Research by Centers

Researchers

Laura Brown, PI, Zhenlin Wang, Co-PI

Sponsor

National Science Foundation, CSR: Small: Collaborative Research

Support

$299,993, 3 years

Abstract

Cloud computing has become a dominant scalable computing platform for both online services and conventional data-intensive computing (examples include Amazon’s EC2, Microsoft’s Azure, IBM’s SmartCloud, etc.). Cloud computing data centers share computing resources among a large set of users, providing a cost effective means to allow users access to computational power and data storage not practical for an individual. A data center often has to over-commit its resources to meet Quality of Service contracts. The data center software needs to effectively manage its resources to meet the demands of users submitting a variety of applications, without any prior knowledge of these applications.

This work is focused on the issue of management of memory resources in a data center. Recent progress in transfer learning methods inspires this work in the creation of dynamic models to predict the cache and memory requirements of an application. The project has four main tasks: (i) an investigation into how recent advancements in transfer learning can help solve data center resource management problems, (ii) development of a dynamic cache predictor using on-the-fly virtual machine measurements, (iii) creation of a dynamic memory predictor using runtime characteristics of a virtual machine, and (iv) development of a unified resource management scheme creating a set of heuristics that dynamically adjust cache and memory allocation to fulfill Quality of Service goals. In tasks (i)-(iii), transfer learning methods are employed and explored to facilitate the transfer of knowledge and models to new system environments and applications based on extensive training on existing systems and benchmark applications. The prediction models and management scheme will be evaluated on common benchmarks including SPEC WEB and CloudSuite 2.0. The results of this research will have broad impact on the design and implementation of cloud computing data centers. The results will help improve resource utilization, boost system throughput, and improve predication performance in a cloud computing virtualization system. Additionally, the methods designed and knowledge they impart will advance understanding in both systems research and machine learning.

Researchers

Laura Brown, PI, Zhenlin Wang, Co-PI

Sponsor

National Science Foundation, CSR: Small: Collaborative Research

Support

$299,993, 3 years

Abstract

Cloud computing has become a dominant scalable computing platform for both online services and conventional data-intensive computing (examples include Amazon’s EC2, Microsoft’s Azure, IBM’s SmartCloud, etc.). Cloud computing data centers share computing resources among a large set of users, providing a cost effective means to allow users access to computational power and data storage not practical for an individual. A data center often has to over-commit its resources to meet Quality of Service contracts. The data center software needs to effectively manage its resources to meet the demands of users submitting a variety of applications, without any prior knowledge of these applications.

This work is focused on the issue of management of memory resources in a data center. Recent progress in transfer learning methods inspires this work in the creation of dynamic models to predict the cache and memory requirements of an application. The project has four main tasks: (i) an investigation into how recent advancements in transfer learning can help solve data center resource management problems, (ii) development of a dynamic cache predictor using on-the-fly virtual machine measurements, (iii) creation of a dynamic memory predictor using runtime characteristics of a virtual machine, and (iv) development of a unified resource management scheme creating a set of heuristics that dynamically adjust cache and memory allocation to fulfill Quality of Service goals. In tasks (i)-(iii), transfer learning methods are employed and explored to facilitate the transfer of knowledge and models to new system environments and applications based on extensive training on existing systems and benchmark applications. The prediction models and management scheme will be evaluated on common benchmarks including SPEC WEB and CloudSuite 2.0. The results of this research will have broad impact on the design and implementation of cloud computing data centers. The results will help improve resource utilization, boost system throughput, and improve predication performance in a cloud computing virtualization system. Additionally, the methods designed and knowledge they impart will advance understanding in both systems research and machine learning.

Researchers

Joseph Burns, PI, Timothy Havens, Co-PI, Brian Thelen, Co-PI, Mark Stuff, Co-PI, Joel LeBlanc, Co-PI, Adam Webb, Co-PI

Sponsor

U.S. Army

Support

$1,238,255

Abstract

The project investigates theory and algorithms for multi-sensor buried target detection that achieve high probability of detection and classification with low false-alarm rate. The primary sensors of interest are handheld GPR and electromagnetic induction sensors.

Researchers

Timothy Havens, PI, TImothy Schulz, Co-PI

Sponsor

U.S. Army Research Office

Support

$285,900 (for the first year out of a potential 3-year project totaling $983,124)

Abstract

This project will investigate theory and algorithms for multi-sensor buried target detection that achieve high probability of detection and classification with low false-alarm-rate. The primary sensors of interest are multisensor FLGPR (i.e., FLGPR plus other sensor modalities, such as thermal video or LIDAR) and acoustic/seismic systems, although our methods will be applicable to other modalities as well.

Researchers

Colin Brooks, PI, Timothy Havens, Co-PI, Kuilin Zhang, Co-PI, Richard Dobson, Co-PI, Tess Ahlborn, Co-PI, A. Mukherjee, Co-PI

Sponsor

Michigan Dept. Transportation (MDOT)

Support

$598,526

Abstract

As unmanned aerial vehicle (UAV) technology has advanced to become more capable at lower cost, it offers transportation agencies a more rapid and safer alternative to collect data for a variety of applications, including condition assessment, traffic monitoring, construction, asset management, operations, and other applications. Through successful research, development, and demonstrations during Phase 1 of this project, the Michigan Tech team was able to test multiple sensors on a Michigan-made multirotor UAV platform, along with other UAVs, enabling the collection of data types such as optical light detection and ranging (LiDAR) and thermal to achieve a detailed view of a bridge deck both on the surface and subsurface. These methods were developed to represent the type of data collected through Michigan Department of Transportation (MDOT) manual inspections. Further development of UAV technology for the use of transportation infrastructure assessment is required in order to fully implement these technologies into MDOT day-to-day operations. By successfully continuing UAV research and development for MDOT, the Michigan Tech team will produce practical applications of large datasets that will support MDOT’s business models and decision making processes. 

Researchers

Timothy Havens, PI

Sponsor

Akela, Inc. / U.S. Army

Support

$83,359

Abstract

In this project researchers examine how unmanned aerial vehicles and terrestrial GPR can coordinate to improve buried explosive hazard detection performance. 

Researchers

Pasi Lautala, PI, Timothy Havens, Co-PI, Philart Jean, Co-PI, Paul Sanders, Co-PI, Thomas Oommen, Co-PI

Sponsor

US Department of Transportation / RITA

Support

$299,966

Abstract

The National University Rail (NURail) Center is a consortium of seven partner colleges and universities offering an unparalleled combination of strengths in railway transportation engineering research and education in North America. The NURail Center is the first USDOT OST-R University Transportation Center dedicated to the advancement of North American rail transportation. The Center is headquartered at the University of Illinois at Urbana-Champaign and includes researchers and educators who are experts and national leaders in railway infrastructure, systems and vehicles from seven prestigious academic institutions in the United States.

Researchers

Timothy Havens, PI, TImothy Schulz, Co-PI

Sponsor

Massachusetts Institute of Technology, Lincoln Laboratory

Support

$15,000

Abstract

The latest research in phased array systems has focused on accommodating multiple functions—radar, communications, electronic warfare—simultaneously on a single array. However, the work has not thoroughly addressed whether or not the partitioning of the antennas and signal generation in the array could be optimized to maximize the performance of the different functions on the array. This work explores these questions.

Researchers

Jeremy Bos, PI

Sponsor

Air Force Office of Scientific Research

Support

$429,886, 3 years

Abstract

This project will explore the nature of imaging in conditions characterized by extreme anisoplanatism.  Under these conditions each point in an image may be affected by a locally unique blurring kernel implying a violation of the linear shift invariance. Bos and his students will use a combination analysis and extensive experimental data to develop new models and new understanding of this phenomenon. Bos has also proposed using angular diversity as a means of mitigating the effects of extreme anisoplanatism on imaging and beam control problems.

