Routing ignores all turn restrictions and road closures

Moderator: Unholy

Re: Routing ignores all turn restrictions and road closures

Postby waynemcdougall » Thu Oct 28, 2010 3:18 pm

Even though @Robin1979's statement is justification of his position is wrong, he has relocked the roads. So the test is off.

@Deego has confirmed he just changed the ramp name and didn't lock the road. So that would mean that, as I thought, the ramp was locked and Waze still gave wrong routing on the Livemap.

But it is working now. So either the map was corrupt in some way (and editing fixed it, or Waze HQ fixed it), or the routing was broken (and Waze HQ fixed it).
Waze User Manuals, Map Editing guide and support at http://www.waze.com/wiki
Wayne McDougall
Make Friends with me on Facebook and we'll compete against each other on the Scoreboard
http://www.facebook.com/wayne.w.mcdougall
waynemcdougall
 
Posts: 517
Joined: Mon Mar 08, 2010 3:59 am
Has thanked: 0 time
Been thanked: 2 times

Re: Routing ignores all turn restrictions and road closures

Postby waynemcdougall » Thu Oct 28, 2010 9:43 pm

robin1979 wrote:There's actually logic in it:
It routes you the opposite of one-way roads that not have been locked, ignoring turn restrictions and all. It does not seem to do this for locked roads.

I can reproduce this now. The problem causing this to 'suddenly' pop up, is that over the last week a lot of freeway junctions have been edited, and they all lost their 'locked' status (either on purpose or due to a bug).


Can you please clarify this report. You say "I can reproduce this now." Was that on the client or on the Livemap or both? On the client did you have road munching preference turned on or not? Did you try it both ways?

I look forward to your reply.
Waze User Manuals, Map Editing guide and support at http://www.waze.com/wiki
Wayne McDougall
Make Friends with me on Facebook and we'll compete against each other on the Scoreboard
http://www.facebook.com/wayne.w.mcdougall
waynemcdougall
 
Posts: 517
Joined: Mon Mar 08, 2010 3:59 am
Has thanked: 0 time
Been thanked: 2 times

Re: Routing ignores all turn restrictions and road closures

Postby waynemcdougall » Fri Oct 29, 2010 8:48 am

The drawing is awesome.

But I fear you're going to hate me when I say this: No *more* questions.

Just the same ones.

1. Was this on the client device, the Live Map or both (and if both, did you get the same results)?
2. If on the client device, was this with road munching preferred turned on, turned off, or did you try both (and if both, did you get the same results)?

I'm sorry if my lack of Dutch is making things too hard.
Waze User Manuals, Map Editing guide and support at http://www.waze.com/wiki
Wayne McDougall
Make Friends with me on Facebook and we'll compete against each other on the Scoreboard
http://www.facebook.com/wayne.w.mcdougall
waynemcdougall
 
Posts: 517
Joined: Mon Mar 08, 2010 3:59 am
Has thanked: 0 time
Been thanked: 2 times

Re: Routing ignores all turn restrictions and road closures

Postby waynemcdougall » Fri Oct 29, 2010 9:43 am

robin1979 wrote:
waynemcdougall wrote:I just wanted to help BrutusNL out. But a routing issue, and how we should edit maps, affects everyone in the world. We have a perfect model example - the closest we can get to a controlled trial here.


I did that test and posted what I have found out. I've have tested this in a lot of different places over the last few days. If you for some reason don't believe me (I'm still not sure what you are trying to proof),


I'm replying to this post, with thanks to @Robin1979 - I hope you can appreciated why I would not see "I did that test and posted what I have found out" previously. It wasn't (to me) at all clear what you had done. So you have done the same tests I would like to have done. Thank you.

robin1979 wrote:please go ahead and mess up the kiwi maps.


Unfortunately we just don't have places where two connecting roads can take 50km to go from one to the other.

robin1979 wrote:We have a large userbase here, and if I can prevent a routing error one way or the other, I will do it. If locking the damn road is the solution, I will go ahead and lock all damn roads.


