I never came upon any of my discoveries through the process of rational thinking.
Albert Einstein

Graduate Student
Electrical and Computer Engineering
Rice University

Office: DH2047
Phone: +1-713-348-2471
Email: davood (at) rice (dot) edu

Mailling Address:
6100 Main St.,
MS-366 ECE Dept.,
Rice University
Houston, TX 77005

In sub-100nm technology, process variation is large compared to the device dimensions. Thus, it has a large impact on the design characteristics. In this project, for the first time, I introduce a fast method to find the gate level process variations of a manufactured chip. To find the process variations, I just measure total power of the chip for a number of random input bit streams. I show that process variation are sparse in wavelet domain. Then, I use the sparsity in wavelet domain to reconstruct process variations with small number of measurements using compressed sensing theory. I show that process variations can be approximated by 10% error with reasonably small number of power measurements. Now, I am working on the best regularizarion for the corresponding norm-one optimization.

Variation in sparse in wavelet domain.

Relaxing localization problem to Semi-Definite Programming (SDP) is addressed in this project. In contrast to the common models that use node's coordinateness as the main variables, I have used connecting vectors of the nodes as the main variables to model the localization problem as an optimization problem. Then, I have relaxed it to a SDP. I illustrate why it is better that previous methods.

I have created the first comprehensive and challenging benchmark data set for the ad-hoc location discovery (LD). The benchmark is a collection of representative reallife distance measurement data that establish a common basis for understanding, characterization, evaluation and comparison of the LD algorithms and solvers. It is constructed using a novel analysis methodology that systematically establishes the difficulty of discovering the locations. There are two major sources of difficulty that were reported in the LD literature and that I study: (1) combinatorial complexity of the LD problem due to the low number of available measurements; and (2) presence of measurement noise that renders the problem difficult even in dense networks. The noise impacts the continuous optimization underlying the LD calculations. I analyze the combinatorial complexity of the network instances with respect to the phase transition of the NP-hardness in sparse graphs. In dense networks, the location calculation is viewed as a continuous optimization problem instance with an objective function and a set of constraints. I devise a number of new metrics that evaluate the difficulty of the continuous optimization based on the data set properties. A fast simulation methodology is devised for rapid analysis of the sensitivity of goodness of the LD optimization with respect to the data set properties. I present a number of applications for the benchmark data and use it for evaluation and comparison of six popular LD algorithms. The LD benchmark is publicly available on the web.

I investigate and develop energy-balancing strategies for wireless ad-hoc networks energy resource allocation and deployment for the purpose of extending the network lifetime. The objective is to find the amount of energy storage (battery) that each node requires for having a balanced energy consumption throughout the network. For a limited set of energy resources in the deployment area, I determine an efficient deployment scenario in which messages are routed across the network using the fastest delivery path. Two ad-hoc architectures are considered: first, where the network is peer-to-peer and all the nodes have the same characteristics; and second, a basestation centric network where a base-station in the center collects the data from the ad-hoc nodes. I study synchronous and asynchronous communication paradigms for both architectures. To address the problems, I first determine the deployment scheme that results in the most comprehensive radio coverage. Next, I find the energy distribution for each network scenario using analytical and convex optimization methods. Then, the derived distributions are extended to randomly deployed networks. I present a thorough analysis and comparison for peer-topeer and base-station architectures, for both synchronous and asynchronous paradigms. My experimental evaluations show that the energy-balancing distributions extend the lifetime of the network by more than 40% when compared to non-balanced networks with no overhead on message delivery time. My evaluations over a range of system parameters including the coverage area, delay, network lifetime, and scaling shed light on many architectural and system design choices for ad-hoc networks.