A Multi-Agent System for Integrating a Large-Scale Project

This paper presents a multiagent-system for integrating AL-ICE: a real-life, distributed, large-scale particle physics experiment of the LHC project at CERN. The work is concerned with solving the large-scale coordination problem using a distributed constraint satisfaction algorithm. We believe that until now, no such algorithm has successfully handled problems of this size. The developed algorithm is a local search algorithm, based on the distributed breakout algorithm and is extended by an incremental problem solving scheme combined with a variable ordering heuristic. By benchmark results we show the algorithm performs extremely well and outperforms the standard version by orders of magnitude. In addition, we develop a constraint based project planning ontology for the knowledge exchange amongst the AL-ICE subproject agents. By using domain variables and constraints for information exchange, ambiguities due to cultural differences are eliminated.