Communications Using Sustainable Energy Sources

The carbon footprint of mobile communications is growing rapidly as Internet-capable but power-hungry handheld devices become more popular. Power consumption by base stations already costs service providers substantial sums of money, and profit margins will only be squeezed further in the future with rising fossil fuel costs and increasingly intensive use of the spectrum by Internet-connected mobile devices. In this project, we address the feasibility of running a mobile communication system entirely on non-fossil fuel energy sources, such as wind and solar energy. Optimization of the system could involve measuring and characterizing the correlation of power availability across cell sites, and subsequently performing inference on the quantum of power that will exist at future points in time across the system. Additional numbers of base stations will be needed to provide redundancy due to the intermittent availability of wind and solar power. However, this could improve connectivity and hence the cost of installing additional base stations will be offset by the gains in network performance. With severe power constraints, the conventional wisdom of spectrum scarcity on earth versus power scarcity in outer space may have to be revisited as well, with an attendant change of views regarding which modulation and coding formats are most suitable for terrestrial wireless communications.

Opportunistic Communications Through Spectrum Sensing

The finite band of electromagnetic spectrum that is suitable for wireless communications is growing more crowded every day, with much of it already dedicated to licensed users such as radio and TV stations, cellular service providers, and emergency services. However the nature of many of these communication systems make it impossible for them to use their assigned bands efficiently, e.g. data communications occurs in bursts with silent periods in between, and TV channels are not used in some parts of the country due to low demand. In this project, we explore the possibility of opportunistic communications through intelligent sensing of the spectrum (e.g. sensors acting cooperatively). In particular, we design schemes for sensing spectrum availability in the shortest possible time, and we calculate the throughput and probability of interfering with the licensed user of various spectrum sensing methods. The outcome of the project will be an understanding of the feasibility of opportunistic communications for various application scenarios, and enhanced spectrum sensing methods to maximize system performance.

Downlink Transmission with Noisy Channel Information

A base station transmitting to multiple users can perform much better when it knows the channel to each user, e.g. it can process the signal it transmits so that each user receives its own signal without interference. However, having accurate channel knowledge is practically impossible, as the channel must be estimated with finite accuracy, and those estimates used after a finite delay during which time the channel may have changed. The characterization of the performance of transmitter precoding in the presence of inaccurate (or noisy) channel information, and the design of precoders that explicitly account for these inaccuracies, are the objectives of this project. Its outcomes will be design guidelines that tell system designers when precoding using noisy channel information is beneficial, and a superior design that always performs better than conventional methods because it models the noise in the channel information explicitly to a reasonably accurate approximation.

CUDA for Wireless Communication Simulations

Wireless communications systems are growing ever more sophisticated in order to provide increasingly demanding users with the service quality they expect. As a result, the signal processing and other functions in a transmitter and receiver are very complex, e.g. it is now standard for systems to adapt their modulation/coding formats to the channel, employ error control codes that are theoretically close to optimal, and so on. In order to design such systems, it is necessary to simulate them in software before prototyping them in hardware. However, due to the system complexity, end-to-end performance simulations using the popular software package Matlab is very time consuming, often requiring a few days of continuous computations for even quite routine simulations. Sometimes it would be simply infeasible to perform these simulations for realistic parameter values and so researchers would resort to simulations of very small systems (two users only for instance). In this project, we have been using Nvidia's CUDA platform to leverage the parallel processing feature of graphical processing units (GPUs) to perform these simulations. The goal is to provide a toolbox of often-used functions that other researchers can apply in their simulations. With speed-up factors up to several hundred over C programs run on conventional serial processors, this approach is very attractive and is fast catching on in many computationally intensive research fields.