Visualize science data as points

[chapulina]

Part of

• Ignition Marker improvements for LRAUV #52
• Generic point cloud visualization gazebosim/gz-sim#1156

Builds on top of

• Convert science data from lat/lon to ENU #98
• Setup profiler #87
• Migrate to Ignition Garden 🔥🌱 #107

Requires

• Marker: support different colors per point gazebosim/gz-msgs#202
• Set marker point size from message gazebosim/gz-gui#317
  • Don't set visual scale for point markers gazebosim/gz-gui#321
• Set per-point color on markers gazebosim/gz-gui#318
• Per-point colors in point clouds gazebosim/gz-rendering#494
• Increase maximum values in ViewAngle widget gazebosim/gz-sim#1221
• Bump in Garden: ign-msgs and dependents gazebo-tooling/release-tools#554
• 6 ➡️ 7 (main) gazebosim/gz-sim#1236
• 6 ➡️ 7 (main) gazebosim/gz-gui#333
• 6 ➡️ 7 (main) gazebosim/gz-rendering#507

Nice to have

• Grid config: set values from startup and improve layout gazebosim/gz-gui#324

---

This PR changes the science data visualization from boxes to points, which brings these advantages:

• No need for performance hacks or hiding part of the data, since points are very efficient
• Therefore, we can see all data at once
• Points are scaled the same no matter how far they are. This allows us to see data that's kilometers away. It can be hidden adjusting the camera clipping distance as needed.

Tasks

• Remove all performance hacks - I'm debugging against all the data we have to see if we can get rid of all hacks. May need to bring some of them back if performance can't be improved without them. No hacks are needed 😄
• Use multiple points in one marker instead of multiple markers, each with one box
• Configure point size
• Set an initial camera clip that visualizes data far away, and add ViewAngle to the world so we can easily change the clip at runtime
• Set color per-point
• The coordinates seem swapped, look into what's happening (edit: those were screenshots from an old branch, the latest branch looks right.

Future

• Visualize currents

Nice-to-have:

• Better mouse control at far distances
• Camera clipping for orthographic
• Ogre 1 per-point coloring
• Look into using a color palette with more colors

Here's the data with the axes added in #104 , left is ENU, right is NED:

[mabelzhang]

The visualization works.

There is, however, a black dot at exactly the origin for all the topics. I see it in your animation too. Is it by design, to help find the origin?

It overlaps with the actual marker at the origin, so the color of the actual marker (see below) isn't visible.

In case it's not by design, I looked in the following places and couldn't find where it comes from.

There isn't a marker with any of RGB components <= 0:

    if ((color.R() < 1e-6) &&
      (color.G() < 1e-6) &&
      (color.B() < 1e-6))
    { 
      ignerr << color << std::endl;
    }

The marker at the origin is not black:

    if ((fabs(*iterX) < 1e-6) &&
      (fabs(*iterY) < 1e-6) &&
      (fabs(*iterZ) < 1e-6))
    { 
      ignerr << color << std::endl;
    }

[GUI] [Err] [VisualizePointCloud.cc:396] 0.530241 0.469759 0 1

If I skip the origin, the black dot is still there:

    if ((fabs(*iterX) < 1e-6) &&
      (fabs(*iterY) < 1e-6) &&
      (fabs(*iterZ) < 1e-6))
      continue;

I looked at whether it has anything to do with marker ID 0 (the DELETE_ALL) one. It doesn't seem to be, in that if I use the same ID for both the marker array and the delete marker, it didn't make any difference.

I tried adding isinf() to the existing isnan() clause, but that didn't get rid of the black dot.

[chapulina]

There is, however, a black dot at exactly the origin for all the topics. I see it in your animation too. Is it by design, to help find the origin?

Ha, glad that you found a use for it. And thanks for the debugging! It was a bug, fixed in gazebosim/gz-rendering#519

[chapulina]

• I broke off a small part of this PR into Grid: modify default grid and add GUI plugin #125 while I look into why CI is hanging on this branch

[chapulina]

I look into why CI is hanging on this branch

• I found what the problem was and fixed in c77869c. Since it was related to the barycentric interpolation and I'm not sure what the proper fix is, I broke those changes into a new PR Remove miniature scale from science data and fix segfault #126. Now this PR builds on top of that. @mabelzhang , mind taking a look at that?

[mabelzhang]

Bye 2021