Closed lrapetti closed 2 years ago
An initial implementation can be found in https://github.com/robotology/human-dynamics-estimation/commit/21bbb5dac4e887f37502740576f2197245a6faca. The implementation has been done in the DynamicalInverseKinematics
class but it is not properly used in the HumanStateProvider
(in that branch the z components only is used for all the orientation targets, instead we should define a new type of target).
In the same branch I am also taking care of selecting a single component for the position targets.
Feature merged onto the main development branch add-different-target-types
, we can close this issue.
In particular, we are interested in aligning the measured gravity vector.
A possible reference for this can be eq.11 from https://www.sciencedirect.com/science/article/pii/S0005109814006104 by @DanielePucci. In a nutshell, the following control law can be used: $$ \omega = -k (z^{meas} \times z) + \omega^{meas} $$ with $k>0$