Co-Evolution of Sensor Morphology and Control on a Simulated Legged Robot

This paper discusses utilizing genetic algorithms to automatically design a suitable sensor morphology and controller for a given task in categories of environments. The type of sensors, the heading angle and the range of the sensor, and the rules the controller, are co-evolved. The described method enables the system to decipher information from the environment to determine that is relevant to completing a given task while configuring a minimal controller and number of sensors, thus increasing the overall efficiency of the robot.