Some convention questions

[edward9503]

Hi @qiayuanliao ,

Thanks for sharing the code!

When I use the code and go through it, I find some questions about the convention of some vectors. More specifically:

1. For the state vector in the Kalman filter (which is called "rbd_state_" in your code), if I understand correctly, it looks like this:

Base_orient_z_global
Base_orient_y_global
Base_orient_x_global
Base_posi_x_global
Base_posi_y_global
Base_posi_z_global
LF_HAA_posi
LF_HFE_posi
LF_KFE_posi
LH_HAA_posi
LH_HFE_posi
LH_KFE_posi
RF_HAA_posi
RF_HFE_posi
RF_KFE_posi
RH_HAA_posi
RH_HFE_posi
RH_KFE_posi
angular_vel_z_global
angular_vel_y_global
angular_vel_x_global
linear_vel_x_global
linear_vel_y_global
linear_vel_z_global
LF_HAA_vel
LF_HFE_vel
LF_KFE_vel
LH_HAA_vel
LH_HFE_vel
LH_KFE_vel
RF_HAA_vel
RF_HFE_vel
RF_KFE_vel
RH_HAA_vel
RH_HFE_vel
RH_KFE_vel

Is this order/convention correct?

2. For the foot contact measurement, is the order like this? "LF_FOOT", "RF_FOOT", "LH_FOOT", "RH_FOOT"

3. When you do the pinocchio::forwardKinematics in "wbc.cpp", I found that the convention of expected measured_v_ is different, as you are using global linear velocity as the linear part, and Euler angle ZYX's derivatives as the angular part. However, as you can see here, a local base linear and angular velocity are expected as the first six elements. So I wonder why you can do that in a different way?

4. Following the idea from 3., previously when I used Pinocchio to load the URDF file of Anymal C, the dimension of model.nq is 18 + 1 = 19, because it uses quaternion to express the orientation. But I notice that you can use Euler angle to present the orientation and model.nq in your case returns 18. I was wondering what did you do to achieve it?

5. Again following the idea from 3. and 4., let's say I use pinocchio::getFrameJacobian(model, data, info_.endEffectorFrameIndices, pinocchio::LOCAL_WORLD_ALIGNED, jac) to get the jacobian jac of the FL_Foot in the global frame. If I wanna calculate the global velocity of the FL_Foot, I need to use xdot = jac * v for the calculation. Then, what is the convention of the v in this case? Is it v = [local_base_velocity_linear, local_base_velocity_angular, joint_velocities] as described here, or v = [global_base_velocity_linear, global_base_ZYX_derivatives, joint_velocities] followed by your convention?

Thanks in advance.

[edward9503]

Hi @qiayuanliao , is there any feedback on this? Thanks

[qiayuanl]

Hi @qiayuanliao , is there any feedback on this? Thanks

Sorry for the delay. I will answer these questions after October 29.

[edward9503]

Ok no worry

[edward9503]

Hi @qiayuanliao ,

is there any feedback on this? Many thanks in advance.

[qiayuanl]

Very sorry for the delay. My TOEFL test was rescheduled to Nov. 1st (yesterday) cause of COVID-19.

Since there are no drafts in the GitHub issue comments. I posted this comment before I finished, and I will update this comment gradually. I suggest that you ask questions after I am finished.

Overall, there are three different representations of state and input in the legged_control which are all brought by OCS2 (not my fault).

• OCS2 centroidal dynamics system input and output defined can be found in the comments of task.info
• Pinocchio generalize coordinate
• Rigid body state following the definition you mentioned.

Answers to your questions:

1. The convention is correct.
2. Defined by OCS2
3. What your are saying is correct. Thank you so much for finding this bug! I will fix it within two days.
4. Check this code

[edward9503]

Hi @qiayuanliao,

I hope you do well in your exam.

According to your reply, I suppose you are answering question 1. 3. and 4. I asked before. But I need to confirm two things:

1. For the variable "rbd_state_" in state estimator, is the convention correct?

 Base_orient_z_global
Base_orient_y_global
Base_orient_x_global
Base_posi_x_global
Base_posi_y_global
Base_posi_z_global
LF_HAA_posi
LF_HFE_posi
LF_KFE_posi
LH_HAA_posi
LH_HFE_posi
LH_KFE_posi
RF_HAA_posi
RF_HFE_posi
RF_KFE_posi
RH_HAA_posi
RH_HFE_posi
RH_KFE_posi
angular_vel_z_global
angular_vel_y_global
angular_vel_x_global
linear_vel_x_global
linear_vel_y_global
linear_vel_z_global
LF_HAA_vel
LF_HFE_vel
LF_KFE_vel
LH_HAA_vel
LH_HFE_vel
LH_KFE_vel
RF_HAA_vel
RF_HFE_vel
RF_KFE_vel
RH_HAA_vel
RH_HFE_vel
RH_KFE_vel

2. For the statement I mentioned in 2.: "When you do the pinocchio::forwardKinematics in "wbc.cpp", I found that the convention of expected measured_v_ is different, as you are using global linear velocity as the linear part, and Euler angle ZYX's derivatives as the angular part.". Is it also correct?

Thanks

[ImFroooost]

Hello, has this problem been fixed in the code? I can't find the relevant commit, thank you! @qiayuanliao

[qiayuanl]

Hello, has this problem been fixed in the code? I can't find the relevant commit, thank you! @qiayuanliao

Hello, I am sure that the original code is correct after checking the code and having a private conversation with @edward9503.

[elpimous]

Hello, could anyone tell me legs order ?? I made front right/front left/rear right and rear left. (and it doesn't work as expected !) Can anyone confirm @qiayuanliao @ImFroooost @edward9503 @LoveThinkinghard Thanks friends Vincent