Constraint Solving by Composition Student :

A structure-based approach to the solution of constraint problems (e.g., tree-based decomposition) typically assumes that knowledge about the problem remains static, and demands extensive resources to pursue the solution. This may explain why such an approach is sometimes infeasible on large-scale, real-world problems. This paper proposes a new approach, called composition that partitions a problem into a disjoint set of star-like subproblems. By design, search for solution on each subproblem is tractable. Composition identifies, solves, and then propagates the result one subproblem at a time. This reformulation makes the original problem model both denser and tighter, and often facilitates search. Initial tests have proved promising.