[Page Update] Detour Prevention Mechanisms

Moderator: Unholy

Re: [Page Update] Detour Prevention Mechanisms

Postby PesachZ » Mon Dec 28, 2015 4:22 pm

This concept is already mentioned on the page with an entire dedicated paragraph.
It is important to understand that Waze uses a penalty-based routing system. It computes the total penalties for a route and selects the one with the least cumulative penalty. Therefore it is possible for big detour routes to occur when the primary route is extremely slow.

Exactly how slow = extremely slow, is proprietary information the devs do not wish to publish, but the concept remains in heavy traffic it is possible to overcome the penalty, and the prevention behaviour.
PesachZ
Wiki Master
Wiki Master
 
Posts: 4512
Joined: Mon Jul 01, 2013 12:51 am
Location: NY, USA (also NJ sometimes) {GC}
Has thanked: 1998 times
Been thanked: 2373 times

Re: [Page Update] Detour Prevention Mechanisms

Postby PesachZ » Wed Jan 20, 2016 6:45 pm

After a recent response from staff which explains several anomalies discovered in routing which appeared to be inconsistent with our understanding of when BDP was applied, I propose to make some changes to the criteria section of the page.

The change basically removes the premise that only one segment missing a name anywhere along the possible detour path is ignored, and does not break name-continuity.

1)
Current wiki text:
Conversely an "uninterrupted name" is defined as a series of segments in middle of a route which do have an identical name on all the segments (with the exception of a single isolated segment among them).

Should be updated to say:
Conversely an "uninterrupted name" is defined as a series of segments in middle of a route which do have an identical name on all the segments.
NOTE: There is no single isolated segment exception.

2)
Current wiki text:
To be considered as a possible detour there must be a Freeway / Highway segment of the same 'Road Type Group', which share any name, both before and after a series of segments which contain a name-discontinuity, and are in a different 'Road Type Group'.

Should be updated to say:
To be considered as a possible detour there must be a Freeway / Highway segment of the same 'Road Type Group', which share any name, both before and after a series of segments (more than one) which contain a name-discontinuity, and are in a different 'Road Type Group'.


If there is no opposition to this change noted, I will make it shortly.
PesachZ
Wiki Master
Wiki Master
 
Posts: 4512
Joined: Mon Jul 01, 2013 12:51 am
Location: NY, USA (also NJ sometimes) {GC}
Has thanked: 1998 times
Been thanked: 2373 times

Re: [Page Update] Detour Prevention Mechanisms

Postby PesachZ » Wed Jan 20, 2016 7:18 pm

DwarfLord wrote:
PesachZ wrote:Should be updated to say:
To be considered as a possible detour there must be a Freeway / Highway segment of the same 'Road Type Group', which share any name, both before and after a series of segments (more than one) which contain a name-discontinuity, and are in a different 'Road Type Group'.

I'm not quite following the language...the second change (above) appears to contradict the first?

Specifically, the phrase "more than one [segment] which contain a name-discontinuity" in the second change appears to contradict the previous change indicating that even one segment's worth of name discontinuity qualifies as an "interrupted name".

Perhaps I missed something in the context of the remainder of the article...I didn't go back to review it.


There are actually two concepts here.
1) In order to be considered a possible detour there must be more than one segment long between the two segments of the same road type group. This means in a freeway example that there is at least 2 segments which are not freeway or major highway between the 2 freeway segment.
2) once assessing a possible detour for name discontinuity, even one segment missing the name will be enough to be considered a discontinuity.

Sent from Android using Tapatalk
PesachZ
Wiki Master
Wiki Master
 
Posts: 4512
Joined: Mon Jul 01, 2013 12:51 am
Location: NY, USA (also NJ sometimes) {GC}
Has thanked: 1998 times
Been thanked: 2373 times

Re: [Page Update] Detour Prevention Mechanisms

Postby PesachZ » Wed Jan 20, 2016 7:29 pm

CBenson wrote:What do the "name-discontinuity" and "uninterruped name" concepts add to understanding the BDP? Would this be correct?
https://wiki.waze.com/wiki/User:CBenson ... Mechanisms

With the old understanding it was necessary to explain the one segment exception not triggering discontinuity. At first glance reading your example draft it seems to apply and make sense, but I did not study it carefully. It is very likely that with this new understanding your draft is more simple and still accurate. I welcome anyone who would try to find whether or not there are discrepancies and thank you for your effort.

