利用强化学习的自适应动态规划中的值迭代和策略迭代方法，神经网络控制方法，LQR状态调节器最优控制方法，实现了三维倒立摆在飞行器上的稳定控制。鲁棒性很强，进行了高斯白噪声的扰动实验。-Reinforcement learning adaptive dynamic programming in value iteration and policy iteration method, neural network control method, LQR state regulator optimal control method to achieve a three-dimensional inverted pendulum on the vehicle stability control. Very robust, conducted experiments Gaussian white noise disturbance.