Risk-Averse Access Point Selection in Wireless Communication Networks

This paper considers the problem of selecting the optimal set of access points and routing decisions in wireless communication networks. We consider networks that are subject to uncertainty in the wireless channel, for example, due to multipath fading effects, and formulate the problem as a risk-averse network flow problem with binary variables corresponding to the status of the sinks, namely, selected or not. Risk measures capture low-probability but high-cost events and, when used for stochastic optimization, they produce solutions that are more reliable compared to mean-value formulations and less conservative than worst-case approaches. By relaxing the integer constraints, we reformulate the problem as a linear optimization problem, which we solve in a distributed way using the accelerated distributed augmented Lagrangian method that was recently developed by the authors to solve optimization problems with convex separable objectives and linear coupling constraints. We present numerical simulations and experimental results using low-power wireless radios that demonstrate the ability of the proposed method to effectively deal with large variations in the quality of the wireless channel.