Hovering Control of ROV Based on Event-Triggered Image Moments Nonlinear Model Predictive Control

Hovering control of remotely operated vehicles (ROVs) is a key fundamental capability for underwater operation tasks, and vision-based ROV hovering control has become a current research focus because of its significant advantages. To address the limited constraint-handling capability and high online optimization frequency in vision-based hovering control, this paper proposes an event-triggered image-moment nonlinear model predictive control method. The proposed method uses image moments as visual feedback features, constructs a prediction model in a reduced control-relevant subspace, and updates the optimal control sequence only when the event-triggering condition is satisfied. Joint simulation experiments were conducted under Ideal, Disturbed, Dynamic-target, and Complex scenarios. The results show that the proposed ETIM-NMPC method achieved effective hovering performance in the joint simulation environment, with translational RMSEs remaining below the simulation-level hovering accuracy reference of 0.15 m in all evaluated directions. Compared with IM-NMPC, ETIM-NMPC maintained comparable control performance while reducing the number of online optimization updates and the cumulative computation time. The trigger rates of ETIM-NMPC ranged from 45.42% to 78.33% across the four scenarios, indicating its on-demand update characteristic under different task conditions.