基于模拟退火-强化学习算法的船舶运动控制

Ship Steering Control Based on SA-Reinforcement Learning Algorithm

船舶运动控制与航行的安全性、可操纵性和经济性密切相关。本文基于模拟退火-强化学习算法提出了一种混合智能控制器,应用于船舶运动航向控制中。这种混合智能控制器发挥了各种智能算法的优势,能够克服风、浪、流等外界的干扰,船舶操纵自身的不确定性,以及干扰或仪器测量误差情况下,精确的训练数据不容易得到的缺陷。最终的仿真结果表明,当存在风浪干扰海况下,船舶航向的控制仍能取得令人满意的效果。

Ship course control is closely related with navigation security, maneuverability and economy. Based on simulated annealing and reinforcement learning algorithm, a hybrid intelligent controller was proposed to ship steering. With the advantages of various intelligent algorithms, the controller can overcome the influence of the wind, wave and flow, the uncertainty of the ship steering, and the limitation that accurate data are hard to get for the disturbance and the error in instrument measuring. Al last, the results of the simulation show that the ship course can be properly controlled when changeable wind and wave exist.