Simultaneous dual-input control

[PeterBowman]

Out of two command/sensor parameters, we only perform closed-loop control on one of them (#3): the polar angle (inclinación). Attemps on replicating this procedure on the azimuthal angle (orientación) lead to undesired behavior, which we believe is rooted in the physical coupling between both coordinates.

Different alternatives have been proposed: MIMO, fuzzy (discouraged), neural networks, single-axis control (a.k.a. decoupling control)...

@jgvictores suggested we just give a try to the classical Resolved Motion Rate Control (RMRC) by Whitney. We embraced this solution at roboticslab-uc3m/kinematics-dynamics, see BasicCartesianControl.

[jgvictores]

Googled Decoupling control rpy, found papers such as "Decoupling control for a three-axis inertially stabilized platform used for aerial remote sensing" (https://doi.org/10.1177%2F0142331214557667).

Yes, the RMRC approach seems like a good first option.

[PeterBowman]

Spoken again with @jgvictores, we concluded the RMRC would not allow us to perform continuous control as it depends on trajectory generation with a well-defined end-point (with zero velocity).

[jgvictores]

Spoken again with @jgvictores, we concluded the RMRC would not allow us to perform continuous control as it depends on trajectory generation with a well-defined end-point (with zero velocity).

Not completely sure about this. Think of the task as a change of pose (in the pos+ori sense, even though rather than SE(3) this C-Space is a much more restricted subset of SO(3)):

• Target is a differential away: use the Jacobian (small enough increment * timestep also works)!
• The distance to the target is larger than a differential (or than a small enough increment): generate a trajectory in the C-Space and use RMRC!