面向乒乓球对弈作业的七自由度仿人臂多目标轨迹规划

Multi-objective trajectory planning of 7–DOF humanoid manipulator for Ping-Pong playing

中文关键词  仿人臂  多目标优化  能量消耗  舒适构型  知识库

英文关键词  humanoid manipulator  multiobjective optimization  energy consumption  comfortable configuration  knowledge database

中文摘要

      人类能对乒乓球拥有快速对弈作业的运动技能, 并能以一种动作能耗小且舒适自然的手臂构型进行动作, 其原因是由于人在长期的学习训练过程中积累了丰富的具有相应运动特性的“知识”信息. 受人臂动作此行为机 制启发, 提出一种七自由度仿人臂面向乒乓球作业的多目标轨迹规划方法. 该方法考虑仿人臂乒乓球作业的机构 物理约束、障碍约束与任务约束条件, 以仿人臂作业轨迹的能量消耗与臂姿构型舒适性为优化准则, 采用多目标粒 子群优化方法优选动作轨迹的决策变量获得一组最优的Pareto解集, 决策者可根据实际决策需求选择其中一非支 配解; 在此基础上利用仿人臂多目标轨迹优选方法对其整个作业任务空间学习训练, 构建仿人臂乒乓球对弈作 业“知识”库. 仿真试验结果表明该方法能较好地逼近真实的Pareto前沿, 实现全局多目标寻优, 且所优选的轨迹能 满足仿人臂乒乓球作业的机构物理约束、障碍约束与任务约束条件, 实际仿人臂动作测试也验证了该方法的有效 性.

英文摘要

      Human being can move his arm rapidly to play the Ping-Pong ball in the manner of small energy consumption and comfortable configuration. Analyze its reason, it is due to the fact that human being accumulates rich“knowledge” information with corresponding motion characteristics in brain during the long training process. Inspired by this mechanism of human being, a multi-objective trajectory planning method of a 7–DOF humanoid manipulator for Ping-Pong playing is proposed. In this method, considered the physical constraint, obstacle constraint and task constraint conditions of the humanoid manipulator, two minimization performance criteria, i.e. energy consumption and configuration comfortable level, are defined to constitute the optimization objective functions, then multi-objective particle swarm optimization (MOPSO) is used to search the trajectory decision variables to obtain the optimal Pareto set. As such, the decision maker can extract a non-dominated solution from the set according to the specific decision requirements. Based on these, a knowledge database of the corresponding operation environment would be constructed through this multi-objective trajectory optimization method training the whole operation space, which can make the 7–DOF humanoid manipulator achieve rapid operation for the Ping-Pong playing. Numerical simulation results show that this method can approximate preferably the real Pareto front and realize the global multi-objective optimization, moreover, the selected trajectory can meet the physical constraint, obstacles constraint and the task constraint of the 7–DOF humanoid manipulator for Ping-Pong playing. Actual humanoid manipulator testing results also demonstrate the effectiveness of this method.