Robot assembly motion planning method based on demonstration trajectory

Abstract

The invention belongs to the field of robot autonomous assembly. The invention particularly relates to a robot assembly motion planning method based on a demonstration track. The method comprises the following steps: a robot demonstration learning stage: obtaining an assembly demonstration track of a robot, wherein the robot learns assembly operation skills based on the demonstration track; a multi-modal representation learning stage: the robot introduces noise and environment interaction by using an imitation learning strategy to obtain multi-modal data and labels, and learns the representation of the corresponding multi-modal data by fully considering the energy consumption and singularity of the robot in the movement process; and a robot reinforcement learning stage: building a robot trajectory planning reinforcement learning process based on the multi-modal representation and demonstration trajectory, and performing reinforcement learning training to obtain a final planning strategy. The method can be applied to a robot system which comprises a demonstration assembly track and needs to realize a flexible assembly task in an uncertain environment, and is used for improving the success rate and efficiency of the assembly task.

Description

technical field [0001] The invention belongs to the field of robot intelligent assembly, in particular to a robot assembly motion planning method based on a demonstration trajectory. Background technique [0002] With the rapid development of intelligent manufacturing, industrial robots with high precision, high speed and strong repeatability are widely used in industrial production. At present, the traditional control methods, such as PID control and impedance control, are a kind of compliant robot control method that uses a small amount of adaptive means to achieve a compliant robot along a preset trajectory. However, due to the machining error of parts, the complex contact state during assembly, the inherent mechanism in the robot assembly process and other factors, many robot assembly tasks require strong adaptive ability, and it is difficult to complete high-precision, multi-contact assembly only by traditional control methods. Task. [0003] In recent years, control ...

AI Technical Summary

Problems solved by technology

It aims to solve the problem of trajectory planning and control of robots in complex, time-varying, and unstructured environments

Images