履带平台无人驾驶系统基于语义信息的模块串联方法

To improve the performance of unmanned tracked platforms in complex off-road environment, this study proposes a semantic information-based method to integrate the environment perception, motion planning and motion control modules of the unmanned system. Through the fusion perception method of camera and lidar, image semantic information and lidar point cloud features are obtained by the environment perception module, and then the traversable area is roughly determined by combining geometric parameters of the platform. Finally, a three-dimensional grid map with different traversability information is generated, after the Gaussian clustering model is used to conduct a fine traversability analysis of the environment in the traversable area. Platform kinematics and dynamic constraints are considered in the motion planning module. On the basis of smooth transition constraints, terrain and road surface attributes are taken into account to generate behavior motion primitives, and the trajectories with behavior semantics are generated using hierarchical online selection of primitives. The motion control module is based on model predictive control which considers the execution ability, tracking deviation, and control stability of the platform drive motor. Based on the semantic attributes of the behavior motion primitives provided by motion planning, the weight coefficient of the control objective function is updated in real time to reduce lateral deviation, heading deviation, and speed deviation of trajectory tracking. Finally, the medium-sized hybrid unmanned tracked platform is used for verification. The results show that the motion planning method proposed in this paper are 46. 1%, 46. 5% and 14. 2% lower in average pitch angle, average roll angle and average curvature than the trajectory planning result using the plane binary grid map. Compared with the motion control method with fixed parameters, the proposed variable-parameter motion control method based on behavior semantic trajectory reduces the lateral deviation, heading deviation, and speed deviation by 17. 1%, 2. 7% and 6. 1%, respectively.