Designing Autonomous Robot Controllers for Planetary Exploration: A Model of a Mars Rover

Navigation Experiment: This research aims to investigate the possibility of using an alternative, infrared based sensor system, for planetary rovers required to operate autonomously in environments with challenging terrain. For this research, computer models of a planetary rover and Martian environment have been constructed using the ODE physics engine. The control system for the robot consists of an artificial neural network trained using evolutionary computation techniques. An adaptive threshold for the infrared sensors is evolved together with the neural control system to allow for the adaptation to various unforeseen environmental conditions. After each evolutionary process, the behavioural properties obtained are tested by measuring the generalisation capabilities of the rover when exposed to new environmental conditions, particularly rough terrain. In addition, the dynamics of the co-evolution between the controller and the threshold are analysed. These analyses show that different pathways are exploited by the evolutionary process in order to adapt the sensing abilities and control system.