User:RonRover View history

m (Acknowledge input from other editors)
Line 44: Line 44:


==== BIG DETOUR PREVENTION (BDP) Explained by Bill473 ====
==== BIG DETOUR PREVENTION (BDP) Explained by Bill473 ====
''The following was presented by Bill473 as a training aid to satisfy the BDP requirement to attain R4 rank for both of us.  Bill has received useful feedback from PesachZ and TerryPurdue but remains responsible for any errors or omissions.''
''The following was presented by Bill473 as a training aid to satisfy the BDP requirement to attain R4 rank for both of us.  Bill has received useful feedback from PesachZ and TerryPurdue but remains responsible for any errors or omissions.  If you have suggestions to make this better, please contact either Ron or Bill.''


This is an introduction to how Waze does detour prevention, how it can go wrong and how to fix it.
This is an introduction to how Waze does detour prevention, how it can go wrong and how to fix it.
Line 55: Line 55:


A very useful user script is [http://www.waze.com/forum/viewtopic.php?f=819&t=173983 WME Show Alt Names].  It shows the names of all selected segments in a table on the edit page.  And it can be turned off and on as needed.  You can use it to select the segments for a LiveMap route or you can select them yourself.
A very useful user script is [http://www.waze.com/forum/viewtopic.php?f=819&t=173983 WME Show Alt Names].  It shows the names of all selected segments in a table on the edit page.  And it can be turned off and on as needed.  You can use it to select the segments for a LiveMap route or you can select them yourself.


I list this now in case you don't have it and want to install it while I continue. I don't have an "exercise" planned where you need it, but it will make diagnosing BDP errors easier in the future.
I list this now in case you don't have it and want to install it while I continue. I don't have an "exercise" planned where you need it, but it will make diagnosing BDP errors easier in the future.
Line 95: Line 94:
#:(The grouping is Freeway/MH in one group, mH in another)
#:(The grouping is Freeway/MH in one group, mH in another)
#:Yes: US-127 N is Freeway, US-36 E is MH so same group. Our ramp "detour" isn't in that group.
#:Yes: US-127 N is Freeway, US-36 E is MH so same group. Our ramp "detour" isn't in that group.
# '''Existence Of Direct Route''' - The idea is there needs to be a direct route before labeling a different route as a "detour" to be avoided. ''I'm having a little trouble with the exact wording because I think it is slighly misstated. (Both the detour and the "direct route" share the first ramp segment.) Can we say "yes, there is a direct route, Waze is showing it in the picture?''
# '''Existence Of Direct Route''' - The idea is there needs to be a direct route before labeling a different route as a "detour" to be avoided.
#: No, the " direct route" doesn't need to be "direct", it just needs be a different route which does not use any segments used by the "detour route" being evaluated. The "direct route" here is not the route wazs is suggesting, rather a different more circuitous option.
 
#: The " direct route" doesn't need to be "direct", it just needs be a different route which does not use any segments used by the "detour route" being evaluated. The "direct route" here is not the route Waze is suggesting, rather a different more circuitous option.
#: The second portion of the "direct route" is that there is another way to approach the segment after the potential detour from a matching road type. For the potential detour here, the alternate approach is coming EB on the Fwy which becomes US-36 E. (This is also the reason the route suggested above is not a detour, the only alternative path to the post-detour segment would be via the ramps, which are not in the same road type group. So no alternate pathway to the post detour segment, means that this route can not be considered a detour.)
#: The second portion of the "direct route" is that there is another way to approach the segment after the potential detour from a matching road type. For the potential detour here, the alternate approach is coming EB on the Fwy which becomes US-36 E. (This is also the reason the route suggested above is not a detour, the only alternative path to the post-detour segment would be via the ramps, which are not in the same road type group. So no alternate pathway to the post detour segment, means that this route can not be considered a detour.)
# '''Minimum Length''' - The possible detour must be more than one segment long.
# '''Minimum Length''' - The possible detour must be more than one segment long.
Line 112: Line 112:
There are, of course, other configurations that will result in BDP errors. But every BDP error I've seen fits the "simple" summary. (Check "Confirmation bias" on the Internet)
There are, of course, other configurations that will result in BDP errors. But every BDP error I've seen fits the "simple" summary. (Check "Confirmation bias" on the Internet)