Bos was a 2016 recipient of the Young Investigator Program (YIP) through the Air Force Office of Scientific Research for this project.

Researchers

Laura Brown, PI, Zhenlin Wang, Co-PI

Sponsor

National Science Foundation, CSR: Small: Collaborative Research

Support

$299,993, 3 years

Abstract

Cloud computing has become a dominant scalable computing platform for both online services and conventional data-intensive computing (examples include Amazon’s EC2, Microsoft’s Azure, IBM’s SmartCloud, etc.). Cloud computing data centers share computing resources among a large set of users, providing a cost effective means to allow users access to computational power and data storage not practical for an individual. A data center often has to over-commit its resources to meet Quality of Service contracts. The data center software needs to effectively manage its resources to meet the demands of users submitting a variety of applications, without any prior knowledge of these applications.

This work is focused on the issue of management of memory resources in a data center. Recent progress in transfer learning methods inspires this work in the creation of dynamic models to predict the cache and memory requirements of an application. The project has four main tasks: (i) an investigation into how recent advancements in transfer learning can help solve data center resource management problems, (ii) development of a dynamic cache predictor using on-the-fly virtual machine measurements, (iii) creation of a dynamic memory predictor using runtime characteristics of a virtual machine, and (iv) development of a unified resource management scheme creating a set of heuristics that dynamically adjust cache and memory allocation to fulfill Quality of Service goals. In tasks (i)-(iii), transfer learning methods are employed and explored to facilitate the transfer of knowledge and models to new system environments and applications based on extensive training on existing systems and benchmark applications. The prediction models and management scheme will be evaluated on common benchmarks including SPEC WEB and CloudSuite 2.0. The results of this research will have broad impact on the design and implementation of cloud computing data centers. The results will help improve resource utilization, boost system throughput, and improve predication performance in a cloud computing virtualization system. Additionally, the methods designed and knowledge they impart will advance understanding in both systems research and machine learning.

Researchers

Laura Brown, PI, Zhenlin Wang, Co-PI

Sponsor

National Science Foundation, CSR: Small: Collaborative Research

Support

$299,993, 3 years

Abstract

Cloud computing has become a dominant scalable computing platform for both online services and conventional data-intensive computing (examples include Amazon’s EC2, Microsoft’s Azure, IBM’s SmartCloud, etc.). Cloud computing data centers share computing resources among a large set of users, providing a cost effective means to allow users access to computational power and data storage not practical for an individual. A data center often has to over-commit its resources to meet Quality of Service contracts. The data center software needs to effectively manage its resources to meet the demands of users submitting a variety of applications, without any prior knowledge of these applications.

This work is focused on the issue of management of memory resources in a data center. Recent progress in transfer learning methods inspires this work in the creation of dynamic models to predict the cache and memory requirements of an application. The project has four main tasks: (i) an investigation into how recent advancements in transfer learning can help solve data center resource management problems, (ii) development of a dynamic cache predictor using on-the-fly virtual machine measurements, (iii) creation of a dynamic memory predictor using runtime characteristics of a virtual machine, and (iv) development of a unified resource management scheme creating a set of heuristics that dynamically adjust cache and memory allocation to fulfill Quality of Service goals. In tasks (i)-(iii), transfer learning methods are employed and explored to facilitate the transfer of knowledge and models to new system environments and applications based on extensive training on existing systems and benchmark applications. The prediction models and management scheme will be evaluated on common benchmarks including SPEC WEB and CloudSuite 2.0. The results of this research will have broad impact on the design and implementation of cloud computing data centers. The results will help improve resource utilization, boost system throughput, and improve predication performance in a cloud computing virtualization system. Additionally, the methods designed and knowledge they impart will advance understanding in both systems research and machine learning.

Researchers

Laura Brown, PI, Zhenlin Wang, Co-PI

Sponsor

National Science Foundation, CSR: Small: Collaborative Research

Support

$299,993, 3 years

Abstract

Cloud computing has become a dominant scalable computing platform for both online services and conventional data-intensive computing (examples include Amazon’s EC2, Microsoft’s Azure, IBM’s SmartCloud, etc.). Cloud computing data centers share computing resources among a large set of users, providing a cost effective means to allow users access to computational power and data storage not practical for an individual. A data center often has to over-commit its resources to meet Quality of Service contracts. The data center software needs to effectively manage its resources to meet the demands of users submitting a variety of applications, without any prior knowledge of these applications.

This work is focused on the issue of management of memory resources in a data center. Recent progress in transfer learning methods inspires this work in the creation of dynamic models to predict the cache and memory requirements of an application. The project has four main tasks: (i) an investigation into how recent advancements in transfer learning can help solve data center resource management problems, (ii) development of a dynamic cache predictor using on-the-fly virtual machine measurements, (iii) creation of a dynamic memory predictor using runtime characteristics of a virtual machine, and (iv) development of a unified resource management scheme creating a set of heuristics that dynamically adjust cache and memory allocation to fulfill Quality of Service goals. In tasks (i)-(iii), transfer learning methods are employed and explored to facilitate the transfer of knowledge and models to new system environments and applications based on extensive training on existing systems and benchmark applications. The prediction models and management scheme will be evaluated on common benchmarks including SPEC WEB and CloudSuite 2.0. The results of this research will have broad impact on the design and implementation of cloud computing data centers. The results will help improve resource utilization, boost system throughput, and improve predication performance in a cloud computing virtualization system. Additionally, the methods designed and knowledge they impart will advance understanding in both systems research and machine learning.

Researchers

Laura Brown, PI, Zhenlin Wang, Co-PI

Sponsor

National Science Foundation, CSR: Small: Collaborative Research

Support

$299,993, 3 years

Abstract

Cloud computing has become a dominant scalable computing platform for both online services and conventional data-intensive computing (examples include Amazon’s EC2, Microsoft’s Azure, IBM’s SmartCloud, etc.). Cloud computing data centers share computing resources among a large set of users, providing a cost effective means to allow users access to computational power and data storage not practical for an individual. A data center often has to over-commit its resources to meet Quality of Service contracts. The data center software needs to effectively manage its resources to meet the demands of users submitting a variety of applications, without any prior knowledge of these applications.

This work is focused on the issue of management of memory resources in a data center. Recent progress in transfer learning methods inspires this work in the creation of dynamic models to predict the cache and memory requirements of an application. The project has four main tasks: (i) an investigation into how recent advancements in transfer learning can help solve data center resource management problems, (ii) development of a dynamic cache predictor using on-the-fly virtual machine measurements, (iii) creation of a dynamic memory predictor using runtime characteristics of a virtual machine, and (iv) development of a unified resource management scheme creating a set of heuristics that dynamically adjust cache and memory allocation to fulfill Quality of Service goals. In tasks (i)-(iii), transfer learning methods are employed and explored to facilitate the transfer of knowledge and models to new system environments and applications based on extensive training on existing systems and benchmark applications. The prediction models and management scheme will be evaluated on common benchmarks including SPEC WEB and CloudSuite 2.0. The results of this research will have broad impact on the design and implementation of cloud computing data centers. The results will help improve resource utilization, boost system throughput, and improve predication performance in a cloud computing virtualization system. Additionally, the methods designed and knowledge they impart will advance understanding in both systems research and machine learning.

