Abstract: Industrial robots are significant equipment for intelligent manufacturing, it is an essential trend for industrial robots to improve their perception and decision-making capability. As the joint stiffness of industrial robots is weak which results in poor dynamic performance, there is deviation between the actual trajectory and the desired trajectory during the working process, and it is generally necessary for the operator to use a teach pendant to debug the robot for a long time according to the actual running trajectory. In order to improve the trajectory accuracy and intelligent decision-making ability of industrial robots, this paper proposes a digital twin model construction method for industrial robots oriented to trajectory dynamic perception and autonomous decision-making, designs a digital twin framework for industrial robots with the ability of trajectory interaction between the real and the virtual, which ensures the dynamic perception and real-time mapping of the twin model on the trajectory of the physical entity by fusing the multi-origin heterogeneous in the process of the robot's operation. A method for evaluating the accuracy of industrial robot trajectory based on the trajectory error tolerance range threshold is proposed, and a polynomial function-based short-time evolution prediction strategy for the end trajectory is established by combining the robot twin model and trajectory error model, which implements an autonomous decision-making process to judge the deviation of the robot trajectory by the digital twin model when the trajectory of the physical robot is about to deviate from the desired trajectory. According to the decision result, a new trajectory is intelligently planned and the robot entity is controlled to execute. Finally, the effectiveness of the method is verified on the UR5 robot, which implements the evolutionary prediction-based autonomous decision-making control of the end trajectory, and improves the level of intelligence and robustness of the trajectory control of industrial robots.

Key words: digital twin, industrial robot, dynamic perception, virtual-reality interaction, autonomous decision-making