Calibrating spatio-temporal network states in microscopic traffic simulation on a global level

on YouTube ↗22m

" summary

Global calibration adjusts network-wide relative mean speed over 24 hours for a downtown Berlin extract with ~500k trips by combining MATSim demand tweaks, reduced urban capacities and velocities, SUMO rerouting and teleport vehicles to relax hotspots. Baseline SUMO only admits ~40% demand; >1,300 SUMO runs with 30 seeds each show combining levers doubles mean speeds in peak hours. A genetic algorithm tunes parameters to match daytime travel times closely but fails to reproduce realistic delays at night.

AI-written summary · about

" chapters (10)

1. 0:00Intro
2. 0:46Calibrating spatio-temporal network states in SUMO to MATSim 1. Motivation - yet another calibration approach?
3. 6:02Calibrating spatio-temporal network states 2. Methodological approach
4. 8:32Calibrating spatio-temporal network states 3. Status quo: What travel demand share goes currently through?
5. 10:244.1 Exploration parameter space
6. 13:364.2 Impact of individual measures
7. 15:124.3 Impact of combined measures (best-ranked levers)
8. 16:445.1 GA Framework design and problem formulation
9. 18:125.2 Discussion best-fit solution
10. 20:34Conclusion and outlook