Inner–Outer Loop Intelligent Morphology Optimization and Pursuit–Evasion Control for Space Modular Robot

Abstract

This paper proposes an inner–outer loop computational framework to address the morphology optimization and pursuit–evasion control problem for space modular robots. First, a morphological design space considering the functional characteristics of different modules is designed. Then, an elite genetic algorithm is applied to evolve the morphology within this space, and a proximal policy optimization algorithm is applied to control the space modular robot with evolved morphology. Considering symmetry, centrality, module cost, and average cumulative reward, a comprehensive morphological assessment is proposed to evaluate the morphology. And the assessment result serves as the fitness of evolution. In addition, by implementing the algorithm on the JAX framework for parallel computing, the computational efficiency was significantly enhanced, allowing the entire optimization process within 17.3 h. Comparative simulation results verify the effectiveness and superiority of the proposed computational framework.