Page 5 of 5

Re: Segment Lengths

PostPosted: Tue Oct 21, 2014 4:14 am
by doctorkb
jasonh300 wrote:So refresh my poor memory of the metric system. 100 km is 100,000 meters, right?


Yes. Makes sense, right? Kilo for thousand, + meter equals one thousand meters is one 1 km. :)

Re: Segment Lengths

PostPosted: Tue Oct 21, 2014 4:16 am
by doctorkb
banished wrote:Although I have a good system, it starts to drag if the segment length is over 70km. I don't know if that's my system or the WME, but you may wish to run your own tests in out-of-the way areas so we can arrive at a number that works for everyone.

Segment length was initially a railroad segment discussion. There aren't too many navigable segments that would reach that length (70km), but there may be a few in the Western US/Canada.


I've driven a lot of the highways around Western Canada and can tell you that, at very least, there is a forest service road that should be junctioned into a highway AT LEAST every 50 km.

I don't think 50km is unreasonable to specify as a maximum length.

Re: Segment Lengths

PostPosted: Wed Oct 22, 2014 6:37 am
by kentsmith9
The other reason we proposed shorter segment lengths was to better indicate where traffic slowdowns were on a long segment. If you are 50km long and at the end of that segment there is a stoppage, you won't see it because the average of 500m at 0 kph will not impact the 50km segment with 80kph average.

The net result for routing is no different though. The routing engine will still account for the time change over the entire length of the segment the same as it would for ten 5km segments with the stoppage at the same point. Since there are no turns off that road for the 50km, its not like you could get around the slowdown.

However, the net result for the visual will be different. The 50km segment will possibly never show slow traffic when you plot the route and look at the initial traffic. Only later will you run into the stopped traffic. With short segments, Waze can warn the driver of a slowdown of traffic because the segments are shorter and will be detected as different from their average.

Re: Segment Lengths

PostPosted: Wed Oct 22, 2014 3:11 pm
by bgodette
kentsmith9 wrote:The net result for routing is no different though. The routing engine will still account for the time change over the entire length of the segment the same as it would for ten 5km segments with the stoppage at the same point.
Routes with fewer segments are preferred by routing over a route with more segments even if the route with more segments seems to save some unspecified amount of time. Some percentage of the secondary route being a faster route is caused by this.

kentsmith9 wrote:However, the net result for the visual will be different. The 50km segment will possibly never show slow traffic when you plot the route and look at the initial traffic. Only later will you run into the stopped traffic. With short segments, Waze can warn the driver of a slowdown of traffic because the segments are shorter and will be detected as different from their average.
FYI we've seen occasional peeks at traffic highlighting in the app showing sub-segment highlighting. Using highlighting as a thin justification for segment splitting is likely to be eliminated "soon".

Re: Segment Lengths

PostPosted: Wed Oct 22, 2014 10:17 pm
by kpouer
@bgodette, that's good news. During Barcelona meetup we were told that they were able to detect traffic on a sub-segment but not highlight a sub-segment yet. If they can now it's a good evolution, maybe we should ask if it is available on all servers.

Re: Segment Lengths

PostPosted: Thu Oct 23, 2014 2:49 pm
by kentsmith9
bgodette wrote:
kentsmith9 wrote:The net result for routing is no different though. The routing engine will still account for the time change over the entire length of the segment the same as it would for ten 5km segments with the stoppage at the same point.
Routes with fewer segments are preferred by routing over a route with more segments even if the route with more segments seems to save some unspecified amount of time. Some percentage of the secondary route being a faster route is caused by this.

I thought we recently discussed this (over the last year and in Vegas) and confirmed that the junction itself only added time when the segments were without traffic data. Then once it had traffic data there was no time added for the junction if data existed.

The logic for adding time because of a junction would be ludicrous if you considered a major roadway through a city like El Camino Real through the SF Peninsula with hundreds of parking lot connections. In the real world, each junction to the parking lot has no affect on the straight through traffic beyond the measured average time to cross that entire segment. Waze repeatedly tells us they want to map everything eventually. They would not have such routing logic if that was the intent.

Re: Segment Lengths

PostPosted: Thu Oct 23, 2014 3:07 pm
by kentsmith9
And I just saw this post from Alan discussing the junctions with RR crossings [underline added by me].
AlanOfTheBerg wrote:Junctions only impact routing until segment speed data is captured, then the junction "penalty" goes away. Waze doesn't need to have a project to identify RR crossings, because with a junction, the Waze speed capture does the rest automatically. Long delays at that junction will be collected and aggregated by the 10min/30min timeslice with all other stop light and stop sign data.