Testing and Benchmarking of Microscopic Traffic Flow Models. Brockfeld, E. & Wagner, P. In WCTR04 - 10th World Conference on Transport Research, volume abstract book I, pages 775–776, 2004.
Testing and Benchmarking of Microscopic Traffic Flow Models [link]Paper  abstract   bibtex   
Microscopic simulation models are becoming increasingly important tools in modelling transport systems. There are large number of available models used in many countries. The important difficult stage in the development and use of such models is the calibration and validation of the microscopic sub-models describing the traffic flow, such as the car following models for example. The aim of this paper is to present recent progress in calibrating more than a dozen microscopic traffic flow models with very different data sets conducted by DGPS-equipped cars (Differential Global Positioning System), loop detectors and human observers. Different approaches to measure the errors the models produce in comparison to reality are compared. It can be stated that from a microscopic point of view errors of about 15-20 % in headway- and travel time-estimation and about 2-7 % in speed-estimation of individual vehicles in the car following process seem to be the minimal reachable level. Furthermore, the larger the simulation horizon is, the smaller the diversity of the analyzed models become in comparison to the diversity in the driver behaviour. Most interesting, no model cold be denoted to be the best and especially highly sophisticated models did not produce better results than very simple ones.
@INPROCEEDINGS{Brockfeld2004,
  author = {Elmar Brockfeld and Peter Wagner},
  title = {Testing and Benchmarking of Microscopic Traffic Flow Models},
  booktitle = {WCTR04 - 10th World Conference on Transport Research},
  year = {2004},
  volume = {abstract book I},
  number = {A-D},
  pages = {775--776},
  abstract = {Microscopic simulation models are becoming increasingly important
	tools in modelling transport systems. There are large number of available
	models used in many countries. The important difficult stage in the
	development and use of such models is the calibration and validation
	of the microscopic sub-models describing the traffic flow, such as
	the car following models for example. The aim of this paper is to
	present recent progress in calibrating more than a dozen microscopic
	traffic flow models with very different data sets conducted by DGPS-equipped
	cars (Differential Global Positioning System), loop detectors and
	human observers. Different approaches to measure the errors the models
	produce in comparison to reality are compared. It can be stated that
	from a microscopic point of view errors of about 15-20 % in headway-
	and travel time-estimation and about 2-7 % in speed-estimation of
	individual vehicles in the car following process seem to be the minimal
	reachable level. Furthermore, the larger the simulation horizon is,
	the smaller the diversity of the analyzed models become in comparison
	to the diversity in the driver behaviour. Most interesting, no model
	cold be denoted to be the best and especially highly sophisticated
	models did not produce better results than very simple ones.},
  journal = {Proceedings of the 10th World Conference on Transport Research},
  keywords = {traffic flow modeling, benchmarking, simulation, GPS, DLR/TS/VM, model
	calibration},
  owner = {dkrajzew},
  timestamp = {2011.09.30},
  url = {http://elib.dlr.de/6709/}
}

Downloads: 0