Resilient redundancy-based control of cyber–physical systems through adaptive randomized switching

A switching based approach using multiple parallel redundant controller implementations is developed to improve resiliency of cyber–physical systems (CPSs). Hardware/software redundancy is known to be a powerful technique for resiliency to mitigate effects of adversaries who infiltrate and maliciously modify a subset of the redundant subsystems. While redundant subsystems are typically combined using fail-over/backup and voting mechanisms, the proposed approach considers a time-division multiplexer using which one of multiple controller implementations is selected at each time instant to drive the input of the controlled system. Through detailed analysis of the switched system, it is shown that time-division multiplexing between redundant controllers can be used to mitigate the impact to stability and/or performance of the closed-loop CPS due to adversarial modifications of subsets of controllers. Additionally, we show that adversarial impact to the closed-loop CPS can be reduced over time by switching among the controllers in a probabilistic manner (rather than round-robin) and by dynamically adapting probabilities of switching to each controller. The efficacy of the proposed adaptive randomized switching algorithm is shown through simulation studies on two illustrative examples: a simple third-order system and a more real-world single-machine-infinite-bus system.