Page 3 of 7

Re: Un-named segments for better navigation

PostPosted: Mon Apr 29, 2013 11:55 am
by CBenson
I think the different road types does cause the problem. Waze does not give a prompt for angle of deflection less than 45 degrees if the route continues on the best continuation. If one choice continues on the same road type and the other choice changes road type, then one that continues on the same road type should be the best continuation and will not receive a navigation instruction (assuming all the segment names are different).

Re: Un-named segments for better navigation

PostPosted: Wed Mar 20, 2013 1:58 pm
by dmcconachie
CBenson wrote:Lastly, I don't understand your caveat that TTS will not read out the name of the following segment, but it does display on the client.

I think this was specifically with regard to the double roundabouts!

Re: Un-named segments for better navigation

PostPosted: Wed Apr 17, 2013 8:42 pm
by foxitrot
For years long, while reading such threads, I'm constantly shaking my had :-| and have a gut feeling, that Waze must be perverse. They would like to obtain an excellent map from their editors, to be eventually able to sell it (or at least some services, based on it), yet they let these editors spend a serious portion of their life with reverse-engineering the routing server's behavior, just in order to draw a correct map, which lets the navigation devices behave like intended :roll:

I understand that this discovery process is funny and interesting at the same time, but not each and every potentially excellent, but often rather desperate editor, might share such feeling... yet years later the situation does not change a bit.

Waze, please, wake finally up!

Re: Un-named segments for better navigation

PostPosted: Fri Apr 12, 2013 9:28 am
by iainhouse
I presume this was moved out of the Champs forum? Can't think how I missed it otherwise.

It seems like a very good idea. My immediate thoughts turned to the junction of the clockwise M25 and A21, where it looks like you may have been testing this.

In the 9 months I have been editing, I have seen URs here almost weekly for "wrong directions". As a noob, I didn't dare touch a motorway junction! By the time I felt experienced enough, there were already several fingers in the pie, so I watched the discussions from the sidelines. Well, it looks like the un-named segment went in at the end of January and I certainly haven't seen a UR there for a long time, so it appears to work nicely.

It should certainly by up for wider discussion - not least because I now realise I may have recently "fixed" several un-named segments between roundabouts without realising they were there for a reason! :shock: :oops: :lol:

I would suggest that any deliberately un-named segment like this should be locked, to try & prevent casual idiots like me coming along & re-naming them.

I'm going to give it a try myself: there's a double-mini roundabout at Brockely Cross near me. I know it gives some long-winded & occasionally confusing instructions, plus it's not well aligned to the GPS & aerials. I'm off to Scotland for a couple of weeks, but I'll make sure I test-drive the rebuild version when I get back. :)

Re: Un-named segments for better navigation

PostPosted: Mon Apr 22, 2013 3:41 pm
by kentsmith9
Something I just thought about. When you create a short segment just before an exit, Waze will lose the ability to distinguish traffic jams that might be exiting one direction and not the other through the actual split beyond this short unnamed segment. All traffic data will be mixed in the unnamed segment and Waze will erroneously route traffic around this area when in fact the other direction is not blocked at all. Or Waze will see 90% of traffic going left at full speed and not notice the 10% of traffic completely stopped going right.

Re: Un-named segments for better navigation

PostPosted: Mon Apr 22, 2013 6:31 pm
by kentsmith9
SegmentTraffic1.png
(14.89 KiB) Downloaded 728 times

Waze separately computes the travel time of:
* Seg4 > Jnct2 > Seg5
* Seg4 > Jnct2 > Seg6

Therefore any traffic building up on Seg4 that turns right to Seg6 will not affect the route timing for the traffic also using Seg4, but instead turning left to Seg5.

The point I was making earlier is given this segment time tracking, if we add short segments like Seg8 between Seg7 and Jnct4, only Seg8 will know the ultimate destination of the traffic to the left (Seg9) or right (Seg10). So if the traffic backs up longer than the length of Seg8 and reaches Seg7, then the average transit time for Seg7 has only a single number from Seg7 > Jnct3 > Seg8. So the backup getting to Seg10 will cause the through traffic from Seg7 to show a decline in average speed and it may try to find an alternate route.

[Edit: this was added to the Wiki under "Segment exit times" and "Turn delays" so far]

Re: Un-named segments for better navigation

PostPosted: Mon Apr 22, 2013 8:30 pm
by kentsmith9
Ah. This is a great example of where the split road is actually reducing the ability of Waze to properly monitor and route traffic, and because of the split intersection it cannot do that.

This is a good case where a bowtie intersection would still enable a split road AND provide the isolated left turn data.

Re: Un-named segments for better navigation

PostPosted: Sun Apr 28, 2013 8:18 am
by kentsmith9
I thought we had a solution to this problem with what is in the Junction Style Guide. As I said earlier, this solution has the typical problem of masking and isolating the data that tells Waze routing what traffic is headed one direction or another.

Re: Un-named segments for better navigation

PostPosted: Mon Apr 29, 2013 2:33 pm
by kentsmith9
I thought we already came up with a solution to this problem clearly detailed in Junction Style Guide - Freeway Splits. I believe this method was completely tested and worked great for this situation. It covered routing, TTS, visual map appearance, and no traffic tracking issues.

What in this setup is not covered in that section?

ncc1701v wrote:segments don't know anything, only the client or route server (collectively "Waze") knows.

I believe AlanOfTheBerg told me from a prior conversation with Waze that the database tracks the average transit times from segment to segment so it can be applied to routing decisions. The problem with a short segment just before the Y-split is only the short segment will have two different average speeds exiting that segment. So when the traffic backs up longer than the short segment, the next segment with only one exit (to the short segment) will only have a single average speed from the two segments beyond it's reach. So yes the routing server knows where the next car is going, but it only has one average speed from the segment before the short one.

Re: Un-named segments for better navigation

PostPosted: Mon Apr 29, 2013 3:19 pm
by kentsmith9
For this particular area I agree with the proposal to simply make that part of the freeway (the two noted segments) a ramp. That will not have the problem of routing I mentioned earlier, e.g., there is no short unnamed segment inserted before the "Y". Yes there is a segment that is unnamed before the "Y", but that is the actual roadway and nothing special should be necessary. Because that segment is as long as possible back to the merge of the two prior ramps, all is as good as you can get it.

There was a prior conversation to insert a short segment before a "Y" to try to control naming that induced me to post my concern for routing. This is not that same "Y".