Efficacy of shared-control guidance paradigms for robot-mediated training

Robot-mediated training has seen a plethora of implementations of shared-control haptic guidance, intended to teach novices to perform dynamic tasks by providing them with real-time visual and haptic feedback from real or virtual experts. The efficacies of these paradigms are difficult to quantify and compare, as the paradigms have typically been developed in an ad-hoc manner to suit specific devices and tasks. This work proposes a novel guidance paradigm taxonomy intended to help classify and compare the multitude of implementations in the literature, as well as a revised proxy rendering model to allow for the implementation of two relatively novel guidance paradigms (in addition to existing paradigms). The efficacies of these two paradigms, plus two more paradigms representing the vast majority of implementations in the literature, are compared in a controlled study with 50 healthy subjects. The results show that none of these paradigms are superior to visual-only guidance (the control condition), and that the two newer paradigms are actually detrimental to training for a target-hitting task. These results contradict many intuitions about how haptic guidance should be implemented and raise doubts about the efficacies of the most commonly-implemented paradigms.