Mean-Field Based Time-Optimal Spatial Iterative Learning Within a Virtual Tube

摘要

Navigating large numbers of autonomous robots through obstacle-dense environments poses a significant challenge in swarm robotics. This letter proposes an innovative approach that merges iterative learning (IL) with mean field feedback to optimize swarm navigation efficiency while aiming to achieve predefined global configurations. The proposed framework utilizes historical data to improve navigation efficiency, ensuring both safe and fast navigation. Key contributions of this letter encompass the extension of macroscopic learning within the virtual tube, the refinement of swarm navigation efficiency, and the expansion of theoretical and practical horizons for IL applications. This letter offers a novel perspective for developing more effective swarm robotics solutions tailored for obstacle-dense environments.