Obstacle avoidance planning for industrial robots based on singular manifold splitting configuration space

Abstract

Obstacle avoidance planning is the primary element in ensuring safe robot applications such as welding, assembly, and drilling. The states in the configuration space (C-space) provide the pose information of any part of the manipulator and are preferentially considered in motion planning. However, it is difficult to express the environmental information directly in the high dimensional C-space, limiting the application of C-space obstacle avoidance planning. This paper proposes a singular manifold splitting C-space method and designs a compatible obstacle avoidance strategy. The specific method is as follows: first, according to the specific structure of industrial robots, arm-wrist separation obstacle avoidance planning is proposed to fix the robot as a 3R manipulator to reduce the dimension of C-space. Next, the C-space is segmented according to the singular manifolds, and the unique domain is delineated to complete the streamlining of the volume of the C-space. Then, with the help of the point cloud, the obstacles are enveloped and mapped to the unique domain to construct the pseudo-obstacle map. Industrial robots' obstacle avoidance planning is completed based on the pseudo-obstacle map combined with an improved Rapidly-Exploring Random Trees (RRT) algorithm. This method dramatically improves the efficiency of obstacle avoidance planning in the C-space and avoids the effect of singularities on industrial robots. Finally, the effectiveness of the method is verified by physical experiments.