Design of a Navigator for the Optimized Path-Tracking of Underwater ROVs using a Nero-Genetic Fuzzy Controller

This paper proposes a sub-time-optimum soft-computing based controller to follow a desired path with a desired velocity by mobile robots with identified dynamical behavior. This method consists of a fuzzy controller where a trained neural network sets its membership functions values in on-line mode. Training of the network is done by a genetic algorithm for various vehicle initial positions and different path convexities in off-line. After the training of the network, during vehicle motion, it retrieves sub-optimized fuzziness values and sends them to the fuzzy control algorithm according to the vehicle position. Meanwhile, the influential of the path convexity is considered in fuzziness of membership functions. This method leads us to make almost the best decision for the mobile robot at each moment. This method is applied to an underwater remotely operated vehicle (ROV) to develop an autopilot for its control system. Simulation results show good performance of the method in this specifics problem.