Evolving predator control programs for a hexapod robot pursuing a prey

Control program learning systems for autonomous robots are important to assist in their development and to allow them to adapt to changes in their capabilities and/or the environment. A common method for learning in robotics is evolutionary computation (EC) and a good problem to demonstrate the effectiveness of a learning system is the predator/prey problem. In previous research, we used a cyclic genetic algorithm (CGA), a form of EC, to evolve the control program for a predator robot with a simple sensor configuration of 4 binary sensors, which yielded 16 possible sensor states. In this paper, we present the use of a CGA to learn control for a predator robot with a more complicated sensor setup, which yields 64 sensor states. The learning system successfully evolved a control program that produced search, chase, and capture behavior in the simulated predator robot.