Online Parameter Optimization in Robotic Force Controlled Assembly Processes

In the high precision robotic assembly processes, the process parameters have to be tuned in order to adapt to variations and satisfy the performance requirements. However, because of the modeling difficulty and low efficiency of the existing solutions, this task is usually performed offline. In this paper, an online parameter optimization method is developed. Gaussian Process Regression(GPR) is utilized to model the relationship between the process parameters and system performance. The GPR surrogated Bayesian Optimization Algorithm(GPRBOA) is proposed to optimize the process parameters. To reduce the risk of converging to a local minimum, a random variation factor is added to the Lower Confidence Bound(LCB) acquisition function to balance the exploration and exploitation processes. To deal with the computational burden of GPR, a switching criterion is proposed to coordinate the optimization process and production process to reduce the computational complexity. Experiments were performed using a peg-in-hole process. The experimental results verify the effectiveness of the proposed algorithm and demonstrate its efficiency and accuracy compared to Design Of Experiment(DOE) methods. The proposed method is the first attempt of model-driven assembly process parameter optimization and will generate big economic impact.