任务约束下七自由度机械臂拟人运动规划

Anthropomorphic motion planning for 7-degrees of freedom manipulator with task constraints

为解决基于随机采样的算法在任务约束下运动规划存在计算量大、效率低且不考虑拟人的问题,提出一种任务约束下七自由度机械臂拟人运动规划方法.本方法在任务空间中采样,首先为机械臂末端规划出一条满足任务约束的无碰撞路径,采用高斯过程学习的方法得到机械臂末端路径点与拟人臂构型间的关系;然后完整描述机械臂自运动流形,为不满足关节限位或碰撞的拟人臂构型对应的末端路径点选取次拟人臂构型,得到任务空间与关节空间的映射关系;最后将此映射关系与任务空间下的无碰撞路径相结合用于机械臂拟人运动规划.实验结果表明:与改进快速扩展随机树(RRT*)和基于投影的方法相比,本方法规划的路径长度分别减少了55%与38%,速度分别增加了39%与68%,满足任务约束,且运动更加拟人.

An anthropomorphic motion planning method for 7-degrees of freedom manipulator under task constraints was proposed to solve the problems of large computation,low efficiency and no consideration of anthropomorphism in motion planning under task constraints based on random sampling algorithm.The proposed method sampled in the task space.First,a collision-free path satisfying the task constraints was planned for the end of the manipulator,and the relationship between the end path point of the manipulator and the configuration of the anthropoid arm was obtained by using Gaussian process regression method.Then,a complete description of the robot arm self-motion manifolds was given,and a sub-anthropomorphic arm configuration was selected for the terminal path points corresponding to the anthropomorphic arm configuration that did not satisfy the joint limit or collision,and a mapping relationship between task space and joint space was obtained.Finally,this mapping relationship was combined with collision-free paths in task space for robotic arm anthropomorphic motion planning.Experiment results show that the proposed method reduces the path length by 55%and 38%,and increases the speed by 39%and 68%than the rapidly-exploring random trees star(RRT*)algorithm and projection-based methods,respectively,which satisfies the task constraint and is more anthropomorphic in motion.