So what's the fix? Let Waze know that this ramp is part of US-36 E. {{Red|You can't touch the first ramp segment since you need the name discontinuity to get the "to US-36 / SR-571 / Greenville / Piqua" instruction. But you can add US-36 E as an alternate name to the second ramp segment.}}
So what's the fix? Let Waze know that this ramp is part of US-36 E. Adding US-36 E as an alternate name to the second ramp segment was sufficient to fix the problem. A more robust solution is to add US-36 E to all ramp segments that carry that route.
 
''You can add the alt name  to both ramp segments, the fact that they are a different type (ramp) that the preceding segment (Fwy) is what dictates that a prompt will be given.''


At one point we believed you could have a one-segment name discontinuity anywhere along a series of ramp segments. We discovered Nov 2015 that the only safe place to have a discontinuity was the first segment (to support the wayfinder function.)
At one point we believed you could have a one-segment name discontinuity anywhere along a series of ramp segments. We discovered Nov 2015 that the only safe place to have a discontinuity was the first segment.  However, unless there is a reason to leave the name off the first segment, it is safer to add it.


<hr>
<hr>
Line 127: Line 125:
And here's the [https://www.waze.com/livemap?lon=-93.5782&lat=41.65459&zoom=15 LiveMap link] to the area so you can see where the segments are.
And here's the [https://www.waze.com/livemap?lon=-93.5782&lat=41.65459&zoom=15 LiveMap link] to the area so you can see where the segments are.


I-80 (E/W) and I-35 (N/S) run concurrently for a while near the center of Des Moines. Where I-35 N leaves I-80 E and heads north was connected by a single ramp segment. Everything was OK at that point because of #4 above ("detours" must be 2 or more segments). It was then cut to Seagull it. Now we have 3 segments and I-35 N was not added as an alternate to the new ramp segments. Within a few days there were numerous URs from commuters wondering why they were no longer being sent the normal way.
I-80 (E/W) and I-35 (N/S) run concurrently for a while near the center of Des Moines. Where I-35 N leaves I-80 E and heads north was connected by a single ramp segment. Everything was OK at that point because of #4 above ("detours" must be 2 or more segments). It was then cut to Seagull it. Now we have 3 segments and I-35 N was not added as an alternate to the new ramp segments. Within a few days there were numerous URs from commuters wondering why they were no longer being sent the normal way.   
 
<i>(Slight update: the current implementation of BDP is that to make sure something is not a detour is to have the name carried through on the last ramp segmentYou can gap any segments before the last.
Probably best to carry names through all segments as that implementation is likely to change)</i>


My feeling is the checking of the last segment only is to make the process faster.  But you certainly can't expect that to always be the case.
Although currently only the last of the three segments is checked for BDP (probably to make checking faster), PesachZ suggests that is likely to change. The best fix is to have I-35 N on allIndeed, if that single ramp segment had I-35 N added earlier, even though not required to prevent a BDP error (since 2 ramp segments required for BDP here), when it was cut into 3 segments no BDP error would have occurred.


The Des Moines error was fixed by adding I-35 N as an alternate to the ramp segments.
The Des Moines error was fixed by adding I-35 N as an alternate to the ramp segments.


It's very easy to cause a BDP problem in a case like this without realizing it. Think "name discontinuity" any time you are working on ramps. As more Seagulling in being done, it's likely other such errors will occur. Again, the most dangerous areas are where two major highways diverge.
It's very easy to cause a BDP problem in a case like this without realizing it. Think "name discontinuity" any time you are working on ramps. As more Seagulling in being done, it's likely other such errors will occur. Again, the most dangerous areas are where two major highways diverge.


