Baxter at the APC: Avoid hard coded joint angles, velocity, and torques

[k-okada]

Dear Baxter Researchers,

I want to take this opportunity to send you an update/reminder concerning your task set up for Baxter at the Amazon Picking Challenge.

The three Rethink Baxter robots that we will supply in Seattle will be slightly different from your own. Each supplied robot will have a different calibration than the robot you've been using in your lab, so hard coding joint angle poses, torques, or velocities is not advisable.

Additionally, each torso has one hand-welded joint per arm, and the transform between torso and arm is measured at the time of manufacture.  Each robot will broadcast this on its param server via the parameters:

/robot_config/jcb_urdf_config/torso_t0/joint/<left/right>_torso_arm_mount/origin/<xyz/rpy>

This parameter replaces the *origin* element in the <side>_torso_arm_mount Fixed Joint in Baxter's urdf:

line 975: <joint name="left_torso_arm_mount" type="fixed">
and
line 1109: <joint name="right_torso_arm_mount" type="fixed">

Hopefully knowledge of these Baxter-quirks will prevent any movement-offset-surprises on the competition field.

If you have any questions or concerns about the above (or anything else for that matter) please let me know.  I'm happy to set up a conf call with you to discuss strategies to mitigate the risk of this causing any set up issues during the challenge.

-Brian

[aginika]

Does this mean we need to prepare some way to calibrate? If the calibration is wrong, I think it will be so hard move the arm to the target position...?

[k-okada]

The problem is how much the joint angle differ for each robots, if that is very small, for example if that is around 1cm we can ignore, if that is large, we'd better to do some visual feedback

◉ Kei Okada

On Sun, Apr 12, 2015 at 7:13 PM, Yuto Inagaki notifications@github.com wrote:

Does this mean we need to prepare some way to calibrate? If the calibration is wrong, I think it will be so hard move the arm to the target position...?

— Reply to this email directly or view it on GitHub https://github.com/start-jsk/2014-semi/issues/198#issuecomment-92028034.

[aginika]

So, we should prepare visual feedback system?

[k-okada]

• put some markers or textures in the custom gripper
• recognize shelf pose

We also ca check how each baxter differs using a robot at nihon bainary

◉ Kei Okada

2015/04/14 22:52、Yuto Inagaki notifications@github.com のメッセージ:

So, we should prepare visual feedback system?

— Reply to this email directly or view it on GitHub.

[wkentaro]

Who should be assigned for this?

[aginika]

特にここに関してこの２日でやる予定はないです。 あるとしたら、本番でキャリブレーションを一度やりましょうというところでしょうか？