An analytical framework and implementation of wireless network inference and optimization

The paper addresses the problem of inference and optimization in wireless networks. An information-geometric framework is presented with scalable solutions to infer the end-to-end rate distributions of stochastic network flows from link rate measurements. Low-complexity solutions are provided by decomposing the network inference problem to smaller-size subproblems that are solved independently. By using the inferred distributions of flow rates, a recursive mechanism is presented to optimize the network performance. The distributions of all links rates are computed first from the inferred distributions of flow rates and then the weighted sum of link outage probabilities is minimized by adapting routing decisions. This approach is iterated between network inference and optimization, and provides link outage and end-to-end throughput gains compared to static approaches with fixed network parameters. The joint network inference and optimization approach is implemented with real radios and the performance is verified with hardware-in-the-loop emulation test results obtained with actual radio transmissions over emulated channels.