That one was pretty dramatic. The tendency is to think "temporary traffic issue" when the URs start coming in. Especially since not every LiveMap route test will show an error. (If you are too close the "detour", there may not be an alternate direct route so LiveMap won't show it.)
That one was pretty dramatic. The tendency is to think "temporary traffic issue" when the URs start coming in. Especially since not every LiveMap route test will show an error. (If you are too close the "detour", there may not be an alternate direct route so LiveMap won't show it.)
Line 160: Line 155:
This is discussed much less because there doesn't seem to be the same type of associated errors. Basically this is meant to prevent doing a U-turn followed by a right turn instead of the more obvious left turn. From the Wiki "While this may save a few seconds over waiting for a long average left turn, it is undesirable. Waze will prevent such detours if there is not a measurable difference in the route times. The exact difference in time required to trigger this prevention is proprietary, and subject to change as needed. "
This is discussed much less because there doesn't seem to be the same type of associated errors. Basically this is meant to prevent doing a U-turn followed by a right turn instead of the more obvious left turn. From the Wiki "While this may save a few seconds over waiting for a long average left turn, it is undesirable. Waze will prevent such detours if there is not a measurable difference in the route times. The exact difference in time required to trigger this prevention is proprietary, and subject to change as needed. "


I do wonder if this is an example where the prevention isn't working:
Here is an example where Small Detour Prevention doesn't seen to be working.


http://i82.photobucket.com/albums/j242/ronbien/BDP-4.png  
http://i82.photobucket.com/albums/j242/ronbien/BDP-4.png  

Revision as of 17:55, 1 May 2016

SAMPLE LETTERS

Sample Editor Outreach Message:

Hi,
I see you are editing around the ________ area. Thanks for helping. I would like to welcome you to editing and offer you to join other editors in Ohio. We communicate in Google Hang Outs (GHO)s and share tips, news and tools that make editing and asking for down-locks easier.
In the meantime please read these two pages:
https://wiki.waze.com/wiki/Waze_Map_Editor/Welcome https://wiki.waze.com/wiki/Edits_to_avoid_(USA)
(You will have to add the closing parenthesis the the above link)

If you would like to pursue joining with us, just read this and request to be added.
https://www.waze.com/forum/viewtopic.php?f=261&t=133427

Hope to see you editing,

Your Name/Editor Name


LINKS TO CITIES AND VILLAGES PDF IN OHIO

http://www.omlohio.org/Municipal%20Resources/listofmunicipalities.pdf


WME HELPFUL LINKS FOR URs

How to remove locations from the Waze App
https://goo.gl/WbYjyQ

How to set or change Home and Work
https://goo.gl/rTbTVx

How to collect logs from the App. (see next item also)
https://goo.gl/1G3Wx3

Form to fill out for Waze reporting to accompany the logs. (see above item also)
https://goo.gl/967r1F

Waze Help Center
https://goo.gl/svEPRF

Waze Bug Report Form
https://goo.gl/967r1F

BIG DETOUR PREVENTION (BDP) Explained by Bill473

The following was presented by Bill473 as a training aid to satisfy the BDP requirement to attain R4 rank for both of us. Bill has received useful feedback from PesachZ and TerryPurdue but remains responsible for any errors or omissions. If you have suggestions to make this better, please contact either Ron or Bill.

This is an introduction to how Waze does detour prevention, how it can go wrong and how to fix it.

First, detour prevention is a good thing. You don't want Waze to recommend getting off an interstate, following some ramps and getting back on even if it saves a few minutes. That is an unneeded detour and Waze has mechanisms to prevent it.

When Waze editors discuss BDP, what they are focusing on is Big Detour Prevention errors – where the attempt to prevent detours goes wrong and actually takes you on a detour.

When checking BDP errors, you need to look at primary and alternate names on segments.

