Secure Distributed State Estimation of an LTI System over Time-Varying Networks and Analog Erasure Channels
This work addresses the critical problem of secure state estimation in networked control systems, which is essential for safety-critical applications like autonomous vehicles or power grids.
The paper presents a distributed state estimation algorithm for LTI systems that is provably correct under arbitrary sensor attacks and time-varying communication networks, including intermittent losses and stochastic packet drops, with mean-square stability guarantees under certain conditions.
We study the problem of collaboratively estimating the state of an LTI system monitored by a network of sensors, subject to the following important practical considerations: (i) certain sensors might be arbitrarily compromised by an adversary and (ii) the underlying communication graph governing the flow of information across sensors might be time-varying. We first analyze a scenario involving intermittent communication losses that preserve certain information flow patterns over bounded intervals of time. By equipping the sensors with adequate memory, we show that one can obtain a fully distributed, provably correct state estimation algorithm that accounts for arbitrary adversarial behavior, provided certain conditions are met by the network topology. We then argue that our approach can handle bounded communication delays as well. Next, we explore a case where each communication link stochastically drops packets based on an analog erasure channel model. For this setup, we propose state estimate update and information exchange rules, along with conditions on the network topology and packet drop probabilities, that guarantee mean-square stability despite arbitrary adversarial attacks.