A new approach for parameter optimization of parallel mechanisms based on Stackelberg game theory and Gaussian process regression

Abstract

Parallel mechanisms-based industrial robots have been widely applied in various industries, such as food, electronics, aerospace, etc. Undoubtedly, the performance of parallel mechanisms plays a crucial role in the working capability of robots. However, in order to achieve a comprehensive optimal performance, more and more performance indices (e.g. kinematic, rigid-body dynamic, stiffness, and elastodynamic performances) and design variables have to be considered, which results in the fact that solving such a multi-variable multi-objective optimization problem becomes an extremely challenging task. To address this problem, this paper proposes a new parameter optimization approach based on Stackelberg game theory and Gaussian process regression. Firstly, the design variables and optimization objectives are categorized into two groups, that is, the leader and the follower, and parameters of each group can be optimized separately, which can effectively reduce the complexity of the optimization problem. Besides, the time-consuming performance indices are assigned to the follower to improve the efficiency of the optimization process. Without loss of generality, an integrated parameter optimization is conducted on Delta robots to verify the effectiveness of the proposed parameter optimization approach.