Talos not correct for pybullet - No link should be left without inertia

Gepetto / example-robot-data

Set of robot URDFs for benchmarking and developed examples.

BSD 3-Clause "New" or "Revised" License

151 stars 56 forks source link

Talos not correct for pybullet - No link should be left without inertia #67

Closed olivier-stasse closed 3 years ago

olivier-stasse commented 3 years ago

In pybullet, links should have an inertia and a mass specified even if this is zero. Otherwise the default mass value is 1 Kg. See https://github.com/robotology/icub-models/issues/12

A PR follows fixing this issue.

jcarpent commented 3 years ago

This is awesome!

olivier-stasse commented 3 years ago

What I should have mentionned is that I did the corrections for TALOS but not for the other robots...

olivier-stasse commented 3 years ago

It looks like that at least the following files would be affected:

kinova.urdf < link name="j2s6s200_end_effector" / >
ur5_joint_limited_robot.urdf: < link name="tool0" / >
ur3_robot.urdf: < link name="tool0" / >
ur10_robot.urdf: < link name="tool0" / >
ur3_joint_limited_robot.urdf: < link name="tool0" / >
romeo.urdf: < link name="l_sole" / >
tiago.urdf: < link name="hand_grasping_frame" / >
tiago_no_hand.urdf: < link name="base_footprint" / >
romeo_laas_small.urdf: < link name="AccelerometerDownTorso_sensor" / >
tiago.urdf: < link name="base_footprint" / >

wxmerkt commented 3 years ago

I performed those fixes for Anymal a while back - but agree we should carry them through all URDFs. I recall at some point some dynamics libraries or simulators did not like "0" inertia either (i.e. crashed) so we went with 1e-6 or 1e-9. I can't recall which where affected though.