Space robots play a crucial role in on-orbit servicing tasks, including spacecraft life extension, on-orbit maintenance, and the removal of failed satellites. Their success in on-orbit capture missions depends not only on precise control strategies but also on good design schemes. Given the diverse nature of space targets, a cooptimization approach for space robot design and control tailored to different mission objectives is urgently needed. This can enhance the cost efficiency, reliability, and intelligence of space robots, ultimately strengthening their on-orbit servicing capabilities. To tackle the challenges of component-level optimization and the high computational cost associated with performance evaluation in this cooptimization process, we introduce a multigranularity evolutionary method. Specifically, we provide a general representation of space robot design and control, utilizing a genetic programming tree structure to effectively articulate component-level design schemes and control parameters. By establishing the multigranularity model, we significantly reduce the computational burden associated with performance evaluation. Additionally, we establish criteria for switching model granularity by quantifying the uncertainty within the multigranularity model, allowing for adaptive transitions during the optimization process. Numerical simulation results demonstrate that the proposed method successfully realizes the cooptimization of space robot design and control for diverse targets. Moreover, in contrast to the method of enumerating all potential morphologies and optimizing them via genetic algorithms, the proposed space robot design and control description method maintains better adjacency among similar space robots, facilitating faster convergence and yielding superior objective function values.
Xie et al. (2026) studied this question.