Decided/calibrate flexibility at the hip on Pyrene

This is a prospective task. Some flex values have to be put in the kinematic model for @tflayols estimator to work. Let's decide which values.

changed title from Decided/calibrate flexibility at the ankle on Pyrene to Decided/calibrate flexibility at the hip on Pyrene

Update: @tflayols and I ran an experiment today to collect data in order to examine the flexibility.

@tflayols, you can find the data from today's experiment in /net/pongo/vol/vol_projects/partage_gepetto/Experiments/Pyrene/2019/2019_06_26/2019_06_26_17_28

Here is the feet distance in [mm] vs the hip torque in [N]

in x-y:

I crosscorelate the two signals to estimate the delay (127ms!). After compensating for it, i got (now in [m ]):

The estimated hip flexibility around the X axis is 1.02753655126 mm/N (to be translated in rad/N with the half-sitting level arm)

The hip torque should be Nm, right? Not N. So it's 1.02753655126 mm/Nm. It seems quite stiff. It corresponds to 973.201390037 Nm/m... Assuming the leg to be 1m long, it would correspond to a torsional stiffness of 973.201390037 Nm/rad. I would have expected less

Yes it's a torque in Nm, my mistake. The K value might not look very flexible but it should be compared with the actual torque at the hip during walking. (60Nm makes a feet distance error of 2cm)

Let's take a torque tracer during the next walking experiment.

For the record : for the experiment, the legs were controlled in position and @tflayols applied manually a force to bring the feet closer.

I am digging a bit this because we are working on Pyrene calibration.

The two figures below represent the angle measured at the motor, the angle measured after the gear ratio (so the real joint angle) and the opposite of the torque (I think at joint level ?) for the left and right hip during the experiments.

@vbonnet We only saved the angles at joint level so we cannot see the effect of the flexibility between the motor and the joint with our measurements.

Hello @jmirabel I'm not exactly sure of what you want to do, but let me precise a little bit what was the test we did at the time. This experiment was to identify the flexibility caused by the supporting structure of the Talos hip. Not only the actuator compliance (gearbox, torque sensor etc..).

To estimate this, we collected the feet distances with the motion capture system. The robot was rigidly controlled in position. By applying a force to the feet, we can deflect the all leg during the experiment. We believed that this flexibility is located not only at the Hip roll-pitch, but at the yaw attachment. Witch meen part of it can not be observed with the joint encoders. (This is why we used the mocap).

We approximate the torque at the hip attachment with the roll-pitch torque measurement, since the level arm between the two is low compared to the leg.

mentioned in commit 5c92aff5