Solution reuse in dynamic constraint satisfaction problems
AAAI '94 Proceedings of the twelfth national conference on Artificial intelligence (vol. 1)
Solving dynamic constraint satisfaction problems by identifying stable features
IJCAI'09 Proceedings of the 21st international jont conference on Artifical intelligence
| Hi-index | 0.00 |
Constraint programming is a successful technology for solving combinatorial problems modeled as constraint satisfaction problems (CSP). An important extension of constraint technology involves problems that undergo changes that may invalidate the current solution. These problems are called Dynamic Constraint Satisfaction Problems (DynCSP). Previous works on dynamic problems sought methods for finding new solutions. Other works developed exploring methods for finding stable solutions. In this paper, we focus on determining, among a set of alternatives, which solution is the most stable. It is well-known that, given a set of solutions for a problem, the user desires to select the most stable solution, that is the solution that is more likely to remain valid after changes that alter the set of constraints or valid assignments. Thus, we present a tool for studying the stability of solutions of a CSP. This tool is composed by two techniques that analyze the solutions and return which solution is more stable under changes. The first technique generates a set of pseudo-random constraints and study how many constraints each original solution satisfies. The second one modifies the value of some variables of the original solutions and determines how many original constraints they satisfy. Thus, we can analyze which solution is able to absorb a large number of disturbances.