A genetic algorithm based nonlinear guidance and control system for an uninhabited surface vehicle

There is an increasing drive to develop uninhabited surface vehicles (USV) as cost effective solutions to a number of naval and civilian problems. In part, the resolution of these problems relies upon such vehicles possessing robust guidance and control (GC) systems. Furthermore, the vehicles need to be operated under tight performance specifications satisfying multiple constraints simultaneously. This requires vehicle nonlinearities and constraints to be explicitly considered in the controller design. Nonlinear model predictive control (NMPC) is well suited to satisfy these requirements. This paper reports the design of a novel GC system based on NMPC for use in a USV named Springer which is benchmarked against a linear proportionalintegral-derivative counterpart. The NMPC combines a recurrent neural-network model and a genetic-algorithm-based optimiser. Common to the two GC systems is a waypoint line-of-sight (LOS) guidance subsystem. The control objective is to guide the vehicle through different waypoints stored in a mission planner. The performances of the guidance and control systems are evaluated and compared in simulation studies with and without appropriate disturbances. From the results presented, it is concluded that the GC system based on NMPC is more efficient and more capable to guide the vehicle through LOS waypoints particularly in the presence of disturbances.