The concept of genetic algorithms determines that a parameter configuration is encoded with a real-valued chromosome; each individual contains one chromosome to define a solution candidate. The quality of a candidate is calculated in three steps: In an initial phase, the social force model is calibrated by encoding the candidate. Subsequently, a scenario is simulated for a determined period of time and finally, the simulated data is evaluated via a target function.

The concept is applied to two different target functions, an elementary function as well as a complex function. The first scenario involves an intersection crossed by several pedestrian flows. The objective of this adjustment: controlling the distance each pedestrian is eager to maintain between himself and other pedestrians on the one hand and between himself and occurring obstacles on the other. Pedestrians prefer to keep this distance as large as possible. The second scenario describes a single lane movement realized as an experiment with real human beings at the Juelich Supercomputing Centre [1]. The objective here: minimizing the deviation between the fundamental diagram describing the experiment and a fundamental diagram created by the respective computer simulation. If this objective is met, pedestrian traffic flow can be described as in close accordance with reality.

The two scenarios are simulated with the tool JWalkerS [2]. In both instances, the concept operates effectively: The calibration of the social force model is successful. In addition, the capacity of the social force model to reproduce a fundamental diagram is demonstrated.

In the future, the genetic algorithm will have to be evaluated and advanced for characteristic scenarios to make the search for optimized model configurations more purposeful and thereby faster. In particular, there is a considerable need for further research in the field of the target function. The function will have to be extended by the addition of supplementary criteria, deduced e.g. from physical limits or rules of pedestrian dynamics.

1

A. Seyfried, B. Steffen, W. Klingsch, M. Boltes, "The Fundamental Diagram of Pedestrian Movement Revisited", Journal of Statistical Mechanics, 2005. doi:10.1088/1742-5468/2005/10/P10002

2

M. Hoecker, "Pedestrian Dynamics Simulation with JWalkerS", Leibniz University Hanover, Institute for Computer Science in Civil Engineering, www.bauinf.uni-hannover.de/427.html, 2009.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £78 +P&P)