Computer Science Department
Current information seeking technologies (i.e., search engines) are designed for large groups of people. SurfAgent adopts a different approach: learning a user's interests and favorite pages so that information gathering can be partially automated and tailored to an individual.
Ph.D. Student: Gabriel Somlo
Recent Paper: "Using
Web Helper Agent Profiles in Query Generation"
(pdf)
To date, no one planner has demonstrated clearly superior performance. Although researchers have hypothesized that this should be the case, no one has performed a large study to test its limits. In this research, we tested performance of a set of planners to determine which is best on what types of problems. The study included six planners and over 200 problems. We found that performance, as measured by number of problems solved and computation time, varied with no one planner solving all the problems or being consistently fastest. Analysis of the data also showed that most planners either fail or succeed quickly and that performance depends at least in part on some easily observable problem/domain features. Based on these results, we implemented a meta-planner that interleaves execution of six planners on a problem until one of them solves it. The control strategy for ordering the planners and allocating time is derived from the performance study data. We found that our meta-planner is able to solve more problems than any single planner, but at the expense of computation time.
Research Assistant: Clayton Daylin
Recent Paper: "A Critical Assessment of Benchmark
Comparison in Planning" (gzip'ed postscript)
The choice of search algorithm can play a vital role in the success of a scheduling application. In our work, we investigate the contribution of search algorithms in solving a variety of synthetic and real-world scheduling problems. We model the search space topology and dynamic performance of local search algorithms for the well known job shop scheduling problem. We have been able to model the performance with high accuracy using Markov models. Additionally, these models have helped explain observations of problem difficulty from the literature and to motivate the development of a simpler algorithm for solving these problems, I-JAR.
For our real application, we have been studying the Air Force
Satellite Control Network scheduling problem. We are comparing performance
of several types of scheduling algorithms: heuristic, genetic algorithms,
local search, tabu search and some hybrids. We have analyzed the
problem and studied several heuristics for it.
Research Assistants: Laura Barbulescu, Mark Roberts, and
Mark Rogers
Recent Papers:
"Scheduling Space-Ground Communications for the Air Force Satellite
Control Network" Accepted to appear in Journal of
Scheduling.
"A
Dynamic Model of Tabu Search for the Job-Shop Scheduling Problem" Accepted to appear in First Multidisciplinary International Conferece on
Scheduling.
Alumni (PhD graduates I used to advise or was on committees for)