Energy Harvesting Communication Networks with System Costs
MetadataShow full item record
This dissertation focuses on characterizing optimal energy management policies for energy harvesting communication networks with system costs. The system costs that we consider are the cost of circuitry to be on (processing cost) at the transmitters, cost of decoding at the receivers, cost of moving to harvest more energy in mobile energy harvesting nodes, and the cost of collecting measurements (sampling cost) from physical phenomena. We first consider receiver decoding costs in networks where receivers, in addition to transmitters, rely on energy harvested from nature to communicate. Energy harvested at the receivers is used to decode their intended messages, and is modeled as a convex increasing function of the incoming rate. With the goal of maximizing throughput by a given deadline, we study single-user and multi-user settings, and show that decoding costs at the receivers can be represented as generalized data arrivals at the transmitters. This introduces a further coupling between the transmitters and receivers of the network and allows us to characterize optimal policies by moving all constraints to the transmitter side. Next, we study the decoding cost effect on energy harvesting cooperative multiple access channels, where users employ data cooperation to increase their achievable rates. Data cooperation requires each user to decode the other user's data before forwarding it to the destination, which uses up some of the harvested energy. With the presence of decoding costs, we show that data cooperation may not be always helpful; if the decoding costs are relatively high, then sending directly to the receiver without data cooperation between the users achieves higher throughput. When cooperation is helpful, we determine the optimum allocation of available energy between decoding cooperative partner's data and forwarding it to the destination. We then study the impact of adding processing costs, on top of decoding costs, in energy harvesting two-way channels. Processing costs are the amounts of energy spent for circuitry operation, and are incurred whenever a user is communicating. We show that due to processing costs, transmission may become bursty, where users communicate through only a portion of the time. We develop an optimal scheme that maximizes the sum throughput by a given deadline under both decoding and processing costs. Next, we focus on online policies. We consider a single-user energy harvesting channel where the transmitter is equipped with a finite-sized battery, and the goal is to maximize the long term average utility, for general concave increasing utility functions. We show that fixed fraction policies are near optimal; they achieve a long term average utility that lies within constant multiplicative and additive gaps from the optimal solution for all battery sizes and all independent and identically distributed energy arrival patterns. We then consider a specific scenario of a utility function that measures the distortion of Gaussian samples communicated over a Gaussian channel. We formulate two problems: one with, and the other without sampling costs, and design near optimal fixed fraction policies for the two problems. Then, we consider another aspect of costs in energy harvesting single-user channels, that is, the energy spent in physical movement in search of better energy harvesting locations. Since movement has a cost, there exists a tradeoff between staying at the same location and moving to a new one. Staying at the same location allows the transmitter to use all its available energy in transmission, while moving to a new one may let the transmitter harvest higher amounts of energy and achieve higher rates at the expense of a cost incurred through the relocation process. We characterize this tradeoff optimally under both offline and online settings. Next, we consider different performance metrics, other than throughput, in energy harvesting communication networks. First, we study the issue of delay in single-user and broadcast energy harvesting channels. We define the delay per data unit as the time elapsed from the unit's arrival at the transmitter to its departure. With a pre-specified amount of data to be delivered, we characterize delay minimal energy management policies. We show that the structure of the optimal policy is different from throughput-optimal policies; to minimize the average delay, earlier arriving data units are transmitted using higher powers than later arriving ones, and the transmit power may reach zero, leading to communication gaps, in between energy or data arrival instances. Finally, we conclude this dissertation by considering the metric of the age of information in energy harvesting two-hop networks, where a transmitter is communicating with a receiver through a relay. Different from delay, the age of information is defined as the time elapsed since the latest data unit has reached the destination. We show that age minimal policies are such that the transmitter sends message updates to the relay just in time as the relay is ready to forward them to the receiver.