Sensor Fusion and Compensation Algorithm for Vehicle Tracking with Front Camera and Corner Radar Sensors

Autonomous vehicle encounters numerous difficulties which need to be solved. Precise detection and perception of surrounding obstacles are the first step required to overcome the problems. Therefore, it is important to properly fuse the sensor data. In this paper, one front camera and four corner radars are used to percept the surrounding obstacles. Two algorithms are introduced; a sensor fusion algorithm and a compensating algorithm which improve the performance of the sensor fusion algorithm. Nearest neighbor (NN) algorithm and linear Kalman filter are used for data association and tracking respectively. Moreover, the obstacles' boundary information which is given by camera and radars is used to estimate the free space. By applying Occupancy Grid Map (OGM) with boundary information, previous fused data can be compensated.