Harnessing energy from environmental sources such as solar and wind is an attractive solution to the critical energy limitation problem in... Show moreHarnessing energy from environmental sources such as solar and wind is an attractive solution to the critical energy limitation problem in wireless sensor networks. Energy harvesting can potentially provide the network with perpetual and sustainable operation, or it can prolong network lifetime even for high consumption applications so as to justify the high cost of deployment. However, in order to efficiently utilize harvested energy, the energy source dynamics need to be incorporated into the network design. One way to do so is to make the network layer routing algorithm energy-harvest-aware. One common property of environmental energy sources is that they are generally only intermittently available. To address this, a storage unit such as a rechargeable battery can be introduced into the system. However, this is only a partial solution due to finite buffer storage capacities that cause harvested energy to be wasted when full. In this work, we aim to maximize the network lifetime by optimizing the energy availability and consumption alignment. To realize this objective, we first show that the minimization of energy wastage is a necessary condition to the maximization of available network energy. We then propose an on-demand routing algorithm that maximizes the total residual network energy by minimizing the energy consumption and wastage. Next, we illustrate the tradeoff between the two objectives of maximizing the total network energy and maximizing the minimum network energy in prolonging network lifetime. Then, we propose a linear-programming routing solution that maximizes a utility objective function based on this tradeoff. Although these routing approaches are shown to achieve high energy utilization, they are still based on deterministic harvest and consumption models. In the last part of this work, we propose a routing algorithm by applying the Semi-Markov Decision Process. Using this method, we are able to incorporate a comprehensive consideration of stochastic solar availability and traffic models, heterogeneous network properties such as non-uniform energy buffer capacities and consumption rates, and the optimization of an analytical formulation for network lifetime. Ph.D. in Electrical and Computer Engineering, December 2014 Show less
