Hello, the randomness of DFRouter is not applied to the vehicle counts but only to the routes (i.e. where vehicles go next after a junction). The only way to get that desired variation in counts is to randomly modify the input data before each run. Likewise, routeSampler will try to match the counts as closely as possible so in a way it's the same issue as with DFRouter. However, if the candidate routes for sampling are not suitable there is more variation in failure. You could try giving more candidate routes and switching on optimization (--optimize) but in the best case you would get the same problem as with DFRouter: closely matching counts each time.
Note, that during the simulation run (when also using sumo option --random), the measured data could vary quite a bit when looking at short time scales. You could try to reduce the aggregation of your validation files to observe this. regards, Jakob Am Do., 1. Okt. 2020 um 07:01 Uhr schrieb Sumo User <sumoquesti...@gmail.com >: > Hello, > > I am trying to run a simulation using detector data collected from a major > highway. I have tried two methods but have not gotten the results that I > wanted. I would like simulations that match the measured data closely but > still have some variation. My measured data is a typical afternoon and I > want to use my measured data as a baseline. So I want my simulations to be > similar to the measured data, but not match it the same way each time. > > First I used DFRouter to create emitter and route files and used them as > the input to sumo. These results came very close to the measured data. > However, when I ran the simulation hundreds of times the validation files > showed that they did not vary much between runs, if at all. This problem > persisted even when I used --random and --randomize-flows. > > So I tried using routeSampler.py instead. This time, I created edgedata > files from my measured data and generated a route file using > routeSampler.py. However, the validation file showed that the measured > data did not match the simulated data at all. For example, the flow rate > was over 1k vehicles per hour smaller in the simulation than the measured > data. Because of this, I did not run the simulation hundreds of times. But > I did run it twice and saw more variation than I did with DFRouter, which > is good. > > Is there any way I can get the accuracy of DFRouter but the variation of > routeSampler (or better)? > > Thank you, > Alex > _______________________________________________ > sumo-user mailing list > sumo-user@eclipse.org > To unsubscribe from this list, visit > https://www.eclipse.org/mailman/listinfo/sumo-user >
_______________________________________________ sumo-user mailing list sumo-user@eclipse.org To unsubscribe from this list, visit https://www.eclipse.org/mailman/listinfo/sumo-user
