Scheduling uncertain links in multihop cognitive relay networks

Uncertainties arise in the actual transmission rates achievable over various channels available on wireless links and in the interference characteristics of those links. In this work, we examine a dynamical learning approach to link scheduling for coping with this uncertainty in multihop cognitive relay networks. We formalize a scheduling with uncertainty problem (SWUP) in which the power received by nodes from a transmitting node may not be known with certainty. The schedule simultaneously should optimize transmissions in the current frame as well as perform measurements to reduce the uncertainty in network parameters. We propose a greedy algorithm to solve the SWUP and demonstrate that it is able to learn network parameters over time in order to improve network efficiency. We also formulate the SWUP as a mixed integer linear program (MILP) in order to assess the optimality of SWUP-Greedy. Extensive numerical simulations demonstrate the effectiveness of our algorithms as compared to non-learning methods.