基于再励模糊神经网络的三轴稳定卫星姿态智能控制

将再励学习引入模糊神经网络的T-S模型,建立了模糊神经网络控制器和控制评估网络的再励学习算法,并应用于三轴稳定卫星的姿态控制。这种再励模糊神经网络不需要精确的卫星数学模型和学习样本,通过再励学习实现控制网络/评估网络参数的在线调节,具有比较强的适应性和学习能力。仿真结果表明,这种智能控制方法可以有效解决卫星的模型不确定性问题,提高了卫星姿态控制的精度和鲁棒性。

Intelligent Attitude Control of Three-axis-stabilized Satellite Based on Fuzzy Neural Network Using Reinforcement Learning

In this paper, the reinforcement learning (RL) is introduced into the T-S model of fuzzy neural network (FNN) , which is then employed in the attitude control of a three-axis-stabilized satellite. The reinforcement learning algorithm separately built up for FNN controller and action evaluation network ( AEN) , realizes the on-line regulating of the parameters of FNN and AEN. Although the satellite ' s mathematical model and learning samples are both not required in the RL algorithm, the method possesses of the advantage of adaptability and learning capability to the environment uncertainty. The simulation results show that such kind of intelligent control method can effectively resolve the model uncertainty and disturbance rejection problems, and greatly improve the accuracy and robustness of the satellite attitude control system.