Faculty Research Seminars
What makes graduate and undergraduate students successful in research?
Dr. Christof Teuscher
February 26th, 2021
In this presentation I will share the outcomes of a multi-year data collection and analysis project of my lab's undergraduate and graduate research students, the MCECS Undergraduate Research and Mentoring Program (URMP), my NSF-funded Research Experience for Undergraduates site, and last summer's altREU program. The project was motivated by several questions, e.g., What are good metrics to predict student success? Are regular meetings and other interventions effective? How does feedback affect student productivity and success? Are there early warning signs that a student may quit the project? Does perceived effort correlate with actual productivity? How much guidance and mentoring leads to success? Do grades predict success in research? How do the size and the difficulty of a research project affect motivation and morale?
Smart Switching Power Electronics
Dr. Mahima Gupta
November 20th, 2020
Power electronic converters comprise of solid-state switching devices and energy storage elements. The power density of converters have drastically increased since the 1970s. This can be attributed to the continuous advancement of power semiconductor device technology that allowed the increase in the switching frequencies of power converters. However, the pace of this advancement has slowed down in the recent past. This talk will present the principles of smart-switching pulse width modulation (PWM) approaches which can enable dramatic reductions in the energy storage requirements for high-density power conversion (0.1-10µF vs. 100-1000µF). Smart switching approaches feature an intelligent switching sequence, accurate duty ratio calculation and robust fast bandwidth controls. Along with a discussion on the main features of smart-switching, several applications of smart-switching PWM approaches will be presented with simulation and experimental results.
Past, present, and future of crystal-free chip-scale wireless sensor nodes
Dr. David Burnett
October 16th, 2020
Millimeter-scale wireless sensor nodes are within reach thanks to recent advances in crystal-free RF communication. Complete wireless systems, usually referred to as "modules," have been limited to sizes of approximately 1cm x 1cm in both industry and academia because they require off-chip components assembled via PCB in addition to the wireless IC. Recent breakthroughs show these components can be eliminated to yield a chip-scale system with incredible applications in implantable medical devices, high-density sensor networks, and animal instrumentation. This talk will describe relevant background pertaining to crystal-free communication and explore the future possibilities these developments enable.
Inverse Problems, Constraint Satisfaction, Reversible Logic, Invertible Logic and Grover Quantum Oracles for Practical Problems
Dr. Marek Perkowski
May 29th, 2020
Material characterization of microwave absorbers and ferroelectrics + what’s going on in engineering education research
Dr. Branimir Pejčinović
April 17th, 2020
In this presentation I will address three different research topics, mainly to pique people’s interest and curiosity. Two deal with the material characterization of a) microwave absorbers, and b) ferroelectrics. For the third one, I will argue that there are many areas of engineering education that are worthy of exploration for which I am seeking collaborators.
A. The shielding of electronic devices is important for reasons ranging from regulatory to proper device functioning, and in applications from defense to consumer electronics. I will discuss the development and characterization of a novel microwave absorber material (MF-RAM) based on micron-sized ferrite beads and mm-length carbon fibers which are deposited electrostatically using the flocking process. The material is thin, lightweight, broadband and customizable with respect to frequency and amount of absorption. It was developed by Tangitek and at PSU we have characterized it in free space and in waveguides. We have demonstrated that MF-RAM exhibits losses associated with both electric and magnetic fields and could be more than competitive with commercially available solutions. I will also explain the basics of material characterization techniques used.
B. Despite their name, ferroelectrics have nothing to do with the magnetic field. Instead, they exhibit a behavior that is reminiscent of magnetic field hysteresis, i.e., they can maintain two different polarization states inside the material, depending on the signal trajectory (or history). Because they can have two different states when the electric field is zero they can be utilized as simple memory elements. However, their scaling to very thin layers is difficult and not well understood. I will present some recent measurement results on structures developed by a local startup.
C. Potential research topics engineering education will be illustrated with a couple of examples: 1. Investigating the time and effort that students spend studying, 2. Investigating how well students are prepared for life-long-learning. These are complex problems and I will discuss some tools and recent observations that could be turned into research projects.
An Introduction to the Research being Conducted in the Portland State University Magnetomechanical Energy Conversion and Controls Laboratory
Dr. Jonathan Bird
February 28th, 2020
In this presentation I will discuss three exciting research topics that are being investigated in the PSU’s laboratory for magnetomechanical energy conversion and control: (1) magnetic gears for improved power conversion, (2) negative stiffness resonant ocean power generators and (3) maglev vehicle transportation using electrodynamic wheels.
Leveraging Signal Structure for Efficient and Adaptive Machine Learning
Dr. John Lipor
January 24th, 2020
Abstract: Many major contributions to signal processing theory have centered around recognizing structure in signals of interest, with prominent examples being the Shannon-Nyquist sampling theorem for bandlimited signals and compressed sensing for signals that are sparse in some domain. In both examples, a key theme is that the structure in these signals can be leveraged to make the most efficient use of the available sampling resources. In this talk, I discuss how this general approach can be extended to the problems of data clustering and level set estimation. In the former case, high-dimensional signals can be effectively sorted (clustered) by recognizing the low-dimensional structure inherent in data from different clusters. In the latter, smoothness in the level set boundary or correlations between points can be harnessed to estimate level sets while minimizing the number of measurements taken, the distance between sampled points, or other costs associated with sampling. In all cases, the underlying structure allows for efficient signal reconstruction with a variety of theoretical guarantees.
Machine Learning for Anomaly Detection in Multi-Dimensional Data
Dr. Stanley R. Rotman, Harold Schnitzer Visiting Scholar from Ben-Gurion University of the Negev
November 22nd, 2019
Abstract: Multi-dimensional signals, such as Hyperspectral or Temporal Synthetic Aperture Radar, have very complicated distributions; machine learning promises to be a reasonable approach to determining structure in the data without any prior assumptions. In this talk, we will consider Non-Negative Matrix Factorization (NNMF) as a method to both determine trends in the data and to significantly reduce the number of redundant dimensions. We will use this transformed data for advanced anomaly detection.
Achieving and Exploiting Fairness in the N-player Ultimatum Game
Dr. Garrison Greenwood
October 18th, 2019
The Ultimatum Game is studied to see how people respond in bargaining situations. In the 2-player version each round a player can be a proposer or a responder. A proposer makes an offer on how to split a monetary amount and responder either accepts or rejects the offer. If accepted, the money is split as proposed; if rejected both players get nothing. N-player models suggest offers decrease over time but are still accepted. However, in human experiments players acted irrationally and rejected offers they deemed unfair. In this talk it is shown a (μ/μ,λ)-evolution strategy can evolve offer and acceptance thresholds that promote fairness. Then a single self-interested player is added who ignores fairness and instead exploits the other players by maximizing his payoffs. Monte Carlo Tree Search (MCTS) is used to adaptively control this self-interested player's offer levels during the game. Our results indicate MCTS can produce payoffs for this player as much as 40% higher than the population average payoff.