Nonlinear Model Predictive Control Applied to an Autonomous Underwater Vehicle

Abstract

This paper presents an experimental analysis of a nonlinear model predictive controller combined with an integral sliding mode control component and an explicit dead-time compensator applied to the closed-loop control of FlatFish, an autonomous underwater vehicle (AUV). The proposed strategy is composed of three control instances that can be integrated to deal with constraints, disturbances, and dead time. The integral sliding mode control instance is considered to attenuate unmeasured disturbance effects, the filtered Smith predictor component is used to compensate for the dead-time effect, and the model predictive control layer aims at systematically optimizing closed-loop performance with constraint satisfaction. The discussion highlights that any of these control layers can be either included or removed from the control chain, which depends on the challenges presented by the AUV problem. The benchmark for such evaluation is the original control chain of FlatFish, which is based on a P cascade controller. Experimental tests were performed with FlatFish in a saltwater tank located in Bremen, Germany, and showed the superior performance of the proposed system. The experimental results show that the proposed control chain may be effectively used to provide improved tracking performance as expected from theoretical analysis.