Research on the humanlike trajectories control of robots based on the k-nearest neighbors

Abstract

At present, industrial robot arms have been widely applied in actual production. Especially for some human-robot cooperation work scenarios, involving the interaction between human and robot in sharing workspace, the requirement of trajectory planning for safety and natural interaction was more demanded. Aiming to solve the problems of the complex planning of trajectory and low security of man-machine cooperation, this paper introduced a control method of human-like trajectory based on the k-nearest neighbors. By using the Kinect equipment, the trajectories of wrist joint of human arm were acquired during manipulation tasks and then the k-nearest neighbors algorithm was adopted to analyze the trajectories. Consequently, the control laws of human-like trajectory of robot end-effecter are extracted based on the collected motion data. The corresponding experiments are conducted to verify the proposed method and this method can generate the human-like trajectory for robot manipulators, avoiding the collision with human arm and obstacles. It also enhances the performance of security and makes robots more human-robot friendly.