Learning optimal measurement and control of assembly robot for large-scale heavy-weight parts

Due to their advantages of high speed, high accuracy, high flexibility, and low cost, assembly robots are widely used in electronics and automotive industries. However, it is still a significant challenge for large-scale, heavy-weight part assembly using industrial robots. First, the deformation and motion errors of industrial robots caused by big payload cannot meet the accuracy requirement of large structure assembly. To solve this problem, an online kinematics compensation method based on Gaussian Process Regression is developed to predict and compensate the deformation and uncertainties of a large structure assembly robot. Second, before the assembly process, the optimal assembly path has to be planned. To this end, we propose an assembly path planning method based on learning from demonstration. Finally, an event-based control method is deployed to achieve optimal assembly cycle time to improve assembly efficiency and performance. An experimental system is developed to validate the proposed algorithm for large structure assembly and the results demonstrate that the proposed method can improve the assembly efficiency by more than 40%.