Research On Human-machine Collaboration Method Driven By Data And Model

With the continuous advancement of scientific and technological levels,the methods of human-machine collaboration have become increasingly widespread,placing higher demands on the intelligence of industrial robots.The academic research on "human-centered" collaborative robots is challenging and significant.However,in the process of human-robot collaboration tasks,there is a problem of insufficient obstacle avoidance capability of collaborative robots,which affects the safety of task execution.To successfully accomplish human-machine collaborative tasks,it is necessary to maintain a safe distance between the collaborative robot and the operator during motion,ensuring the efficiency of the collaborative robot’s work.To address the current issue of inadequate obstacle avoidance capability of collaborative robots in human-machine collaborative scenarios and to broaden the applicability of collaborative robots in practical applications,this study focuses on the path planning for obstacle avoidance of a UR5 six-degree-of-freedom collaborative robot.Specifically,the research objectives are as follows:Firstly,to enhance the performance of deep reinforcement learning algorithms and broaden their applicability,a novel framework of data/model hybrid-driven algorithm is proposed.By combining model-driven and data-driven approaches,the learning efficiency and performance of deep reinforcement learning algorithms are accelerated.Secondly,to address the issue of insufficient obstacle avoidance capability of collaborative robots in human-machine collaborative scenarios,an IGWO-DDPG algorithm is proposed,combining the data/model hybrid-driven algorithm framework.The proposed algorithm utilizes a sixth-degree polynomial for path planning of the collaborative robot and optimizes it using the IGWO algorithm,storing the superior data in a pre-training library.The DDPG algorithm,after learning from the pre-training library,interacts with the environment,benefiting from prior knowledge.This approach aims to enhance the performance of the DDPG algorithm.Finally,Simulation experiments are conducted in a human-machine collaborative setting to validate the obstacle avoidance capabilities of the proposed algorithms.Based on a simplified human-machine collaborative automatic oiling system,two types of human-machine collaborative environments,namely static and dynamic,are constructed within simulation software.Through extensive testing,the effectiveness of the proposed algorithms is validated.