Here's the thing. I don't think the solution is good enough. The problem is that locked roads do not stay locked. Waze server will unlock them automatically if it thinks there is something wrong. And since ramps are close to freeways with multidirectional traffic, I think they're highly subject to this unlocking. You'll end up playing whack-a-mole - do you keep on checking them , or wait for another BrutusNL to make a report. I really don't think that's acceptable.

Worse, we've been told that locking only makes it less likely to go down that corner. There may be an even more perverse case where locking will not help. Now it may be that the weighting is such that this is massively unlikely. Or it may just be rare. I'm not happy that that may be the case.

It seems clear to me that Waze is taking this route because it is a short cut. We can imagine the algorithm says

I can go left or right.
Right is a restricted turn -1000 penalty
Left is 18km so that's -1800 penalty
Bang - illegal routing, versus

I can go left or right.
Right is a restricted turn *from a locked segment to a locked segment -2000 penalty
Left is 18km so that's -1800 penalty
Bang - legal routing

until the case of

I can go left or right.
Right is a restricted turn *from a locked segment to a locked segment -2000 penalty
Left is 21km so that's -2100 penalty
Bang - illegal routing again.

It all depends on the weighting. I think the weighting must be high, but is it high enough. But still we have locking being unstable.

One can imagine that road munching does something like

I can go left or right.
Right is a restricted turn -1000 penalty
Left is 9km so that's -900 penalty
But Right is the least frequently travelled road by a factor of 5, so 500 bonus taking it to -500
Bang - illegal routing

It is noble of you to make the solution that you can. But IMO the routing weighting is flawed.
It's not like it is a feature. It's not making you explore, or test options. It only kicks in when the one-way alternative is so huge that it overflows the weight. The forums are full of people who complain waze won't route *at all* because it won't go through the default restricted turns on a base map. So it really is an underestimate of the tortuous nature of the freeways in the Netherlands (that is intended as a compliment because you have to drive them).

So that's my argument. I'd like to make the case to Waze HQ that restricted turns need to be given much higher weighting - high enough to force, what, 60 km? a nice round 100km? detour - basically it should be the computer equivalent of impossible - let's not quibble - so weighing against a restricted turn, whatever the source, so that even with a 60km detour, and both road segments unlocked and road munching turned on, that it STILL won't route through a restricted turn.

Now this is NEVER going to affect my country. I believe it will affect your large userbase, that you care about, and lead. And I value your opinion, and I'd like your support.

Is there anything wrong with my reasoning? Anything you disagree with? Can you support the proposition?

Note that I am NOT saying you should not lock your segments. Nor am I saying it should be a high priority for the Waze team although I don't think it could be that hard - it ought to be changing a few constants. Should be easier than making a special exception for ramps in the Netherlands.

But I do think it is a bug that needs to be fixed. I agree with
BrutusNL wrote:I'd prefer if:
- it doesn't encourage illegal behavior, so no instructions for illegal turns, no matter what.


But you're a schmott guy @Robin1979 - I'd really like your comments. You don't have to join my crusade, but if you have any opposition I'd like to hear it and see if it can be addressed. You've done the great diagnosing work I wanted. I think the patient needs to be treated, and you're only allowed to dispense sticking plasters. Waze HQ has the penicillin.

Sorry for going on.
Waze User Manuals, Map Editing guide and support at http://www.waze.com/wiki
Wayne McDougall
Make Friends with me on Facebook and we'll compete against each other on the Scoreboard
http://www.facebook.com/wayne.w.mcdougall
waynemcdougall
 
Posts: 517
Joined: Mon Mar 08, 2010 3:59 am
Has thanked: 0 time
Been thanked: 2 times

Re: Routing ignores all turn restrictions and road closures

Postby waynemcdougall » Sun Oct 31, 2010 8:57 am

Atlas85 wrote:"From my experience (and trust me, it's a lot): Waze will never send you down the wrong direction of a 'locked' road or restricted turn on a node attached to a road. It will abort with an error that it can't find a suitable road (in the client, on the live map it fails silently)."

Indeed - you remember the A50-case with the wrong directionality. If Waze would work with a penalty that is not very high, it would definitely have "ordered" me to go just through it. Instead, it planned me around the error: through the A2 (15km detour) or even through cities (Wijchen, Uden, Veghel), delaying the total route for more than 30 minutes. So it does seem to me that locking the segment disables routing over it all together.