Researchers

Laura Brown, PI, Zhenlin Wang, Co-PI

Sponsor

National Science Foundation, CSR: Small: Collaborative Research

Support

$299,993, 3 years

Abstract

Emerging nonvolatile memory (NVM) technologies, such as PCM, STT-RAM, and memristors, provide not only byte-addressability, low-latency reads and writes comparable to DRAM, but also persistent writes and potentially large storage capacity like an SSD. These advantages make NVM likely to be next-generation fast persistent storage for massive data, referred to as in-memory storage. Yet, NVM-based storage has two challenges: (1) Memory cells have limited write endurance (i.e., the total number of program/erase cycles per cell); (2) NVM has to remain in a consistent state in the event of a system crash or power loss. The goal of this project is to develop an efficient in-memory storage framework that addresses these two challenges.

This project will take a holistic approach, spanning from low-level architecture design to high-level OS management, to optimize the reliability, performance, and manageability of in-memory storage. The technical approach will involve understanding the implication and impact of the write endurance issue when cutting-edge NVM is adopted into storage systems. The improved understanding will motivate and aid the design of cost-effective methods to improve the life-time of in-memory storage and to achieve efficient and reliable consistence maintenance.

Researchers

Laura Brown, PI, Zhenlin Wang, Co-PI

Sponsor

National Science Foundation, CSR: Small: Collaborative Research

Support

$299,993, 3 years

Abstract

Caches, such as distributed in-memory cache for key-value store, often play a key role in overall system performance. Miss ratio curves (MRCs) that relate cache miss ratio to cache size are an effective tool for cache management. This project develops a new cache locality theory that can significantly reduce the time and space overhead of MRC construction and thus makes it suitable for online profiling. The research will influence system design in both software and hardware, as nearly every system involves multiple types of cache. The results can thus benefit a wide range of systems from personal desktops to large scale data centers. We will integrate our results into existing open source infrastructure for the industry to adopt. In addition, this project will offer new course materials that motivate core computer science research and practice.

The project investigates a new cache locality theory, applies it to several caching or memory management systems, and examines the impact of different online random sampling techniques. The theory introduces a concept of average eviction time that facilitates modeling data movement in cache. The new model constructs MRCs with data reuse distribution that can be effectively sampled. This model yields a constant space overhead and linear time complexity. The research is focused on theoretical properties and limitations of this model when compared with other recent MRC models. With this lightweight model, the project seeks to guide hardware cache partitioning, improve memory demand prediction and management in a virtualized system, and optimize key-value memory cache allocation.

Researchers

Laura Brown, PI, Zhenlin Wang, Co-PI

Sponsor

National Science Foundation, CSR: Small: Collaborative Research

Support

$299,993, 3 years

Abstract

Cloud computing has become a dominant scalable computing platform for both online services and conventional data-intensive computing (examples include Amazon’s EC2, Microsoft’s Azure, IBM’s SmartCloud, etc.). Cloud computing data centers share computing resources among a large set of users, providing a cost effective means to allow users access to computational power and data storage not practical for an individual. A data center often has to over-commit its resources to meet Quality of Service contracts. The data center software needs to effectively manage its resources to meet the demands of users submitting a variety of applications, without any prior knowledge of these applications.

This work is focused on the issue of management of memory resources in a data center. Recent progress in transfer learning methods inspires this work in the creation of dynamic models to predict the cache and memory requirements of an application. The project has four main tasks: (i) an investigation into how recent advancements in transfer learning can help solve data center resource management problems, (ii) development of a dynamic cache predictor using on-the-fly virtual machine measurements, (iii) creation of a dynamic memory predictor using runtime characteristics of a virtual machine, and (iv) development of a unified resource management scheme creating a set of heuristics that dynamically adjust cache and memory allocation to fulfill Quality of Service goals. In tasks (i)-(iii), transfer learning methods are employed and explored to facilitate the transfer of knowledge and models to new system environments and applications based on extensive training on existing systems and benchmark applications. The prediction models and management scheme will be evaluated on common benchmarks including SPEC WEB and CloudSuite 2.0. The results of this research will have broad impact on the design and implementation of cloud computing data centers. The results will help improve resource utilization, boost system throughput, and improve predication performance in a cloud computing virtualization system. Additionally, the methods designed and knowledge they impart will advance understanding in both systems research and machine learning.

 

Researchers

Laura Brown, PI, Zhenlin Wang, Co-PI

Sponsor

National Science Foundation, CSR: Small: Collaborative Research

Support

$299,993, 3 years

Abstract

Cloud computing has become a dominant scalable computing platform for both online services and conventional data-intensive computing (examples include Amazon’s EC2, Microsoft’s Azure, IBM’s SmartCloud, etc.). Cloud computing data centers share computing resources among a large set of users, providing a cost effective means to allow users access to computational power and data storage not practical for an individual. A data center often has to over-commit its resources to meet Quality of Service contracts. The data center software needs to effectively manage its resources to meet the demands of users submitting a variety of applications, without any prior knowledge of these applications.

This work is focused on the issue of management of memory resources in a data center. Recent progress in transfer learning methods inspires this work in the creation of dynamic models to predict the cache and memory requirements of an application. The project has four main tasks: (i) an investigation into how recent advancements in transfer learning can help solve data center resource management problems, (ii) development of a dynamic cache predictor using on-the-fly virtual machine measurements, (iii) creation of a dynamic memory predictor using runtime characteristics of a virtual machine, and (iv) development of a unified resource management scheme creating a set of heuristics that dynamically adjust cache and memory allocation to fulfill Quality of Service goals. In tasks (i)-(iii), transfer learning methods are employed and explored to facilitate the transfer of knowledge and models to new system environments and applications based on extensive training on existing systems and benchmark applications. The prediction models and management scheme will be evaluated on common benchmarks including SPEC WEB and CloudSuite 2.0. The results of this research will have broad impact on the design and implementation of cloud computing data centers. The results will help improve resource utilization, boost system throughput, and improve predication performance in a cloud computing virtualization system. Additionally, the methods designed and knowledge they impart will advance understanding in both systems research and machine learning.

Researchers

Laura Brown, PI, Zhenlin Wang, Co-PI

Sponsor

National Science Foundation, CSR: Small: Collaborative Research

Support

$299,993, 3 years

Abstract

Cloud computing has become a dominant scalable computing platform for both online services and conventional data-intensive computing (examples include Amazon’s EC2, Microsoft’s Azure, IBM’s SmartCloud, etc.). Cloud computing data centers share computing resources among a large set of users, providing a cost effective means to allow users access to computational power and data storage not practical for an individual. A data center often has to over-commit its resources to meet Quality of Service contracts. The data center software needs to effectively manage its resources to meet the demands of users submitting a variety of applications, without any prior knowledge of these applications.

This work is focused on the issue of management of memory resources in a data center. Recent progress in transfer learning methods inspires this work in the creation of dynamic models to predict the cache and memory requirements of an application. The project has four main tasks: (i) an investigation into how recent advancements in transfer learning can help solve data center resource management problems, (ii) development of a dynamic cache predictor using on-the-fly virtual machine measurements, (iii) creation of a dynamic memory predictor using runtime characteristics of a virtual machine, and (iv) development of a unified resource management scheme creating a set of heuristics that dynamically adjust cache and memory allocation to fulfill Quality of Service goals. In tasks (i)-(iii), transfer learning methods are employed and explored to facilitate the transfer of knowledge and models to new system environments and applications based on extensive training on existing systems and benchmark applications. The prediction models and management scheme will be evaluated on common benchmarks including SPEC WEB and CloudSuite 2.0. The results of this research will have broad impact on the design and implementation of cloud computing data centers. The results will help improve resource utilization, boost system throughput, and improve predication performance in a cloud computing virtualization system. Additionally, the methods designed and knowledge they impart will advance understanding in both systems research and machine learning.

