poutnikl / Brouter-profiles

My profiles for Brouter offline routing on Android platform
76 stars 17 forks source link

Turncost for main roads #9

Closed utack closed 7 years ago

utack commented 7 years ago

I wanted to ask about turncost for main roads

assign turncost       if  ismainroad  then 90    else if ispaved then 60   else 30  

Maybe 90 is a bit high.
Often when I go down from a mainroad, I have better vision of what is happening, than when I have to turn in small residential sideroads, and don't have to slow down much more
The tecking-dry also produces something odd here
bildschirmfoto vom 2017-01-12 15-07-26

Just a discussion, I am not entirely certain it is a good idea, but I tried many routes and none got worse, and this problem above was fixed
Also the argument from experience above would tell me that turncost might be the same on small and large roads, in average

poutnikl commented 7 years ago

Turncost BRouter parameter addresses any physical turning, as an estimated loss of distance due slowing down. Turning on main roads are considered more costly, as you are supposed going faster, and the distance loss by breaking, going on lower speed and accelerating back is bigger.

Note also that BRouter default Trekking profile uses turncost 90 for all roads, ańd I decreased it for small roads. OTOH, it sets turncost=0 and costfactor=1.0 for cycleroutes, unless ignore cycleroutes is chosen. ( actual_turncost = turncost * ( 1 - cos(alfa)) .

Not sharp, round-like turning is mapped in OSM as a serie of small turns, so effect is like e.g. 4 turncost ( 1 - cos(22.5)), instead of turncost.

Trekking-dry selected smaller roads mainly because secondary is not liked by the profile, and it was avoiding the traffic light.

What I may reconsider is initialcost management, increasing initialcost for small ways. ( Default Trekking does not use initialcost for ways)

You may easily follow the evaluation of way segments on the data tab of BRouter web.

I admit the profile is focussed more on rural area trekking than on city biking. It is possible city would need different tuning. There is need of balanced costfactors and initialcosts.

poutnikl commented 7 years ago

When I changed at assign initialcost statement the ending else 0 to else 30, the result routing was the same as for the Trekking profile, e.g. along the main roads.

I may also reconsider the lowering cost of traffic lights.

poutnikl commented 7 years ago

V2.5.8-changed initial costs for minor ways

utack commented 7 years ago

Thanks!
I hope it leads to consistently better results

utack commented 7 years ago

Initialcost 10 also fixes the "many small ways" thing, but it does not influence some larger routes I tried

poutnikl commented 7 years ago

The point is, in the dense way network as there is in cities, there is needed a trade-off between

utack commented 7 years ago

Yes that is exactly what solves the problem I found, much better than messing with turncosts