Nope, we know for a fact, as Waze HQ has told us, that locking just makes it less likely to route against a turn restriction, and to route against directionality. So conclusions based on what it seems to you are invalid.

Atlas85 wrote:IF Waze would unlock segments automatically, as assumed by Wayne,


Nope, wrong again. If I had assumed something, I would have said so. If I was to make an assumption, I would have assumed that a lock would stay locked. But again Waze HQ have said that is not the case. I was very surprised to read that. Mind you, I was equally surprised that Waze would ever route against a turn direction or directionality. IMO it is a stupid decision.

Atlas85 wrote: it would have done there, because something was clearly wrong. It doesn't. Locking is for disabling Waze-interference. If locking is being made undone by Waze, there is no point in locking anything.


I agree. Apart from delaying a problem. So there is something we can agree upon. Locking is a temporary fix. There needs to be a better solution. Agreed?

Atlas85 wrote:And I haven't seen any locked segment being unlocked automatically. And you know I'm a regular user ;-)


I have.

And I think the very first road in this thread that I gave the permalink for in the report to support was unlocked too. @Deego is a meticulous careful user. He locks every road as a matter of habit. The road had not been edited by anyone else since June. And yet it was unlocked.

Now @Deego could have made a mistake. Or a ramp with many spurious GPS signals around it may have become unlocked as Waze HQ has said can happen.

But I have seen other roads that I know I have locked, that show I was the last to edit, and become unlocked months later.
Waze User Manuals, Map Editing guide and support at http://www.waze.com/wiki
Wayne McDougall
Make Friends with me on Facebook and we'll compete against each other on the Scoreboard
http://www.facebook.com/wayne.w.mcdougall
waynemcdougall
 
Posts: 517
Joined: Mon Mar 08, 2010 3:59 am
Has thanked: 0 time
Been thanked: 2 times

Re: Routing ignores all turn restrictions and road closures

Postby waynemcdougall » Sun Oct 31, 2010 9:16 am

robin1979 wrote:
waynemcdougall wrote:
robin1979 wrote:please go ahead and mess up the kiwi maps.


Unfortunately we just don't have places where two connecting roads can take 50km to go from one to the other.


We don't have those as well. In NL there's an off ramp at least every 10kms.


By connecting road, I meant a road that is connected, even if the turn is illegal. So you do have those. And we don't.

robin1979 wrote:
waynemcdougall wrote:
robin1979 wrote:We have a large userbase here, and if I can prevent a routing error one way or the other, I will do it. If locking the damn road is the solution, I will go ahead and lock all damn roads.


Here's the thing. I don't think the solution is good enough. The problem is that locked roads do not stay locked. Waze server will unlock them automatically if it thinks there is something wrong. And since ramps are close to freeways with multidirectional traffic, I think they're highly subject to this unlocking. You'll end up playing whack-a-mole - do you keep on checking them , or wait for another BrutusNL to make a report. I really don't think that's acceptable.


Well, what is the purpose of locking if the Waze server automatically unlocks them? I have not seen this behavior (of automatically unlocking) yet, and I doubt I will appreciate it much when Waze starts doing this. Locking is there just for the purpose to disable automatic adjustments!


You can see my response above. It's pretty clear that Locking is not well defined.
There was a very strong view expressed for a long time that it stopped people of lower rank from editing roads. That's not true. So eventually it was made clear that it does TWO things.
Firstly, locking makes it less likely (a lot less likely), that Waze will change the layout (geometry) of the road, and much less likely that it will change the directionality.
Secondly, it gives a boost (a big boost), to honouring the directionality of roads, and turn restrictions on junctions.

I think it is a mistake to have locking do the two different functions. Indeed, I don't think locking should be used at all when considering routing.

robin1979 wrote:
waynemcdougall wrote:Worse, we've been told that locking only makes it less likely to go down that corner. There may be an even more perverse case where locking will not help. Now it may be that the weighting is such that this is massively unlikely. Or it may just be rare. I'm not happy that that may be the case.


