Ubiquitous and Embedded Computing Systems Lab

Accessible Learning Technologies and Precollege Engineering Curricula

Engineering professions face a unique set of challenges including the underrepresentation of women and minority groups, as well as a lack of interest or poor understanding of these professions among younger audiences. Moreover, schools have limited resources to obtain learning technologies and curriculums for engineering education, as well as face a shortage of teachers experienced in these areas. While not every student is expected to become an engineer, introducing engineering concepts and curriculums at the middle and high school level are recognized to cultivate valuable skill sets regardless of future educational and career goals. Research has shown that middle school students are at a critical age where exposure to STEM fields can greatly impact their career goals.

In this project we focus on the development and integration of low-cost, accessible learning technologies for resource constrained schools. In addition, we have sought to alleviate many of the roadblocks faced in the integration of engineering curriculums by researching and developing curriculums to enhance low-cost, low-tech, learning technologies that take into account classroom management issues and support for out-of-field teachers.

Dynamic Profiling and Optimization (DPOP) for Distributed Sensor-Based Platforms

Sensor network platforms consist of a distributed set of sense and compute nodes and promise numerous application possibilities, including but not limited to, environmental, structural, medical, industrial, and military monitoring. One of the main disadvantages of the existing technology is the complexity of implementing a single application, with each application composed of a unique set of requirements. A wide range of expertise is needed, including software development, hardware design, sensor interfacing, operating systems, networking, and optimization. Given these challenges, the accessibility and feasibility of these sensor network platforms are limited. Users of these platforms are not necessarily engineers with extensive programming or engineering experience, rather scientists, biologists, or teachers. We referred to these users as application experts.

To alleviate much of the complexity associated with sensor-based platform use and optimization, the DPOP framework was introduced - consisting of an innovative suite of dynamic profiling, analysis, and automated optimization methods combined with intuitive methods for defining application requirements that can be readily and effectively utilized by application experts. A configurable Profiler Module was developed to enable an accurate view of an application's execution behavior at run-time within the deployment environment, while ensuring low overheads are incurred. In addition, Overhead Estimation and Profile Data Management Modules were integrated to assist a developer in choosing a customized profile methodology, as well as the development of a Code Generator Module to easily integrate these mechanisms within an application. In terms of optimization, interfaces that utilize fuzzy logic were introduced to enable application experts to more intuitively specify design tradeoffs.

For more information on this project, visit the project page.

Integrated Optimization and Monitoring Platform for Hierarchical Smart Grid Applications

Unlike most consumable goods, energy consumption is invisible and abstract, leading to a poor understanding of energy, as well as misconceptions in how to conserve energy. While the tools to monitor energy usage are becoming more prevalent, understanding how behavior and decision have long term impacts on energy consumption and cost is challenging for an end user. Thus the principles of computer-aided design and user-interface design are applicable within this domain to present data in a meaningful way to a variety of end users as well as provide guidance in how to effectively utilize and optimize energy usage.

We consider various approaches to tackle this problem. In one research thrust we look at a data-driven, hierarchical simulation-based planning and control approach as the basis for a management framework to optimize energy usage of applicances based on existing and predicted resource availability. The framework strives to look at optimization and control from the house to community level, mandating the need for large-scale distributed coordination between users and resources.

For more information on this project, visit the project page.

In another research thrust we focus on the heating and cooling subsystem and present novel management framework which incorporates a light-weight data-driven prediction component to dynamically learn the data relationships of the underlying system and estimate its future behavior within a given time horizon. This technology is in the early stages of commercialization, more information can be found at Acomni, LLC.

Sensor Network Architectures for Compute-Intensive Applications

Sensor network platforms are slowly facilitating the realization of numerous ubiquitous computing environments. As these platforms mature, application possibilities have similarly matured from simple monitoring applications that may track motion or temperature to robust surveillance applications that combine cameras with image processing algorithms to automatically extrapolate data without constant human intervention. As these sensor network applications continue to evolve, the existing node level hardware struggle to balance the large amounts of computation and communication needed with limited energy resources.

For more information on this project, visit the project page.