Researchers

Scott Kuhl, Keith Vertanen

Sponsor

ECE Alumnus Paul Williams '61

Support

$44,000, 1 year

Abstract

What if an everyday surface, like a table, could be transformed into a rich, interactive surface that can remotely operate things like computers, entertainment systems, and home appliances?

That’s what Michigan Tech Institute of Computing and Cybersystems (ICC) researchers Keith Vertanen and Scott Kuhl set out to do with a $44K seed grant from Electrical and Computer Engineering alumnus Paul Williams ’61.

Vertanen, assistant professor of computer science, and Kuhl, associate professor of computer science, are members of the ICC’s Center for Human-Centered Computing, which integrates art, people, design, technology, and human experience in the research of multiple areas of human-centered computing. They were assisted in this research by PhD candidate Siva Krishna Kakula, Computer Science, and undergraduate Zachary Garavet, Computer Engineering.

The team’s research goals were threefold: to create machine learning models that can precisely locate a user’s taps on a surface using only an array of inexpensive surface microphones; demonstrate the feasibility and precision of the models by developing a virtual keyboard interface on an ordinary wooden table; and conduct user studies to validate the system’s usability and performance.

The researchers are working on a related technical conference paper to present to their peers. Their outcomes included a prototype virtual keyboard that supports typing at rates comparable to a touchscreen device; possibly the first-ever acoustic sensing algorithm that infers a continuous two-dimensional tap location; and novel statistical models that quickly adapt to individual users and varied input surfaces.

Further, their results, hardware, and data sets can be applied to future collaborative work, and were used in the researchers’ $500K National Science Foundation proposal, “Text Interaction in Virtual and Augmented Environments,” which is under review.

Future applications of the research include enriched interactions in Virtual Reality (VR) and Augmented Reality (AR), compared to existing vision-only based sensing; and on-body interaction, like using your palm as an input surface.

Vertanen and Kuhl plan to continue this research, working to improve the accuracy of tap location inference, build richer interactions like swiping or tapping with multiple fingers, develop wireless sensor pods that can be quickly and easily deployed on any flat surface, and explore the display of virtual visual content on surfaces via Augmented Reality smartglasses.

Printable flyer

Researchers

Myounghoon "Philart" Jeon, PI

Sponsor

KATRI (Korea Automotive Testing & Research Institute)

Support

$450,703, 4 years

Abstract

The goal of the project is to design and evaluate intelligent auditory interactions for improving safety and user experience in the automated vehicles. To this end, we have phased plans for four years. In year 1, we prepare for the driving simulator for the automated driving model. In year 2, we estimate and model driver states in automated vehicles using multiple sensing approaches. In year 3, we design and evaluate discrete auditory alerts for safety purpose, with a focus, specifically, on take-over scenarios. In year 4, we develop real-time sonification systems for overall user experience and make guidelines.

The proposed work will significantly advance theory and practice in the design of natural and intuitive communication between a driver (or occupant) and an automated/autonomous vehicle. The results of the proposed research will not only contribute to design of the current vehicles, but also guide directions of social interaction in the automated vehicles in the near future. We will additionally obtain a theoretical framework to estimate and predict driver state and driving behavior. A more comprehensive understanding of the relationship between multiple sensing (e.g., neurophysiological and behavioral) data and driver states will ultimately be used to construct a more generic driving behavior model capable of combining affective elements and cognitive elements to positively influence safer driving. The proposed work will specifically contribute to the body of current literature in using auditory user interfaces for the automated vehicle contexts. Subsequently, this work will significantly advance theory and practice in interactive sonification design, affective computing, and driving psychology.

Applying novel in-vehicle auditory interactions to facilitate take-over process and eco-driving, and to mitigate distractions has the high potential to significantly decrease driving accidents and carbon footprints. Moreover, the entire program of the proposed research can be further developed for other vehicle situations. The proposed work offers an exciting simulated driving platform to integrate research with multidisciplinary STEM (Science, Technology, Engineering & Math) education. The principal investigator of the project will train graduate students, who will mentor undergraduates in the Mind Music Machine Lab at Michigan Tech. Driving simulators will be used for diverse hands-on curricula and courses. The PI will design driving simulation activities to teach courses as part of Michigan Tech’s Summer Youth Programs (SYP). Michigan Tech’s SYP has a strong longitudinal history of recruiting women, rural students from the Upper Peninsula of Michigan, and inner city minority students from the Detroit and Grand Rapids area. They also have a decade of assessment data that demonstrate SYP alumni are more likely to pursue STEM college degrees. The PI will also try to develop close partnerships and collaborations with other universities of the consortium and KATRI in Korea. Students and researchers can come to Michigan Tech and conduct and experience research projects together using cutting edge technologies. Research and education outcomes will be disseminated by the team, through the planned workshop on “New opportunities for in-vehicle auditory interactions in highly automated vehicles” at the International Conference on Auditory Display and AutomotiveUI Conference.

Researchers

Keith Vertanen, PI

Sponsor

Google Faculty Research Award

Support

$47,219

Abstract

While there have been significant improvements to text input on touchscreen mobile devices, entry rates still fall far short of allowing the free-flowing conversion of thought into text. Such free-flowing writing has been estimated to require input at around 67 words-per-minute (wpm) [1]. This is far faster than current mobile text entry methods that have entry rates of 20–40 wpm. The approach we investigate in this project is to accelerate entry by allowing users to enter an abbreviated version of their desired text. Users are allowed to drop letters or entire words, relying on a sentence-based decoder to infer the unabbreviated sentence. This project aims to answer four questions: 1) What sorts of abbreviations, a priori, do users think they should use? 2) How do users change the degree and nature of their abbreviations in response to recognition accuracy? 3) Can we train users to drop parts of their text intelligently in order to aid the decoder? 4) Can we leverage the abbreviation behaviors observed to improve decoder accuracy?

To answer these questions, we adopt a data-driven approach; collecting lots of data, from many users, over long-term use. To this end, we will extend our existing multi-player game Text Blaster, deploying the game on the Android app store. Text Blaster’s game play encourages players to type sentences both quickly and accurately. Players adopting successful abbreviation strategies will gain a competitive advantage. Text Blaster provides a platform to not only investigate abbreviated input but also a host of other open research questions in text entry. Data collected via Text Blaster will be released as a public research resource.

References

[1] Kristensson, P.O. and Vertanen, K. (2014). The Inviscid Text Entry Rate and its Application as a Grand Goal for Mobile Text Entry. In MobileHCI 2014, 335-338.

Researchers

Philart Jeon, PI

Sponsor

Hyundai Motor Company

Support

$130,236, 1 year

Abstract

To investigate effectiveness of an in-vehicle gesture control system and culture-specific sound preference, Michigan Tech will design a prototype of in-air gesture system and auditory displays and conduct successive experiments using a medium fidelity driving simulator.

Touchscreens in vehicles have increased in popularity in recent years. Touchscreens provide many benefits over traditional analog controls like buttons and knobs. They also introduce new problems. Touchscreen use requires relatively high amounts of visual-attentional resources because they are visual displays. Driving is also a visually demanding task. Competition between driving and touchscreen use for visual-attentional resources has been shown to increase unsafe driving behaviors and crash risk

[1]. Driving researchers have been calling for new infotainment system designs which reduce visual demands on drivers [2]. Recent technological advances have made it possible to develop in-air gesture controls. In-air gesture controls, if supported with appropriate auditory feedback, may limit visual demands and allow drivers to navigate menus and controls without looking away from the road. Research has shown that accuracy of surface gesture movements can be increased with addition of auditory feedback [3]. However, there are many unanswered questions surrounding the development of an auditory supported in-air gesture-controlled infotainment system: What type of auditory feedback do users prefer? How can auditory feedback be displayed to limit cognitive load? What type of menu can offer an easily navigable interface for both beginners and experienced users? More importantly, do these displays reduce the eyes-off-road time and frequency of long off-road glances? Does the system improve driving safety overall when compared to touchscreens or analog interfaces? These are among the many questions that we attempt to address in this research project. Moreover, we want to explore if there is any cultural difference in auditory perception. As a starting point, HMC and MTU will design in-vehicle sounds and conduct an experiment with American populations. 

