Evolved neural networks based on cellular automata for sensory-motor controller

Constructing the controller of a mobile robot has several issues to be addressed: how to automate behavior generation procedure, how to insert available domain knowledge effectively, and how to hybrid these methods in an integrated manner. There has been extensive work to construct an optimal neural network for controlling a mobile robot by evolutionary approaches such as genetic algorithm, genetic programming, and so on. However, evolutionary approaches have a difficulty to design the controller that conducts complex behaviors. In order to overcome this shortcoming, we propose an incremental evolution method for neural networks based on cellular automata and a method of combining several evolved modules by a rule-based approach. The incremental evolution method evolves the neural network by starting with simple environment and gradually making it more complex. The multi-modules integration method can make complex behaviors by combining several modules evolved or programmed to do simple behaviors. Simulation results show the potential of the incremental evolution and multi-module integration methods as sophisticated techniques to make the evolved neural network to do complex behaviors. In this paper, we attempt to investigate the applicability of cellular automata-based neural networks and propose sophisticated techniques for the generation of high-level behaviors.