Cooperative Robot Swarm Locomotion Using Genetic Algorithms

This paper focuses on the design of a decentralized controller for cooperative locomotion in a swarm of robotic agents. Each agent is assumed to be able to locomote over even terrain, grasp other robots, and climb onto a stack of robots. Genetic algorithms and genetic programming methods are applied to develop locomotion controllers for single units as well as decentralized controllers intended to allow the swarm of units to pass over uneven terrain that no single unit can surmount. The goal is to develop a system that provides a high probability of large numbers of units passing through sets of test obstacles. The results are shown through simulation exercises.