I would like to get some ideas on how to measure the force and torque between body and body. #1998

2-woo-10 · 2024-04-16T07:54:59Z

2-woo-10
Apr 16, 2024

Hi, I'm a student and I'm trying to use MuJoCo for estimation of the state of holding an object/

I'm looking for some help with how you can measure sensor values

Here is a model which explains my question:

Fingers with the following sensors were modeled.

Currently, the orange part is the force/torque sensor.
I would like to use mujoco simulator to measure Force and torque in simulator like real environment.

Is there a way to implement this using mujoco simulator?
I would really appreciate it if you could tell me how.

Here is a screenshot / video, illustrating my question:

Balint-H · 2024-04-16T14:07:35Z

Balint-H
Apr 16, 2024
Collaborator

Hello,

If I understand correctly, you want to measure the force/torque between the parent and child bodies (the two "phalanges" in this case)? There are sensors to do exactly that! https://mujoco.readthedocs.io/en/stable/XMLreference.html#sensor-force

If you add them to the MJCF model of your hand, their readings will be available to you each simulation step.

0 replies

2-woo-10 · 2024-04-18T08:22:05Z

2-woo-10
Apr 18, 2024
Author

Hello,

Thank you for letting me know the good features.

As you said, I simplified the model and experimented with it, can you tell me why the force and torque are taken abnormally?

I will show you my model and mujoco simulator sensor data.

XML:

    <body name="Test_body_1" pos="0.1 0.1 0.1">
        <geom name="Test_body_1" size="0.05 0.05 0.05" type="box" rgba="1 0 0 0.5"/>
        <site name="ft_sensor" type="box" pos="0 0 0.05" />
        <body name="Test_body_2" pos="0.0 0.0 0.1">
            <geom name="Test_body_2" type="box" size="0.05 0.05 0.05" rgba="0 0 1 0.5"/>        
        </body>
    </body>


</worldbody>

<sensor>
    <force name="force_sensor" site="ft_sensor"/>
    <torque name="torque_sensor" site="ft_sensor"/>
</sensor>

0 replies

Balint-H · 2024-04-18T08:41:05Z

Balint-H
Apr 18, 2024
Collaborator

Hello, the site should be on the child body, check out the "site" part of the force sensor documentation for details

…

________________________________ From: rororofo2 ***@***.***> Sent: Thursday, April 18, 2024 9:22:27 AM To: google-deepmind/mujoco ***@***.***> Cc: Hodossy, Bálint K ***@***.***>; Comment ***@***.***> Subject: Re: [google-deepmind/mujoco] I would like to get some ideas on how to measure the force and torque between body and body. (Issue #1597) This email from ***@***.*** originates from outside Imperial. Do not click on links and attachments unless you recognise the sender. If you trust the sender, add them to your safe senders list<https://spam.ic.ac.uk/SpamConsole/Senders.aspx> to disable email stamping for this address. Hello, Thank you for letting me know the good features. As you said, I simplified the model and experimented with it, can you tell me why the force and torque are taken abnormally? I will show you my model and mujoco simulator sensor data. XML: <body name="Test_body_1" pos="0.1 0.1 0.1"> <geom name="Test_body_1" size="0.05 0.05 0.05" type="box" rgba="1 0 0 0.5"/> <site name="ft_sensor" type="box" pos="0 0 0.05" /> <body name="Test_body_2" pos="0.0 0.0 0.1"> <geom name="Test_body_2" type="box" size="0.05 0.05 0.05" rgba="0 0 1 0.5"/> </body> </body> </worldbody> <sensor> <force name="force_sensor" site="ft_sensor"/> <torque name="torque_sensor" site="ft_sensor"/> </sensor> image.png (view on web)<https://github.com/google-deepmind/mujoco/assets/109202740/69880ea6-2a48-4053-a906-b3a34ee339c1> — Reply to this email directly, view it on GitHub<#1597 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AJZVOK3YODXH7GZEE37ZJ2LY557EHAVCNFSM6AAAAABGIZTVXCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDANRTGMYDKMJQGY>. You are receiving this because you commented.Message ID: ***@***.***>

0 replies

2-woo-10 · 2024-04-18T09:31:06Z

2-woo-10
Apr 18, 2024
Author

**Hello, Is this what you said correct?

    <body name="Base" pos="0.1 0.1 0.1">
        <geom name="Base"  mass="1"  size="0.05 0.05 0.05" type="box" rgba="1 0 0 0.5"/>
        <body name="Parent_Body" pos="0.0 0.0 0.055">
            <geom name="Parent_Body"  mass="10e-6"  size="0.05 0.05 0.005" type="box" rgba="0 1 0 0.5"/>
            <site name="ft_sensor" type="box" pos="0 0 0.0" />
            <body name="Child_Body" pos="0.0 0.0 0.055">
                <!-- <site name="ft_sensor" type="box" pos="0 0 0.0" /> -->
                <geom name="Child_Body" type="box" mass="1" size="0.05 0.05 0.05" rgba="0 0 1 0.5"/>        
            </body>
        </body>
    </body>**

0 replies

Balint-H · 2024-04-18T09:36:26Z

Balint-H
Apr 18, 2024
Collaborator

Yeah, except remove the site from the parent body and uncomment it in the child body? Or does that give you an error?

