Topology control for time-varying contested environments

Abstract

We consider a problem of topology design and control when some of the links may become unavailable. When a network operates in a contested environment, the operator may wish to alter the network topology to enhance robustness. This may come at the expense of operational costs. We consider a dynamic programming framework where the goal is to minimize time average costs, where costs are the sum of network property cost (e.g. eccentricity), edit costs and maintenance costs. We consider practical low-complexity control algorithms which focus on instantaneous network states. We introduce a horizon-aware modified myopic policy which attempts to reduce average cost by benefiting future costs. We also provide decision rules for different algorithms and characterize expected costs for random hostility models. Our numerical results demonstrate that our modified myopic scheduler outperform other schedulers in terms of average sum cost for various settings.