Abstract

A new era of Cyber-Physical Systems (CPSs) is emerging due to the vast growth in computation and communication technologies. A fault-tolerant and timely communication is the backbone of any CPS to interconnect the distributed controllers to the physical processes. Such reliability and timing requirements become more stringent in safety-critical applications, such as avionics and automotive. Future networks have to meet increasing bandwidth and coverage demands without compromising their reliability and timing. Ethernet technology is efficient in providing a low-cost scalable networking solution. However, the non-deterministic queuing delay and the packet collisions deny low latency communication in Ethernet. In this context, IEEE 802.1 Time Sensitive Network (TSN) standard has been introduced as an extension of the Ethernet technology to realize switched network architecture with real-time capabilities. TSN offers Time-Triggered (TT) traffic deterministic communication. Bounded Worst-Case end-to-end Delay (WCD) delivery is yielded by Audio Video Bridging (AVB) traffic. In this thesis, we are interested in the TSN design and verification.

TSN design and verification are challenging tasks, especially for realistic safety-critical applications. The increasing complexity of CPSs widens the gap between the underlying networks' scale and the design techniques' capabilities. The existing TSN's scheduling techniques, which are limited to small and medium networks, are good examples of such a gap. On the other hand, the TSN has to handle dynamic traffic in some applications, e.g., Fog computing applications. Other challenges are related to satisfying the fault-tolerance constraints of mixed-criticality traffic in resource-efficient manners. Furthermore, in space and avionics applications, the harsh radiation environment implies verifying the TSN's availability under Single Event Upset (SEU)-induced failures. In other words, TSN design has to manage a large variety of constraints regarding the cost, redundancy, and delivery latency where no single design approach fits all applications. Therefore, TSN's efficient employment demands a flexible design framework that offers several design approaches to meet the broad range of timing, reliability, and cost constraints.

This thesis aims to develop a TSN design framework that enables TSN deployment in a broad spectrum of CPSs. The framework introduces a set of methods to address the reliability, timing, and scalability aspects. Topology synthesis, traffic planning, and early-stage modeling and analysis are considered in this framework. The proposed methods work together to meet a large variety of constraints in CPSs. This thesis proposes a scalable heuristic-based method for topology synthesis and ILP formulations for reliability-aware AVB traffic routing to address the fault-tolerance transmission. A novel method for scalable scheduling of TT traffic to attain real-time transmission. To optimize the TSN for dynamic traffic, we propose a new priority assignment technique based on reinforcement learning. Regarding the TSN verification in harsh radiation environments, we introduce formal models to investigate the impact of the SEU-induced switches failures on the TSN availability. The proposed analysis adopts the model checking and statistical model checking techniques to discover and characterize the vulnerable design candidates.