Hello Nengo Forums!
I have been wanting to implement the basal ganglia-thalamo-cortical pathway into the REACH model (published in a paper by Travid DeWolf et al. (2016)) but have been struggling with how I can do this as it seems (to me at least) that there is one dynamic movement primative which controls some specified trajectory (i.e., all 8 reaches in order). However, with the basal ganglia and thalamus pathway acting as an action selection mechanism, I was wanting to have each one of the 8 reaches individually defined and then some way to select a specific reach.
I think having the PMC generate ALL possible trajectories makes sense. Then some visual input causes the selection of the appropriate trajectory (within PMC) which is then sent (in task space) via BG and THA to M1 where it is then transformed to joint space and executed. Without a selected trajectory the excitation from THA to M1 is absent, only when a motor program is selected is there disinhibition of THA thus excitation of M1 (while trajectory information is sent to M1).
There could be an entirely new pathway implemented which goes: PMC → BG → THA → M1 where now the trajectory is defined and given to the PMC in the same way but the action selection process decides on whether to “go” (and execute this trajectory) or “nogo” (and inhibit the abiltiy of M1 to execute that trajectory), sort of like in response inhibition models.
Two different approaches, neither of which seem obvious to be as to how to actually implement them, despite trying. If anyone has any advice on which approach is better and/or how to actually implement this bg-tha-M1 pathway, that would be greatly appreciated!!