机构地区: 北京信息科技大学自动化学院
出 处: 《北京信息科技大学学报(自然科学版)》 2017年第4期10-17,共8页
摘 要: 研究使用神经网络结构的近似动态规划(Approximate Dynamic Programming,ADP)方法解决欠驱动刚体航天器姿态运动的最优控制问题。首先,依据多体系统动力学理论建立研究对象的姿态运动方程,并将其转化为控制系统的状态方程,然后利用ADP方法,通过评价网络逼近代价函数,执行网络逼近最优控制,并且给出效用函数的具体表达式,最后通过神经网络的在线学习,实现对研究对象姿态运动规划的最优控制。数值仿真结果验证了ADP方法能够有效求解欠驱动航天器姿态运动规划最优控制问题的有效性。 The optimal control of the attitude motion of an underactuated rigid spacecraft is discussed in this paper,using approximate dynamic programming(ADP) approach based on neural network. First of all,according to the theory of multi-body system dynamics,the equation of the underactuated spacecraft motion is established and converted into state equation. By using ADP approach,critic network and action network are used to approximate the cost function and the optimal control respectively. In addition,a concrete expression of the utility function is given. Lastly,the optimal control of the underactuated spacecraft is realized by neural network online learning. The effectiveness of the application of ADP approach is verified by the numerical simulation experiment.