The emergence of miniature sensors with low-power wireless transceivers holds the promise of a new phase in the wireless revolution. Sensor networks have the ability to collect and transmit environmental data through the ubiquitous deployment of inexpensive wireless devices. This project addresses some of the most important current issues in wireless sensor networks. It seeks to develop an integrated framework for the combined analysis of sampling, distributed signal processing, and data dissemination in the context of delay-sensitive applications. Understanding the interplay between data gathering, resource allocation, and overall performance in wireless sensor networks necessitates a new mindset and a global system perspective. To this end, this project introduces a methodology for the analysis and the design of delay-sensitive sensor networks based, in part, on large deviations, queueing theory, and network calculus.
Systems composed of hundreds or thousands of wireless sensors make the dense sampling of stochastic environments possible. The amount of data generated by such large numbers of devices is equally vast, and it creates new challenges for the processing and transmission of observed data. Novel methods of analysis are required to provide insight into the efficient design and conception of sensor networks. As part of this research project, techniques from large-deviation theory and asymptotic analysis will be extended and applied to distributed sensing. These techniques will be used to: derive guidelines for node placement in correlated fields, propose novel paradigms for communication over multihop sensor networks with statistical delay guarantees, develop access control strategies that utilize the information content of individual packets, and characterize the interplay between in-network signal processing and traffic profiles for event-driven applications.
As wireless systems become an increasingly important part of everyday life, it is necessary to better understand their operation. This project studied distributed sensing and it focused on three particular aspects of sensor networks: data packets have different values depending on the information they carry, latency is critical when it comes to decision making, and the most efficient way to reach the Internet may be through multiple wireless hops. A first outcome of this project is an analysis framework that deepens our understanding of delay-constrained communications. This framework can be employed to select optimal system parameters for wireless sensor networks. This project also offers a new perspective on quantization for the purpose of decision making. In particular, all bits are not created equal and, in some cases, sending a few bits per sensor is sufficient to make excellent decisions. Finally, this project offers a new way to prevent congestion in wireless networks that support multiple packet types. In particular, packets that convey crucial information are given priority over messages that can wait, without being unfair to the various devices connected to the network. Altogether, the findings associated with this project will help design better wireless sensing systems in the future. Another important benefit of this project is the training of qualified engineers who are now working in the United States, contributing to its economy, and shaping the digital landscape of tomorrow.
