中文摘要：

对于具有未知参数的LQG（Linearquadratic Gaussian）问题，提出了一种次优对偶控制方法，用Kalman滤波处理过程噪声和测量噪声，用前一时刻的后验概率对Cost．to-go进行线性近似，然后，用动态规划获得了次优控制律．最后，用一个例子说明了本文设计的控制器的实施过程．结果表明，该控制律具有良好的对偶性质，并能在学习和控制之间实现较好平衡．

英文摘要：

For the LQG problem with unknown parameters, a novel suboptimal duM control approach is proposed in this paper. First, Kalman filter is used to deal with the noises of process and measurement and posterior probabilities at the pre- vious moment are used to linearly approximate the cost-to-go at the present moment. Then dynamic programming is adopted to obtain a suboptimal control law. Finally, an example is pre- sented to illustrate the implementation process of the developed controller. The result shows that this control law has good dual property and achieves a better balance between learning and control.