Osmand: non-needed keep left/right

[Carles Pina i Estany]

Hi,

We are in a trip in Romania and we are observing something new for us. I
don't know if it's a recent Osmand update problem, a mapping problem,
etc.

Basically in a road with bends Osmand says every now and then to keep
right/left but I don't see any reason in the road for this indications.

I'm looking now at openstreetmap.org routers and they don't have this
problem.

I'm also looking at the osm data (in case that the road should be part
of a relation) but I don't see any problem either.

Example:
Starting point: 45.1498337,25.9894218
Destination: 45.188782,25.953548

Osmand generated directions:
1. Head DJ102 Strada Principală (2.84 km, 3:24)
2. Keep left and go DJ102G (37 m, 0:03)
3. Turn slightly right and go DJ102G (69 m, 0:05)
4. Keep right and go DJ102G (268 m, 0:19)
5. Keep right and go DJ102G (149 m, 0:11)
6. Keep right and go DJ102G (610 m, 0:44)
7. Keep right and go DJ102G (72 m, 0:05)
8. Keep left and go DJ102G (1.46 km, 1:50)
9. (0 m, 0:00)

The directions generated by Car (GraphHopper):
1. Continue 20m
2. Turn sharp left 10m
3. Turn right onto Strada Principală, DJ102 3km
4. Keep right onto DJ102 1000m
5. Arrive at destination

(see the .gpx file attached as well)

Osmand version:
OsmAnd+ 3.4.8 (with Romania Osmand live enabled)

Why is osmand generating too many keep right and go DJ102G turnings?

Thank you very much!

--
Carles Pina i Estany

[Florian Lohoff]

Hi,

On Sun, Oct 06, 2019 at 06:58:40PM +0300, Carles Pina i Estany wrote:
>
> Hi,
>
> We are in a trip in Romania and we are observing something new for us. I
> don't know if it's a recent Osmand update problem, a mapping problem,
> etc.
>
> Basically in a road with bends Osmand says every now and then to keep
> right/left but I don't see any reason in the road for this indications.

These are topology issues. Streets dont join up in 90° Angles but in
sharp angle. So the navigational messages see a split of the street and
tell you to either use the right or left road.

> I'm looking now at openstreetmap.org routers and they don't have this
> problem.

They should issue the same commands although the threshold when to
genereate a "keep left/right" to detecting a straight may be different
angles.

So OSMAnd may decide to issue a "keep xxx" command from 120° on, others
may decide to do so from 135°.

So you will need to fix all these junctions to be more like a 90° which
in reality they will be on the last 5-10m as junction will typically
be designed to let vehicles line up in 90° angles to have a better view
of the road.

Flo
--
Florian Lohoff f...@zz.de
UTF-8 Test: The 🐈 ran after a 🐁, but the 🐁 ran away

[Carles Pina i Estany]

Hi,

On Oct/06/2019, Florian Lohoff wrote:
> Hi,
>
> On Sun, Oct 06, 2019 at 06:58:40PM +0300, Carles Pina i Estany wrote:
> >
> > Hi,
> >
> > We are in a trip in Romania and we are observing something new for us. I
> > don't know if it's a recent Osmand update problem, a mapping problem,
> > etc.
> >
> > Basically in a road with bends Osmand says every now and then to keep
> > right/left but I don't see any reason in the road for this indications.
>
> These are topology issues. Streets dont join up in 90° Angles but in
> sharp angle. So the navigational messages see a split of the street and
> tell you to either use the right or left road.

Some of the examples (see screenshot from Osmand to see the arrows and
from openstreetmap.org) don't have any streets joining.

In openstreetmap.org I don't see any element joining.

And actually it's the same line (way) with the proper name and Osmand
knows that it's "correct" as it says

But Osmand knows that it's the same road (as it says in the messages all
the time "keep right and go to DJ102G)

> > I'm looking now at openstreetmap.org routers and they don't have this
> > problem.
>
> They should issue the same commands although the threshold when to
> genereate a "keep left/right" to detecting a straight may be different
> angles.
>
> So OSMAnd may decide to issue a "keep xxx" command from 120° on, others
> may decide to do so from 135°.
>
> So you will need to fix all these junctions to be more like a 90° which

For many of the examples there are no junctions that I can see (two
clear on the screenshot, there is one in the screenshot with a street
jorining the road... but osmand knows that I'm in DJ102G and still says
to go to DJ102G.

Is there anything else that I could verify?

Thank you!

