Dynamic obstacle detection based on multi-sensor information fusion

Abstract

Dynamic obstacle detection is the key to ensure the agricultural robots could move autonomously in the non-structural environments. In this study, a method of dynamic obstacle detection based on multi-sensor information fusion is presented by selecting a Compass Equipment, an Inertial Measurement Unit and a 2D laser scanner as the system's external sensors. A method based on Kalman filter to fuse data from a Compass Equipment and an Inertial Measurement is presented to obtain the position of agricultural machinery. 2D laser scanner has the feature of scanning widely and getting the angle and distance of each obstacle directly. On this foundation, the absolute position and motion state of the obstacle is obtained by the transformation of the relative coordinates. After filtering, clustering and segmentation of laser data by using the method of Voxel grid method and Euclidean method, the absolute position of the same obstacle in the adjacent sampling period is analyzed to distinguish the static and dynamic obstacles. The experiment verify the effectiveness of the algorithm and have certain significance for the realization of autonomous mobile robot.