复杂环境建模与机器人避障规划研究

Research on complex environmental modeling and robot operation planning

针对复杂环境下的机械臂作业问题,提出了一种基于空间凸多面体集模型的避碰路径规划方法。首先,给出一种利用空间凸多面体模型的复杂物体建模方法,该方法具有一套树型递归结构框架,能精确逼近任意形状的物体,并支持子物体之间的自由度连接,从而能恰当表示如多关节机械臂和电脑桌等带有运动自由度(关节转动、柜门开合、抽屉拉出等)的物体;然后,设计了一种基于直线扫掠体序列的碰撞检测方法,它能快速、准确地进行连续运动的机械臂与空间物体间的碰撞检测;最后,采用遗传算法实现了机械臂的避碰路径规划。在桌面上摆放多个复杂形状物体后,机械臂顺利完成了避障作业规划。仿真结果验证了算法的有效性,表明了其实用价值。

For the operation of a manipulator in a complex environment,a collision avoidance path planning method on the basis of spatial convex polytope model was developed. First,a complex object modeling method utilizing spatial polytope model was proposed. The approach was characterized by a framework with a tree structure. It can precisely approximate the outline of an arbitrary object and support DOF linking among the affiliated bodies. Therefore, a moving objective with certain DOF （joint rotation, opening/ closing of cabinet door, pull-out of drawer, etc.）, such as a multi-joint robot arm and a computer table, can always be represented appropriately. Then, a collision detection method featuring a sequence of linear-swept volumes was designed. It permits efficient and accurate collision detections between the continuously moving manipulator and the environmental objectives involved. Finally,a collision avoidance path planning was realized by using the genetic algorithm. In simulations, the operational planning was implemented successfully for the case where some tasks are required to be linished on a desktop with mul-tiple complex-shaped objects settled. Experimental results demonstrate the effectiveness and utility value of the proposed method.