Optimization for energy-efficient communication in wireless distributed embedded systems

This dissertation explores the problem of optimizing network lifetime in low-power wireless ad-hoc networks of distributed embedded systems. Due to the energy-constrained nature of these systems, which limits network lifetime, energy-efficient communication is a fundamental design goal. From the network structure perspective, this study examines two areas: purely ad-hoc and hierarchical network structures. In the first part, this dissertation investigates crucial properties for balancing energy drain levels between nodes and introduces several techniques to optimize network resources in purely ad-hoc networks. In the second part, it considers the problem of constructing minimal infrastructure to reduce energy burden on low-power nodes, resulting in prolonging network lifetime in hierarchical network structure.;The proposed techniques in the first part are implemented as centralized and localized algorithms. A localized algorithm is developed using data-driven and non-parametric statistical models. An experimental comparison between the proposed localized and the centralized routing protocols is followed, because analyzing how well the localized algorithm using only partial information can achieve capability comparable to centralized schemes with global knowledge is critical to the performance criterion of the localized scheme. In addition, behaviors of an optimal routing protocol are studied to apply these behaviors to the localized schemes. To describe the optimal routing behaviors, a linear program formulation and non-parametric statistical modeling techniques are used, and several probabilistic models are presented. Using these presented models, the localized probabilistic routing scheme is realized.;Furthermore, the local awareness about the neighbors' resources has been assumed until now but not been examined well. In this dissertation, latency of information sharing between one-hop neighboring nodes is defined as operational latency and a methodology to optimize it is proposed. Subsequently, a methodology of maximizing network lifetime by adding relaying nodes is considered. To this end, beneficial network properties are studied and used: to examine how to strategically deploy new relaying nodes in an existing network; and to study to what extent each relaying node can prolong the network's lifetime.;The last part of the dissertation presents a method of building and operating a minimal number of backbone nodes for efficient communication and throughput optimization. It is often desirable to construct infrastructure in ad-hoc networks to achieve the goals of special tasks; on the other hand, the deployment and management cost of backbone nodes in such hierarchical networks is significant. Thus, it is critical to construct infrastructure minimally, while offering high communication quality between deployed nodes and the infrastructure.