Efficient Data Delivery And Scheduling Mechanisms For Cyber-Physical Systems

Cyber-Physical Systems (CPS) are integrations of computation, networking, and physical processes. CPS enable the monitoring and control of the physical world in a secure, depend-able and real-time fashion, and they have extensive application prospects. The recent advances in wireless sensor networks (WSN), bio-medical sensors, and cloud computing are making CP-S to be widely employed for health care applications, namely medical CPS. Medical CPS are networked intelligent systems of medical devices that collect biometric information and dis-pense medical therapy, where the information interaction between different units is achieved via network communication. They enable the interconnection, sharing and collaboration of med-ical resources. Based on the fusion of CPS and medical systems, this dissertation focuses on efficient data delivery and scheduling mechanism for CPS. The main contribution of this dis-sertation consists of three aspects, namely secure and reliable data collection (inter-user inter-ference reduction and privacy protection of sensitive data), real-time and reliable data transmis-sion (degree-bounded real-time data aggregation), and mixed-criticality data processing (mixed-criticality real-time scheduling and probabilistic response time analysis).(1) Secure and reliable data collection. To address the inter-user interference due to the mobility, resource limitation and selfishness of body sensor networks (BSN) in medical CPS, we model the selection of channel and transmission power as a noncooperative game between different BSNs, and propose a no-regret learning algorithm to derive Nash equilibrium in the game process. To ensure the privacy and security of sensitive data collected by BSN with low power data transmission, the sensitive data is compressed losslessly with delta coding first to reduce the number of embedded bits of sensitive data, and then the compressed sensitive data is embedded into different kinds of ordinary data within the combined packet to protect sensitive data by using a lightweight data hiding algorithm.(2) Real-time and reliable data transmission. To ensure the real-time, reliable and low power wireless data transmission in medical CPS, bottleneck path and degree constraint are introduced in the data aggregation tree. The bottleneck path is used to minimize the maximum delay on data aggregation trees, and thus the bounded real-time communication can be ensured. Degree constraint is used to control the amount of processed data flow of each aggregation node, and hence the network lifetime can be prolonged by balancing resource utilization. Moreover, the system reliability can also be enhanced by introducing a degree constraint to limit the damage caused by a single node failure. We show that this problem is NP-hard, and give a learning automata based solving method.(3) Mixed-criticality data processing. To meet the mixed-criticality requirement of data processing tasks in medical CPS, we associate each high-criticality task with a subset of low-criticality tasks and encapsulate them in a task group. Within each task group, the tasks are scheduled with a server-based scheduling strategy, and different task groups are scheduled under the Earliest Deadline First (EDF) scheduling policy. Based on the schedulability constraints, we present a mixed integer nonlinear programming (MINLP) formulation for the task grouping and the scheduling parameter calculation. For the multiprocessor platform, we propose a task pack-ing algorithm that takes into account the demand of both low-criticality tasks and high-criticality tasks during deployment. For the medical CPS task characterized based on probability, we ab-stract its probabilistic workload as a probabilistic request function. Based on the cumulative request distribution of all tasks with higher priority, the rough worst case response time distri-bution is first obtained for each task, and then the exact worst case response time distribution is refined by a refinement algorithm based on request increase distributions.