Study Of Active Safety Strategies For Robots During Human-robot Collaboration In Industrial Environments

Modern manufacturing processes emphasize customization and flexibility,and human-robot collaboration can combine the flexibility of workers with the benefits of industrial robots,making it ideal for modern manufacturing.However,ensuring operator safety is the primary challenge in human-machine collaboration tasks,and autonomous dynamic obstacle avoidance of robots is the key to achieving safety in human-machine collaboration.In this paper,we combine the robot safety standard SSM to transform the human-robot collaboration safety problem into a safe dynamic obstacle avoidance planning problem for robotic arms by applying the minimum safe distance.In this paper,a real-time safe collision-free motion planning method for human-machine collaborative robots based on the Go-Explore algorithm is proposed to address the shortcomings of traditional path planning methods and deep reinforcement learning methods,which accomplishes the specified task using simple and clear sparse rewards.In order to simplify the calculation of the shortest distance between the robot arm as a whole and the obstacles,the OBB wraparound box technique is used to wrap the seven-degree-of-freedom robot arm;to improve the algorithm’s ability to predict the motion trajectory of obstacles,the GRU network structure is designed;for the experiments in this paper,appropriate cell representations,state space and action space are designed according to the proposed algorithm,etc.,and the algorithm flow framework is given.Unlike previous robotic arm dynamic obstacle avoidance tasks based on deep reinforcement learning methods that are mainly implemented on reach base tasks,this paper designs an end-to-end dynamic obstacle avoidance task for a seven-degree-of-freedom robotic arm on pick-place base tasks based on the powerful exploration performance of the Go-Explore algorithm,and this innovation improves the applicability of reinforcement learning methods in practical complex tasks of human-machine collaboration This innovation improves the applicability of reinforcement learning methods in practical complex tasks of human-machine collaboration.Finally,an experimental simulation environment based on the Mujoco physics engine is built to test the proposed approach.Two experimental tasks are designed: in the pick-place task of the robotic arm,the effectiveness and excellent exploration performance of the Go-Explore algorithm are demonstrated by comparing with the counting-based PPO algorithm,and the importance of the SIL algorithm is demonstrated;in the safe dynamic obstacle avoidance task of the robotic arm,the motion rules of the obstacles are designed,and the experimental results show that the robotic arm can accomplish the original basic task goal while ensuring the safety of human-machine collaboration.