Efficient Human-Robot Safety Framework Based on Human Hand Trajectory Prediction and Path Replanning

To achieve safe and efficient human-robot interaction(HRI), it is essential to predict how operators will move during the execution of tasks by both humans and robots. Existing research in human-robot interaction often focuses solely on human hand predictions, frequently overlooking the significance of robotic arm movements in maintaining safety. Addressing safety concerns in dynamic and unstructured environments, this paper introduces a Predictive Obstacle-Aware Rapidly-exploring Random Trees Connect(POA-RRT-Connect) approach. This method utilizes a Hidden Markov Model (HMM) to predict human motion over forthcoming intervals. Based on these predictions, the results are forwarded to the robotic arm's collision detection algorithm, which checks for potential collisions and plans an optimal collision-free path if none are detected. Experimental results show that compared to our proposed method, the Lazy RRT algorithm reduces planning time by approximately 71.19%, and by about 72.82% when compared to the RRTextend algorithm. In terms of path efficiency, it saves approximately 36.87% of the path length compared to our proposed method, and about 7% compared to RRT*. Consequently, the method proposed in this paper is capable of identifying an optimal collision-free path within a shorter timeframe for human-robot interaction tasks.