search

viamrobotics / motion-testing

Apache License 2.0
2 stars 3 forks source link

RSDK-1889 Remove interaction space from scene 7 #4

Closed spieswl closed 1 year ago

spieswl commented 1 year ago

I removed the interaction space and instead put two walls where the sides of the I-Space were.

The walls are specified to have non-zero thickness, so that changes the Y coordinates for origin and box width, but the faces closest to the robot should be in the exact same locations as the former I-Space. I expect this to have a small impact on performance for this scene, due to collision checking against +1 geometry now (2 Obstacle boxes versus 1 I-Space box). Below is a picture of what that looks like in Ray's visualizer.

image

biotinker commented 1 year ago

I believe that zero-thickness walls should in fact work (acknowledging that the original design did not have them in mind; but they work for capsules).

Our internal testing repo might be a good place to give that a shot; that way if there's a radical change in Scene 7 availability we know that that behavior changes.

spieswl commented 1 year ago

Before I make that change, do we have a way to force the Motion CI tests to run for this repo somehow? It would be handy to test the current incarnation, make the zero-thickness change, and then re-run and compare.

raybjork commented 1 year ago

Since the behavior of zero thickness walls runs the risk of objects not being in collision if they appear on the other side, I think its unecessarily risky to model them as such and we should do the original suggestion of non-zero thickness.

do we have a way to force the Motion CI tests to run for this repo somehow?

Yes this repo is the motion CI suite itself and is designed to run standalone. The CI just calls two of the go tests that are exposed. TestDefault will generate the results TestScores will generate a markdown file of the results.

EDIT: although I should note that those scripts by default use flags to create and read from the appropriate folders. This is not documented anywhere and is something that I will try to add information on later today.

spieswl commented 1 year ago

OK awesome, I had attempted to run the same tests but ran into problems locally, so I was probably missing those flags you mentioned.