Predictor-based model reference adaptive control of an unmanned underwater vehicle

Unmanned Underwater Vehicles (UUVs) are being deployed in advanced applications that require precise manoeuvring close to complex underwater structures such as oilrigs and subsea installations or moving objects such as ships and submarines. The effect of vehicle’s hydrodynamic parameter variations is significant in such scenarios and in extreme conditions the UUV may experience loss of control. In addition, external disturbances and actuator failures degrade the performance of the UUV. Adaptive control has been identified as a promising solution that can improve the performance in such situations. However, adaptive control is not widely used in UUVs mainly due to the trade-off between fast learning and smooth control signals. The latter can be guaranteed at low learning rates but require additional input to improve learning. The Predictor Model Reference Adaptive Control (PMRAC) is one such method that uses a prediction error to improve learning. In this paper, the performance of PMRAC in UUV applications is investigated and compared to standard Model Reference Adaptive Control (MRAC) at low learning rates under normal operational conditions, partial actuator failure, and under the influence of external disturbances. Simulation results show that PMRAC significantly reduces the tracking error compared to MRAC. In addition, PMRAC is less affected and recovers quickly from actuator failure and external disturbances, while generating smooth control signals with less oscillation compared to MRAC.