The system that is demonstrated by live demo software is an agent platform developed for the DEAL (Distributed Engine for Advanced Logistics) project. The primary goal of this project is to enhance the utilisation of trucks for transportation companies. This is to be realised by a distributed system that connects all trucks, orders, planners, and customers. Each of these entities is modelled by an agent. In a logistics setting, the main problem is to nd the best truck for a container while nding appropriate routes (plans) for all trucks. In this project a feasible solution for this problem is constructed by distributed, communicating agents. For example, a container agent may request its transportation from a truck agent, and a truck agent can refuse such a request if it may be able to construct a more ecient route without this container. A similar system is introduced in [DC04][FMPS95], with the dierence that they have not modelled the orders by agents, instead the company and truck agents take care of all aspects. Modelling the system by agents has the benet that the load can be distributed among several computer, and by putting the emphasis on coordination of agents exceptional situations like truck breakdown, or traffc jams, accidents can be modelled and handled more efficiently.

The agent platform used in the project is designed to study agent coordination. In the logistics case auctions are used to coordinate the agents. An order agent can initiate an auction where truck agents can bid to get this order. The demo will show how such auctions can lead to a solution to the logistical problem.

The planning for all trucks is achieved incrementally by auctioning one order at a time. In the demo the plans of the trucks are visualised textually through the agent attributes and graphically by connecting cities which should be visited by the trucks by lines on the map of The Netherlands. Step by step, as the auctions go, the lines indicating the tours are growing longer and longer. In the end all the orders are sold in the auctions and all the trucks have a list of cities to visit.

references

[DC04] K. Dorer and M. Calisti. Agent-based dynamic transport optimization. Technical Report WT-2004-05, Whitestein Technologies, 2004.

[FMPS95] Klaus Fischer, Jörg P. Muller, Markus Pischel, and Darius Schier. A model for cooperative transportation scheduling. In Proceedings of the First International Conference on Multiagent Systems., pages 109--116, Menlo park, California, June 1995. AAAI Press / MIT Press.