Characterization of security levels for the dynamics of autonomous vehicle networks

Abstract

Increasingly, teams of self-coordinating autonomous vehicles are being used in lieu of manned transport in hostile environments, and hence characterizing the security and robustness of these autonomous vehicle networks (AVNs) from uncertainty and adversarial conduct is becoming paramount. Using a canonical double-integrator network model, we study security concepts for AVN dynamics; specifically, we characterize the ability of an adversary to estimate the full network dynamics, from noisy measurements of local vehicles' dynamics. We obtain several spectral characterizations of the estimability or security level, that highlight the role of the network's communication topology and control in security.