Sent from Android using Tapatalk
PesachZ
Wiki Master
Wiki Master
 
Posts: 4512
Joined: Mon Jul 01, 2013 12:51 am
Location: NY, USA (also NJ sometimes) {GC}
Has thanked: 1998 times
Been thanked: 2373 times

Re: [Page Update] Detour Prevention Mechanisms

Postby PesachZ » Wed Jan 20, 2016 7:56 pm

DwarfLord wrote:
CBenson wrote:Would this be correct?
https://wiki.waze.com/wiki/User:CBenson ... Mechanisms

If I am understanding the clarification, the following paragraph from the above link may make an incorrect implication:

CBenson's draft wiki article wrote:If the ramp does not also carry a simple alternate name of "I-1234", then there is a discontinuity of the highway name between the concurrency and the continuation highway. If there are two ramp segments like this, it will trigger Big Detour Prevention unexpectedly, and Waze will create an unneeded penalty for continuing on I-1234.

Based on the clarification, I think the discontinuity of the highway name for even as much as a single segment will trigger the unexpected penalty. The above paragraph leaves the strong impression that there need to be two segments with the name discontinuity to trigger the unwanted penalty.

This may be a big deal, actually, as we are not always able to set ramp alternate names to reflect concurrent highway names without unwanted side-effects on BC and thus navigation instructions.

That example is copied from the original page.
I think the reason that example mentions "another ramp like this", is because a single segment alone cannot trigger the BDP penalty. One of the criteria is the possible detour must be longer than one segment.

Sent from Android using Tapatalk
PesachZ
Wiki Master
Wiki Master
 
Posts: 4512
Joined: Mon Jul 01, 2013 12:51 am
Location: NY, USA (also NJ sometimes) {GC}
Has thanked: 1998 times
Been thanked: 2373 times

Re: [Page Update] Detour Prevention Mechanisms

Postby PesachZ » Wed Jan 20, 2016 9:35 pm

DwarfLord wrote:Still, it may be more helpful to reword to something like "If there are two or more ramp segments for the concurrent highway, but even one of the segments does not carry the simple highway name as either primary or alternate, Waze will impose an unexpected detour penalty for continuing on I-1234".

Provided I'm understanding the mechanism myself, that is.

The understanding is correct. Now try to make it fit nicely in the paragraph, so the overall paragraph reads well.

Sent from Android using Tapatalk
PesachZ
Wiki Master
Wiki Master
 
Posts: 4512
Joined: Mon Jul 01, 2013 12:51 am
Location: NY, USA (also NJ sometimes) {GC}
Has thanked: 1998 times
Been thanked: 2373 times

Re: [Page Update] Detour Prevention Mechanisms

Postby PesachZ » Wed Jan 20, 2016 11:46 pm

CBenson wrote:Regarding the Criteria section, the significant differences that I see between my draft and your proposals are:

1) I've eliminated the "to be considered as a possible detour" paragraph. Even if the routing algorithm operates with a two step process, I'm not sure it matters to us if as long as "detours" are a subset of "possible detours." My reasoning assumes they are. That is there are no "detours" that meet the criteria for "detours" but are not penalized because they weren't evaluated because they failed to qualify as a "possible detour."

2) The current wiki defines "name-discontinuity" as a series of segments that . . . . The possible detour statement refers to "a series of segment that contain a name-discontinuity." I've read this to mean that the possible detour can be comprised entirely of segments without the relevant street name or can include some segments that have the relevant street name as long as one is missing the relevant street name. This fits with my understanding of the BDP, so I'm pretty sure this is correct.

3) There has been some confusion regarding the road type limitation. The wiki states: "The possible detour is not composed of the same 'Road Type Group' as the continuation after the possible detour." The first question is does "not composed" mean does not include any segments of the same 'Road Type Group' or does it mean does not include all segments of the same 'Road Type Group.' My experience tells me this should mean "does not include all segments." In other words the detour must include at least one segment that is not in the same 'Road Type Group.' The second question is that the 'Road Type Group' limitation is also mentioned in the possible detour paragraph and seems to modify the series of that contains the "name-discontinuity" series. If this is intended to mean that a segment with the 'Road Type Group' discontinuity must also have the "name-discontinuity," then I do not believe I've included such a requirement in my draft. (It is usually the case, for example a ramp segment with the name discontinuity, but I don't know if its required or not).

