Two New Mexico State University doctoral students' theories could help robots solve their own problems without much information. NMSU computer science students Khoi Nguyen and Vien Tran have been awarded "Best Student Paper" at the International Conference on Automated Planning and Scheduling for their work with software programs that solve problems and accomplish things on behalf of humans. The two will present their paper at a special conference of the ICAPS conference June 25-29 in Brazil.
"I feel honored to receive this award, Tran said. "It is great that we got the best from what we strike to achieve. It also shows that the planning and scheduling community recognizes the value and contribution of our work."
"I think it's a good thing that we're going to Brazil with the best student paper," Nguyen said. "In the science community, when you go to a conference, they always ask you if you have a paper. This time, I can say we had the best student paper."
ICAPS is a forum for exchanging news and research results on theory and applications of intelligent planning and scheduling technology. The NMSU students' paper, "On Computing Conformant Plans Using Classical Planners: A Generate-And-Complete Approach," focuses on intelligent agents reaching solutions with incomplete information, such as a robot navigating in an unknown terrain on a different planet.
"This type of problem is very hard to solve," said Enrico Pontelli, computer science department head in the College of Arts and Sciences. "What Khoi and Vien did is to discover that the problem can be ingeniously reduced to solving simpler versions of the problem, by making some assumptions about the missing knowledge. Their approach is simply brilliant in its simplicity."
Nguyen said the original approach was established in 2002, but was abandoned because researchers believed other approaches were superior. Nguyen and Tran revised the method using the reduction techniques.
"At the beginning, we had ideas on how to approach some of problems, but we were not clear on how to generalize the solution for the whole class of problems," Tran said. "It took us sometime to figure it out. The generalized solution turned out very effective in helping to reduce the searching time and to increase the scalability in solving problems."
The duo originally submitted the paper last year, but the committee rejected it for incomplete information. Nguyen said they resubmitted after further explaining their reasons for success.
Of the 132 papers submitted this year, the ICAPS committee accepted only 45. Nguyen and Tran took home one of the top three awards.
"This is an enormous achievement - the conference peer-reviews and accepts a small percentage of the submissions, thus being accepted to present at ICAPS is in itself a distinguished accomplishment; being able to receive the best paper award is simply outstanding and demonstrates the top quality research the students have conducted," Pontelli said.