Adapt ReferenceLine ST definition to avoid transformation ambiguity and discontinuities

[ReinhardBiegelIntech]

This addresses #645

Current definition of s,t coordinates has some (well documented) downsides, which we would like to mitigate with this proposed change. With the updated s,t definition, x,y <-> s,t associations are unique inside the lane geometries and there are no gaps in the s coordinate anymore.

This approach is used and tested by openPASS .

Checks will follow (but the links to howtocontribute.html seem to be dead... where to get this information?)

• My suggestion follows the style and contributors guidelines.
• I have taken care about the documentation.
• I have done the DCO signoff.
• My changes generate no errors when passing CI tests.
• I have successfully implemented and tested my fix/feature locally.
• Appropriate reviewer(s) are assigned.

[ArunDas1]

@tbleher @ThomasNaderBMW @jdsika Here is our proposal regarding the discussed issue with computation of s-t-coordinates.

[tbleher]

s/instersection/intersection/

[tbleher]

This change removes the \c formatting markers. I found these helpful for readability when reading the rendered document (as will be displayed on the website). I suggest you check how it looks in Doxygen, and probably re-add the \c formatting markers.

[ReinhardBiegelIntech]

Agree, these should be added again.

[tbleher]

Change looks good to me.

[ReinhardBiegelIntech]

As discussed in the ASAM OSI Other Models Workgroup meeting on Monday, we updated the PR to clarify the constraints on the T axis.

[lemmer-fzi]

After reading the documentation and having a look at the provided diagrams i have a question:
Consider a reference line that only has two points (straight line). How would you calculate the (s,t) coordinates for a point that lies on a line perpendicular to the last point of the segment. As far as i can, using the proposed definition of the t axis their is no way to define a t-axis that crosses the point and still maps it to a valid s-coordinate (a value of s smaller than the maximal value specified by the last point on the reference line).
In large, this question also applies to the last segment of a specified reference line. Could you maybe elaborate on this?

[weissdavid]

After reading the documentation and having a look at the provided diagrams i have a question: Consider a reference line that only has two points (straight line). How would you calculate the (s,t) coordinates for a point that lies on a line perpendicular to the last point of the segment. As far as i can, using the proposed definition of the t axis their is no way to define a t-axis that crosses the point and still maps it to a valid s-coordinate (a value of s smaller than the maximal value specified by the last point on the reference line). In large, this question also applies to the last segment of a specified reference line. Could you maybe elaborate on this?

You can always define the t-axis to be the line perpendincular to the segment. In this case your point, which is perpendicular to the last point lies exactly on this t-axis and therefore has the same s coordinate (the t-axis consists of all points with equal s coordinate). Furthermore all s coordinates are valid (even negative values).

[ReinhardBiegelIntech]

@ThomasNaderBMW
We updated the PR as discussed last week (addition of a reference line type enum).

[tbleher]

I didn't have time to participate in the discussion, so the following review is purely based on reading the patchset. This seems fine for keeping backward compatibility (but also means that all the old documentation will have to be kept), but also means that for agent models, it now becomes MUCH harder to support this, since they will have to implement TWO modes for ST calculations instead of one (simulation environments need only support one, but agent models need to support two). So I don't really like this idea.

Personally, I would be OK with breaking backward compatibility here (considering this is a very new feature, and probably noone has implemented it fully), but I don't know if this will gain consensus.

If the two modes are kept, then please at least mark one of the modes as deprecated, and advise agent models to only support the newer modes. If some agent models only support the first mode, and some only support the second mode, then no interoperability is achieved!

[weissdavid]

OSI_s-t_calculation.pdf
As discussed in the CCB meeting, we prepared an exemplary calculation.

[thempen]

From WG Harmonization: Please add a hint at the beginning, that the ST coordinate definition should not be misunderstood as the ST definition in OpenDRIVE and OpenSCENARIO 1.x.

The current proto definition for the enum fields "other" and "unknown" are missing, so it does not follow the coding convention guidelines: https://opensimulationinterface.github.io/osi-documentation/#_contributing_changes
To be determined by CCB if this is an exceptional situation and can be accepted.

[ReinhardBiegelIntech]

@thempen Added a note as discussed. Should this be tagged as ReadyForCCBReview again?

[pmai]

OSI CCB 2022-11-21: Can be merged as is with necessary changes to resolve technical conflicts. Merge and cleanup done by @pmai.