous vehicles. Multi-sensor fusion is a reliable way to obtain sufficient object information compared with detection by only a single sensor. In this paper a two-stage fusion scheme is proposed to deal with object detection and state estimation simultaneously. Three typical sensors, radar, camera and LiDAR, are studied as the data sources. Compared with the existing work, not only detection results but also state estimation accuracy are analyzed. A simulation in PanoSim is performed to verify the effectiveness of our method. The test results illustrate the output of fusion can meet the requirement of motion decision, where both overtaking and adaptive following tasks are conducted as expected.