GPIO-based optimal containment control for autonomous underwater vehicles with external disturbances

This paper concerns on an optimal containment control problem for the autonomous underwater vehicles (AUVs) with unmeasurable velocity information and external disturbances. Compared with the existing AUVs system, a containment control strategy based on the generalized proportional integral observer (GPIO) is developed, which can predict polynomial-type disturbances because of the consideration of information on disturbances and their derivatives. Moreover, the reinforcement learning algorithm is employed to optimize the control performance to decrease the energy consumption when AUVs execute tasks. The modification term is added to the actor-critic neural network learning algorithms applied for performing the reinforcement learning, which can ensure that the constructed positive function is strictly positive definite for avoiding the training termination of the learning algorithms. The outputs of follower AUVs can converge to the convex hull formed by leader AUVs and all signals are semiglobally uniformly ultimately bounded. Finally, the simulation results are provided to describe the effectiveness of the designed control strategy.