The forces read with the joint force sensor are unreasonably high

Hi,

I’m trying to perform simulation of a real process where forces on the tcp of the robot are being recorded. I found out that on Isaac SIM, although the shape of the force is apparently correct, I am obtaining values for the forces that are excessively high compared with the real setup. Attached you can find two graphs, one with the real data (bottom) from the robot and the other with simulated data (top).

The forces that I am reading are the ones on articulation 10, which I assumed are the ones on the wrist of the robot like the sensor the UR5e has.

The application is basically trying to insert a piece into another while using force control.

Are the values of forces given by get_articulation_force_sensor given in Newtons?

Thank you in advance

Hi @adapa20 - You should use following docs to tune the robots to reflect forces correctly.

https://docs.omniverse.nvidia.com/isaacsim/latest/features/robots_simulation/ext_omni_isaac_gain_tuner.html

https://docs.omniverse.nvidia.com/isaacsim/latest/features/sensors_simulation/isaac_sim_sensors_physics_articulation_force.html#articulation-joint-sensors

May I ask if you are using version 2022.2.1? As far as I know, version 2023 seems to abandon the articulation force sensor, I also want to install a force sensor on the end-effector of the robotic arm to detect the six-dimensional force changes during the assembly process, but the force readings don’t always feel right, how did you add the force sensor? Thanks for the answer!

Please use the latest release

Hi Olive,

I used the last release 2023.1.1. I don’t think this verision is actually abandoning the articulation force sensor because the function it’s still on the Articulation class as @rthaker mentioned.

However, the way I managed the force sensor in the robot was reading the forces on the articulations of the robot (you can read the force at the TCP for instance), but please note that the force is read on the axes of the articulation frame, you may need to transform frames.

Since I developed my program on a Standalone application, I noticed that the key feature to align the behaviour with the real world was essentially the physics_dt (and the rendering). I had to read forces and possitions, so my physics loop was quite slow, therefore it definitelly didn’t match with the real world.

Hi adapa20.

I followed “add” → “physics” on version 2023.1.1 and didn’t see "articulation force sensor ".

Is this how you added it? Actually, I don’t know if I have the sensor mounted in the right place, I created an xform under the joints and added the force sensor as follows (where the “fixed joint” is connected to “tool0” and a peg): ! image|270x500

Is this correct? Or should we just add sensors to the peg rigid body? I feel like the experiment we’re working on should be more similar, it’s an assembly task, what you installed should be correct, please help me out, thanks a lot!

Please use Isaac Sim 4.0

Unfortunately I have not found the articulation force sensor in version 4.0 either, are you sure it is there in this version? I can easily find it in version 2022.2.1, this page on the official website will hopefully be updated if it has changed location or is no longer available.


Articulation force sensors were deprecated in the previous version so they are removed.
Alternative APIs that are supported now are get_measured_joint_forces and get_measured_joint_efforts. See this page for more information on how to use them

I see.Thank you very much!

Have you found a way to implement an end effector force sensor with these functions?

Hi, I would like to ask you how this force curve in your graph is generated?I was able to save the data and then visualise it, but I would prefer to observe the force changes in real time, like an oscilloscope, can this be done in the isaac sim UI?