EFFICIENT DATA AGGREGATION SCHEMES FOR WIRELESS SENSOR NETWORKS

Abstract

The advancements in distributed computing and technologies such as the Internet of Things have led to the large-scale distribution of low-power wireless sensor networks over various applications. In many such applications, battery-powered tiny wireless sensors are geographically distributed to capture the measurements of the underlying physical phenomenon. The sensor measurements are to be efficiently aggregated and transferred to a control center. Cluster-based network architectures exploit the sensor data correlations effectively, and transmission in multiple hops leads to lesser energy consumption as per the first-order radio model. Therefore, cluster-based multihop networks are preferred for data aggregation and energy efficiency. In such networks, the sensor data is transmitted to the control center via intra-cluster and inter-cluster transmissions. To enable different types of transmissions, sensors can be equipped with either continuous variable or different discrete transmission ranges. In conventional networks, each sensor is assigned a continuously variable transmission range, where the transmission power of the sensor is controlled according to the position of the transmitter and the receiver. However, in low-power tiny sensors with limited functionalities, precise power control in the medium access control layer is difficult. Therefore, networks with sensors of different discrete transmission ranges are preferred for building the ecosystem for future-ready distributed networks. In such systems, sensors operate with one of the transmission ranges according to their role. However, the discrete transmission ranges assigned to the sensors introduce connectivity constraints in transmitting data to the control center. The conventional wireless sensor network protocols are not governed by the aforementioned connectivity constraints. This motivates us to investigate efficient data aggregation strategies in such connectivityconstrained networks and propose solutions.This thesis proposes six methods to address the data aggregation problem in connectivityconstrained networks, broadly under two heads; mathematical programming approaches and approximation methods. The mathematical programming approach starts with a multilinear program representing the data aggregation problem under network constraints. However, no method exists to solve the program because of its multilinear constraints. Therefore, a geometric programming approach and an integer linear programming method are proposed. This work proposes three approximation methods to address the data gathering problem in largescale networks. In the first method, a graphical framework that captures the connectivity constraints of the aforementioned networks is introduced, and an approximation technique based on the graphical framework is designed to solve the data aggregation problem. The second method relies on a mathematical framework and an approximation technique based on it. The integer program is relaxed in the third method, and the fractional solution is converted to an integer solution by devising an appropriate rounding technique. The proposed methods are evaluated by incorporating efficient data compression schemes based on compressed sensing.