Synthesis of Safety-Critical Real-Time Systems

Abstract: Modern safety-critical real-time systems are becoming more and more complex, due to sophisticated applications such as advanced driving assistance, automated driving, advanced infotainment, and applications involving machine learning and deep learning. This has led to increased requirements for the communication infrastructures. Real-time bus-based communication techniques, such as CAN and FlexRay, have been widely adopted for decades, due to their low cost and reliable communication capability. However, the bandwidth provided by these technologies is often not enough for modern safety-critical systems. Time-Sensitive Networking (TSN) is a promising technique that can handle the increasing bandwidth requirements, while meeting real-time constraints and providing Ethernet compatible solutions. We have studied the synthesis of schedules and routes for TSN, in order to fulfill timing and reliability requirements for safety-critical systems. Functional safety is an important goal for such systems, to ensure that no unreasonable risks are taken. This involves handling random and systematic faults, both of which are considered in this work. We synthesize schedules and routes for TSN so that the probability of faulty transmission due to random faults is below a certain threshold.ASIL Decomposition, introduced in the automotive industry, is applied to handle systematic faults, while achieving overall cost minimization. In order to improve schedulability, preemption support in TSN has also been studied. Heuristic algorithms are proposed for all the above contributions to address scalability issues characterized for the constrained synthesis and optimization problem addressed.Traditional designs for safety-critical systems usually deploy a federated architecture, where several processors are available and each processor implements one dedicated function. An important goal is to achieve fault containment. However, due to the increasing complexity of modern safety-critical systems, this architecture is no longer scalable. Therefore, several tasks with different criticality levels are usually integrated on the same computing platform. A key aspect for such systems is to achieve the required independence between tasks at different criticality levels and to guarantee that they do not interfere each other. We have developed a partitioned scheduling technique for mixed-criticality systems to achieve temporal independence, while minimizing the CPU usage.