再励学习在卫星姿态控制中的仿真研究

Simulation of Reinforcement Learning for Satellite Attitude Control

为了满足卫星姿态控制系统对控制精度、抗干扰和鲁棒性要求的不断提高,将模糊神经网络结合再励学习算法应用到卫星姿态控制系统中,即可以在不需要被控卫星的精确数学模型的前提下解决网络参数在线调整的问题,又可以在无需训练样本的前提下实现控制器的在线学习。最后同传统PID控制相比的仿真结果表明,基于再励学习的三轴稳定卫星姿态控制系统不仅可以达到卫星姿态控制任务对控制精度的要求,还可以有效地克服干扰,从而达到了在线学习的目的。

To meet the higher requirements such as accuracy, disturbance rejection ability and robustness in satellite attitude control system, a fuzzy neural control approach based on reinforcement learning applied to the three - axis stabilized satellite is presented to solve the online learning problem without the satellite mathematic model and online training samples. The simulation results compared with the traditional PID control method showed that the system could not only meet the requirements through online learning but also give the proof of the feasibility of reinforcement learning process to deal with the system disturbance.