Wiedemann 99 model has 10 calibration components CC0 - CC9 in VISSIM driver behavior car-following options for freeways. Can we use vehicle trajectory data e.g. those available on NGSim project's websit to estimate those parameters? Are means of those variables more suitable, or modes?
Unfortunetly I tried that but doesn't work Umair. I calibrate VISSIM for brazilian freeways and the main parameters that you should variate for better and faster freeway calibration is CC0, CC1, CC4 and CC5.
Hi Marcelo,
Thanks for your response. According to some studies CC0 is correlated with CC8 and CC4/CC5 with CC6. Therefore, I think it is important to vary those as well. I have tried varying all 10 simultaneously (using estimations from NGSiM data) and got a better cumulative average speed distribution of vehicles compared to the default CC. For the people who understand the theory behind WIedemann 99 model, I want to know if this is an appropriate method for calibration?
Umair, attached you'll find one of my articles about VISSIM calibration for freeways. It is in portuguese but you can take a look in one chart that present a bibliographic review about the CC parameters variations. Hope helps you.
Thanks for the explanation and references, Tobias. What about parameter CC6? Is it theoretically possible to estimate it from trajectory data? Since it is the dependency of speed oscillation on the distance to lead vehicle. For each 'following' oscillation loop (on the plot of relative speed vs distance to lead vehicle), if we determine the max and min distance and determine the standard deviation of following vehicle speed in that loop and then compare the same with other loops for the same driver, won't this be an estimate of change in speed oscillation with change in distance (definition of CC6)? Is this a meaningful estimate?
"change in speed oscillation with change in distance" is, to me, a strange definition. Which aspect of speed oscillations should change? amplitude, frequency, both, neither? So, I stick to the math telling the truth: According to the Wiedemann99 model as presented in Aghabayk et al., CC6 is the main factor defining the CLDV and OPDV thresholds, e.g., the CLDV threshold:
dv_CLDV=CC6*gap**2-CC4
where CC4 (e.g., -0.35 m/s) has only a small contribution. The higher CC6, the later the driver reacts to speed differences when closing in (CLDV) or leaving (OPDV) the interaction zone. So, CC6 is proportional to the width of the indifference zone with respect to speed differences, for sufficiently large gaps. In ideal situations, this means that the oscillation amplitude (in terms of dv) increases, and the frequency decreases, with CC6. In practical situations, however,this is overlain by macroscopically non-steady-state traffic situations, inter-and intra-driver variations etc. Moreover, during one oscillation, the driver moves of the order of 0.5 km, i.e., through the whole measuring region of NGSIM. Finally, there is some important confounder: stop-and-go waves observed in some NGSIM data are collective effects and have nothing to do with CLDV-OPDV oscillations. So, calibrating CC6 by some observed oscillations of NGSIM data is clearly impossible.
The only chance to calibrate CC6 by NGSIM data would be (i) trying to detect discrete changes in acceleration while closing in/leaving the interaction zone (i.e., directly observing the thresholds defined by CC6), (ii), more realistically, observing some macroscopic stability consequences: The higher CC6, the more unstable the traffic flow. However, this is also confounded by other parameters influencing stability.
- Badges
- Science topic