24thEUROConferenceon.PDFVIP

24th EURO Conference on Operational Research Lisbon, July 11-14 2010 On the multi-agent job shop scheduling problem C. Briand and T. Lehaux (briand@laas.fr, tlehaux@laas.fr) LAAS-CNRS 7 avenue du colonel Roche, F-31077 Toulouse, France Introduction  Scheduling – Assigning start times to a set of interdependent tasks respecting a set of constraints (precedence constraints, resource capacity constraints, etc) in order to optimize a criterion (makespan, maximum lateness, …) – A global decision function / there exists a “supervisor” having a total decision power  However : resources or/and tasks are often distributed among a set of agents having their own decisional autonomy – Ex : supply chain management, project management, timetabling, grid computing – Multi-Agent Scheduling Problem (MASP)  Agent’s knowledge is restricted – Every agent’s organization has to be robust towards environment fluctuations  ANR project no. 08-BLAN-0331-02 called “ROBOCOOP C. Briand – EURO’2010 2 Outlines  Introduction  Multi-agent scheduling problems  The multi-agent job shop scheduling problem  A centralized solving approach  Conclusion C. Briand – EURO’2010 3 Multi-agent scheduling problems  Tasks and resources are distributed among agents  Agents have  a restricted knowledge  their proper objective Task A2 A 3 Time constraint A 1 Resource Use A4