Study On Task Scheduling For Cyber-Physical Systems

Cyber-Physical System (CPS) is a new intelligent hybrid system which integrates computation, communication and physical process. Task scheduling is of vital impor-tance for increasing resource utilization and performance of CPS. However, tasks in CPS are of various types because of system complexity and heterogeneity. Traditional task scheduling algorithms can not satisfy the overall performance of CPS. This paper started from CPS architecture and proposed three task scheduling schemes for network, con-troller, and computing center of CPS respectively. The proposed scheduling schemes are aimed for the overall performance of CPS. The main works and achievements of this paper are as follows:1. This paper proposed a dynamic multi-priority scheduling (DMPS) scheme for large scale wireless sensor network (WSN) based CPS based on traditional priority schedul-ing. The proposed scheme combined preemptive and non-preemptive scheduling strategy to ensure system real-time performance and using waiting threshold mech-anism to ensure fairness between different priority tasks. This paper also analyzed algorithm performance theoretically and experimentally. The result shows DMPS outperforms traditional preemptive and non-preemptive priority scheduling.2. This paper modeled CPS from the perspective of control and designed the con-troller with communication delay and dropout compensation, then analyzed how communication delay and dropout affect the stability of CPS controller using ma-trix technology. A control scheduling co-design scheme is proposed for CPS which can enhance system robustness and flexibility via information feedback between schedule, controller and physical device.. Then three task scheduling algorithms, stability first scheduling (SFS), communication performance first scheduling (CPF-S) and comprehensive scheduling are proposed aimed for CPS applications of d-ifferent performance requirement. Simulation results show that SFS has a better performance than fixed sample period scheduling in terms of steady state error and instability rate, CPFS has a better performance than DMPS in terms of total trans-mission delay, comprehensive scheduling outperforms SFS and CPFS on some con- dition.3. This paper proposed the basic concept and architecture of Cyber-Physical Cloud. Cloud service is used to enhance computing power of big data driven CPS. For pub-lic cloud based CPS, we use priority queue with negative arrivals to analysis how negative arrival rate affects system performance. A minimum cost first scheduling (MCFS) is proposed and simulation results show that MCFS has a better perfor-mance than fixed resource scheduling. For private cloud based CPS, we use M/D/C queueing system to analysis system performance. A dynamic resource allocation scheduling algorithm (DRAS) to balance system performance and computing cost.