From my experience (and trust me, it's a lot): Waze will never send you down the wrong direction of a 'locked' road or restricted turn on a node attached to a road. It will abort with an error that it can't find a suitable road (in the client, on the live map it fails silently).


Yup, I agree, which is why I was so disbelieving of BrutusNL's report.

But see the conclusion:
1. We need to lock every road to make sure routing works correctly. If you have to lock every road, why not just change the routing algorithm to consider every road locked? Is that really a difficult proposition?
2. Since roads will unlock, you will get problems, without any prior warning.

robin1979 wrote:But, also, sometimes blocks (which are hard restrictions to nodes and links, not high-cost like in your example) are used, and that is how 'locked' roads seem to function. For unlocked roads you are probably right, it is merely a story of increasing costs for going the opposite direction.


See, I told you you were a schmott guy (not an insult) - it's smart, I've been reading too much Girl Genius - it's a compliment http://www.google.co.nz/search?q=schmott+guy&ie=utf-8&oe=utf-8&aq=t&rls=org.mozilla:en-GB:official&client=firefox-a

BUT Waze HQ have told us that locks are NOT blocks - they are higher weights. Maybe they are high enough so they are effectively blocks.

IF they were high enough and IF locks didn't disappear, then I would be satisfied.

But Locks do disappear. And I don't know if the blocks are high enough.

robin1979 wrote:
waynemcdougall wrote:Note that I am NOT saying you should not lock your segments. Nor am I saying it should be a high priority for the Waze team although I don't think it could be that hard - it ought to be changing a few constants. Should be easier than making a special exception for ramps in the Netherlands.


Since I don't believe it works the way you describe it (from my experience), I don't think there need to be any changes just yet (maybe a higher cost to roads to go against the direction of the road, to prevent cases like that of BrutusNL).


Fair enough. I've elaborated on the points you didn't know about. You may change your opinion. Since your objections aren't based on anything substantive, I'll just continue directly with Waze HQ.

robin1979 wrote:
waynemcdougall wrote:But I do think it is a bug that needs to be fixed. I agree with
BrutusNL wrote:I'd prefer if:
- it doesn't encourage illegal behavior, so no instructions for illegal turns, no matter what.



Well, I do agree with that. For trying to make that happen, locking is the answer. On unverified parts of the map (the non-locked roads), it is hard to do that, since Waze doesn't and can't know what restrictions are in effect.


For unverified parts of the map, it's not relevant, because there are no working routes anywhere. Waze doesn't just say "oh, try this". It just gives up completely. Much to the disgust of all the noobs complaining in the forums that they have a base map but can't route anywhere.

And locking is not the answer, as I've said, because roads don't stay locked.

robin1979 wrote:And by the way: it would be much better when one of the Waze cracks chimed in here, instead of having a big discussion with you. They should be able to clear things up, instead of us speculating about it.


We don't need to speculate. They have already told us how it works. You'e done the tests I wanted. I can resume talking with waze HQ. I'm just trying to include everyone. Waze HQ really avoid wading into the forums.

I still value anything further you want to discuss or add. In particular, whether you still don't think there is a problem if locked roads do go unlocked.

And you say locking is a solution - there are 87 million roads in Waze - most are not locked. More are added all the time. Even if they did stay locked, on what possible basis could you object to the routing algorithm ignoring the locked status?

When would you EVER lose a correct route because Waze assumed the map was correct? And even if you could, remember that there are 1000s of correct routes not taken because Waze usually assume the map is correct when it is not, and it doesn't disregard the map.

If locking is the solution, what possible harm would there be to lock every road in the database?
And if you were to do that, isn't it better to change the algorithm, and leave the locking for directionality and geometry updates.
Waze User Manuals, Map Editing guide and support at http://www.waze.com/wiki
Wayne McDougall
Make Friends with me on Facebook and we'll compete against each other on the Scoreboard
http://www.facebook.com/wayne.w.mcdougall
waynemcdougall
 
Posts: 517
Joined: Mon Mar 08, 2010 3:59 am
Has thanked: 0 time
Been thanked: 2 times

Previous

Return to Navigation & Routing

Who is online

Users browsing this forum: No registered users