Intelligent Decision Making Supporter Design For Dynamic Positioning System Based On Machine Learning

As human exploration and development of the ocean gradually moves toward the deep sea,the marine positioning technology of marine structures has become an important field in marine engineering research.In recent years,with the rapid development of communication technology,computer technology and artificial intelligence,the development of unmanned systems has become a hot spot and frontier field of modern technology.On the basis of providing motion control system solutions,dynamic positioning systems need to further adapt to the needs of unmanned and intelligent marine structures,and continue to broaden the research and application in the field of large-scale cooperative control at sea.Based on the dynamic model of marine structure,mooring system,and the mathematical model of each subsystem in dynamic positioning system,the time domain simulation program of dynamic positioning system based on ROS is constructed.Aiming at the problem of setpoint selection of thruster-assisted position mooring system,deep reinforcement learning method is adopted to construct an intelligent decision making supporter.Firstly,the model-free Q-laerning method is adopted to design the intelligent supporter,and it is validated in the time domain simulation program.Secondly,a model-based reinforcemenr learning supported based on Dyna method is constructed.The nonlinear support vector regression method is used to model and train the reward function.Lastly,a deep deterministic policy gradient(DDPG)method is adopted to design the intelligent decision making supporter.The numerical simulation results show that the modelfree decision supporter can effectively solve the problem of setpoint selection of thruster-assisted position mooring system,while the modelbased supporter can effectively improve the learning speed of the process,and the DDPG based supporter can solve the problem in a continuous action space,ensuring the learning speed and stability.In summary,this paper conducted research in terms of the intelligent decision making supporter design for the thruster-assisted position mooring system.The conclusions will provide some useful guidance for offshore operations.