Robot-assisted acquisition of a motor skill: Evolution of performance and effort

Abstract

Robots are widely used to promote the neuromotor recovery of stroke survivors, but it is unclear whether robots might be useful to facilitate the acquisition of novel motor skills. In principle, robots could be used to guide a trainee to experiment the correct movements and/or by preventing him/her from performing incorrect ones (the `guidance' hypothesis). Here we investigate whether and on what circumstances physical interaction with a robot may facilitate the acquisition of a novel motor skill. We focused on a simulated putting task, consisting of gently hitting an object (e.g. a `ball') by means of a tool (the `pad', e.g. the golf putter) to move it to a desired final position. A virtual environment, created through a planar robot manipulandum and a computer screen, was used to simulate the physics involved in the putting task. Putting is a redundant task, as the same final position of the ball can be obtained by different combinations of pad velocity, acceleration and point of impact. Two groups of subjects were analyzed: in an `assisted' group, the robot guided subjects toward the correct movement, whereas a control group performed the task without assistance. In both groups we looked at the subjects' performance and its evolution with exercise at several levels of description, namely: (i) final error (distance between final ball position and center of the target area); (ii) ball velocity just after impact; and (iii) hand position and velocity just before impact. In all cases, we looked at both mean value and variability (variance). We found that guidance is helpful in decreasing longitudinal error (a matter of speed accuracy), but not directional error (a matter of position accuracy). These results are consistent with the notion that guidance can help with the dynamic, but not the geometric components of a task