[jareklupinski]

with machine vision, it might be feasible to change the learning step to where instead of using joysticks to manipulate the robot through the steps, it can watch you 'do it once' then extrapolate the kinematics for its body just based on the end states of your hands or something