[Florian Lohoff]

On Sun, Oct 06, 2019 at 10:41:56PM +0300, Carles Pina i Estany wrote:
> And actually it's the same line (way) with the proper name and Osmand
> knows that it's "correct" as it says
>
> But Osmand knows that it's the same road (as it says in the messages all
> the time "keep right and go to DJ102G)

How does OSMAnd know its the same road driving "straight on"? Not by
name - Roads may turn 90° left/right and carry the name with them. So
there is no tag which says "this is the same road - straight on is here"

The only thing which matters in "keep left" "keep right" messages
is the topology. So when the topology looks like a fork in the street
e.g. a junction with one of the streets in a steep angle OSMAnd
assumes there is no predominant "straight on" but it needs to
tell the user which of the two roads to take. So it generates
"keep left", "keep right" messages.

> For many of the examples there are no junctions that I can see (two
> clear on the screenshot, there is one in the screenshot with a street
> jorining the road... but osmand knows that I'm in DJ102G and still says
> to go to DJ102G.

But thats an arbitrary ref/name not used to identify its the same road.
Its just used for annotating the message on where to go.

> Is there anything else that I could verify?

Topology, Topology, Topology.

Whenever a router/navigation software generates instruction messages
its because of topology in the base data. So if you get stray messages
about keeping left/right although for you as a human the "straight on"
is obvious the junction is most likely broken topology wise.

[Carles Pina i Estany]

Hi,

On Oct/06/2019, Florian Lohoff wrote:

> On Sun, Oct 06, 2019 at 10:41:56PM +0300, Carles Pina i Estany wrote:
> > And actually it's the same line (way) with the proper name and Osmand
> > knows that it's "correct" as it says
> >
> > But Osmand knows that it's the same road (as it says in the messages all
> > the time "keep right and go to DJ102G)
>
> How does OSMAnd know its the same road driving "straight on"? Not by
> name - Roads may turn 90° left/right and carry the name with them. So
> there is no tag which says "this is the same road - straight on is here"

When I was reading your message I realised that I had forgotten to
attach the screenshot. I've attached now the screenshot. Osmand is
saying "keep left and go to DJ102G 0:03" but as I said I can't see any
other option than following the road, no junction / street / footpath /
etc. joining at this point.

So, in the screenshot there is the yellow arrow with the orange
indicator where it said "keep left and go..." and short after this one
there is another one with the same case (the third one yes, it could be
the street joining, I agree with this one and I'm ignorning it for now).
There are no junctions in the visible map, no junctions in the
openstreetmap.org data.

> The only thing which matters in "keep left" "keep right" messages
> is the topology. So when the topology looks like a fork in the street

there is no fork that I can see in the example (that I've now attached)

> > For many of the examples there are no junctions that I can see (two
> > clear on the screenshot, there is one in the screenshot with a street
> > jorining the road... but osmand knows that I'm in DJ102G and still says
> > to go to DJ102G.
>
> But thats an arbitrary ref/name not used to identify its the same road.
> Its just used for annotating the message on where to go.

Oh, I had assumed that the ref/name OR the fact that lines are part of a
relation were used for the directions. It's good to know that it's not
used. But I don't see why in this case it would still say "keep left..."

> > Is there anything else that I could verify?
>
> Topology, Topology, Topology.
>
> Whenever a router/navigation software generates instruction messages
> its because of topology in the base data. So if you get stray messages

Please see the screenshot (that now I've attached), as I don't
understand why Osmand would indicate "keep left" when there is no
junction that I can see.

> about keeping left/right although for you as a human the "straight on"
> is obvious the junction is most likely broken topology wise.

I'd be happy to understand and then fix at least some of them... but see
the screenshot (that is now available, not like before :-) )

Thank you very much,

[Florian Lohoff]

On Mon, Oct 07, 2019 at 10:27:17AM +0300, Carles Pina i Estany wrote:
> Hi,
>
> On Oct/06/2019, Florian Lohoff wrote:
> > On Sun, Oct 06, 2019 at 10:41:56PM +0300, Carles Pina i Estany wrote:
> > > And actually it's the same line (way) with the proper name and Osmand
> > > knows that it's "correct" as it says
> > >
> > > But Osmand knows that it's the same road (as it says in the messages all
> > > the time "keep right and go to DJ102G)
> >
> > How does OSMAnd know its the same road driving "straight on"? Not by
> > name - Roads may turn 90° left/right and carry the name with them. So
> > there is no tag which says "this is the same road - straight on is here"
>
> When I was reading your message I realised that I had forgotten to
> attach the screenshot. I've attached now the screenshot. Osmand is
> saying "keep left and go to DJ102G 0:03" but as I said I can't see any
> other option than following the road, no junction / street / footpath /
> etc. joining at this point.

I had a look - At the exact position of the marker - there is no
junction. There are junctions within 200m in both direction which
qualify for a steep angle.

> > The only thing which matters in "keep left" "keep right" messages
> > is the topology. So when the topology looks like a fork in the street
>
> there is no fork that I can see in the example (that I've now attached)

I cant see it either.

