2004 Fiscal Year Final Research Report Summary
A Study on Analyzing Phase Transition Phenomena for Constraint Satisfaction Problems
| Project/Area Number |
14380134
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
計算機科学
|
|---|
| Research Institution | University of Tsukuba |
|---|
Principal Investigator |
NISHIHARA Seiichi University of Tsukuba, Graduate School of Systems and Information Engineering, Professor, 大学院・システム情報工学研究科, 教授 (50026168)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
FUKUI Yukio University of Tsukuba, Graduate School of Systems and Information Engineering, Professor, 大学院・システム情報工学研究科, 教授 (80311596)
KATO Nobuko Tsukuba College of Technology, Department of Information Science and Electronics, Associate Professor, 電子情報工学科, 助教授 (90279555)
|
|---|
| Project Period (FY) |
2002 – 2004
|
| Keywords | constraint satisfaction problem / search / phase transition / heuristics / graph coloring problem / NP-complete / computational complexity / knowledge representation |
|---|
| Research Abstract |
The purpose of our study is to observe phase transition (PT) phenomena for constraint satisfaction problems (CSP) and analyze the factors of PTs. Based on the purpose, (1) we have proposed the method that systematically generates very hard problem instances with combinatorial complexity and (2) we have developed a constraint satisfaction system for solving real-world and large-scale problems.
In (1), using the graph 3-colorability problem (3COL) which is one of the typical CSPs, we found the original minimal unsolvable structures to generate very hard 3COL instances. By embedding one of the structures repeatedly, we can stably generated arbitrarily large 3COL instances. We demonstrated that the computational time of our generated 3COL instances is of an exponential order for some experiments using well-known efficient search algorithms and coloring programs.
In (2), we defined the search problem that decides a work schedule for rearranging office layouts as the CSP. Rearranging floor layouts includes time-consuming task, where both of planning and scheduling are solved simultaneously. We can solve the problem efficiently by unifying the constraints for requiring both of planning and scheduling.
|
