On the spatial predictability of wireless channels and robust networked cooperation in mobile sensor networks

The wireless sensor network revolution has created the possibility of exploring and controlling the environment in ways not possible before. The vision of a multi-agent network cooperatively learning and adapting in harsh unknown environments to achieve a common goal is closer than ever. In such networks, communication plays a key role in the overall performance of the network as each mobile agent improves its knowledge by processing the information received from others. Therefore, proper prediction of the communication signal strength and fundamentally understanding the spatial predictability of a wireless channel, based on only a few measurements, become considerably important. The first contribution of this thesis is to propose a framework for predicting the spatial variations of wireless channels and to fundamentally understand wireless channel predictability. This framework can have a significant impact on intelligent connectivity maintenance in mobile sensor networks. More specifically, in Chapter 2, we develop a probabilistic framework for predicting the channel spatial variations, based on a small number of measurements. By using this framework, we then propose a mathematical foundation for understanding the impact of different environments, in terms of their underlying parameters, on wireless channel predictability. Furthermore, we show how sampling positions can be optimized to improve the prediction quality. Inspired by the recent results in non-uniform sampling theory, we then pursue a different path in Chapter 3 and show how the sparsity of the wireless channel in the frequency domain can be exploited in order to estimate channel variations based on a small number of measurements. The sparsity-based estimator is model-free and independent of the underlying channel parameters. Along this line, we then demonstrate the underlying tradeoffs between these two frameworks and propose an integrated approach which takes advantage of both channel compressibility in the frequency domain and probabilistic characterization in the spatial domain. All the theoretical results are validated with experimental measurements using our robotic testbed. The second contribution of the thesis is to consider different cooperative network operations with imperfect local sensing and under realistic modeling of communication links. More specifically, we consider the group agreement problem, where the cooperative network is trying to reach consensus on the occurrence of an event, by communicating over fading channels. This problem has received little attention in the literature as compared to the estimation consensus problem. However, it can find several applications such as networked fire detection and cooperative spectrum sensing in cognitive radio networks. Thus, another contribution of this dissertation is to fundamentally understand the behavior of such a cooperative network operation under imperfect communication links. To do so, we propose a novel consensus-seeking protocol that utilizes information of link qualities and noise variances to improve the performance and increase the robustness of the network to local sensing limitations. We mathematically characterize the impact of fading, noise, network connectivity and time-varying topology on consensus performance, which becomes challenging due to all the introduced uncertainties. We consider two different strategies, in terms of using the available transmissions: fusion and diversity, and shed light on the underlying tradeoffs in terms of speed of convergence and memoryless asymptotic behavior. Motivated by our analysis, we then propose an integrated framework, which keeps the benefits of both diversity and fusion approaches. We mathematically analyze the proposed framework and show how it achieves accurate consensus asymptotically.Electrical EngineeringDoctoralUniversity of New Mexico. Dept. of Electrical and Computer EngineeringMostofi, YasaminPereyra, Maria CristinaSen, PradeepSanthanam, Bal