Skip to Main content Skip to Navigation
New interface
Reports (Research report)

Calibrating Traffic Simulations as an Application of CMA-ES in Continuous Blackbox Optimization: First Results

Dimo Brockhoff 1, * Anne Auger 1 Fabrice Marchal 2 
* Corresponding author
1 TAO - Machine Learning and Optimisation
LRI - Laboratoire de Recherche en Informatique, UP11 - Université Paris-Sud - Paris 11, Inria Saclay - Ile de France, CNRS - Centre National de la Recherche Scientifique : UMR8623
Abstract : Estimating car traffic is crucial in many big cities around the world to provide users with good alternatives for their travels but also to help decision makers while evaluating the impact of road pricing, new roads, or short-term road works. It will, in the best case, reduce traffic jams drastically resulting in a significantly smaller production of atmospheric greenhouse gases. The estimation of traffic relies thereby on a precise monitoring of the current traffic as well as on a simulation which can reproduce the observed data reliably. To match the observed traffic and the simulation outputs, an optimization of parameters of both the model and the simulator itself is necessary. Within this paper, we present a first study of how state-of-the-art evolutionary computation approaches can be employed in such a scenario. In particular, we use the well-known Covariance Matrix Adaptation Evolution Strategy (CMA-ES) to calibrate the dynamic car traffic simulator Metropolis. As the calculation of the objective function is expensive, a parallel evaluation of the fitness is implemented. First experiments for a simplified city network of Sioux Falls, SD, USA show that the approach is working in principle but also that the objective function contains noise. An easy way to deal with noise within CMA-ES is to simply increase the population size|experiments on easy-to-calculate noisy test functions support this impact exemplary. Additional traffic calibration runs with larger population size, however, do not support this impact but make us believe that numerical precision problems within the simulator are the reason for the noise observed.
Document type :
Reports (Research report)
Complete list of metadata

Cited literature [16 references]  Display  Hide  Download
Contributor : Dimo Brockhoff Connect in order to contact the contributor
Submitted on : Wednesday, June 30, 2010 - 8:56:14 PM
Last modification on : Wednesday, October 26, 2022 - 8:16:44 AM
Long-term archiving on: : Monday, October 4, 2010 - 12:48:05 PM


Files produced by the author(s)


  • HAL Id : inria-00496596, version 1


Dimo Brockhoff, Anne Auger, Fabrice Marchal. Calibrating Traffic Simulations as an Application of CMA-ES in Continuous Blackbox Optimization: First Results. [Research Report] RR-7304, INRIA. 2010. ⟨inria-00496596⟩



Record views


Files downloads