> > Whenever a router/navigation software generates instruction messages
> > its because of topology in the base data. So if you get stray messages
>
> Please see the screenshot (that now I've attached), as I don't
> understand why Osmand would indicate "keep left" when there is no
> junction that I can see.
>
> > about keeping left/right although for you as a human the "straight on"
> > is obvious the junction is most likely broken topology wise.
>
> I'd be happy to understand and then fix at least some of them... but see
> the screenshot (that is now available, not like before :-) )

What makes me wonder is that the calculated route does not match the
rendered map. So something has changed in between. Maybe the route
is calculated on a different map dataset than the map beeing rendered?

[Carles Pina i Estany]

Hi,

On Oct/07/2019, Florian Lohoff wrote:

> > I'd be happy to understand and then fix at least some of them... but see
> > the screenshot (that is now available, not like before :-) )
>
> What makes me wonder is that the calculated route does not match the
> rendered map. So something has changed in between. Maybe the route
> is calculated on a different map dataset than the map beeing rendered?

This device has OsmandLive. Later on I'll check the same route using the
same Romania map version in a device without OsmandLive. In case that
all of this is caused by some "interference" of osmandlive.

I'll let you know what happens.

Thanks for your help,

[Grzegorz Adamiak]

I have experienced exactly the same behavior on my recent trip in Poland. Osmand generated a lot of "keep left/riht" or "turn slightly  left/right" on a curvy road. That actually resulted in TTS audible spam so I finally turned it off to let my hearing relax. I'm keeping it silent for that reason as it hurts more than helps.

I suspect it to result from map data after all. In few cases a similarly curvy road had no direction arrows and thus no TTS messages. Instead there was only a "continue for X km", which one would expect when staying on the same road for prolonged distance. This is how Google Maps navigates and it even ignores joining roads, which is natural IMO. I'd expect the same from Osmand navigation.

It is a bit of a stalemate situation. You are not able to fix or edit all junctions you may possibly travel through. The only sensible option IMO would be to provide fine tuning for Osmand directions, and in result for voice messages or solely for it, to set thresholds on sharpness and/or road type or similar. I'd definitely want lane assist on highways but not on a local curvy road.

--

Grzegorz

[Xavier]

On Mon, Oct 07, 2019 at 10:27:17AM +0300, Carles Pina i Estany wrote:
>On Oct/06/2019, Florian Lohoff wrote:
>> On Sun, Oct 06, 2019 at 10:41:56PM +0300, Carles Pina i Estany wrote:
>> > And actually it's the same line (way) with the proper name and
>> > Osmand knows that it's "correct" as it says
>> >
>> > But Osmand knows that it's the same road (as it says in the
>> > messages all the time "keep right and go to DJ102G)
>>
>> How does OSMAnd know its the same road driving "straight on"? Not
>> by name - Roads may turn 90° left/right and carry the name with
>> them. So there is no tag which says "this is the same road -
>> straight on is here"
>

>I'd be happy to understand and then fix at least some of them... but
>see the screenshot (that is now available, not like before :-) )

Looking at just the screenshot (as I do not know where this is to
access the OSM data) I notice that the marker is aligned with an edge
of the landcover area adjacent to the road.

Which makes me wonder if the landcover area is actually connected to
the road way at this spot? I can't tell from the screen shot (because
the marker icon is obscuring where the two might connect, if they do.

If the two are connected, you might try disconnecting them and moving
the edge of the landcover slightly off from the road way (which is
actually more accurate, given that the OSM way denotes the road
centerline, and the land cover does not extend across the pavement to
the center of the road).

[Paul Johnson]

On Mon, Oct 7, 2019 at 2:27 AM Carles Pina i Estany <car...@pina.cat> wrote:

Hi,

On Oct/06/2019, Florian Lohoff wrote:
> On Sun, Oct 06, 2019 at 10:41:56PM +0300, Carles Pina i Estany wrote:
> > And actually it's the same line (way) with the proper name and Osmand
> > knows that it's "correct" as it says
> >
> > But Osmand knows that it's the same road (as it says in the messages all
> > the time "keep right and go to DJ102G)
>
> How does OSMAnd know its the same road driving "straight on"? Not by
> name - Roads may turn 90° left/right and carry the name with them. So
> there is no tag which says "this is the same road - straight on is here"

When I was reading your message I realised that I had forgotten to
attach the screenshot. I've attached now the screenshot. Osmand is
saying "keep left and go to DJ102G 0:03" but as I said I can't see any
other option than following the road, no junction / street / footpath /
etc. joining at this point.

Are you using Osmand Live?  I've noticed this comes up a lot when Osmand Live differs from the monthly map.