References

Horrey, and C. Wickens, “In-vehicle glance duration: distributions, tails, and model of crash risk” Transportation Research Record: Journal of the Transportation Research Board, vol. 2018, pp. 22-28, 2007.
Green, “Crashes induced by driver information systems and what can be done to reduce them,” In Sae Conf. Proc. SAE; 1999, 2000.
Hatfield, W. Wyatt, and J. Shea, “Effects of auditory feedback on movement time in a Fitts task,” Journal of Motor Behavior, vol. 42, no. 5, pp. 289-293, 2010.

Researchers

Philart Jeon, PI

Sponsor

National Institutes of Health through the National Robotics Initiative

Support

$258,362

Abstract

The purpose of the research is to design novel forms of musical interaction combined with physical activities for improving social interactions and emotional responses of children with autism spectrum disorder (ASD). It is well known that a subject with ASD shows deficiency in emotional and social interactions. We propose to address two aspects of this issue: physio-musical stimulus for initiating engagement and empathizing for deepening interaction and thus enhancing a child’s emotional and social interactions. People with or without ASD between the ages of 5 and 10 may join this study.

In the United States, the rapid increase in the population of children with autism spectrum disorder (ASD) has revealed the deficiency in the realm of therapeutic accessibility for children with ASD in the domain of emotion and social interaction. There have been a number of approaches including several robotic therapeutic systems, but most of these efforts have been centered on speech interaction and task-based turn-taking scenarios. Unfortunately, the spectral diversity of ASD is so vast that many current approaches are, as novel and intriguing as they are, still insufficient to provide parameterized therapeutic tools and approaches.

To overcome this challenge, state-of-the-art techniques must still be developed to facilitate the autonomous interaction methods for robots to effectively stimulate the emotional and social interactivity of children. We focus on the recent studies that reveal strong relevance in premotor cortex among neural domains for music, emotion, and motor behaviors. We propose that musical interaction and activities can provide a new therapeutic domain for effective development in the children’s emotion and social interaction. Of key importance within this proposed work are providing capabilities for the robotic system to monitor the emotional and interactive states of children and provide effective musical and behavioral stimuli with respect to the emotion and social interaction of children. Major research questions include (1) What kinds of music-based signals and music-coordinated activities can play effective roles in encouraging emotional interaction and social engagement?, (2) How can robotic learning of human behaviors during musical activities increase the interaction between human and robot and reinforce the emotional and social engagement?, and (3) What metrics can be designed to effectively measure and evaluate the changes in emotional and social engagement through musical interaction activities?

Designing and evaluating core techniques to fuse music, emotion, and socio-physical interaction should be invaluable to advancing affective computing, robotics, and engineering psychology theory as well as providing guidelines in developing an effective therapeutic robot companion. With this research endeavor, we will identify the most crucial features of musical components in stimulating emotional and interactive intentions of children and the most effective way to correlate those musical components with motion behaviors to maximize the children’s social engagement and development. The findings of the proposed work will also contribute to the design of interactive scenarios for natural and creative therapy with an individualized and systematic approach.

The successful development of our framework with the music-based approach has the capability of creating a new domain of pediatric therapy that can tremendously increase the abilities of robots to interact with children in a safe and natural manner. The novelty and significance of this approach correlates to therapeutic design for children with ASD, but the foundation for our interactive and adaptive reinforcement scheme can be extended to other pediatric populations and developmental studies. We plan to incorporate these knowledge and approaches into courses designed for robotics and affective computing. Furthermore, we plan to encapsulate many of these ideas into an “outreach” workshop for underrepresented students. Undergraduate research projects and demonstrations are expected to inspire the next generation of engineers and scientist with a new robot-coexistent world. 

We have developed an emotion-based robotic motion platform that can encapsulate spatial components as well as emotional dimensions into robotic movement behaviors. The Romo robot and DARwin-OP are first used for our project (Figure 1). The robotic characters and interfaces are also newly designed to accommodate characteristics of children with ASD while satisfying software design specifications. We have also developed a multi-modal analysis system to monitor and quantify physical engagements and emotional reactions of children through facial expression recognition app (Figure 2), Kinect based movement analysis system, voice analysis app, and a music analysis software. All systems are designed for mobile computing environments, so our developed therapy sessions can be installed in any place with proper space and connectivity. We have implemented a sonification server (Figure 3) with 600 emotional sound cues for 30 emotional keywords and a real-time sonification generation platform. We have also carried out extensive user study with Americans and Koreans to validate our sounds and compare cultural differences. The results show that there are some commonalities in terms of emotional sound preferences between the two countries and we can narrow down our emotional sound cues for further research. We have created robotic scenarios for a pilot study (Typical day of Senses & Four Seasons), and will expand the scenarios with diverse genres of music and motion library sets.

Publications

• Robotic Sonification for Promoting Emotional and Social Interactions of Children with ASD.
• Musical Robots for Children with ASD Using a Client-Server Architecture

Researchers

Charles Wallace, PI, Leo Ureel, Co-PI, Shreya Kumar, Co-PI

Sponsor

National Science Foundation (IUSE: Improving Undergraduate STEM Education)

Support

$218,735, September 2015 – August 2018

Abstract

Among software professionals, the quality of team communication is widely acknowledged to be a key factor in the success or failure of software projects. The inherent mutability and intangibility of software, coupled with the intense rate of change in the software workplace, demand attentiveness and precision in oral and written communication. Students in CISE programs need training from communication specialists and practice in the particular genres common to their future profession. Successful communication in the workplace, however, requires more than technical mastery of common genres presented in isolation; while a software process model can provide guidelines, there is no comprehensive, rote communication workflow. Developers must make strategic communication decisions, and they must be agile — flexible, proactive, and creative — in these decisions.

We envision an enhanced undergraduate curriculum for CISE programs that promotes agile communication through practice in inquiry, critique and reflection, grounded in authentic software development settings. In the proposed project, we will target early courses in our undergraduate computer science/software engineering curriculum: our first-year introductory programming sequence, our second-year data structures course, and our third-year team software project course. Agile development skills will be introduced gradually over the three years: test-driven development in Year 1; user stories in Year 2; full agile development using Scrum in Year 3.

Communication intensive activities will be woven through this curriculum in a variety of ways. The POGIL framework will provide a structured approach to inquiry. Automated feedback on test coverage, programming style and code documentation will be provided through CanvasTA, a novel tool integrated into the Canvas LMS, supplementing instructor feedback with continual critique of code and documentation. A program of guided inquiry through real case studies of software communication will prepare students for their team software activities, and a series of reflective exercises will lead them to focus on their own team communication practices.

In our two-year project, we plan to build a model communication-oriented curriculum for CISE programs nationwide. We will develop our communication-intensive course materials and use them in three offerings of each targeted course. We will assess and refine these materials through formative and summative external evaluation, and we will increase awareness and promote adoption through a broad dissemination effort.

Integrating communication activities into introductory CISE courses is part of a transformative effort to instill software engineering skills early in the curriculum. The PIs bring a significant amount of experience and original research to bear on this topic. We are also leveraging results from recent NSF-funded activity on POGIL and integrating communication into CISE. An broad-based advisory board consisting of software engineering and writing studies faculty as well as software professionals will provide formative and summative feedback.

