【摘要】 本文实现了一个基本的机器人规划系统。它能自动生成一系列避免与障碍物发生碰撞的机器人动作轨迹。其视觉子系统可获取环境知识。算法对可能引起碰撞的障碍物进行从直角坐标空间到机器人关节坐标空间的转换。并采用分级式方式对自由空间进行先“粗”后“细”的两级描述。合理的数据结构既减少了存储量又给出了充分的启发信息。在此基础上又采用了先“全局”后“局部”的两级路径优化,从而能以较快速度决策出一能避免碰撞且运行时间短的优化路径。实验表明,该系统可在一般 PC 机上实现。更多还原

【Abstract】 This basic robot planning systcm chooscs a motion path to avoid collision with obstaclcs inworkspace.With a camera above the workspace the computer vision can give a simple dcscription of theworld model.The presented algorithm transfers a cartesian description of obstacles into the coordinatcs ofthe first three joints of a manipulator.The free space is described hierarchically by a two-level representa-tion,and a two-level optimization is used for the path planning.Experiments show that the whole systemcan be realized by an IBM—PC with very small memory.The average computation time for one obstacle is10s.更多还原