2010 | OriginalPaper | Chapter
A Genetic Algorithm Module for Spatial Optimization in Pedestrian Simulation
Authors : Lukas Kellenberger, Ruedi Müller
Published in: Pedestrian and Evacuation Dynamics 2008
Publisher: Springer Berlin Heidelberg
Activate our intelligent search to find suitable subject content or patents.
Select sections of text to find matching patents with Artificial Intelligence. powered by
Select sections of text to find additional relevant content using AI-assisted search. powered by
Regarding pedestrian simulation applications, technologies to optimize the built-up environment apart from pure analysis of pedestrian flows, and based on simulation results, are of crucial importance for the wider acceptance of pedestrian simulation. Apart from conventional pedestrian analysis measures such as density maps, flow rates and travel times, optimization of spatial configurations, leading to congestion or travel time reduction, promises an additional benefit for users of the simulation. Spatial optimization therefore delivers specific solutions for the application of pedestrian simulation in general.
Here we present a genetic algorithm optimizer module, a prototype created for the pedestrian simulation software SimWalk. Based on CAD plans, the module allows optimizing plans and objects (walls, obstacles, etc.) automatically. The user defines and marks a plan section for optimization where, for example, pedestrian density problems occur. Additionally, the user defines which changes of the built-up environment are allowed, based on boundary conditions predefined by his or her architectural or engineering knowledge. After having defined these boundary conditions, the evolutionary process performed by the genetic algorithm gets started, and a first generation of plans and predefined populations is generated. Every succeeding plan shows random variations of the selected obstacles. To evaluate the fitness of each generation, density maps and travel times generated by the software are used to optimize the selected environment. The ultimate goal consists in finding plan configurations with low densities and shorter travel times. If the first generation is established, the best plans can be identified. Based on “elite selection” (“survival of the fittest”), the next generation then gets started, using various GA operators like random generators, selectors, recombination and mutation to generate new plan variations. Every generation, in optimal cases, results in a better plan configuration. A main topic of the research project consisted in mapping the scalability of plan obstacles to the chromosomes of an already existing GA framework of the research institute. To get trend information during the software development, it was necessary to develop a graphical user interface (GUI). It made it possible to edit and prepare plans for optimization, and additionally to select interim solutions for simulation with different parameters, boundary values, population sizes and operators. Statistical tests have shown that with the existing operator set and favorably chosen parameters, after a few generations a significantly improved plan can be achieved. With this prototype, a first result for the optimization of spatial environments in pedestrian simulation regarding congestion and travel times has been accomplished.
Further research will include an extended operator framework to find better results in a shorter time. Additionally, the application workflow will be improved for more intuitive work.