I'm looking for the mathematical formulation of Wiedemann 99 car-following model, but I can only find a formula and not the algorithm or its way of working.
Can you suggest the articles where I can find this?
Unfortunately, there was never a publication of the 99 model by Wiedemann himself. The paper Alessandro recommended above contains probably all the information you can get.
In case you are interested in the history, here is a small piece of text from a presentation about VISSIM's history I gave at the TRB meeting 2014:
...To be more flexible when calibrating freeway traffic, we added a slightly different version of Wiedemann’s car-following model and made more of its parameters available in the user interface. The “new” Wiedemann model was the result of some private research Rainer Wiedemann did after his retirement. Compared to the original Wiedemann-74 model the model we called Wiedemann-99 added less stochastic noise and was simpler, e.g. the action threshold for a driver changing from free driving to approaching was now defined by a fixed time-to-collision, whereas before it had used a root function of the speed difference. One problem of Wiedemann-99 was that there has never been a publication about it; instead we got the information from Rainer Wiedemann in the form of some printed pages of Basic programming code. So one of the often asked questions in the hotline from academic users was if we can provide a reference for W-99, and we could not. The other problem was that users were not sure which of the two models to choose. For some years there was a recommendation around to use W-74 for urban traffic and W-99 for freeways, but actually it is difficult to justify that by the model itself.
I am not sure. I remember that we were searching for them some years ago, for nostalgic purposes, and did not find them. I will have a second look in my files, and probably ask Lukas Kautzsch to do so as well.
What we really should do is "reengineering" our Vissim source code and extract and publish the model in a readable form. The reason why we never did it was not to keep it confidential, but simply because we never found the time, and the kernel of the car following model was not so interesting for most of the commercial users of Vissim, as long as the model did behave well in calibration. Besides that, there are so many extensions to the model for tactical driving behaviour, that the impact of the pure car-following is not as much as one might expect.
From a bibliographic analysis more detailed of the models Wiedemann 74 and 99, I have a problem. I have seen that the formulations of the car-following proposals in the various articles, are not really present in the algorithm, why?
I finally found some pages which probably have been our source of W99. I attach a scan. But I am not sure if you really get from them what you want. At least it might be historically relevant...
Here I finally have a scan of Wiedemanns thesis from 1974. It is rather low quality, and I hope I will get a better one soon. And it is in German, of course. But the original ALGOL source code is included.
I will have a look on your questions, but it will take some more time.
This is my first time to read the wiedemann thesis from 1974. Thank you for Peter Vortisch. What I know about the Wiedemann model is how to calibrate the parameter of behavior threshold curve. The formula posted by Vincenzo is the speed taken at the next time? Do you know whether there is a formula about the space headway delta x and speed difference delta v? thank you