An underactuated ship parallel control method based on data physics-driven reinforcement learning

By constructing a three-degree-of-freedom underactuated ship mathematical model with unknown dynamic characteristics and a recurrent neural network, and combining reinforcement learning and optimal control methods, a path-tracking parallel feedforward controller is designed. This solves the problems of high cost due to reliance on physical models and weak physical interpretability of data-driven methods in existing technologies, and achieves efficient and reliable path-tracking control for underactuated ships.

CN121956587BActive Publication Date: 2026-06-26DALIAN MARITIME UNIVERSITY
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
DALIAN MARITIME UNIVERSITY
Filing Date
2026-03-31
Publication Date
2026-06-26

AI Technical Summary

Technical Problem

Existing control schemes rely on accurate ship dynamics physical models, which are costly and difficult to match actual working conditions. Purely data-driven methods have weak physical interpretability and cannot meet the comprehensive requirements of high safety, high reliability, and low cost.

Method used

A mathematical model of a three-degree-of-freedom underactuated ship with unknown dynamic characteristics is constructed. By combining recurrent neural networks and real ship data, a path-following parallel feedforward controller is designed. The reinforcement learning optimal control method is adopted to realize data-physical driven parallel control.

Benefits of technology

It significantly improves the generalization ability and operational reliability of the control system, reduces the power consumption of the propulsion system and the frequency of rudder action, balances navigation accuracy and economy, lowers the threshold for technology implementation, and promotes the unmanned and intelligent upgrading of ships.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121956587B_ABST
    Figure CN121956587B_ABST
Patent Text Reader

Abstract

The application discloses a kind of data physical drive reinforcement learning underactuated ship parallel control methods, it is related to ship motion control technical field, including, for the three degrees of freedom underactuated ship of dynamic characteristics and disturbance unknown, establish its mathematical model, fusion offline data and recurrent neural network are constructed data physical drive ship model;Based on this model and online ship state information, a virtual ship system is established to map the real ship;Using backstepping method and dynamic surface control technology, a path tracking parallel feedforward controller is designed on the virtual system;Combined with the optimal control method of reinforcement learning, an approximate parallel optimal controller is designed;Integrate the outputs of the two controllers, generate the actual control command after constraint processing, realize the parallel tracking control of ship path;The application can dynamically adapt to ship loading changes and complex navigation conditions, improve the generalization ability and operation reliability, reduce equipment overconsumption and energy waste, solve the inherent limitations of single target optimization of traditional control method.
Need to check novelty before this filing date? Find Prior Art