The Effect of Flag Introduction on the Explosion of Nogood Values in the Case of ABT Family Techniques

Abstract

Starting from the algorithm of asynchronous backtracking, a unifying framework for some of the asynchronous techniques has recently been suggested. Within this unifying framework, several techniques have been derived, known as the ABT family. The asynchronous backtracking technique is characterized by an explosion of nogood values for the large dimensions problem. The storing of these values, needed within the asynchronous backtracking technique needs a very large quantity of space. Recently, a derivate technique called asynchronous backtracking with flags allowed the reduction of nogood values explosion. In this article, a solution of flag technique application is suggested for the ABT family. It implies the indexing of local values in the domain of each flagged agent. This eliminates the nogood value storing and, therefore, for great dimension problems, allows avoiding the nogood value explosions. The allocated memory space is considerably reduced in the case of flag technique. Solutions are being suggested for the derivation of the common kernel in order to reach two known techniques: Asynchronous Backtracking and Distributed Dynamic Backtracking.