Asynchronous Backtracking variant applied to a rolls replacement process within a wire rod mill

Abstract

There is a wide number of application problems included in multi-agent systems and distributed artificial intelligence which can be described as the issue of finding a consistent combination satisfying all the existing inter-agent constraints, that is, they are distributed constraint satisfaction problems (DisCSPs). In this paper, a new technique for solving such kind of issues when they can be modelled as a succession of interrelated elements with constraints among an agent and its immediately following, having one of the agents the ability to notify either that the problem has been solved or that there are not any possible solutions, is presented. It has been developed using Asynchronous Backtracking, one of the most known algorithms for solving DisCSPs as a basis. To validate its applicability, it has been tested in a wire rod rolling mill for rolls replacement decision making process. It was implemented using the PADE multi-agent framework. The results demonstrate the method is valid to solve this kind of problems usually solved with recursive algorithms such as backtracking.