Hybridized GP and self-adaptive DE for morphology-controller co-evolution of heterogeneous modular robots

This paper explores the use of hybridized Genetic Programming and self-adaptive Differential Evolution algorithm to automatically design and co-evolve both the controller and morphology of heterogeneous swarm modular robots. A novel tree-based structure is proposed and implemented for the modular robot structure and ANN representation, which allows the evolutionary optimization of the co-evolution of morphology and controller to be carried out simultaneously. Two different simulations are conducted in this paper with the aim of co-evolving the multi-branching modular robots to develop moving behavior and the snake-like modular robots to crawl through a narrow path. The results from the simulation show the promise of this work and illustrate the importance of co-evolving both the robots´ bodies as well as their controllers.