Through this enhanced curriculum, CISE students will be prepared for the communication-intensive realities of the software workplace and better able to compete in a global software economy. Software developers with strong preparation in agile communication skills will produce higher quality software that generates greater customer satisfaction; being more engaged in their profession, these developers will be less prone to frustration and burnout. Furthermore, we anticipate that the emphasis on communication in early courses will attract and retain a broader group of students in CISE programs, particularly those who are excited by the concept of software development as a social activity.

Researchers

Zhaohui Wang, Nina Mahmoudian

Sponsor

ECE Alumnus Paul Williams, '61

Support

$50,000, 1 year

Abstract

Underwater acoustic communication has been in use for decades, but primarily for military applications. In recent years, private sectors such as environmental monitoring, off-shore oil and gas exploration, and aquaculture have become interested in its possibilities.

But existing research about underwater acoustic communication networks often relies on human-operated surface ships or cost-prohibitive autonomous underwater vehicles (AUVs). And these cost barriers can limit academic research evaluation to computer simulations, constraining research innovation towards practical applications.

Recognizing the above gap, Michigan Tech Institute of Computing and Cybersystems (ICC) researchers Zhaohui Wang, assistant professor, Electrical and Computer Engineering, and Nina Mahmoudian, adjunct professor, Mechanical Engineering-Engineering Mechanics,  saw an opportunity to combine their areas of expertise: for Wang, underwater acoustic communications, for Mahmoudian, low-cost marine robotics and AUVs.

Also part of the research team were PhD student Li Wei, Electrical and Computer Engineering, and post-doc research engineer Barzin Moridian, Mechanical Engineering-Engineering Mechanics. The team also collaborated with scientists at Michigan Tech’s Great Lakes Research Center.

With a $50K seed grant from Electrical and Computer Engineering alumnus Paul Williams ’61, the team took the research beneath the surface to develop a low-cost marine mobile infrastructure and investigate the challenges and possible solutions in engineering a leading-edge AUV communication network.

They broke it down into three areas: the development of low-cost, high-modularity autonomous surface vehicles (ASVs), each equipped with a collection of sensors and serving as surrogates for AUVs; equipping each ASV with an acoustic modem and implementing communication and networking protocols to facilitate underwater communication among the vessels; and conducting field experiments to collect data about the fundamental challenges in mobile acoustic communications and networking among AUVs.

The team’s outcomes included two low-cost, autonomous, on-the-water boats; an experimental data set, data analysis, and preliminary results; a technical paper presented at the 2018 IEEE OES Autonomous Underwater Vehicle Symposium; and a marine mobile wireless networking infrastructure for use in continued research.

Just half of their seed grant has been used, and this summer Wang and Mahmoudian will work to improve the boats and the communications system, and conduct more field research. In addition, they are planning to write two National Science Foundation proposals to take their research even further.

Printable Flyer

Researchers

Laura Brown, PI, Zhenlin Wang, Co-PI

Sponsor

National Science Foundation, CSR: Small: Collaborative Research

Support

$299,993, 3 years

Abstract

Cloud computing has become a dominant scalable computing platform for both online services and conventional data-intensive computing (examples include Amazon’s EC2, Microsoft’s Azure, IBM’s SmartCloud, etc.). Cloud computing data centers share computing resources among a large set of users, providing a cost effective means to allow users access to computational power and data storage not practical for an individual. A data center often has to over-commit its resources to meet Quality of Service contracts. The data center software needs to effectively manage its resources to meet the demands of users submitting a variety of applications, without any prior knowledge of these applications.

This work is focused on the issue of management of memory resources in a data center. Recent progress in transfer learning methods inspires this work in the creation of dynamic models to predict the cache and memory requirements of an application. The project has four main tasks: (i) an investigation into how recent advancements in transfer learning can help solve data center resource management problems, (ii) development of a dynamic cache predictor using on-the-fly virtual machine measurements, (iii) creation of a dynamic memory predictor using runtime characteristics of a virtual machine, and (iv) development of a unified resource management scheme creating a set of heuristics that dynamically adjust cache and memory allocation to fulfill Quality of Service goals. In tasks (i)-(iii), transfer learning methods are employed and explored to facilitate the transfer of knowledge and models to new system environments and applications based on extensive training on existing systems and benchmark applications. The prediction models and management scheme will be evaluated on common benchmarks including SPEC WEB and CloudSuite 2.0. The results of this research will have broad impact on the design and implementation of cloud computing data centers. The results will help improve resource utilization, boost system throughput, and improve predication performance in a cloud computing virtualization system. Additionally, the methods designed and knowledge they impart will advance understanding in both systems research and machine learning.

Researchers

Xinli Wang, PI, Guy Hembroff, Co-PI

Sponsor

National Science Foundation

Support

$199,934, 4 years

Abstract

The goal of this research is to enhance the security component in undergraduate IT education to meet the strong demand for security professionals in IT fields.
It has been widely admitted by researchers, educators and students that the benefits of hands-on lab experiments are threefold in IT security education:

• They expose students to the real-world challenges of computer and network security.
• They help students consolidate and gain in-depth understanding of the knowledge presented in class lectures.
• These hands-on activities help students to be better prepared for their careers in industry.

This project will develop a collection of instructional hands-on laboratories for undergraduate IT security education to achieve the following objectives:

• To provide students with an active-learning environment by challenging them with real-world problems in the field of IT security.
• To provide students with the opportunity to learn from experiences with advanced technologies and well developed tools which will make them better prepared for their careers in industry.
• To help instructors prepare and deliver security courses more effectively and efficiently by making the lab experiments publicly accessible through the Internet.

This project is a collaborative effort between Michigan Technological University and the University of Washington Tacoma.

Researchers

Laura Brown, PI, Zhenlin Wang, Co-PI

Sponsor

National Science Foundation, CSR: Small: Collaborative Research

Support

$299,993, 3 years

Abstract

Cloud computing has become a dominant scalable computing platform for both online services and conventional data-intensive computing (examples include Amazon’s EC2, Microsoft’s Azure, IBM’s SmartCloud, etc.). Cloud computing data centers share computing resources among a large set of users, providing a cost effective means to allow users access to computational power and data storage not practical for an individual. A data center often has to over-commit its resources to meet Quality of Service contracts. The data center software needs to effectively manage its resources to meet the demands of users submitting a variety of applications, without any prior knowledge of these applications.

This work is focused on the issue of management of memory resources in a data center. Recent progress in transfer learning methods inspires this work in the creation of dynamic models to predict the cache and memory requirements of an application. The project has four main tasks: (i) an investigation into how recent advancements in transfer learning can help solve data center resource management problems, (ii) development of a dynamic cache predictor using on-the-fly virtual machine measurements, (iii) creation of a dynamic memory predictor using runtime characteristics of a virtual machine, and (iv) development of a unified resource management scheme creating a set of heuristics that dynamically adjust cache and memory allocation to fulfill Quality of Service goals. In tasks (i)-(iii), transfer learning methods are employed and explored to facilitate the transfer of knowledge and models to new system environments and applications based on extensive training on existing systems and benchmark applications. The prediction models and management scheme will be evaluated on common benchmarks including SPEC WEB and CloudSuite 2.0. The results of this research will have broad impact on the design and implementation of cloud computing data centers. The results will help improve resource utilization, boost system throughput, and improve predication performance in a cloud computing virtualization system. Additionally, the methods designed and knowledge they impart will advance understanding in both systems research and machine learning.

