TY - GEN

T1 - Constrain relaxation in distributed constraint satisfaction problems

AU - Yokoo, Makoto

PY - 1993

Y1 - 1993

N2 - The distributed constraint satisfaction problem (DSCP) formulation has recently been identified as a general framework for formalizing various Distributed Artificial Intelligence problems. In this paper, we extend the DCSP formalization by introducing the notion of importance values of constraints. With these values, we define a solution criterion for DCSPs, that are over-constrained (where no solution satisfies all constraints completely). We show that agents can find an optimal solution with this criterion by using the asynchronous incremental relaxation algorithm, in which the agents iteratively apply the asynchronous backtracking algorithm to solve a DCSP, while incrementally relaxing less important constraints. In this algorithm, agents act asynchronously and concurrently, in contrast to traditional sequential backtracking techniques, while guaranteeing thee completeness of the algorithm and the optimality of the optimality. Furthermore, we show that, in this algorithm, agents can avoid redundant computation and achieve a five-fold speed-up in example problems by maintaining the dependencies between constraint violations (nogoods) and constraints.

AB - The distributed constraint satisfaction problem (DSCP) formulation has recently been identified as a general framework for formalizing various Distributed Artificial Intelligence problems. In this paper, we extend the DCSP formalization by introducing the notion of importance values of constraints. With these values, we define a solution criterion for DCSPs, that are over-constrained (where no solution satisfies all constraints completely). We show that agents can find an optimal solution with this criterion by using the asynchronous incremental relaxation algorithm, in which the agents iteratively apply the asynchronous backtracking algorithm to solve a DCSP, while incrementally relaxing less important constraints. In this algorithm, agents act asynchronously and concurrently, in contrast to traditional sequential backtracking techniques, while guaranteeing thee completeness of the algorithm and the optimality of the optimality. Furthermore, we show that, in this algorithm, agents can avoid redundant computation and achieve a five-fold speed-up in example problems by maintaining the dependencies between constraint violations (nogoods) and constraints.

UR - http://www.scopus.com/inward/record.url?scp=0027867092&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0027867092&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:0027867092

SN - 0818642009

T3 - Proceedings of the International Conference on Tools with Artificial Intelligence

SP - 56

EP - 63

BT - Proceedings of the International Conference on Tools with Artificial Intelligence

A2 - Anon, null

PB - Publ by IEEE

T2 - Proceedings of the 5th International Conference on Tools with Artificial Intelligence TAI '93

Y2 - 8 November 1993 through 11 November 1993

ER -