As mobile and machine-to-machine traffic is expected to grow exponentially in the next decade, tools for the design and optimization of agile and heterogeneous wireless networks are of great interest. Indeed, network design and operation have enormous complexity, due to the huge state space, the lack of global network state information at the decision units, the decentralized operation and resource constraints of wireless devices, thus requiring a holistic approach for network control and design. In this talk, I will present a principled framework for joint distributed sensing, estimation and control in wireless networks, which captures the interplay between state estimation and control and accounts for cross-layer factors such as the cost of acquisition of state information and the shared wireless channel.
The framework will be applied to a spectrum sensing-scheduling application, where a network of secondary users (SUs) attempts to opportunistically access portions of the spectrum left unused by a licensed network of primary users (PUs). Adaptive spectrum sensing and scheduling schemes are jointly optimized so as to maximize the SU throughput, subject to constraints on the PU throughput degradation and the sensing-transmission cost incurred by the SUs. I will show how low-complexity can be achieved by exploiting a large network approximation, a two-stage decomposition of the dynamic programming algorithm, as well as sparsity of network dynamics enabling efficient state estimation via sparse recovery techniques. Additionally, I will present a novel multiscale approach for spectrum sensing in large wireless networks, by which SUs maintain fine-grained estimates of the spectrum occupancy of nearby cells but coarse-grained estimates of that of distant cells. The cellular network is arranged into a hierarchy of increasingly coarse macrocells and SUs fuse local spectrum observations up the hierarchy. A probabilistic framework for spectrum sensing and information exchange is defined, which balances optimally improvements in spectrum estimation against energy costs.