Markov控制过程基于性能势仿真的并行优化

Parallel Optimization for Markov Control Processes Based on Performance Potentials Simulation

Markov控制过程是研究随机离散事件动态系统性能优化问题的一个重要模型,并在许多实际工程问题中有着广泛的应用。在Markov性能势理论的基础上,我们讨论了一类连续时间Markov控制过程在紧致行动集上的性能优化仿真问题。由于实际系统的状态空间往往非常巨大,通常的串行仿真算法,可能耗时过长,也可能由于硬件限制而无法实现,故我们提出了一种基于性能势的并行仿真优化算法,来寻找系统的最优平稳策略。一个仿真实例表明该算法有较好的运行效率。该算法可应用于大规模实际系统的性能优化。

A Markov control process is an important model for performance optimization in stochastic discrete event dynamic systems, and is widely used in many practical engineering problems. Based on the theory of Markov performance potential, the problems of performance optimization simulation for a class of continuous-time Markov control processes are studied. Since the state space of an actual system is often very large, when applying traditional serial simulation algorithms, long time is possibly spent, or it is impossibly realized because of hardware. A parallel simulation optimization algorithm based on performance potentials is proposed to find the optimal stationary policy of a system. A simulation example shows that the algorithm can achieve high speedup. The algorithm can be used in optimization for large-scale practical systems.