Foundation for power management in battery powered wireless sensor networks

Vivek Agarwal, Purdue University

Abstract

Design and technical advancements in sensing, processing, and communicating capabilities of small, portable, and battery powered devices, known as sensor nodes, have drawn extensive research attention. These sensor nodes when interconnected wirelessly, regardless of topology, form a network, known as wireless sensor network (WSN).Wireless sensor networks have found vast applications in science, engineering, and in new consumer applications. Most WSN applications require operation over extended periods of time beginning with their deployment. Therefore, power management to increase network lifetime is a key characteristic of any battery powered (i.e., energy constrained) WSN. Some of the main contributions of this thesis are computation of (i) deterministic bounds on the lifetime of a schedule-driven sensor node, (ii) an expression for the expected value and the variance of a schedule-driven sensor node lifetime using a renewal theory, and (iii) an expression for the expected sensor network lifetime. These results and the details of their development are the key to power management control strategies. As a strong functional dependency exists between the sensor network lifetime and the sensor node lifetime; we initially develop a new battery model to estimate the load dependent battery lifetime. Different operations of a schedule-driven sensor node are categorized into six distinct operational states. Each state has a certain associated power cost which discharges the sensor node battery. The amount of time and energy a sensor node spends in a state before transitioning to the next state is a random variable (each governed by a probability distribution). The expected value and the variance of the random time and energy that a sensor node spends in each state are computed. This allows us to compute (i) the expected value and the variance of the energy consumed by a schedule-driven sensor node per cycle, and (ii) the expected cycle lifetime and the expected (time) lifetime of a schedule-driven sensor node. The accuracy of the schedule-driven sensor node lifetime is validated via a MATLAB simulation. At the sensor network level, for a simple three sensor node homogeneous network with a base station, we initially obtain deterministic bounds on the sensor network. Later, we show that for a basic forwarding protocol, the expected value of the energy consumption by a sensor node and the expected sensor node lifetime is readily extended to a network scenario. As expected the expected sensor network lifetime depends strongly on the expected lifetime of the individual sensor nodes. The findings in this thesis lay the foundation for the development of a power management strategy to extend the lifetime of a sensor node and a sensor network.

Degree

Ph.D.

Advisors

DeCarlo, Purdue University.

Subject Area

Computer Engineering|Electrical engineering

COinS