Resumen de Algorithms for energy-efficient adaptive wireless sensor networks
In this thesis we focus on the development of energy-efficient adaptive algorithms for Wireless Sensor Networks. Its contributions can be arranged in two main lines. Firstly, we focus on the efficient management of energy resources in WSNs equipped with finite-size batteries and energy-harvesting devices. To that end, we propose a censoring scheme by which the nodes are able to decide if a message transmission is worthy or not given their energetic condition. In order to do so, we model the system using a Markov Decision Process and use this model to derive optimal policies. Later, these policies are analyzed in simplified scenarios in order to get insights of their features. Finally, using Stochastic Approximation, we develop low-complexity censoring algorithms that approximate the optimal policy, with less computational complexity and faster convergence speed than other approaches such as Q-learning. Secondly, we propose a novel diffusion scheme for adaptive distributed estimation in WSNs. This strategy, which we call Decoupled Adapt-then-Combine (D-ATC), is based on keeping an estimate that each node adapts using purely local information and then combines with the diffused estimations by other nodes in its neighborhood. Our strategy, which is specially suitable for heterogeneous networks, is theoretically analyzed using two different techniques: the classical procedure for transient analysis of adaptive systems and the energy conservation method. Later, as using different combination rules in the transient and steady-state regime is needed to obtain the best performance, we propose two adaptive rules to learn the combination coefficients that are useful for our diffusion strategy. Several experiments simulating both stationary estimation and tracking problems show that our method outperforms state-of-the-art techniques in relevant scenarios. Some of these simulations reveal the robustness of our scheme under node failures. Finally, we show that both approaches can be combined in a common setup: a WSN composed of harvesting nodes aiming to solve an adaptive distributed estimation problem. As a result, a censoring scheme is added on top of D-ATC. We show how our censoring approach helps to improve both steady-state and convergence performance of the diffusion scheme.