[Carles Pina i Estany]

Hi,

On Oct/07/2019, Carles Pina i Estany wrote:
>
> Hi,
>
> On Oct/07/2019, Florian Lohoff wrote:
>
> > > I'd be happy to understand and then fix at least some of them... but see
> > > the screenshot (that is now available, not like before :-) )
> >
> > What makes me wonder is that the calculated route does not match the
> > rendered map. So something has changed in between. Maybe the route
> > is calculated on a different map dataset than the map beeing rendered?
>
> This device has OsmandLive. Later on I'll check the same route using
> the same Romania map version in a device without OsmandLive. In case
> that all of this is caused by some "interference" of osmandlive.

I've disabled OsmandLive in this device. Osmand asked if I wanted to
delete the Osmand live updates, I said yes. Then the problem doesn't
happen.

I've enabled Osmand live updates and updated and then the problem is
here again.

I've tried disabling it again: this time Osmand didn't ask me if I
wanted to delete the files and the problem still happens (but I have the
osmandlive files).

So, like Paul Johnson has said: it seems to be a probem with Osmandlive.
The timings make sense: I enabled Osmand live a few weeks ago and I
realised of the problem then... but it's a new area for me, so I wasn't
sure if it's a mapping problem (it didn't occur to me initially to think
that it's an Osmandlive problem).

I'll keep testing (and with the other device) and if I can reproduce the
problem without Osmandlive I'll share it here.

Thanks all for the feedback / questions / ideas :-)

[Peter B]

Please report to github.

[Carles Pina i Estany]

Hi,

On Oct/08/2019, Peter B wrote:
> Please report to github.

I'll do it (later this week) (with the example, screenshot, etc.)

Cheers,

[Martin Trautmann]

On 19-10-07 10:59, Grzegorz Adamiak wrote:
> I have experienced exactly the same behavior on my recent trip in
> Poland. Osmand generated a lot of "keep left/riht" or "turn slightly 
> left/right" on a curvy road. That actually resulted in TTS audible spam
> so I finally turned it off to let my hearing relax. I'm keeping it
> silent for that reason as it hurts more than helps.

yes, I had the same issue this weekend - osmand kept talking to "links
einordnen" (or right) all the time, while staying on the same road for
many kilometers. It was wrong and annoying for 99.9% of its announcements.

It is that bad, I want an option to turn off that kind of "advice".

Schönen Gruß
Martin

[Martin Trautmann]

On 19-10-07 21:23, Carles Pina i Estany wrote:

> I've disabled OsmandLive in this device. Osmand asked if I wanted to
> delete the Osmand live updates, I said yes. Then the problem doesn't
> happen.

I never used OsmandLive, but still have that problem - but probably
since the recent update only.

Schönen Gruß
Martin

[vzteklik]

I have the same happening many times as well and my guess, since I do
not know the details of map creation, is that in the map when there is a
intersection, one must tag somehow the 'main' road. If you then drive on
this 'main' road only, no turn instructions are provided.

Obviously if there is an intersection and the road connectiong to the
one you are driving on is tagged as 'main', you will be instructed to
turn.

So I would be looking at how the roads are tagged in OSM first. Would it
make sense?

V.



------ Původní zpráva ------
Od: "Martin Trautmann" <tr...@gmx.de>
Komu: osm...@googlegroups.com; "Carles Pina i Estany" <car...@pina.cat>
Odesláno: 28/07/2020 16:33:04
Předmět: Re: Osmand: non-needed keep left/right

>--
>You received this message because you are subscribed to the Google Groups "OsmAnd" group.
>To unsubscribe from this group and stop receiving emails from it, send an email to osmand+un...@googlegroups.com.
>To view this discussion on the web visit https://groups.google.com/d/msgid/osmand/ed431ab2-1c88-baf6-2db2-648bd9cb2778%40gmx.de.

[Majka]

As far as I know, OsmAnd gives announcements depending on the road geometry. You can get announcement to "keep right/left", even if there is no other way to go.

The tagging has nothing to do with it.

[Poutnik Fornntp]

If one uses BRouter as the 3rd party routing service, it provides navigation hints based on true crossings/junctions.

The downside is, the current OsmAnd router API doesn't ( or did not ) support the Brouter hints, so they are available only if the route is generated by BRouter application or Web as GPX file which is then followed by OsmAnd.

Dne 9. října 2020 10:28:39 Majka <majk...@gmail.com> napsal:

--

You received this message because you are subscribed to the Google Groups "OsmAnd" group.
To unsubscribe from this group and stop receiving emails from it, send an email to osmand+un...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/osmand/680ad2a7-608e-4cd7-b1ad-2b9675dc8f09o%40googlegroups.com.