Resolving CSP with naming games

Abstract

In constraint satisfaction problems (CSP) we consider N variables x 1,x 2, ...x N , their definition domains D 1,D 2, ...,D N and a set of constraints on the values of these variables; solving the CSP means finding a particular value for the variables that satisfies the constraints. In the distributed CSP (DCSP) as defined by Makoto Yokoo [1], the variables of the CSP are distributed among the agents. These agents are able to communicate between themselves and know all the constraint predicates relevant to their variable. The agents through interaction find the appropriate values to solve the CSP. The naming game describes a set of problems in which a number N of agents bootstrap a commonly agreed name for an object. Each naming game is defined by an interaction protocol/algorithm. An important aspect of the naming game is the hierarchy-free agent architecture. For other references on the naming game see the work of Steels [3] and Baronchelli et al. [2]. © 2008 Springer Berlin Heidelberg.