A Backjumping Search Algorithm for a Distributed Memory Multicomputer

Abstract

Solving Constraint Satisfaction Problems (CSPs) has been subject to intense study by earlier researchers because CSPs can be used to model a whole variety of practical real world problems. This work involves investigation of the performance of parallel backtracking algorithms for solving CSPs. Two classes of backtracking search algorithms are considered: i) chronological backtracking, and ii) dependency directed backtracking, called backjumping. Results show that the new parallel backjumping algorithm retains the efficiency that the sequential algorithm exhibits, while realizing twice the speedup of parallel backtracking for random constraint networks.