In this thesis single input - single output and multiple input - multiple output stochastic systems are discussed respectively . innovations are introduced to reconstruct the original minimum variance control problem of stochastic system , which is unsolvable by means of dynamic programming . so it can be converted into multiple single - step control problems , in which kalman filter is used to estimate unknown system parameters 本文分别针对单输入单输出和多输入多输出的随机系统进行了研究，通过引入系统的新息对原不可解的动态规划问题进行重构，将系统参数随机变化的最小方差控制问题转化成为多个基于新息的单步控制问题。
