Implement HumanControlBoard device to send human data to simulink about human-dynamics-estimation HOT 7 CLOSED

If I understood what @Yeshasvitvs meant, the new HumanControlBoard will have the support of attaching to IHumanState and IHumanDynamics interfaces. This means that it will run within the same yarprobotinterface process (that means same xml file).

This conforms to the architectural choice of HDE v2. As we discussed for all the HDE v2 devices, as soon as we need network transparency we can develop the pairs wrapper-client for each interface. This would allow to run HumanControlBoard on a different process / different machine.

from human-dynamics-estimation.

HumanControlBoard is added to devices. I implemented only the core methods to access data in matlab simulink through controlboardremapper. Accordingly, we get joint positions and joint velocities from IHumanState interface and joint torques from IHumanDynamics interace. Ideally, joint accelerations should also be available through IHumanDynamics interface but currently they are not set in HumanDynamicsEstimator. So, I am setting them to zero in HumanControlBoard device.

On reading <HumanControlBoard>/stateExt:o data we can read correctly the joint positions, velocities, accelerations and torques indicated by [ok] at the end of the data. But the other methods display [fail] as they are not implemented in HumanControlBoard and are not necessary to get human joint data in simulink.

This test is done with offline wearable data. I will test this correctly in simulink controller next with both offline and online data next.

from human-dynamics-estimation.

I added the tests for HumanControlBoard device here and ran the test HumanSimulinkController with offline wearable data from xsens suit device.

Accordingly, with the data set of left knee movement of 90 degrees, I see the correct joint angles in simulink scope

For the right knee movement of 90 degrees, I see the correct joint angles in simulink scope

Also, the joint torques values are read correctly on simulink side from the HumanControlBoard device

NOTE: The joint torque values may not be reasonable with respect to the knee joint angles. This is because the time synchronization differences in the playback of xsens werable data and the ftshoes wearable data.

A final test with online data from suit and shoes with a human subject has to be done before closing this issue. @lrapetti and I will do this test soon and update the issue.

from human-dynamics-estimation.

🔝

Ideally, joint accelerations should also be available through IHumanDynamics interface

From IHumanState.

from human-dynamics-estimation.

During this testing, https://github.com/dic-iit/component_andy/issues/193#issuecomment-461117811 @lrapetti and I tested the new HumanControlBoard device with the human movements as shown below

• A set of arms to T pose 5 times
• A set of two feet squat 5 times

The corresponding joint angle data in degree seen in Matlab scope is shown below

The corresponding joint torques seen in Matlab scope is shown below

So, the simulink controller gets the human joint data faithfully from HumanControlBoard device.

One key details a new user has to be aware of is that in case of feet or knee values and direction being wrong in the Matlab scope. The problem could be feet or knee distortion as observed in this case
https://github.com/dic-iit/component_andy/issues/188#issuecomment-461118542.

We can close this issue now and when we perform another test of HDEv2 with another human subject and the corresponding urdf model, we can update this issue with the Matlab plots from the test.

@DanielePucci @lrapetti

from human-dynamics-estimation.

Discussing with @diegoferigo, I understood this task may be blocked by the HumanDynamicWrapper which for what I understood has not been developed yet. (also I don't know the state of HumanStateWrapper)

from human-dynamics-estimation.

Not necessarily. The first one was designed as a module to work with modules but subscribing to the output ports streaming joint data in thrift format. Now I want to change it to a device called HumanControlBorad that will read joint position, velocity through IHumanState interface and then subscribe to joint torques output port from HumanDynamicsWrapper. Later when we implement HumanDynamicsWrapper to expose data through IHumanDynamics we can update from the logic of reading joint torques port to reading IHumanDynamics interface.

from human-dynamics-estimation.