 

Researchers

Laura Brown, PI, Zhenlin Wang, Co-PI

Sponsor

National Science Foundation, CSR: Small: Collaborative Research

Support

$299,993, 3 years

Abstract

Cloud computing has become a dominant scalable computing platform for both online services and conventional data-intensive computing (examples include Amazon’s EC2, Microsoft’s Azure, IBM’s SmartCloud, etc.). Cloud computing data centers share computing resources among a large set of users, providing a cost effective means to allow users access to computational power and data storage not practical for an individual. A data center often has to over-commit its resources to meet Quality of Service contracts. The data center software needs to effectively manage its resources to meet the demands of users submitting a variety of applications, without any prior knowledge of these applications.

This work is focused on the issue of management of memory resources in a data center. Recent progress in transfer learning methods inspires this work in the creation of dynamic models to predict the cache and memory requirements of an application. The project has four main tasks: (i) an investigation into how recent advancements in transfer learning can help solve data center resource management problems, (ii) development of a dynamic cache predictor using on-the-fly virtual machine measurements, (iii) creation of a dynamic memory predictor using runtime characteristics of a virtual machine, and (iv) development of a unified resource management scheme creating a set of heuristics that dynamically adjust cache and memory allocation to fulfill Quality of Service goals. In tasks (i)-(iii), transfer learning methods are employed and explored to facilitate the transfer of knowledge and models to new system environments and applications based on extensive training on existing systems and benchmark applications. The prediction models and management scheme will be evaluated on common benchmarks including SPEC WEB and CloudSuite 2.0. The results of this research will have broad impact on the design and implementation of cloud computing data centers. The results will help improve resource utilization, boost system throughput, and improve predication performance in a cloud computing virtualization system. Additionally, the methods designed and knowledge they impart will advance understanding in both systems research and machine learning.

 

Researchers

Laura Brown, PI, Zhenlin Wang, Co-PI

Sponsor

National Science Foundation, CSR: Small: Collaborative Research

Support

$299,993, 3 years

Abstract

Cloud computing has become a dominant scalable computing platform for both online services and conventional data-intensive computing (examples include Amazon’s EC2, Microsoft’s Azure, IBM’s SmartCloud, etc.). Cloud computing data centers share computing resources among a large set of users, providing a cost effective means to allow users access to computational power and data storage not practical for an individual. A data center often has to over-commit its resources to meet Quality of Service contracts. The data center software needs to effectively manage its resources to meet the demands of users submitting a variety of applications, without any prior knowledge of these applications.

This work is focused on the issue of management of memory resources in a data center. Recent progress in transfer learning methods inspires this work in the creation of dynamic models to predict the cache and memory requirements of an application. The project has four main tasks: (i) an investigation into how recent advancements in transfer learning can help solve data center resource management problems, (ii) development of a dynamic cache predictor using on-the-fly virtual machine measurements, (iii) creation of a dynamic memory predictor using runtime characteristics of a virtual machine, and (iv) development of a unified resource management scheme creating a set of heuristics that dynamically adjust cache and memory allocation to fulfill Quality of Service goals. In tasks (i)-(iii), transfer learning methods are employed and explored to facilitate the transfer of knowledge and models to new system environments and applications based on extensive training on existing systems and benchmark applications. The prediction models and management scheme will be evaluated on common benchmarks including SPEC WEB and CloudSuite 2.0. The results of this research will have broad impact on the design and implementation of cloud computing data centers. The results will help improve resource utilization, boost system throughput, and improve predication performance in a cloud computing virtualization system. Additionally, the methods designed and knowledge they impart will advance understanding in both systems research and machine learning.

 

Researchers

Laura Brown, PI, Zhenlin Wang, Co-PI

Sponsor

National Science Foundation, CSR: Small: Collaborative Research

Support

$299,993, 3 years

Abstract

Cloud computing has become a dominant scalable computing platform for both online services and conventional data-intensive computing (examples include Amazon’s EC2, Microsoft’s Azure, IBM’s SmartCloud, etc.). Cloud computing data centers share computing resources among a large set of users, providing a cost effective means to allow users access to computational power and data storage not practical for an individual. A data center often has to over-commit its resources to meet Quality of Service contracts. The data center software needs to effectively manage its resources to meet the demands of users submitting a variety of applications, without any prior knowledge of these applications.

This work is focused on the issue of management of memory resources in a data center. Recent progress in transfer learning methods inspires this work in the creation of dynamic models to predict the cache and memory requirements of an application. The project has four main tasks: (i) an investigation into how recent advancements in transfer learning can help solve data center resource management problems, (ii) development of a dynamic cache predictor using on-the-fly virtual machine measurements, (iii) creation of a dynamic memory predictor using runtime characteristics of a virtual machine, and (iv) development of a unified resource management scheme creating a set of heuristics that dynamically adjust cache and memory allocation to fulfill Quality of Service goals. In tasks (i)-(iii), transfer learning methods are employed and explored to facilitate the transfer of knowledge and models to new system environments and applications based on extensive training on existing systems and benchmark applications. The prediction models and management scheme will be evaluated on common benchmarks including SPEC WEB and CloudSuite 2.0. The results of this research will have broad impact on the design and implementation of cloud computing data centers. The results will help improve resource utilization, boost system throughput, and improve predication performance in a cloud computing virtualization system. Additionally, the methods designed and knowledge they impart will advance understanding in both systems research and machine learning.

 

Researchers

Laura Brown, PI, Zhenlin Wang, Co-PI

Sponsor

National Science Foundation, CSR: Small: Collaborative Research

Support

$299,993, 3 years

Abstract

Cloud computing has become a dominant scalable computing platform for both online services and conventional data-intensive computing (examples include Amazon’s EC2, Microsoft’s Azure, IBM’s SmartCloud, etc.). Cloud computing data centers share computing resources among a large set of users, providing a cost effective means to allow users access to computational power and data storage not practical for an individual. A data center often has to over-commit its resources to meet Quality of Service contracts. The data center software needs to effectively manage its resources to meet the demands of users submitting a variety of applications, without any prior knowledge of these applications.

This work is focused on the issue of management of memory resources in a data center. Recent progress in transfer learning methods inspires this work in the creation of dynamic models to predict the cache and memory requirements of an application. The project has four main tasks: (i) an investigation into how recent advancements in transfer learning can help solve data center resource management problems, (ii) development of a dynamic cache predictor using on-the-fly virtual machine measurements, (iii) creation of a dynamic memory predictor using runtime characteristics of a virtual machine, and (iv) development of a unified resource management scheme creating a set of heuristics that dynamically adjust cache and memory allocation to fulfill Quality of Service goals. In tasks (i)-(iii), transfer learning methods are employed and explored to facilitate the transfer of knowledge and models to new system environments and applications based on extensive training on existing systems and benchmark applications. The prediction models and management scheme will be evaluated on common benchmarks including SPEC WEB and CloudSuite 2.0. The results of this research will have broad impact on the design and implementation of cloud computing data centers. The results will help improve resource utilization, boost system throughput, and improve predication performance in a cloud computing virtualization system. Additionally, the methods designed and knowledge they impart will advance understanding in both systems research and machine learning.

 

Researchers

Yu Cai, PI

Sponsor

National Security Agency

Support

$149,184, 1 year

Abstract

Recent high-profile cyber breaches indicate that cyber attacks are becoming more common, sophisticated and damaging. People with cybersecurity skills are in great demand as the threat environment increasingly becomes more complex and challenging. The need to have well-trained and well-prepared cybersecurity workforce is a pressing issue. The goal of this project is to develop a hands-on cybersecurity curriculum with real-world case analysis. The proposed curriculum includes six cybersecurity related courses: 1. Cyber Ethics; 2. Cyber Security I; 3. Scripting for Automation and Security; 4. Wireless System Administration; 5. Cyber Security II; 6. Digital Forensics. This curriculum is designed for CS and IT students who are interested in cybersecurity. 