4) The final difference that I see is that I have changed criteria 5 regarding the one segment rule to refer to the possible detour rather than the name continuity. I'm fairly certain that this is the basic change that needs to me made even if my other suggestions are not adopted.

For 3) to be penalized as a detour, the path must not be in the same road type group, and must not carry a continuous name.
In other words if the possible detour has a matching name, or a matching Road type, it will not be penalized. Ideally this should be applied to every segment of the possible detour, so that if even one segment of the possible detour does not match the road type and the name there is the potential for the penalty to be applied. However in practice today the process is not always applied exactly precisely due to server processing limitations. Therefore to be safe and future proof if you want to ensure a specific behaviour, then you should ensure compliance from all the segments of the possible detour.

Sent from Android using Tapatalk
PesachZ
Wiki Master
Wiki Master
 
Posts: 4512
Joined: Mon Jul 01, 2013 12:51 am
Location: NY, USA (also NJ sometimes) {GC}
Has thanked: 1998 times
Been thanked: 2373 times

Re: [Page Update] Detour Prevention Mechanisms

Postby PesachZ » Thu Jan 21, 2016 7:19 pm

edsonajj wrote:If I'm understanding the proposal correctly it seems to imply that if on a route that takes you out of a road and soon after it takes you back, name discontinuity on a single segment would cause detour prevention to kick in.

That would contradict what I see on this route since it takes you out of a road and back again with a name discontinuity in the middle.

Could someone tell me if I misunderstood the text of the proposal?

You are understanding correct in theory however in practice it is not 100% as I explained above.

PesachZ wrote: However in practice today the process is not always applied exactly precisely due to server processing limitations. Therefore to be safe and future proof if you want to ensure a specific behaviour, then you should ensure compliance from all the segments of the possible detour.



Sent from Android using Tapatalk
PesachZ
Wiki Master
Wiki Master
 
Posts: 4512
Joined: Mon Jul 01, 2013 12:51 am
Location: NY, USA (also NJ sometimes) {GC}
Has thanked: 1998 times
Been thanked: 2373 times

Re: [Page Update] Detour Prevention Mechanisms

Postby PesachZ » Thu Jan 21, 2016 8:18 pm

CBenson wrote:I don't know. I don't think its good practice to name cross streets between a divided highway with the name of the highway in order ensure compliance with a shifting BDP implementation.

We wouldn't have to since cross streets between the divided highway are only one segment long and therefore wouldn't meet penalty criteria as a possible detour regardless of name or type.

Edit: under your criteria section you could clarify this to make it stand out more.
"possible detour" is a series of segments (more than one) in the middle of a route which meet the minimum criteria to trigger an evaluation.

Sent from Android using Tapatalk
PesachZ
Wiki Master
Wiki Master
 
Posts: 4512
Joined: Mon Jul 01, 2013 12:51 am
Location: NY, USA (also NJ sometimes) {GC}
Has thanked: 1998 times
Been thanked: 2373 times

Re: [Page Update] Detour Prevention Mechanisms

Postby PesachZ » Thu Jan 21, 2016 8:54 pm

edsonajj wrote:But what if it is, like on my example, a crossing with an overpass? According to the article I should put name A on the connecting ramps and in one segment of road B.
Mind you, that would very probably suppress the instruction to exit right on the ramp.

In you example yes the overpass segment should also have the alternate name of "Av. Paseo Constituyentes".
In the USA this would not have been an issue at all, since our guidance here is to always add a cardinal direction to end of a divided freeway name. Such that in the USA the two directions of this fwy would have been named "Av. Paseo Constituyentes N" and "Av. Paseo Constituyentes S". One of the reasons we do that is precisely for such U-turns not to trigger BDP.

You don't need to worry that adding the alt name to the overpass PS will eliminate the instruction to the ramp because of a BC name match, since continuing straight on the PS is also a name match. Therefore the BC algorithm will check for a type match; continuing straight on the PS will be BC, taking the ramp will get an instruction even with the alt name.
PesachZ
Wiki Master
Wiki Master
 
Posts: 4512
Joined: Mon Jul 01, 2013 12:51 am
Location: NY, USA (also NJ sometimes) {GC}
Has thanked: 1998 times
Been thanked: 2373 times

PreviousNext

Return to Wiki Updates and Discussion

Who is online

Users browsing this forum: No registered users