研究集多进多出MIMO技术、认知网技术、多信道并行与移动云计算技术于一体并创立有关归一化的随机系统与最优控制的理论与方法框架体系。首先，在不完全信息的信道基础上，利用最优滤波技术来确定信道的随机演化过程，提炼出集MIMO与认知网频谱、容量速率等资源配置与性能优化为一体的具有外部环境因子的随机最优控制系统，并在用户输入过程为一般双随机更新过程时，设计出有关的控制策略并利用随机扩散逼近证明所设计的策略的确具有渐进最优性；其次，将用户输入信号速率过程是由多个随机偏微分方程藕合描述的彩图过程的情况融入到上述归一化的体系中，并研究随机环境下的反射随机偏微分方程解的存在唯一性、以反射随机偏微分方程为极限过程的扩散逼近及有关的随机渐近最优控制；再者，将上述讨论应用到具有多类别用户多服务池且每个服务池具有并行服务器云计算网络的资源-性能最优配置并得到由随机偏微分方程驱使的极限性能分析与渐近最优控制体系。

We will integrate MIMO technology, cognitive radio network technology, parallel channels and mobile cloud computing technology together concerning related stochastic systems and optimal controls to form a unified theory and methodology framework to conduct control policy-making and analysis. Firstly, under imperfect channel information, we will use optimal filtering technologies to determine the channel evolving processes, and then we will combine the determinations of spectrum, capacity rate, etc. in MIMO and cognitive radio networks together to form a unified stochastic optimal control system with external environmental factors. Moreover, under the condition that user input processes are doubly stochastic renewal processes, we will design related control policy and prove it indeed to be asymptotically optimal via the method of stochastic diffusion approximation. Secondly, we will incorporate the case that user input signal processes are multi-channel color image processes whose rates determined by coupling stochastic partial differential equations into the above unified framework. In this case, we need to study the existence and uniqueness of solution to related reflecting stochastic partial differential equations under random environment and study the diffusion approximation and asymptotic optimal control based on the limiting processes driven by reflecting stochastic partial differential equations. Thirdly, we will integrate the above discussions into optimal resource-performance allocation for cloud computing network with multiple class users and multiple service pools of parallel servers, and moreover, we will obtain the unified framework for limiting performance analysis and asymptotic optimal control.