This thesis presents our study of the constraint structure of general constraint satisfaction problems (CSPs). It begins with a review of our previous work on developing two constraint-directed CSP algorithms that solve general CSPs directly without having to transform a CSP to an equivalent binary one. Although this previous work has shown that these two algorithms are more efficient than some related CSP solvers, the complexity of these algorithms are still exponential in the number of variables. This thesis shows that characterizing, analyzing and exploiting the structure of the constraints of the given problem leads to improvement.
To characterize the structure of the constraints of a general CSP, we introduce the ý-graph , which raptures the structure of the constraints more precisely than other related representative graphs, such as the join graph , and thus provides a more effective tool for analyzing and solving CSPs.
Next, we define ý-consistency , which is applicable to both binary and non-binary constraints. Because ý-consistency is stronger than some related forms of consistency, such as pairwise consistency and hyper consistency , it can be used to simplify CSP solving by enforcing a greater degree of consistency.
The thesis goes on to show how the ý-graph and ý-consistency can be used to identify a class of tractable CSPs which properly contains tractable CSPs identified using previously reported methods.
Finally, we present new constraint-directed CSP algorithms which incorporate the ý-graph. We show that if the cyclicity of a CSP's ý-graph is less than a fixed number, these algorithms can solve the CSP in polynomial time.
Cited By
- Pang W and Goodwin S A graph based backtracking algorithm for solving general CSPs Proceedings of the 16th Canadian society for computational studies of intelligence conference on Advances in artificial intelligence, (114-128)
- Pang W and Goodwin S Consistency in general CSPs Proceedings of the 6th Pacific Rim international conference on Artificial intelligence, (469-479)
Recommendations
Periodic Constraint Satisfaction Problems: Tractable Subclasses
We study a generalization of the constraint satisfaction problem (CSP), the periodic constraint satisfaction problem . An input instance of the periodic CSP is a finite set of "generating" constraints over a structured variable set that implicitly ...
Periodic constraint satisfaction problems: polynomial-time algorithms
CP'03: Proceedings of the 9th International Conference on Principles and Practice of Constraint ProgrammingWe study a generalization of the constraint satisfaction problem (CSP), the periodic constraint satisfaction problem. An input instance of the periodic CSP is a finite set of "generating" constraints over a structured variable set that implicitly ...
ICHEA for discrete constraint satisfaction problems
AI'12: Proceedings of the 25th Australasian joint conference on Advances in Artificial IntelligenceConstraint satisfaction problem (CSP) is a subset of optimization problem where at least one solution is sought that satisfies all the given constraints. Presently, evolutionary algorithms (EAs) have become standard optimization techniques for solving ...