Multiobjective Joint Optimization for Trajectory and Controller Parameters to Enhance Hydraulic Legged Robot Jumping Performance

Abstract

Vertical jumping performance of the limb leg unit (LLU) significantly impacts the movement of legged robots. In this article, a multiobjective joint optimization method for the trajectory generator and controller coupling parameters is proposed. This method constructs a quasi-realistic simulation model with physical feasibility constraints, ensuring an accurate representation of the physical prototype performance. Subsequently, the design variables and objectives are defined considering the inherent coupling relationships and complex evaluation metrics of the LLU. By leveraging data from the proposed model, the empirical objective functions are formulated in the form of a response surface model and used to analyze the influence of variables on different objectives. An improved genetic algorithm with weight selection is proposed to prioritize maximizing the jump height while satisfying the optimization objectives. Finally, the accuracy of the quasi-realistic model is validated by experiments and the results demonstrate an increase of 11.95% in jump height and a decrease of 6.60% in force ratio compared with the conventional method. The joint optimization method can serve as a valuable reference for parameter selection in other multivariable, multiobjective, and multiconstraint problems.