Control of a fleet of vehicles using Computer Vision, Cellular Automaton and Genetic Trained Behaviour

Multirobot system control is faced on this paper. Five control layers have been identified and developed: motion control, computer vision, path following, trajectory planning and behaviour. Computer vision algorithms obtain environmental information through a zenith camera. Path following is based on inverse kinematics of the vehicle. A cellular automaton is implemented to generate the trajectory avoiding obstacles. Behaviour control is developed using a state machine trained with genetic algorithms. Functions of the state machine are neural networks. These five layers have been implemented on a demonstrator system to evaluate the efficiency of the algorithms