Genetic fuzzy system lets robot teams haul cargo on planetary terrain

Researchers Daegyun Choi and Donghoon Kim have introduced a new approach to multi-robot coordination for planetary missions, detailed in a paper submitted to arXiv (2608.12755) and presented at the 2021 AAS/AIAA Space Flight Mechanics Meeting. The method, called a genetic fuzzy system (GFS), enables a decentralized multi-robot system (MRS) to collaboratively transport objects across unstructured environments while minimizing the total path length and avoiding obstacles. The system first performs terrain traversability analysis on an elevation map, using slope information to identify non-traversable areas and converting the map into a 2D traversability grid, which simplifies path planning and obstacle avoidance.

During training, the fuzzy inference systems (FISs) that generate velocity commands for each robot are optimized using a genetic algorithm across multiple challenging scenarios, including local minima, targets positioned close to obstacles, and cluttered environments. The trained FIS models are then applied to test environments created from the traversability maps and validated across several test cases. By combining fuzzy logic's interpretability with genetic optimization's search power, the approach demonstrates a practical path toward decentralized, adaptive coordination for robot swarms operating on rough, alien terrains—key for future planetary exploration and habitat construction missions.