Implementation and Calibration of a Large Scale Mesososcopic Dynamic Traffic Model of Grenoble, France



Implementation and Calibration of a Large Scale Mesososcopic Dynamic Traffic Model of Grenoble, France

Authors

Fabien TSHITEYA, Cerema - Central-Eastern Division, Nicolas DITCHI, Cerema - Western Division

Description

This paper aims to report the ongoing efforts on calibration of a large-scale mesoscopic dynamic traffic model of Grenoble.

Abstract

Traffic simulation is an important tool for modelling the operations of dynamic traffic systems. While microscopic simulation models provide a detailed representation of the traffic process, mesoscopic models capture traffic dynamics of large networks, in lesser details, but with also lesser problems of application and calibration.
Large-scale mesoscopic models are starting to spread although the lack of knowledge from public authorities about their application field and reliability.
This paper aims to report the ongoing efforts on calibration of a large-scale mesoscopic dynamic traffic model of Grenoble.

After importing the static travel demand model (STDM) of Grenoble from VISUM environment into the Aimsun environment, we checked the coherence between these different software’s static affectation results.
The lack of accurate data in the STDM (which is normal for static modelling) made us complete manually parameters such as priority signs, number of lanes or designated lanes, essentials for Aimsun mesoscopic environment. This work lead to an operational network consisting of 232 centroids, 4 078 links, 2 006 nodes (216 of which are control plans). No fundamental diagram calibration for each link has yet been made, in order to spare time and to see the results with a simple “ raw ” implementation.

For the demand calibration, creating a three hour time-dependent OD demand matrix from the one hour static matrix of the STDM was the principal aspect and we chose to test three tools suggested by the Aimsun environment :
- “Matrix Adjustment”
- “Static OD departure adjustment”
- “Dynamic OD adjustment”
The “ matrix adjustment ” is a procedure for calibrating an OD matrix, from an a priori matrix, using available section/detector traffic counts and/or turn traffic counts. The solution algorithm is based on a bi-level model solved heuristically by a gradient algorithm, and includes a static assignment at each iteration. We produced six 30-min matrix with this tool.
The “ static OD departure adjustment ” is an alternative procedure to create a profiled dynamic demand from a static demand. To obtain this profiled demand, the original static demand is distributed through smaller intervals over the time period. The objective is to reproduce the observed traffic counts specified in the real data per interval, staying as close as possible to the original number of OD pair trips for the whole period.
The “dynamic OD ajustment” tool is similar to the “matrix adjustment” tool but is based on dynamic mesoscopic modelling at each iteration. We used it on the six matrix derived from the matrix adjustment described above. This tool was on beta-test during the calibration of our model.

We followed this process on six periods of thirty minutes and, for each period, obtained R² of 95 % and above for the comparison between measured and estimated flow on the 198 detectors.
Those results allowed us to appraise some traffic management scenario with the model.
However some problems in speed calibration remain, and it seems now inevitable to perform a rigorous fundamental diagram calibration for main links.
This further calibration will occur in 2017 after the expansion of the model wished by Grenoble Metropolis.

Publisher

Association for European Transport