Robot Arm Path Planning with Adaptive Obstacle Avoidance for Man–Robot Collaboration

Abstract Robot arms have been widely used in various production factories. They are able to complete desired tasks, such as picking and placing, with good repeatability. However, robots cannot completely replace human workers due to many different reasons. Human workers can complete delicate tasks more effectively with their skillful hands. Robots could be human workers helpers in terms of picking and placing items, delivering items to humans, lifting items for humans, etc. However, the risk of harming human workers greatly increases as the robots get closer to them. Recently, researchers began to develop advanced technologies for human–robot collaboration. In this paper, a novel system will be presented. A spatial-temporal graph network was used to identify human motions, and the random forest model was used to evaluate the danger factor between the human and the robot in the robot’s moving path. A Lagrangian minimization was used to determine a new robot’s moving trajectory to keep a safe distance from humans. The safety distance could be adaptively shortened as the robot moves closer to humans for specific man–robot collaboration missions.