Random Generalized Coordinates: Normalization of quaternion

[iamchoking]

I understood that the objective is to write a function for the LH foot position given any gc value.

Can we assume that the randomized gc will have a normalized quaternion (gc[3] ~ gc[5])?
Or, should we also follow certain convention for normalizing them?