Configuration Design and Planning Algorithms for Modular Robots in On-Orbit Assembly

In low-gravity environments, modular selfreconfigurable robots are attractive for tasks such as satellite reconfiguration, structural expansion, truss assembly, and robotic arm deployment, yet planning remains highly challenging due to the NP-hard nature of module assignment and the requirement to maintain both connectivity and reachability throughout reconfiguration. To tackle these difficulties, we establish an integer programming model with penalty functions that enforce connectivity and reachability, and we design a hierarchical mechanism to decouple global assignment from local motion planning. At the upper level, the proposed CMGA (Convolution-MST-Genetic Algorithm with Gaussian mutation) efficiently addresses the NP-hard assignment problem by integrating structural preprocessing, refined initialization, and Gaussian mutation to accelerate convergence and improve global exploration. At the lower level, an A* search generates feasible, connectivity- and reachability-preserving paths for individual module relocations. Experiments on truss-assembly and satellite-reconfiguration scenarios demonstrate that this layered framework significantly outperforms conventional methods in convergence speed and movement cost. Overall, the results highlight the effectiveness of combining a hierarchical optimization mechanism with cooperative spherical module design, offering practical computational tools for autonomous modular robotic reconfiguration in future on-orbit missions.