Project Title:

Evaluating the Architectural Design Quality of Object-Oriented Software
[2nd yr of project]

Principal Investigator:

James Bieman, Associate Professor
Department of Computer Science
Colorado State University

Collaborating Companies:

Storage Technology

Company Representative:

Frank Molnar

We will develop techniques to evaluate the quality of the architectural designs of object-oriented software systems. Our focus is on identifying design structures and patterns that will make the software easier to adapt and test. We will develop our techniques through design pattern recognition and measurement, and an examination of a specific software system developed by the collaborating company. We will also develop and evaluate methods for reconstructing software architectural designs from implementations, based, in part, on design-pattern recognition.

Project Title:

Real-Time 3D Navigation Using Video

Principal Investigator:

Christian Debrunner, Assistant Professor
Division of Engineering
Colorado School of Mines

Collaborating Company:

Omnitech Robotics, Inc.

Company Representative:

Sid Gluckman, Vice President of Research & Development

Many robotic vehicle applications would benefit from a vision-based sensor which provides accurate three-dimensional location of the vehicle relative to its environment as well as the three-dimensional structure of the environment. Other sensors providing similar capabilities are either too expensive, too slow, or too inaccurate. Recent developments in computer vision in the area of projective geometry, in particular the formulation of the trifocal tensor, provide the mathematical tools needed to develop such a capability. We propose to develop algorithms providing such capabilities and to demonstrate a robot vehicle that navigates using the algorithms.

Project Title:

Carrier Interference Pulse Shaping for Improved TDMA Capacity and Performance(CIMA)

Principal Investigator:

Carl Nassar, Assistant Professor
Electrical Engineering
Colorado State University

Collaborating Companies:

Idris Communications

Company Representative:

Steve Shattil, Chief Data Architect

The researchers design and develop an innovative software simulator. The simulator determines the capacity (measured in bits/sec/TDMA). CI/TDMA refers to TDMA where a novel pulse shaping (based on carrier interference) is applied. The simulator determines capacity and performance results of CI/TDMA under a diverse range of practical communication environments (those environments in which TDMA is or is likely to become standard). In addition to a simulator determining the performance results of CI/TDMA, the researchers provide complete sets of simulated performances and capacities, and a detailed analysis of these results, including comparisons with traditional TDMA physical layer performances.

Project Title:

Evaluating Machine Learning Methods for Data Mining

Principal Investigator:

Darrell Whitley, Professor
Department of Computer Science
Colorado State University

Collaborating Companies:

RML Technologies, Inc.

Company Representative:

Frank Francone

Data Mining has become increasingly important to many industries. Machine Learning methods are widely used for data mining, including neural networks, decisions trees, and more recently, genetic programming. This research will 1) evaluate these technologies on a number of test problems from the UIC repository and on data supplied by the collaborating company. We will also 2) develop and test new hybrid methods that emphasize problem decomposition as a strategy for improving generalization. In particular, we will develop and embed these technologies and strategies within a genetic programming learning system. The collaborating company, RML, markets the tools and services for the AIM-Genetic Programming system and has special interest in data mining applications.

Project Title:

Customer Transactions Analysis in Electronic Commerce

Principal Investigator:

Xindong Wu, Associate Professor
Department of Mathematical and Computer Sciences
Colorado School of Mines

Collaborating Company:

U S West Advanced Technologies

Company Representative:

Marcia Derr

A market basket in electronic commerce is a customer transaction that records the products purchased by an individual buyer. An electronic commerce system accumulates large collections of such market baskets over time. This project will start with a synthetic transaction database of market baskets and analyze buying patterns across vendors by applying data mining techniques (viz. association analysis and cluster analysis). The discovered buying patterns can be used to identify opportunities for cross advertising, and to better serve registered buyers. The essential research issue in this project is data mining with large databases in the electronic commerce scenario. The number of market baskets can be too large to be processed at one time, especially when there are different vendors and buyers from different geographic regions. Therefore, while putting all transactions together might amass a huge transaction database for centralized processing, segmenting transactions by buyers' geographic region, mining association rules at different regions and sharing the rules (rather than the original transactions) provides a feasible way to deal with it. We will design a model to aggregate association rules from different collections of transactions.

Project Title:

Reinforced Soil Design: Integration of Digital Image Analysis, Numerical Modeling and Limit Equilibrium
[2nd yr of project]

Principal Investigator:

Jorge Zornberg, Assistant Professor
Department of Civil, Architectural and Environmental Engineering
University of Colorado - Boulder

Collaborating Company:

Bowman Construction Supplies, Inc.

Company Representative:

David Bowman

Soil reinforcement is now a highly attractive alternative for highway embankment and retaining wall projects because of the economic benefits it offers in relation to conventional retaining structures. Reinforced soil design is typically performed using computational techniques (limit equilibrium), which require significant assumption regarding strain distributions within the reinforcements. However, to date, the assumed strain distributions have not been fully validated against monitored results. One consequence of this is a perceived overconservatism in design.

The overall objective of this CASI Technology Transfer plan is to generate and integrate, via advanced software technology, strain data that will provide a definite answer regarding reinforcement strain distributions to be adopted in design. Specifically, this project will integrate the results from: (i) A digital processing effort involving images of reinforced soil models videotaped in-flight during testing in a geotechnical centrifuge; (ii) a numerical (finite element) modeling validation of the digital data collected in the previous phase; and (iii) a limit equilibrium investigation that will incorporate reinforcement strain distributions drawn from the previous two phases into a still practical, though more accurate reinforced soil design methodology.