Locus Map - forum

Support => Troubles & Questions => Topic started by: Graf Geo on August 23, 2021, 13:46:57

Title: Route planner: Difference in quantity and quality of navigation commands
Post by: Graf Geo on August 23, 2021, 13:46:57
Hello dear forum,

I have another question about navigation instructions and navigation along a route created with the Locus Route Planner.

In planning mode, I can choose whether or not I want to create navigation commands when planning a route (screenshot 1).
Since I always had the impression that 1. produces the worse results, I experimented today.

I created the exact same test route in Berlin's Charlottenburg Palace Park once with and once without navigation instructions.

In the first case, 17 navigation waypoints (including start and destination) are created during planning. In the process, numerous junctions are ignored or omitted and also some incorrect or pointless instructions are created. If I were to rely on this navigation, I would hardly reach my destination.

In the second case, no navigation waypoints are created during planning. However, when navigating, navigation commands are generated according to the change of direction in the route. Now I can display the route details and 28 navigation waypoints (incl. start and destination) are then listed there.
Practically no important direction changes are omitted and the commands are all coherent.

Now that's ok for me, I always prefer the second variant anyway. Nevertheless, I wonder why variant 1 works so unreliably that it is practically unusable. Do you have similar experiences?

Here are 2 groups of screenshots to illustrate.

In the first case, I have outlined the 15 navigation instructions created during planning in red for better recognition (without start and destination). You can see that no navigation instructions were created at several junctions and that, for example, the 6th instruction "Turn around if possible (Drehe nach Möglichkeit um)" is complete nonsense.

In the second case, no navigation instructions can be shown on the map because none are created during planning. In the route details you then see a list of 28 navigation waypoints and if you compare them with the course of the route, you see that 99% of the time it fits perfectly. 

Even though there are two "superfluous" instructions in a curved section of the route, the instructions are also much more precise.

Under what circumstances then can variant 1 be useful at all?

Addition: Both variants were created with the LM4 app on the smartphone.
If I create the same route with the Web Planner, 26 navigation commands are generated (including some superfluous straight-ahead commands), but also with errors and gaps.
Title: Re: Route planner: Difference in quantity and quality of navigation commands
Post by: poutnikl on August 23, 2021, 16:11:07
Interesting. I ask myself the opposite question - When is navigation based on geometry useful ( aside of off-road navigation through free terrain ).

I used to it in past for precalculated routes, when using OSMand, but I have later switched to Brouter based navigation hints, never looking back.

But one has to understand, how and when Brouter creates the hints ( or does not ).
E.g. If passing a town via a secondary or tertiary road, Brouter will never raise a hint, no matter what turns are there, unless there is a crossroad with a road of the same or higher priority ( profile priority, not traffic priority). OTOH, If you are to go straight ahead to the town residential network, leaving the major road turning away, it would tell you.

Look - in Brouter context - for terms PriorityClassifier and TurnInstructionCatchingRange

For more see https://GitHub.com/poutnikl/Brouter-profiles/wiki/Glossary

The priority classifier
-----------------------

Priorityclassifier is a BRouter numerical parameter calculated for ways and used for generation of pictogram/voice navigation instructions. Higher values means the more significant (noticeable) way, as far as it can be predicted from OSM data.

To avoid a navigation instruction flood, it was decided that the instructions are provided only if:

1/ You are supposed to turn at a crossroad/junction
    and some other leaving options have the same or higher Priorityclassifier value.
OR
2/ You are supposed to go straight ahead  and some other leaving options have the higher Priorityclassifier value.


----

assign turnInstructionCatchingRange = 20    # %turnInstructionCatchingRange% | Within this distance (in m) several turning instructions are combined into one and the turning angles are better approximated to the general direction | number

This parameter considers all crossroads/junctions within the range as a single one, providing rather a cartographic wider picture relative direction. So it takes all the potential crossraod maze within the range as a "crossroad black box". If you approach this black box e.g. from south, and if you are to leave it to northwest ( in a broader picture ), it would say - "turn slightly left". If you are to go to southeast, you get "turn sharply right".

Reasons for this are (my guess, not official explanation ):
- sometimes confused mapping of the road connections and their angle geometry
- sometimes confused precise road connection locations
- sometimes confused precise user location due position error
- sometimes confused user position timing, misleading the user he is elsewhere than reported position itself. ( Like the Locus reports there is 70 m to the crossroads, at the time when the physical distance is just 40 m. It may lead to taking a wrong turn.

Title: Re: Route planner: Difference in quantity and quality of navigation commands
Post by: 0709 on August 24, 2021, 09:46:31
@ Graf Geo.
Test. Turn instructions generated by the RouteYou web Planner.
https://www.routeyou.com/nl-de/route/view/9599031/alle-routes/charlottenburg-wilmersdorf
@ download: Select Locus gpx.
Title: Re: Route planner: Difference in quantity and quality of navigation commands
Post by: Graf Geo on August 24, 2021, 12:19:38
Thanks 0709.
Navigation commands created by routeyou.com seems to be excellent.  :)

But replacing Locus functions with other tools is not the solution. Locus has a basically good and easy-to-use route planner and should be able to produce reasonable and reliable navigation instructions. At least that is what I would expect.