Motion System Optimisation for Direct-Drive Selective Compliance Assembly Robot Arm

A direct-drive selective compliance assembly robot arm (DDSCARA) poses difficulties in system optimisation due to numerous design parameters and strong coupling between components. This article presents a new optimisation framework based on the characteristic parameters surrogate model to solve the computation burden of the DDSCARA motion system optimisation. The framework divides the optimisation problem into three sub-optimisations. Firstly, we construct a characteristic parameters surrogate model of the direct-drive motor (DDM) by multi-objective optimisation to reduce the training dataset size at the component level. Secondly, in the system-level optimisation, we take the cost and reliability indicators as the optimisation objectives and obtain the optimal characteristic parameters of the DDMs, motion trajectory parameters, and design parameters of other components. A pre-optimisation of the DDM characteristic parameters using gradient descent is used to accelerate the convergence and improve optimisation results. Thirdly, the closest point method and Bayesian optimisation are used to recover the DDM design parameters. For the given optimisation problem, the new framework saves 97.7% computation time compared to the traditional framework. We design an optimised prototype and conduct comparative experiments with the original prototype. The optimised prototype achieves 33% and 12.5% improvements in reliability and unit production per hour, respectively.