Informatica

In this article we analyzed the way in which a technique of obtaining efficient nogood values could combine with a technique of storing these values. In other words we try combining the resolvent-based learning technique introduced by Yokoo with the nogood processor technique in the case of asynchrounous weak-commitment search algorithm (AWCS). These techniques refer to the possibility of obtaining efficient nogoods, respectively to the way the nogood values are stored and the later use of information given by the nogoods in the process of selecting a new value for the variables associated to agents. Starting from this analysis we proposed certain modifications for the two known techniques.

We analyzed the situations in which the nogoods are distributed to more nogood processors handed by certain agents. We proposed a solution of distributing the nogood processors to the agents regarding the agents' order, with the purpose of reducing the storing and searching costs. We also analyzed the benefits the combining of nogood processor technique with the resolved-based learning technique could bring to the enhancement of the performance of AWCS technique. Finally, we analyzed the behavior of the techniques obtained in the case of messages filtering.