Sensor fusion based on the particle filter and multi-rate kalman filter

Abstract

Assigning the correspondences among moving objects, which are measured from two different sensors such as radar and vision, is difficult when the ego vehicle is changing its lane. In this paper, we propose a robust and reliable decision method of the correspondences between radar and vision when the ego vehicle is changing its lane. A particle filter that consider the yaw rate of the ego vehicle and a convex hull method are proposed to improve the object ID association. In addition, the overall architecture of sensor fusion is described with the proposed association method. The proposed methods have been validated in actual vehicle experiments. The test results show that the proposed algorithm can effectively associate the objects from two different sensors on Yeoju smart highway with the minimum curvature radius of 1,124m and a lane width of 3.5m in the lane-changing situation. ? ICROS 2017.