Operational space pseudoinverse doesn't actually use regularization parameter

[dmalexa5]

Hi all,

Thanks for making such a clean set of controllers! Very helpful.

Noticed this little bug: the task-space inertia matrix pseudoinverse doesn't actually use the operational_space_regularization parameter, instead only using the default epsilon = 1e-4

    Mx_inv = J * data_.Minv * J.transpose();
    Mx = pseudo_inverse(Mx_inv);

should be

    Mx_inv = J * data_.Minv * J.transpose();
    Mx = pseudo_inverse(Mx_inv, params_.operational_space_regularization);

Is there a reason for this? I see that the default operational_space_regularization parameter is 0.01 and the range is capped to 0.001. But the default epsilon of 1e-4 works just fine on our franka arm.

[danielsanjosepro]

Thanks for catching this, you are right, it should be parametrized as well! This will be fixed in #40!