An Adaptive State Observer for the Autonomous Underwater Vehicles With Uncertainties by Utilising Neural Network

ABSTRACT The article proposes a novel adaptive observer utilising a two‐hidden‐layer neural network to estimate the states, including positions, rotations and velocities in the North‐East‐Down (NED) reference frame, of Autonomous Underwater Vehicles (AUVs). The studied AUVs are characterised by uncertain and nonlinear mathematical models, as well as the effects of unknown environmental disturbances. In addition to the novelty of the observer's structure, a salient feature of the proposed observer lies in its ability to estimate accurately unmeasured states irrespective of the strong nonlinear output‐to‐state mapping of the AUVs and inaccurate output measurement. In terms of designing output‐feedback controllers, instead of proposing a particular controller, the class of output‐feedback controllers for the AUV system has been systematically investigated such that the stability of the closed‐loop AUV system is ensured when combined with the proposed observer. The efficacy of the proposed observer has been verified rigorously through theoretical analysis and simulation results.