Maliput road networks inspection

[hidmic]

Description

Currently, our maliput_viewer can render Maliput road geometries for visual inspection. However, it does not provide similar capabilities for road network rule books.

This ticket tracks progress as delphyne-guis capabilities are augmented to fill that gap.

Stages

Work's split in two stages.

1. Consolidate inspection procedures

This stage serves two main purposes:

• To verify the practical usefulness of the inspection outcome and how it is presented (or even help defining how it should be presented).
• To implement the inspection logic itself, which is the same regardless of how I/O is handled.

To this end, the existing malidrive_query command line tool can be used as grounds to work on each POC.

Tasks

• Implement rules' (starting with right-of-way) inspection for a given lane region.
  • https://github.com/ToyotaResearchInstitute/malidrive/pull/270
• Implement rules' state and region of influence inspection for a given phase.
  • https://github.com/ToyotaResearchInstitute/malidrive/pull/271

2. Support UI-based inspection

This stage deals with actually adding road network inspection support to delphyne-guis visualizer. It does so incrementally, in two substages, using ignition-guis plug-in widget infrastructure.

2.1 Text-only I/O

The simplest transition from malidrive_query command line interface. This may include listing intersections, phases and/or rules in list or tree widgets, input boxes for region selection, dialogs, etc., suitable for each inspection procedure.

Tasks

• Show rules for a given lane.
  • Added basic support for visualizing rules for clicked lanes #221
• Filter rules for a given lane by phase.

2.2 Graphical I/O

Once text I/O is in place, inspection support can be extended to provide graphical I/O as well. This may include include highlighting a rule's region of influence and/or providing visual cues for it [1], rendering traffic lights, rendering floating labels, supporting point-n-click lane and/or intersection selection [2], etc., suitable for each inspection procedure.

[1] Ignition rendering struggles somewhat with large road networks. Road overlays may be too high a burden for it. Road outlines may help use here, but we might hit a performance limitation.
[2] Click position-to-scene coordinates transformation appears to be non-trivial in ignition-rendering / ogre. To be confirmed.

Tasks

• Enable point-n-click selection for lanes
  • Get lane id from clicked lane #217
  • Added basic support for visualizing rules for clicked lanes #221
• Implement a hovering/highlighting device for select lanes.
  • Add arrow to selected lanes #223
• Show srh and xyz coordinates for a clicked point.
• Add sliders for s lane region definition.

[hidmic]

@agalbachicar PTAL, I think it correctly reflects what we discussed.

[agalbachicar]

Agreed. It does so. @andrewbest-tri would you like to take a look? After your approval, I think we can kick off different issues in particular repositories to fulfill the functionality.

[andrewbest-tri]

Looks good to me. Can we add a note that we want to inspedt *Rules not just right of way? Vehicle usage, direction usage, speed limit, etc...

but we can start with right-of-way

[liangfok]

Looks good. Longer term, can the visual GUI be extended to enable editing and save-as in case we want to change something?

[hidmic]

@liangfok hmm that's a much larger task. Not only because of edition support, but also because our UI so far has been backend agnostic. So, yes, it'd be possible, but it'd be a long long term goal.

[liangfok]

Okay, let's just keep it in mind.

[stonier]

[2] Click position-to-scene coordinates transformation appears to be non-trivial in ignition-rendering / ogre. To be confirmed.

That seems really odd. Even in simple frameworks, like libqglviewer, there is simple api to go from cursor to the visible scene point underneath the cursor. TLDR; must be something there in Ogre at least, maybe not exposed via ignition?

[hidmic]

TLDR; must be something there in Ogre at least, maybe not exposed via ignition?

That may be the case. I've opened #215 to follow up on this.