Flow field prediction-based formation surrounding algorithm for autonomous underwater vehicles

The formation surrounding control of underwater targets aims to achieve close-range and all-round monitoring tasks for the targets by autonomous underwater vehicles (AUVs). However, due to the limitations of the unknown flow velocity field and the model uncertainties of AUVs, it is difficult for AUVs to form the reliable and stable formations. Therefore, the flow field prediction-based formation surrounding algorithm for AUVs is studied. First, a distributed velocity field parameter estimator is designed, which achieves real-time prediction of the velocity field using the AUVs' actual and estimated trajectories. Based on this, a neural network-based dynamic model estimator is proposed that combines the state and environmental information of AUVs. Then, a guidance vector field based target-formation surrounding algorithm is designed to achieve target surrounding control while avoiding obstacles. Finally, the theoretical derivations of the guidance vector field and the weight update rate for the model estimator are provided. The simulation and experimental results show that the proposed velocity field prediction method can overcome the dependence on grid partitioning, and the proposed algorithm can complete target surrounding tasks in environments with obstacles and unknown velocity fields.