INTELLIGENT FAULT TOLERANT CONTROL SCHEMES FOR AUTONOMOUS UNDERWATER VEHICLES

Abstract

The area of autonomous underwater vehicles (AUVs) is an increasingly important area of research, with AUVs being capable of handling a far wider range of missions than either an inhabited underwater vehicle or a remotely operated vehicle (ROV). One of the major drawbacks of such vehicles is the inability of their control systems to handle faults occurring within the vehicle during a mission. This study aims to develop enhancements to an existing control system in order to increase its fault tolerance to both sensor and actuator faults. Faults occurring within the sensors for both the yaw and roll channels of the AUV are considered. Novel fuzzy inference systems (FISs) are developed and tuned using both the adaptive neuro-fuzzy inference system (ANFIS) and simulated annealing tuning methods. These FISs allow the AUV to continue operating after a fault has occurred within the sensors. Faults occurring within the actuators which control the canards of the AUV and hence the yaw channel are also examined. Actuator recovery FISs capable of handling faults occurring within the actuators are developed using both the simulated annealing and tabu search methods of tuning FISs. The fault tolerance of the AUV is then further enhanced by the development of an error estimation FIS that is used to replace an error sensor. It concludes that the novel FISs designed and developed within the thesis provide an improved performance to both sensor and actuator faults in comparison to benchmark control systems. Therefore having these FISs embedded within the overall control scheme ensure the AUV is fault tolerant to a range of selected failures.