Researchers

Min Song, PI

Sponsor

National Science Foundation

Support

$221,797, 3 years

Abstract

This project investigates fundamental understanding and challenges of inter-vehicle networking, including theoretical foundation and constraints in practice that enable such networks to achieve their performance limits. This is a collaborative research project with Professor Wenye Wang at North Carolina State University.
Vehicle networks have been playing an increasingly important role in promoting mobile applications, driving safety, network economy, and people’s daily life. It is predicted that there will be over 50 million self-driving cars on the road by 2035; the sheer number and density of vehicles have provided non-negligible resources for computing and communication in vehicular environments. In addition, vehicular communications are also driven by the demands and enforcement of intelligent transportation system (ITS) and standardization activities on DSRC and IEEE 802.11p/WAVE. Many applications, either time-sensitive or delay-tolerant have been proposed and explored, such as cooperative traffic monitoring and control, and recently extended for blind crossing, prevention of collision, real-time detour routes computation, and many others as defined by Car2Car Communication Consortium (C2CCC). Along with the popularity of smart mobile devices, there is also an explosion of mobile applications in various categories, including terrestrial navigation, mobile games, and social networking, through Apple’s App store, Google Play, and Windows phonestore etc. Each aforementioned application seemingly is well-suited for either vehicle-to-vehicle (V2V) ad hoc networks or vehicle-to-infrastructure (V2I) communications. Therefore, vehicular networks have been playing an increasingly important role in promoting mobile applications, driving safety, network economy, and people’s daily life.

In this project, a systematic investigation of vehicular networking properties, which is so called ontology of inter-vehicle communications, will be carried out to acquire in-depth scientific understanding and engineering guidelines that are critical to achieving theoretical performance limits and desirable services. This research includes four key innovative contributions: (i) the discovery of inter-vehicle networks composition by using spectrum cognition in finite and large-scale of V2V and V2I networks, (ii) the space and time domains correlations of vehicles on the move, and development of a set of dissemination strategies using a new constrained mobility model, (iii) detection and identification algorithms to achieve fast neighbor discovery using reinforcement learning, and case-based reasoning scheme; and (iv) theoretical limits of the coverage of messages by following the message trajectory in vehicle networks and schemes to achieve the maximum message coverage in both V2V and V2I networks. The results will advance the knowledge of opportunistic communications and facilitate engineering practice for much-needed applications in vehicular environments.
The intellectual merit of the project centers on the development of theoretical and practical foundations for services using inter-vehicle networks. The project starts from the formation of such opportunistic networks, and then moves on to the coverage of messages, with respect to Euclidean distance and time to stop. Given that an inter-network is in present, the project further studies how resilient such a network under network dynamics, including vehicular movements, message dissemination, and routing schemes. The broader impacts of the proposed research are timely yet long-term, from fully realistic setting of channel modeling, to much-needed applications in vehicular environments, and to transforming performance analysis and protocol design for distributed, dynamic, and mobile systems. Therefore, the proposed research outcome will advance knowledge and understanding not only in the field of vehicular networks, but also mobile ad-hoc networks, cognitive radio networks, wireless sensor networks, and future 5G networks.

Researchers

Xiaohua Xu, PI

Sponsor

National Science Foundation

Support

$244,808, 2 years

Abstract

Cognitive Radio Network is considered as a promising paradigm for the future networks. To significantly improve spectrum utilization, we conduct optimal or near-optimal joint spectrum allocation and scheduling in cognitive radio networks. We address critical and practical challenges for spectrum allocation and scheduling in cognitive radio networks, in particular multi-hop cognitive radio networks, such as dynamic traffic demands and pattern, unpredictable primary user activity, wireless interference, and coexistence. We develop creative models and algorithms in the framework of restless multi-armed bandit where the problem for spectrum allocation and scheduling in cognitive radio networks is formulated as a partially observable Markov decision process. The proposed methodology is novel in that it intelligently combines the networked multi-armed bandit modeling, graph theory, and communication scheduling theories. The developed algorithms, models, and protocols significantly improve spectrum utilization in future wireless communication systems and advance the fundamental knowledge and understanding of cognitive radio networks. The proposed algorithms, protocols, and models enable future wireless systems to design, deploy, and operate much more efficiently than today’s systems, which will result in significant economical, societal, and public safety impacts
The objective of this project is to significantly improve spectrum utilization through conducting optimal or near-optimal joint spectrum allocation and scheduling in cognitive radio networks. The PIs address critical and practical challenges for spectrum allocation and scheduling in cognitive radio networks, in particular multi-hop cognitive radio networks, such as dynamic traffic demands and pattern, unpredictable primary user activity, wireless interference, and coexistence. A test-bed will be set up to extensively evaluate the designed algorithms and protocols.
This project significantly improves the design, deployment, and operation of future wireless communication systems. The proposed algorithms, protocols, and models enable future wireless systems to share spectrum much more efficiently than today’s systems, which will result in significant economical, societal, and public safety impacts. In addition, the proposed research is integrated into education and training for both undergraduate and graduate students. This project also significantly broadens the participation of underrepresented minority groups, e.g., the Native Americans in South Dakota.

Researchers

Xinli Wang, PI, Guy Hembroff, Co-PI

Sponsor

National Science Foundation

Support

$199,934, 4 years

Abstract

The goal of this research is to enhance the security component in undergraduate IT education to meet the strong demand for security professionals in IT fields.
It has been widely admitted by researchers, educators and students that the benefits of hands-on lab experiments are threefold in IT security education:

• They expose students to the real-world challenges of computer and network security.
• They help students consolidate and gain in-depth understanding of the knowledge presented in class lectures.
• These hands-on activities help students to be better prepared for their careers in industry.

This project will develop a collection of instructional hands-on laboratories for undergraduate IT security education to achieve the following objectives:

• To provide students with an active-learning environment by challenging them with real-world problems in the field of IT security.
• To provide students with the opportunity to learn from experiences with advanced technologies and well developed tools which will make them better prepared for their careers in industry.
• To help instructors prepare and deliver security courses more effectively and efficiently by making the lab experiments publicly accessible through the Internet.

This project is a collaborative effort between Michigan Technological University and the University of Washington Tacoma.

Researchers

Zhuo Fang, PI

Sponsor

National Science Foundation

Support

$450,000, 3 years

Abstract

This research is motivated by investigations on scalable methods for design simplifications of nanoscale integrated circuits (ICs). This is to be achieved by extending the associated spectral graph sparsification framework to handle Laplacian-like matrices derived from general nonlinear IC modeling and simulation problems. The results from this research may prove to be key to the development of highly scalable computer-aided design algorithms for modeling, simulation, design, optimization, as well as verification of future nanoscale ICs that can easily involve multi-billions of circuit components. The algorithms and methodologies developed will be disseminated to leading technology companies that may include semiconductor and Electronic Design Automation companies as well as social and network companies, for potential industrial deployments.

Spectral graph sparsification aims to find an ultra-sparse subgraph (a.k.a. sparsifier) such that its Laplacian can well approximate the original one in terms of its eigenvalues and eigenvectors. Since spectrally similar subgraphs can approximately preserve the distances, much faster numerical and graph-based algorithms can be developed based on these “spectrally” sparsified networks. A nearly-linear complexity spectral graph sparsification algorithm is to be developed based on a spectral perturbation approach. The proposed method is highly scalable and thus can be immediately leveraged for the development of nearly-linear time sparse matrix solvers and spectral graph (data) partitioning (clustering) algorithms for large real-world graph problems in general. The results of the research may also influence a broad range of computer science and engineering problems related to complex system/network modeling, numerical linear algebra, optimization, machine learning, computational fluid dynamics, transportation and social networks, etc.