A very useful user script is WME Show Alt Names. It shows the names of all selected segments in a table on the edit page. And it can be turned off and on as needed. You can use it to select the segments for a LiveMap route or you can select them yourself.

I list this now in case you don't have it and want to install it while I continue. I don't have an "exercise" planned where you need it, but it will make diagnosing BDP errors easier in the future.

Here's a real-life example of a BDP error (fixed now, of course):


http://i82.photobucket.com/albums/j242/ronbien/BDP-1.jpg


And here's the LiveMap link to the area so you can see where the segments are.


The "obvious" way to get to the destination is possible. Going NB, take the exit and drive down to the bow-tie on US-36 W. Turn left and drive to the destination.

So why didn't Waze recommend that? Because it thought that would be a detour!

Let me try to explain the "detour" before getting into all the details. US-127 N has an alternate name of US-36 E (because both highways share the road before the exit.) Our destination is on US-36 E. Once we leave the freeway, we are taking 2 ramp segments to get to US-36 E. Neither (at the time) have the (primary or alternate) name US-36 E.

Visually we can see that path makes sense, but Waze sees this as leaving US-36 E, driving on two ramp segments and getting back on US-36 E. "Clearly" a detour to be avoided. (The Waze recommendation shown in the picture goes to US-36 W which is not considered a detour due to more advanced criteria discussed later.)

(As an aside, the old guidance to disable the "straight-through" at the end of interstate ramps was to prevent the type of detour where you leave a highway, travel a few ramps and return to the highway. However, BDP (working properly) should keep that from happening. For this reason that old guidance is old, and no longer recommended.)

  • If we are OK with my description here, we can take a look at the Wiki to see that the exact definition is met.

The global Wiki entry on Detour Prevention is Detour Prevention Mechanisms, however is it rather vague, and partially inaccurate leading to confusion.

A more detailed version written for the US is Detour Prevention Mechanisms/USA

This latter link has five criteria that must ALL be met to be considered a detour. Let's check to make sure my summary above is right:

  1. Road Name Discontinuity - The freeway/highway segments immediately before and immediately after the possible detour must share at least one street name
    Yes: US-36 E on both. (Primary or alternate, makes no difference)
  2. Road Type Discontinuity - The freeway/highway segments immediately before and immediately after the possible detour must be from the same 'Road Type Group' as defined in the chart below. The possible detour must include at least one segment not in that same 'Road Type Group.'
    (The grouping is Freeway/MH in one group, mH in another)
    Yes: US-127 N is Freeway, US-36 E is MH so same group. Our ramp "detour" isn't in that group.
  3. Existence Of Direct Route - The idea is there needs to be a direct route before labeling a different route as a "detour" to be avoided.
  1. The " direct route" doesn't need to be "direct", it just needs be a different route which does not use any segments used by the "detour route" being evaluated. The "direct route" here is not the route Waze is suggesting, rather a different more circuitous option.
    The second portion of the "direct route" is that there is another way to approach the segment after the potential detour from a matching road type. For the potential detour here, the alternate approach is coming EB on the Fwy which becomes US-36 E. (This is also the reason the route suggested above is not a detour, the only alternative path to the post-detour segment would be via the ramps, which are not in the same road type group. So no alternate pathway to the post detour segment, means that this route can not be considered a detour.)
  2. Minimum Length - The possible detour must be more than one segment long.
    Yes: 2 (ramp) segments.
  3. Maximum Length - The possible detour must be shorter than the threshold length
    Yes: Threshold for this group is 5km.

Whew! Let's take a breath. The complexity of the Waze routing algorithm and the difficulty in putting it into words is why (I think) some view BDP as very hard to understand.

Here is the basic idea that I think makes this "simple".

You have 2 highways (A and B) running together for some distance. The highways then diverge with B exiting on ramps. If the ramp segments aren't labeled to indicate they are part of highway B, Waze will see that as a detour and "penalize" the way you should be driving...which results in Waze recommending a strange detour.

