Machine Learning and Control Advancements in Marine Robotics

The field of marine robotics is undergoing a significant transformation with the integration of machine learning and data-driven intelligence into control strategies. Traditional model-based methods are being complemented or replaced by innovative approaches that can better handle the complexities and uncertainties of marine environments. Notable advancements include the development of nonlinear feedback controllers that account for actuator constraints and disturbances, as well as the application of machine learning algorithms for real-time prediction and control. These innovations are enabling more precise and efficient control of marine robots, including unmanned surface vessels (USVs) and autonomous underwater vehicles. The use of data-driven methods is allowing for more accurate modeling and prediction of complex marine systems, and is paving the way for the development of more autonomous and adaptable marine robots. Some noteworthy papers include: The application of machine learning to the motion response prediction of floating assets, which achieved mean prediction errors of less than 5% for critical mooring parameters. The development of a nonlinear guidance scheme for interceptor-equipped seekers with bounded input, which effectively accounts for seeker field-of-view and actuator limitations. The review of recent progress in marine robot control, which highlights notable achievements in data-driven control and provides a roadmap for future research.

Sources

An application of machine learning to the motion response prediction of floating assets

Trajectory tracking control of USV with actuator constraints in the presence of disturbances

Time-Constrained Interception of Seeker-Equipped Interceptors with Bounded Input

Control of Marine Robots in the Era of Data-Driven Intelligence

Active Disturbance Rejection Control for Trajectory Tracking of a Seagoing USV: Design, Simulation, and Field Experiments

Built with on top of