Autonomous Navigation In Outdoor Semi-Structured Environment For Unmanned Ground Vehicle

Effective navigation is the key technology in the field of mobile robot. This paper considers the issues of autonomous location, navigation, and control problem to improve unmanned ground vehicle’s (UGU) driving safety and stability in outdoor semi-structured environments (e.g., campus scene) with weak signal of the global position system.In order to achieve high-accuracy localization of UGV, this paper proposes a road edge extraction algoritlim based on geometry properties of3D laser data to realize the read-time pose estimation of UGV. And then integrates with lane line from image to enhance the locating reliability and stability on the road.Our approach combines the priori map to plan global path using well-known A*research algorithm and produce a piecewise continuous curve that passes through all way points in each intersection utilizing cubic spline interpolations. Based on the shape of roads, we generates local paths in the local3D map which is built online. According to traffic rules, collision avoidance, and kinematic constraints of vehicle, we select optimal path. Then, combining the optimal path and UGV’s needs, we improve the Pure Pursuit method which is among the most common approaches to the path following problem for mobile robots from two aspects. One is goal point selection on the path ahead of the vehicle, and another is speed limitation by path curvature. Our method can effectively eliminate the "Cutting Corners" which is the biggest shortcoming of the Pure Pursuit method and guarantee the safety of UGV. The capabilities of our approach are demonstrated in the experiments of some typical campus road scenarios.