Dynamic Resilient Network Games With Applications to Multiagent Consensus

A Rolling Horizon Game Considering Network Effect in Cluster Forming for Dynamic Resilient Multiagent Systems

“…Our previous works [19,20] considered related games in continuous time, where the timings for launching attack/defense actions are also part of the decision variables. This aspect complicated the formulation, making it difficult to study games over a time horizon.…”

“…In this section, we state some results when a = 0, i.e., when the players do not consider the agents' states but agent-group index in determining their strategies so that the defender (resp., attacker) has higher (resp., lower) utility when more agents belong to the same group. Similar to the analysis in [20], here we explore some possible optimal strategy candidates for the players in a game. However, since a game consists of several steps in this formulation, the subgame perfect equilibrium is more involved to characterize, compared to the case of a game consisting of one step as in [20].…”

“…In this paper, we consider a security problem in a twoplayer game setting between an attacker, who is motivated to disrupt the communication among agents by attacking communication links, and a defender, who attempts to recover some of the attacked links. We formulate the problem based on [6,20], which use graph connectivity to characterize the game and the players' strategies. The game in this paper is played repeatedly over discrete time in the context of multiagent consensus.…”

“…Consensus protocol is considered due to its simple characterization, where all agents should converge in the case of no attack. More specifically, in comparison to [6,20], our contribution is threefold: (i) We introduce more options for the attacker's jamming signal strengths; (ii) the game consists of multiple attack-recovery actions, resulting in more complicated strategies; and (iii) we consider a rolling horizon approach for the players so that their strategies may be modified as they obtain new knowledge of the status of the system. More specifically, it is now possible for the attacker to disable links with stronger intensity of attack signals so that the defender is unable to recover those links (the decision on which edges are to be attacked with stronger attack signals is made at the same time as the decision on which edges are to be attacked with normal attack signals); this feature is motivated by [32,33].…”

Resilient Consensus with Multi-hop Communication

“…Based on consensus algorithms, various applications and distributed algorithms have been developed to solve different industrial problems, e.g., clock synchronization [3], energy management [4], distributed state estimation [5], distributed optimization [6], [7], and so on. As concerns for cyber security have rised in general, consensus problems in the presence of adversarial agents creating failures and attacks have attracted much attention; see, e.g., [8]- [11]. One class of interdisciplinary problems that has been studied in both control and computer science is that of resilient consensus [9], [12], [13].…”