Penguin Huddling inspired Distributed Boundary Movement for Group Survival in Multi-robot Systems using Gaussian Processes

Traditionally, research on mobile robot adaptability to damaging external stimuli has only focused on making individuals capable of surviving on their own, which is often very expensive and lacks robustness. In this paper, we present our initial work on an Emperor Penguin huddling-inspired group survival methodology for a multi-robot system exposed to a damaging directional external stimulus. Individuals on the stimuli side are able to relocate to the safest location on the leeward side of the group in turns without any communication requirements or prior knowledge of the group size or shape, for prolonged survival of the group as a whole. A distributed boundary movement method is proposed with Gaussian Processes machine learning that allows individuals to relocate to the robot health-loss-rate global minima around the group boundary only using stimuli measurements. Simulation results validate our proposed methodology of successfully relocating individuals. The robustness of the proposed method was tested under different formation sizes and shapes. In all cases, successful relocation was obtained. As an effect, significant improvement in survivability of the robot group as a whole was obtained.