Model predictive control of an autonomous underwater vehicle with a fuzzy objective function optimized using a GA

Abstract

Recently, unmanned underwater vehicles (UUVs) have emerged as a viable tool for ocean exploration and for military purposes. This is due to the inability of human divers to reach deep sea and the hostile nature of underwater environment. UUVs are of two types, namely, remotely operated vehicles (ROVs) and autonomous underwater vehicles (AUVs). This paper is concerned with the control of an AUV. A model predictive controller is developed herein where the traditional cost function has been replaced by a fuzzy performance index which represent the goals and constraints of the problem. Since fuzzy logic is basically derived from knowledge of human expertise, it is therefore more intuitive than a conventional cost function. Moreover, the choice of aggregation operator can lead to significant reduction in tuning time which is essential for a quadratic objective function. A genetic algorithm (GA) is used as an optimization tool to evaluate the control inputs by minimization of the performance index represented by fuzzy membership values. The resulting controller is applied to an AUV simulation model obtained from system identification techniques on test trials data. Simulation results are presented that demonstrate the efficacy of the approach.