摘要
针对机械臂应用环境状况较复杂、不确定条件较多,文中使用基于Markov对策的算法对二维机械臂进行路径规划。二维机械臂路径规划是三维多关节机器人规划的基础。首先根据实际的工作环境设定机械臂的运动范围并选择经常出现的动作组合作为机械臂运动的基本行为集,给出各种情况可能获得的报酬,依据多智能体Q值学习算法更新每个关节的报酬值,反解出对应最大报酬值的动作组合。文中仿真绘制最佳动作组合时的运动轨迹,分别仿真绘制机械臂运动环境中无障碍与放置圆形障碍物时的二维运动轨迹,并确定轨迹的误差。
Use the algorithm based on Markov games to do path-planning for 2D robot arm in connection with its complex application en- vironment and more uncertain conditions. Path-planning for 2D robot arm is the basis of 3D path-planning. According to the actual work environment, set the ann's range of motion and select the common movements as the basic behavior set. Under kinds of conditions, the profits were shown. Then the profit of each joint was updated by the multi-agent Q-learning algorithm,and the formulas of movement's inverse kinematics are obtained. So the complexity of the algorithm is also reduced. It shows the best combination of trail, respectively, to draw the 2D motion trail in the case of barrier-free and a round obstacle,and then confirm the error of trail.
出处
《计算机技术与发展》
2012年第5期57-59,63,共4页
Computer Technology and Development
基金
国家自然科学基金(60674100)
南京航空航天大学基本科研业务费专项科研项目(NS2010069)