CAREER: Capacity-Driven Design of Large-Scale Wireless Sensor Networks

Regents Of The University Of Michigan - Ann Arbor, Ann Arbor MI

Investigators

Abstract

The proposed research centers on a design methodology for large-scale wireless sensor networks used for data gathering that uses fundamental capacity limit studies as guidelines. The motivation is that a large sensor network can potentially consist of thousands or tens of thousands of sensors densely populated, with strict energy and complexity constraints. Therefore the design of any protocol or algorithm should come with precision and a quantifiable measure, and be based on a good understanding of the ultimate scalability limit. This proposal aims at bridging the gap between the fundamental limit studies and practical protocol design. The goal of the proposed research is to (1) derive capacity limits critical to the large class of data gathering sensor network applications; (2) develop practical distributed algorithms that use these limits as guidelines and can approach these limits; and (3) examine the actual achievable performance of these algorithms in a real sensor testbed setting. Consequently, there are three parts to the proposed research: fundamental capacity limits, distributed algorithms, and testbed experiments. A central theme of the proposed research is to bridge the gap between the theoretical achievability of fundamental limits and the optimality of practical network designs. The proposed design methodology is thus driven by capacity limit studies via novel modeling techniques. It provides a novel and powerful tool in the study of network scalability and feasibility. Under fundamental capacity limits, will study two capacity notions: the throughput capacity, defined within the context of many-to-one communication as the maximum achievable throughput when all nodes are communicating with a single receiver via either a single hop or multiple hops; and the lifetime capacity, defined as the maximum amount of data deliverable by a sensor network until the first sensor dies (due to energy depletion) or till a pre-specified percentage of sensors die. The study of these two capacity notions has direct implications on organizing communications within a network, e.g., whether clustering should be used and how big a cluster should be, how many data collecting base stations should there be and where should they be placed. This study will progress from simple idealized scenarios to increasingly more realistic and complex. Under distributed algorithms, will apply the capacity analysis to the design of distributed algorithms of energy efficient data dissemination, optimal clustering and efficient sensor sleep schedules. These algorithms will be designed to approach or approximate network capacities. Under the proposed research will also develop an experimental wireless sensor testbed for implementation and measurement purposes. The proposed research has a strong education aspect and involves collaboration with two research centers at the University of Michigan. Will seek close collaboration with the Wireless Integrated Micro-systems (WIMS) Center, an NSF ERC at the University of Michigan, and University of Michigan's Wu Manufacturing Research Center (WuMRC). Will incorporate state-of-the-art MEMS sensors currently being developed at WIMS and intelligent infotronics agent sensors being developed for automation systems by WuMRC into our sensor testbed. Collaboration with them will allow the PI to apply design methodology to different application contexts with realistic physical devices.