【摘要】 研究了一种带有CMAC神经网络的再励学习 (RL)控制方法 ,以解决具有高度非线性的系统控制问题 .研究的重点在于算法的简化以及具有连续输出的函数学习上 .控制策略由两部分构成 :再励学习控制器和固定增益常规控制器 .前者用于学习系统的非线性 ,后者用于稳定系统 .仿真结果表明 ,所提出的控制策略不仅是有效的 ,而且具有很高的控制精度 .更多还原

【Abstract】 A novel technique which intergrates the cerebellar model articulation controller (CMAC) into the reinforcement learning (RL) control scheme developed by Barto et al is presented to tackle the control problems of nonlinear systems with huge uncertainties. The emphasis is placed on the simplification of integrated algorithm and functions learning with continuous outputs. The control strategy is composed of two elements. The constant gain feedback controller is used to stabilize the system. The second one is developed based on the RL algorithm and the CMAC neural network to learn the system nonlinearity. Simulation results show that the presented method is not only effective but also has high control accuracy.更多还原