Jack De Havas,

I don't know if there are specific ways of detecting these differences from a motor control perspective, but one very simple statistical approach may be to check the variance.  Control can imply a reduction in the variance of the quantity being monitored in the feedback loop.  So, if position data show large variance, whereas velocity data are tightly grouped, you might infer that velocity is being controlled.

I hope that is helpful.

Sean Osis

Thanks Sean, I'll take a look at the variance, hopefully its not too correlated.   

Hallo Jack De Havas,

 as Sean T Osis said looking at the variance is a good idea. There exists an esteblished methodoligy to do this called "unconstrained manifold method". Its idea is to implement a forward model for caluclating the quatity you are interested in if it is controlled or not. The forward model is often based on joint angles. So one idea is to build up such a model for the end effector position and also for the endeffector velocity. Than you can determine in a first step if position or velocity is controlled by the joint angles which is of course the case. Then you can compare and you can also compare for positions before reaching the end position.

Look in the literatur for

UCM = unconstrained manifold method

My experience is that you need at minimum 15 trials to make any variance analysis, so also for UCM you need at minimum 15 trials better more.

Hope that helps.

Oliver Rettig

Thanks Oliver, very helpful. I'll try to make it work with my low number of trials.