0 replies

2-woo-10 · 2024-04-18T10:26:13Z

2-woo-10
Apr 18, 2024
Author

Yeah, The sensor value is still strange

Can't I use the mocap to apply external force to the geom?

<body name="Base" pos="0.1 0.1 0.1">
    <geom name="Base"  mass="1"  size="0.05 0.05 0.05" type="box" rgba="1 0 0 0.5"/>
    <body name="Parent_Body" pos="0.0 0.0 0.055">
        <geom name="Parent_Body"  mass="10e-6"  size="0.05 0.05 0.005" type="box" rgba="0 1 0 0.5"/>

        <body name="Child_Body" pos="0.0 0.0 0.055">
            <site name="ft_sensor" type="box" pos="0 0 0.0" />
            <geom name="Child_Body" type="box" mass="1" size="0.05 0.05 0.05" rgba="0 0 1 0.5"/>        
        </body>
    </body>
</body>

0 replies

Balint-H · 2024-04-18T10:49:25Z

Balint-H
Apr 18, 2024
Collaborator

Trying your MWE gives correct results as expected. The huge values show on your figure makes me think there's some collision issue. Maybe a plane oriented the wrong way? You can visualize contact points in Simulate, or try disabling collisions in the Physics options.

0 replies

2-woo-10 · 2024-04-28T12:25:41Z

2-woo-10
Apr 28, 2024
Author

Hello, I'm sorry for the late thank you for your response.

It helped me a lot through sincere answers.

I have an additional question, so I'd like to ask you, is it possible?

I added the ft sensor, and even if the weight of the object changes while the gripper is closed, the data measured in the ft sensor appears the same.
For example, if a 1kg object holds and lifts an object when the joint is free, and if there is no joint and the object does not move, it will slip. I want to know why the data appear the same.

0 replies

Balint-H · 2024-04-28T12:36:31Z

Balint-H
Apr 28, 2024
Collaborator

Glad you are making progress. Could you please rephrase your point 1? I had trouble understanding it.

Could you please share the xml, or describe the hierarchy and where the sensor is?

0 replies

2-woo-10 · 2024-04-28T12:49:18Z

2-woo-10
Apr 28, 2024
Author

Currently, a force/Torque sensor has been attached to that part.

The experiment I want to proceed with is that when I hold the object with the gripper and the gripper moves up and down, I want to compare the ft sensor data.

Case 1 :

<body pos="0.13 0 0.08">
  <joint type="free"/>
  <geom name="object" type="box" mass="5" size="0.015 0.015 0.08" rgba="0.1 0 0.5 0.1"/>
</body>

The object was declared as follows, and the data were obtained when the gripper held the object stably and moved upward.

Case 2 :

<body pos="0.13 0 0.08">
  <geom name="object" type="box" mass="5" size="0.015 0.015 0.08" rgba="0.1 0 0.5 0.1"/>
</body>

This time, the object is fixed, and I got the data that appears when the gripper holds and moves the object. (In this case, the object does not move and only the gripper moves -> Slip)

In this case, the data appears the same, and I wonder if it is measured correctly..

Additionally, is it correct that the data appear similar even if the experiment is conducted in the same way by changing the mass of the object?
For example, when there are 1kg and 5kg objects, holding the gripper with the same grasping force and moving the gripper upward, the data on the f/t sensor is the same.

0 replies

Balint-H · 2024-05-07T09:54:16Z

Balint-H
May 7, 2024
Collaborator

Well, if the object is fixed, you expect the same data measured with any mass (as the slip conditions would stay the same). Do you still see the same measurements when the object is free as well and lifted?

0 replies

2-woo-10 · 2024-05-10T06:32:01Z

2-woo-10
May 10, 2024
Author

Yes.
In the case of a free object, it also shows the same measurement values.

0 replies

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

I would like to get some ideas on how to measure the force and torque between body and body. #1998

{{title}}

Replies: 12 comments

{{title}}

{{editor}}'s edit

{{editor}}'s edit

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

{{editor}}'s edit

{{editor}}'s edit

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

Select a reply

I would like to get some ideas on how to measure the force and torque between body and body. #1998

2-woo-10 Apr 16, 2024

Replies: 12 comments

Balint-H Apr 16, 2024 Collaborator

2-woo-10 Apr 18, 2024 Author

Balint-H Apr 18, 2024 Collaborator

2-woo-10 Apr 18, 2024 Author

Balint-H Apr 18, 2024 Collaborator

2-woo-10 Apr 18, 2024 Author

Balint-H Apr 18, 2024 Collaborator

2-woo-10 Apr 28, 2024 Author

Balint-H Apr 28, 2024 Collaborator

2-woo-10 Apr 28, 2024 Author

Balint-H May 7, 2024 Collaborator

2-woo-10 May 10, 2024 Author

2-woo-10
Apr 16, 2024

Balint-H
Apr 16, 2024
Collaborator

2-woo-10
Apr 18, 2024
Author

Balint-H
Apr 18, 2024
Collaborator

2-woo-10
Apr 18, 2024
Author

Balint-H
Apr 18, 2024
Collaborator

2-woo-10
Apr 18, 2024
Author

Balint-H
Apr 18, 2024
Collaborator

2-woo-10
Apr 28, 2024
Author

Balint-H
Apr 28, 2024
Collaborator

2-woo-10
Apr 28, 2024
Author

Balint-H
May 7, 2024
Collaborator

2-woo-10
May 10, 2024
Author