Floating base robot q and v

[xinsongyan]

Hi,

I am using pinocchio to calculate a floating base robot kinematics and dynamics, but I am not sure about the convention of q and v?

I used RBDL before, the q and v in RBDL are defined as:

q = [global_base_position, global_base_quaternion_x, global_base_quaternion_y, global_base_quaternion_z, joint_positions, global_base_quaternion_w]
v = [global_base_velocity_linear, local_base_velocity_angular, joint_velocities]

I tried a little bit pinocchio, it seems that it takes:

q = [global_base_position, global_base_quaternion, joint_positions]
v = [local_base_velocity_linear, local_base_velocity_angular, joint_velocities]

I didn't find any documentation about this. Anyone can confirm this?
Thanks a lot!

[jcarpent]

You're correct. The base translation part is expressed in the parent frame (here the world coordinate system) while its velocity is expressed in the body coordinate system.

[edward9503]

You're correct. The base translation part is expressed in the parent frame (here the world coordinate system) while its velocity is expressed in the body coordinate system.

Hi @jcarpent ,

as you mentioned here, for the floating base robot, the first 6 elements of the generalized velocity vector v are always the linear and angular velocity of the robot's base represented in the floating base frame for the pinocchio's convention.

Then, I have following two questions:

1. For instance, I want to use the pinocchio::ccrba to compute the centroidal momentum matrix. This function expects a joint velocity vector v as input. So for the first 6 elements of this vector, I always need to change the 6d spatial velocity of robot's base to be represented in the base frame, right? That's what I understand from this section.
2. I've learned that the centroidal momenta h_com in the CoM frame which is aligned with the world frame can be computed by pinocchio::computeCentroidalMomentum. And, pinocchio::ccrba computes the centroidal momentum matrix A. If I had a velocity vector v as mentioned in the first question, where the first 6 elements of v are always the linear and angular velocity of the robot's base represented in the base frame, then is "h_com = A * v" correct?

Many thanks in advance.

[FenglongSong]

Hi Justin @jcarpent , thanks for your nice explanation. I'd like to ask about the case when the orientation of the floating-base frame is expressed by ZYX Euler angle.

1. By my understanding, when creating the PinocchioInterface, if I use a free flyer joint, the base orientation will be automatically represented by a quaternion, because I found model.nq = model.nv+1. But if I use a JointModelTranslation() and a JointModelSphericalZYX() as the 6-DOF root joint rather than using a free flyer joint, I got model.nq=model.nv. So I guess now the Pinocchio model orientation is represented by ZYX Euler angles. In this case, are the following interpretation of q and v correct?
   q = [global_base_position, global_base_ZYX_euler_angles, joint_positions]
v = [local_base_velocity_linear, local_base_velocity_angular, joint_velocities]
2. If I use aba() to calculate the forward dynamics, how is the returned acceleration vector defined? Is it qdd=[local_base_acceleration_linear, local_base_acceleration_angular, joint_accelerations] (nothing but the time derivative of velocity vector) ?