# Understanding contact sensor data

I am receiving contact sensor measurements that I am having trouble making sense of. One example is the following:

Normal vector:      (-0.347998,  0.937495,  0)
Body1 wrench force: (-0.270598,  1.06599,   0.546762)
Body2 wrench force: (-0.797889, -0.838441, -0.410959)


From what I understood by going through the description of the contact sensor and per this answer by @hsu is that these forces describe the correction force applied to the CG of corresponding links to correct for the interpenetration.

In my case body 1 is a free object and body 2 is a robot hand whose base is fixed to the world. I would like to understand the following:

1. Does the fact that the robot is fixed to world influence the physical interpretation of the results? E.g. for a static object the contact forces that it feels are always 0.
2. Shouldn't the contact forces for body1 and body2 be the same with a different sign?
3. In case 1. is not meaningful, would taking the negative of the force off of the second body make sense? If not, how could I obtain this contact force? What happens in case of multiple contacts from different sides of the object, e.g. during grasping?
4. Why do the directions of the forces differ so much from the normal?
5. What would be the physical interpretation of this sensor in general? If I had a glove with a haptic sensor, would it sense something similar to the body wrench force, or something similar to the contact normal, e.g. contact normal scaled with penetration depth?

Any insights on this would be much appreciated. Thank you!

edit retag close merge delete

Sort by ยป oldest newest most voted

I got some new intuition:

1. This, I think, does not influence the physical interpretation of the results.

The difference in sign might be true for the forces in the world frame. The forces given by the contact sensor are given in the link frame. I still get difference to the normal when I project the forces to the world frame, and am not really sure where this comes from. My guess is that the normal is the surface normal, and the force vector is a vector within the friction cone.

Also, I found that the HAPTIX sensor simulation processes these forces, as shown here. Is this closer to what the actual sensor would sense?

more