A Resilient Real-Time Traffic Control System

This paper describes a resilient control system operating in a critical infrastructure. The system is a real-time weather responsive system that accesses weather information that provides near-real-time atmospheric and pavement observation data that is used to adapt traffic signal timing to increase safety. Since the system controls part of a safety critical application survivability and resilience considerations must be an integral part of the system architecture. In order to provide adaptation to system behavior as the result of faults or malicious acts an architecture is presented that monitors itself and adapts its behavior in real-time. The main theoretical contributions are the combination and extension of approaches introduced in previous work. The theory of certifying executions is extended by three concepts: the detection of dependency violations, exceptions triggers, and sensor analysis are considered; a dual-bound threshold approach for detecting off-nominal executions is introduced; profiling is augmented with the concept of behavior sets. Extensive evidence of the effectiveness of the solutions based on a one-year observation of the system in action is presented.