There are, of course, other configurations that will result in BDP errors. But every BDP error I've seen fits the "simple" summary. (Check "Confirmation bias" on the Internet)

So what's the fix? Let Waze know that this ramp is part of US-36 E. Adding US-36 E as an alternate name to the second ramp segment was sufficient to fix the problem. A more robust solution is to add US-36 E to all ramp segments that carry that route.

At one point we believed you could have a one-segment name discontinuity anywhere along a series of ramp segments. We discovered Nov 2015 that the only safe place to have a discontinuity was the first segment. However, unless there is a reason to leave the name off the first segment, it is safer to add it.


Des Moines Iowa

http://i82.photobucket.com/albums/j242/ronbien/BDP-2%20DesMoines.png


And here's the LiveMap link to the area so you can see where the segments are.

I-80 (E/W) and I-35 (N/S) run concurrently for a while near the center of Des Moines. Where I-35 N leaves I-80 E and heads north was connected by a single ramp segment. Everything was OK at that point because of #4 above ("detours" must be 2 or more segments). It was then cut to Seagull it. Now we have 3 segments and I-35 N was not added as an alternate to the new ramp segments. Within a few days there were numerous URs from commuters wondering why they were no longer being sent the normal way.

Although currently only the last of the three segments is checked for BDP (probably to make checking faster), PesachZ suggests that is likely to change. The best fix is to have I-35 N on all. Indeed, if that single ramp segment had I-35 N added earlier, even though not required to prevent a BDP error (since 2 ramp segments required for BDP here), when it was cut into 3 segments no BDP error would have occurred.

The Des Moines error was fixed by adding I-35 N as an alternate to the ramp segments.

It's very easy to cause a BDP problem in a case like this without realizing it. Think "name discontinuity" any time you are working on ramps. As more Seagulling in being done, it's likely other such errors will occur. Again, the most dangerous areas are where two major highways diverge.

That one was pretty dramatic. The tendency is to think "temporary traffic issue" when the URs start coming in. Especially since not every LiveMap route test will show an error. (If you are too close the "detour", there may not be an alternate direct route so LiveMap won't show it.)

Are we OK with the Des Moines example and fix?

Here's an example that show this is really Big Detour Penalty rather than Big Detour Prevention:


http://i82.photobucket.com/albums/j242/ronbien/BDP-3.jpg


And here's the LiveMap link to the area so you can see where the segments are.

The obvious path is still listed, but listed second so it won't be chosen by default. (Again, the problem is caused by 2 ramp segments, the second one didn't have SR-5 for name continuity.) The recommended route is 6 minutes, the second (penalized) one is 3 minutes. But the penalty (not publicly revealed. changeable without notice and possibly different for different configurations) is enough to make it listed second. Also, depending on traffic conditions for the first choice, the second route with the BDP error could actually be listed first so BDP errors could (theoretically) come and go.

(This is also a "favorite" of mine because I lived it. I took the 6 minute drive since I trusted Waze that there was a problem at the SR-82/SR-5 intersection. The problem was BDP.)

I have other examples of BDP errors I've seen in the last year, but they are all the same flavor. When you see a strange detour, after checking Shift-Z to make sure there aren't invalid turn restrictions, take a look at name discontinuities.

If we're OK with BDP, there is also there is also Small Detour Prevention.

This is discussed much less because there doesn't seem to be the same type of associated errors. Basically this is meant to prevent doing a U-turn followed by a right turn instead of the more obvious left turn. From the Wiki "While this may save a few seconds over waiting for a long average left turn, it is undesirable. Waze will prevent such detours if there is not a measurable difference in the route times. The exact difference in time required to trigger this prevention is proprietary, and subject to change as needed. "

Here is an example where Small Detour Prevention doesn't seen to be working.

http://i82.photobucket.com/albums/j242/ronbien/BDP-4.png

It seems that at certain times of the day, the obvious left-turn takes long enough for Waze to recommend continuing and doing a U-turn.

And here's the LiveMap link to the area.