Multi-Joint Active Collision Avoidance for Robot Based on Depth Visual Perception

Dear editor, Human-robot collaboration is a research topic that has numerous potential applications, such as in smart cities. An important safety consideration human-robot collision avoidance [1]. Many studies have prioritized of the robot end-effector. However, multi-joint (whole-body) also very complex working scenarios. Several achieved by estimating distance between obstacles and control points placed on body, which adds computational effort to collision-detection process. Therefore, we combine skeleton with depth visual perception achieve fast response. This method can be divided into detection strategies, illustrated Fig. 1. In detection, use statistical filtering algorithm denoise data, propose real-time obstacle distance-estimation based skeleton. strategy, repulsive force model established artificial field, motion strategy developed obtain collision-free trajectory. Real-time experiments 7-DOF collaborative are conducted verify our approach. The experimental results show enables capability active for both end-effector whole body. avoid complete desired tasks smooth