Regensburg 2004 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Downloads | Hilfe

AKSOE: Physik sozio-ökonomischer Systeme

AKSOE 10: Mikro-ökonomische Modelle und Multi-Agenten-Systeme

AKSOE 10.6: Vortrag

Donnerstag, 11. März 2004, 17:00–17:30, H8

Simulated Trading and Working — •Johannes J. Schneider — Department of Physics, Johannes Gutenberg University, Staudinger Weg 7, D-55122 Mainz

Simulated Trading is a multi agent approach to solve combinatorial optimization problems: the overall system shall be optimized by a set of agents who buy and sell parts of the problem, which are currently in their possessions, from or to each other [1,2]. The agents try to solve their system parts in a rather optimum way – but this way is only locally optimum. The results achieved with this approach depend strongly on the overall number N of agents and on the fraction q of agents taking part in a trading move. In order to improve results, the agents shall not only trade but also work in this new approach. This means that improvements are not only achieved with inter-agent trading moves but also with intra-agent local moves. I will show results for an application to the Traveling Salesman Problem and will compare this new approach to the original approach and also to global optimization algorithms like Simulated Annealing.

[1] A. Bachem, W. Hochstättler, M. Malich (1993), The Simulated Trading Heuristic for Solving Vehicle Routing Problems, Report No. 93.139, Institute of Mathematics, University of Cologne, Germany

[2] A. Bachem, W. Hochstättler, M. Malich (1994), Simulated Trading – A New Parallel Approach for Solving Vehicle Routing Problems, Parallel Computing: Trends and Applications: Proceedings of the International Conference ParCo93, Grenoble, France, 7–10 September 1993