A Stochastic Games based Resource Allocation Method for Cloud based Safe Computing Platform.

Safe computing platform is a safe architecture for the wayside equipment of CBTC, which keeps train operation control safely and efficiently in urban rail transit system. With the expansion of the subway line and the increase of passenger flow, the cloud computing technology is introduced into transit system to cope with the increase of maintenance cost and accident risk probability caused by the increase of physical equipments of CBTC. However, the architecture of safe computing platform must be considered when transplant the physical wayside equipment to cloud servers, which is hard to allocate the resources for the wayside equipment softwares to realize a good load-balance value while meeting the safety requirements defined by the safe computing platform. As one of the standard analytical tools, game theory has been applied in various fields to analyze resource allocation and decision-making problems. And stochastic games is good at modeling multi-agent decision-making problems, which is suitable for the problem we face. In this paper, a stochastic games is utilized to formulate the resource allocation of the cloud servers among the safe computing platform softwares and a SARSA based resources allocation algorithm with safety requirements is proposed to solve it. We compare SRAS with other similar algorithms in some simulation scenarios. The simulation results show that the SRAS method can get a better convergence performance and load-balance value of the cloud servers.