Learning-Based Event-Triggered Containment Maneuvering of USVs via Integral Concurrent Learning and Deep Neural Networks

This paper investigates the distributed containment maneuvering problem of multiple underactuated unmanned surface vehicles (USVs) in the presence of uncertain nonlinear dynamics, actuator saturation, external disturbances, and communication constraints. A nonlinear state transformation is developed to reformulate the original USV dynamics into an equivalent second-order nonlinear system expressed in terms of transformed surge force and yaw moment control inputs. The transformation accumulates all unknown dynamics into a single purely nonlinear lumped uncertainty, making conventional adaptive methods inapplicable. To address this challenge, a deep neural network (DNN)-based integral concurrent learning (ICL) adaptation law is proposed to estimate the lumped uncertainty directly from recorded input-state data without requiring state-derivative measurements. In addition, a curvature-aware maneuvering law is proposed to improve the smoothness of the virtual-leader trajectories, and a distributed dynamic event-triggered communication mechanism is incorporated to reduce communication burden while guaranteeing Zeno-free operation. Rigorous Lyapunov analysis establishes the uniform ultimate boundedness of all closed-loop signals. Comparative simulations validate the proposed approach and demonstrate its superior containment accuracy, robustness to nonlinear uncertainties, and communication efficiency.