An Unknown Input Multiobserver Approach for Estimation and Control Under Adversarial Attacks

We address the problem of state estimation, attack isolation, and control discrete-time linear time-invariant systems under (potentially unbounded) actuator sensor false data injection attacks. Using a bank unknown input observers, each observer leading to an exponentially stable estimation error (in attack-free case), we propose observer-based estimator that provides exponential estimates system despite Exploiting redundancy, scheme is guaranteed work if sufficiently small subset sensors actuators attack. proposed estimator, provide tools for reconstructing isolating attacks, capable stabilizing closed-loop dynamics by switching off isolated actuators. Simulation results are presented illustrate performance our tools.