Correct Atlas model?

[Steven Gray]

Is this the robot model that you guys use internally in the controller? ihmc_ros / ihmc_models / urdf / atlas_unplugged_v5_no_hands.urdf

I noticed @dljsjr commented on DRCSim issue #492, but atlas_unplugged_v5_no_hands.urdf doesn't reflect all the changes there. Its masses disagree with the atlas_v5_simple_shapes.urdf found in the associated DRCSim pull request #536. It also doesn't help that atlas_v5.urdf and atlas_v5_simple_shapes.urdf disagree on that pull request - only the simple shapes version has the updated masses.

For instance,
l_clav is 4.466kg in atlas_v5_simple_shapes.urdf, but 3.45kg in atlas_unplugged_v5_no_hands.urdf
l_foot is 2.410kg in atlas_v5_simple_shapes.urdf, but 2.05kg in atlas_unplugged_v5_no_hands.urdf
l_uarm is 4.386kg in atlas_v5_simple_shapes.urdf, but 3.388kg in atlas_unplugged_v5_no_hands.urdf

What is the correct robot model we should be using to match what you use internally?

Thanks, Steve

[DouglasS]

@stgray This is an error on our side, we will address it in a future release. However, this will not impact our control algorithm. The mass and kinematic model used by the controller is internal to the Java distribution (we use .sdf instead of .urdf). The vendored URDF is the model used to populate the ROS Robot State Publisher. Feel free to override this using the description_model:= roslaunch argument.

The .sdf we are using internally is generated from the "correct" URDF as outlined in the pull request with the increased masses. Not sure where the discrepancy came from.

[DouglasS]

Need to upload a new/correct URDF of Atlas v5

[DouglasS]

Fixed the .urdf model. See issue #30.

→ <<cset 267bbf8c550e>>

[Steven Gray]

@dljsjr Awesome, I was just about to ask about this. Thank you!

[Steve Peters]

I think 267bbf8c550e happened just before drcsim pull request 536, which fixed the left arm kinematics. I'll open another pull request to update the arm kinematics.

[Steve Peters]

see pull request #30