Approaches for resolving the dilemma between model structure refinement and parameter calibration in agent-based simulations

ABSTRACT

Agent-based simulations form a valuable tool for learning about real world societies and global behaviors of systems emerging from microscopic relationships. Calibration of model parameters for detailed agent-based models is a big problem for standard calibration techniques, due to the large parameter search spaces, long simulation run times, uncertainties in the structural model design and different observation levels upon which the model needs to be calibrated. In this paper we propose several methods to improve the calibration process of agent-based simulations.

REFERENCES

• 1: Benoit Calvez and Guillaume Hutzler. Automatic tuning of agent-based models using genetic algorithms. In Luis Antunes and Jaime Semao Sichman, editors, Proceedings of the 6th International Workshop on Multi-Agent-Based Simulation 2005, 2005.
• 2: Jim Doran, Agent-based modelling of ecosystems for sustainable resource managment, Mutli-agents systems and applications, Springer-Verlag New York, Inc., New York, NY, 2001
• 3: Manuel Fehler, Franziska Klügl, and Frank Puppe. Techniques for Analysis and Calibration of Multi-agent Simulations. In M.-P. Gleizes, A. Omicini, and Franco Zambonelli, editors, ESAW 04, LNAI 3451, pages 305--321. Springer-Verlag Berlin Heidelberg 2005, 2005.
• 4: Michael C. Fu, Feature Article: Optimization for simulation: Theory vs. Practice, INFORMS Journal on Computing, v.14 n.3, p.192-215, July 2002 [doi>10.1287/ijoc.14.3.192.113]
• 5: Douglas C. Montgomery, Design and Analysis of Experiments, John Wiley & Sons, 2006
• 6: Jorge Nocedal and Stephen J. Wright. Numerical Optimization. Springer Verlag New York, Inc., 2000.
• 7: Carles Sierra , Jordi Sabater , Jaume Agustí , Pere Garcia, Evolutionary programming in SADDE, Proceedings of the first international joint conference on Autonomous agents and multiagent systems: part 3, July 15-19, 2002, Bologna, Italy [doi>10.1145/545056.545114]

INDEX TERMS

Primary Classification:
  I. Computing Methodologies
  I.6 SIMULATION AND MODELING

General Terms:
Algorithms, Experimentation, Measurement

Keywords:
agent-based simulation, simulation calibration