search

eclipse-sumo / sumo

Eclipse SUMO is an open source, highly portable, microscopic and continuous traffic simulation package designed to handle large networks. It allows for intermodal simulation including pedestrians and comes with a large set of tools for scenario creation.
https://eclipse.dev/sumo
Eclipse Public License 2.0
2.56k stars 1.43k forks source link

revisit radial drag coefficient #14896

Open behrisch opened 5 months ago

behrisch commented 5 months ago

The energy model currently applies an additional power resulting form the centripetal force when the vehicle drives a curve. While the total amount of the force seems correct, it probably does not result in the work (e.g. power consumption) used in the calculations. Turning the vehicle does not mean that all parts get an additional displacement but rather some parts get less and others more than in straight motion. So the effect is vastly overestimated in the energy model. Probably it should just be an additional factor on the rolling resistance, see https://www.mdpi.com/2075-1702/11/5/570

theDivj commented 5 months ago

Just an observation: The power calculation in src/utils/emissions/HelpersEnergy.cpp does seem to be treating radial drag coefficient/SUMO_ATTR_RADIALDRAGCOEFFICIENT as an additional factor to the rolling resistance in the way you suggest?

A year or so ago I was trying to correlate my observed EV performance with that predicted by SUMO.

I found it very difficult to get any correlation with the SUMO defaults at that time and eventually determined that radial drag was far too high and the efficiencies too low. (The internalMomentOfInertia/SUMO_ATTR_ROTATINGMASS had negligible impact in my experiments - I might guess that was intended to reflect flywheel/gyroscopic impacts of the buses in the original model?)

As the existing figures were derived from a bus model I'd assumed that the radial drag figures used were associated with the much greater tread contact of larger tyres, doubled on rear axles etc. The low efficiencies were just not reflecting improvements in battery/motor technology.

The paper cited is computing for two axle passenger cars - I might wonder if the multiple rear axles as found on trucks/buses have a non-linear relationship with radial drag - eg a simple doubling of the transverse/longitudinal ratio feels wrong? (Vehicles with Steering rear wheels might offer even more fun!)

I found that adjusting those values gave me real world comparable behaviour but I had to use realistic road networks to get consistent results across long and short routes, ie The many junctions, curves roundabouts etc. of my real world routes, as retrieved from OSM, were necessary to let me get within better than 10% of observed.