In summer 2012, the Department of Electrical & Computer Engineering at Duke University will host undergraduate students from around the country in our research laboratories.  These students will work with a faculty member and their research group to tackle an innovative research project.  Students admitted to the program receive a competitive monthly research stipend as well as arranged on-campus housing and a travel allowance.

When to Apply

The application has closed for 2012. Please check back early 2013.

Eligibility

All applicants must be United States citizens or permanent residents. The program is designed for students who are juniors in Spring 2012, but exceptional sophomores will also be considered.  A major in electrical and computer engineering is helpful but not necessarily required.  

Dates

Selected students should expect to hear of their acceptance to the program by April 1st. Student participants will be on site from late May to late July. 

REU Compensation:

• 9 weeks housing on Duke's campus
• $4,200 stipend
• Up to $400 towards travel costs 

Research Opportunities for 2012

The following three projects are available for the coming summer. Interested students are encouraged to apply. Questions about any of the projects or the REU program in general should be directed to Samantha Morton (samantha@ee.duke.edu). To be considered for any project, students must apply online through the link above.

Heterogeneous Datacenter Design and Deployment

Demand for computing capacity is driven by the data deluge. Over the past 45 years, computer engineers have transformed exponentially increasing transistor density into exponentially increasing capacity. At present, energy costs jeopardize further scaling. The US Environmental Protection Agency estimates datacenters already consume 1.5% of total nationwide electricity, which is comparable to the consumption of 5.8M US households. No combination of existing datacenter architectures can improve computing capacity by the desired three orders of magnitude within datacenter power budgets, which are already at megawatt scales.

This project examines the design and deployment of heterogeneous datacenter architectures that improve efficiency by 10x. Heterogeneity deploys a mix of specialized hardware for a mix of software needs, improving efficiency as unnecessary hardware resources are eliminated. To build heterogeneous datacenters, we explore design spaces for processors, memory, network, and storage using techniques in statistical inference and machine learning. To deploy heterogeneous datacenters, we use multi-agent markets in which applications bid for heterogeneous architectures, maximizing utility.

REU students participating in this project may participate in data collection and analysis. Responsibilities may include (1) analyzing performance and power for a variety of processor and memory designs, (2) simulating future processor and memory designs, (3) performing data analysis and design optimization. While not required, some knowledge in computer architecture and a major programming language (e.g., C, C++, Java) is helpful.

Faculty Contact: Prof. Benjamin Lee (benjamin.c.lee@duke.edu)

Thickness Variation in Polymer Thin Films Deposited by Resonant Infrared Matrix-Assisted Pulsed Laser Evaporation

Resonant infrared matrix-assisted pulsed laser evaporation (RIR-MAPLE) is a promising deposition technology for the fabrication of conjugated polymer-based optoelectronic devices for two primary reasons: i) the ability to control film morphology, and ii) the ability to deposit multi-layered heterostructures. The Stiff-Roberts group has developed a variation of RIR-MAPLE that uses emulsified targets of organic solvents and water such that the incident laser wavelength (Er:YAG at 2.9 µm) is resonant with hydroxyl (O-H) bonds in the host matrix, which are absent from the guest material. The novelty of the approach lies in the fact that while most polymers of interest and many compatible solvents do not resonantly absorb the laser energy at 2.9 ?m, the emulsion with water enables high-quality, thin-film deposition with minimal photochemical and structural degradation for almost any polymer of interest. In order to fabricate polymer-based optoelectronic device heterostructures, careful control over film thickness across a substrate is required. In this project, atomic force microscopy (AFM) will be used to characterize film thickness of polymer thin films across an entire substrate as a function of RIR-MAPLE growth parameters. The goal is to determine the thickness uniformity of the thin films for application to optoelectronic devices.

Faculty contact: Prof. Adrienne Stiff-Roberts (adrienne.stiffroberts@duke.edu)

Wide-Area Biometric Vital-Sign Monitoring Radar

The goal of this project is to develop a radar device which will permit the measurement of respiration and/or heart functioning of people over a wide area, such as in a room of a house. Important applications of this technology would be eldercare and infant care, e.g. to provide non-contact alerts for sudden infant death syndrome (SIDS). This project concerns development of signal processing algorithms to enable wide-area biometric monitoring using a low-power, Wi-Fi band radar. A key technical challenge in this effort is the estimation of weak respiration and/or heartbeat signals in the presence of noise and radar clutter. Stand-off life-sign monitoring radars exploit pulse-to-pulse variations of the radar return from a particular range to discriminate movement due to respiration and heart beating. Prof. Krolik’s research group at Duke has been involved in the development of a high quality, first-of-its-kind, indoor laboratory microwave radar, and this system is ideal for studying wide-area vital sign monitoring. The challenge to wide-area biometric monitoring in typical environments is the presence of high noise levels, clutter, and signal attenuation at long ranges. The goal of this project, therefore, is to develop signal processing methods that will facilitate wide-area biometric monitoring in more realistic, everyday settings. We currently have a functioning 16-element receive array and 4 transmit channel microwave radar system that can detect respiration of individuals at close range. The signal processing for the radar is entirely written in MATLAB. Analysis of STRADAR data can be performed off-line or in real-time, depending on the computational requirements. Currently, we have a basic real-time demo which performs respiration detection in the simplest of stand-off life-sign detection scenarios. The undergraduate project proposed here will involve developing MATLAB algorithms and displays for this radar to expand our ability to monitor the respiration and heart-rate of people at distances between 1 to 10 meters range. Two functions of particular interest are:

1. Analysis of algorithms and displays used for life sign monitoring of a patient in different poses relative to the radar (e.g. lying, sitting, standing) in noiseless environments.
2. Planning and execution of radar experiments for wide-area vital sign monitoring in realistic, everyday settings. To suggest improvements to signal processing methods which will enhance signal-to-noise and signal-to-clutter performance.

Faculty contact: Prof. Jeff Krolik (jk@ee.duke.edu)