A planning framework of environment detection for unmanned ground vehicle in unknown off-road environment

This paper describes a planning framework of environment detection for unmanned ground vehicle (UGV) in the completely unknown off-road environment, which is able to quickly guide the UGV with nonholonomic constraints to detect the environmental information as much as possible. The contributions of this paper contain four fold. First, due to the sensor characteristics of camera and lidar, we present a two-layer combined detection map which can accurately represent the detected and undetected area. Second a frontier extraction algorithm based on RRT considering information acquisition and nonholonomic constraints of UGV is used to extract the target pose. Third, we use a search path planning method based on motion primitive which is able to handle obstacle constraints of environment, nonholonomic constraints of UGV. Fourth the heuristic fusion is proposed to guide the extension of motion primitives to generate a kinodynamically feasible and collision-free trajectory in real-time. And it works well in both simulation and real scene.