SO-101 Support #1171
SO-101 Support #1171
Conversation
| @@ -337,11 +337,11 @@ | |||
|
|
|||
| <!-- Joint from gripper to moving_jaw_so101_v1 --> | |||
| <joint name="gripper" type="revolute"> | |||
| <origin xyz="0.0202 0.0188 -0.0234" rpy="1.5708 -0.909 -1.41553e-15"/> | |||
| <origin xyz="0.0202 0.0188 -0.0234" rpy="1.5708 -5.24284e-08 -1.41553e-15"/> | |||
There was a problem hiding this comment.
this e-08 value should just be 0?
| <parent link="gripper_link"/> | ||
| <child link="moving_jaw_so101_v1_link"/> | ||
| <axis xyz="0 0 1"/> | ||
| <limit effort="0" velocity="10" lower="-1.13778" upper="1.16081"/> | ||
| <limit effort="10" velocity="10" lower="-0.174533" upper="2.0944"/> |
There was a problem hiding this comment.
we don't use effort but just checking these values of 10 are copied from the original so101 urdf, the new calibrated one here right https://github.com/TheRobotStudio/SO-ARM100/blob/main/Simulation/SO101/so101_new_calib.urdf
| @@ -398,7 +398,7 @@ | |||
| <!-- Joint from upper_arm to lower_arm --> | |||
| <!-- Note: 5-degree calibration offset applied to joint limits --> | |||
| <joint name="elbow_flex" type="revolute"> | |||
| <origin xyz="-0.11257 -0.028 1.73763e-16" rpy="-3.63608e-16 0.03 1.5708"/> | |||
| <origin xyz="-0.11257 -0.028 1.73763e-16" rpy="-3.63608e-16 8.74301e-16 1.4408"/> | |||
| @@ -438,7 +438,7 @@ | |||
|
|
|||
| <!-- Joint from base to shoulder --> | |||
| <joint name="shoulder_pan" type="revolute"> | |||
| <origin xyz="0.0388353 -8.97657e-09 0.0624" rpy="3.14159 4.18253e-17 -3.14159"/> | |||
| <origin xyz="0.0388353 -8.97657e-09 0.0624" rpy="3.14159 4.18253e-17 -3.20159"/> | |||
There was a problem hiding this comment.
e-17 should be 0.
I noticed the official urdfs from the robotstudio used the e-17 values, but we should use 0 (and I'll tell them they should use 0 as well).
| from lerobot.common.motors.motors_bus import MotorNormMode | ||
| from lerobot.common.robots.robot import Robot | ||
| from lerobot.common.utils.robot_utils import busy_wait | ||
| from cameras.camera import Camera |
There was a problem hiding this comment.
this seems like a wrong import
| @@ -58,6 +58,8 @@ def set_target_qpos(self, qpos: Array): | |||
| # NOTE (stao): It seems the calibration from LeRobot has some offsets in some joints. We fix reading them here to match the expected behavior | |||
| if self.real_robot.name == "so100_follower": | |||
| qpos["elbow_flex.pos"] = qpos["elbow_flex.pos"] + 6.8 | |||
| elif self.real_robot.name == "so101_follower": | |||
There was a problem hiding this comment.
this is not needed the elif. In another PR we will also remove the SO100 one.
| @@ -113,6 +115,8 @@ def get_qpos(self): | |||
| # NOTE (stao): It seems the calibration from LeRobot has some offsets in some joints. We fix reading them here to match the expected behavior | |||
| if self.real_robot.name == "so100_follower": | |||
| qpos_deg["elbow_flex"] = qpos_deg["elbow_flex"] - 6.8 | |||
| elif self.real_robot.name == "so101_follower": | |||
There was a problem hiding this comment.
this elif is also not needed
| <?xml version="1.0" ?> | ||
| <robot name="so101"> | ||
|
|
||
| <!-- TODO: Check if this is correct --> |
There was a problem hiding this comment.
can you verify this? Basically run the demo robot code in cpu sim, and open the contact tab and check there aren't any impossible self collisions being detected or considered
|
Is there a reason why the SO101 task is not set to be at the "normalized" 0 position? For consistency we want to reset to this "right angle, 0" position (even though in radians the joints wont be a multiple of np.pi/2). This way in the real world it also easier for users to do alignment as measuring multiples of 90 degrees is much easier. We will do the same treatment for SO100 later as well once we fix that URDF rpy and limits as well 
|
|
Another is we can decompose the mesh for the moving jaw a bit more on the SO101? To achieve the same quality as the SO100 one. You can even use the same collision mesh file probably. Here is SO100 overlaid 
Here is SO101 overlaid at the moment. Red circles mark some discrepancies 
|
|
@Viswesh-N looking at the PR again. The collision meshes are still too off. For the moving jaw you may need two stl files to describe it and add both of them into the urdf, we load these both individually to form a non convex shape. SO100 modelling does this as well. SRDF also needs updating. wrist_link and the moving_jaw_so101_v1_link are marked as potentially colliding when they can't actually collide. |
finally this rest position seems off and not at the expected right angles. Can those be fixed as well?
|
|
One more thing, some of the values that should be 0 are really just those smaller than 1e-16 (at that precision it was probably a typo it was that small). Other values (like inertias which often are 1e-5 to 1e-8 range) should be kept the same. |
No description provided.