基于约束采样RRT的机械臂运动规划

Constrained sampling method based RRT algorithm for manipulator motion planning

鉴于快速搜索随机树(RRT)方法进行机械臂运动规划时,存在节点利用率低、拓展导向性差、路径粗糙等问题,提出一种基于约束采样的RRT算法。改进了稀疏节点产生机制,通过减少重复性采样提高了全局搜索效率;提出动态采样域策略,根据采样情况动态调整采样区域大小来提高拓展方向的确定性,从而减少无用节点;采用贪婪策略提高节点的利用率,减少迭代次数。搜索完成后,剔除路径中的冗余节点,并用B样条曲线进行轨迹平滑,提高了轨迹质量。通过二维地图、Robotic System Toolbox仿真实验,以及机器人操作系统中的机械臂避障测试和真机实验,验证了该算法的可行性和有效性。

Aiming at the problems of Rapidly-exploring Random Tree(RRT)for manipulator motion planning such as low nodes utilize efficiency,poor expansion guidance and poor path quality,a constrained sampling method based RRT algorithm was proposed.The sparse node generation mechanism was improved,which significantly enhanced the global search efficiency of RRT through reducing the repetitive sampling.Besides,a dynamic sampling region strategy was proposed to dynamically adjust the size of the sampling region and improve certainty of growth direction,so as to reduce the useless nodes.The greedy strategy was applied to improve the utilization rate of nodes and reduce iterations.After searching DSSP-RRT,β-spline curve was used to smooth the trajectory and improve its quality.Through the obstacle-avoidancing simulation under different situations and real UR5 experiment,the effectiveness of the proposed algorithm was proved.