Human learning dynamics during teleoperation of surgical robots

Abstract

Robot-assisted minimally invasive surgery requires operators to learn precise motor control. In this paper, we examined how task complexity and the human-robot motion scaling ratio affect teleoperation learning in a ring-through-rail task implemented on the da Vinci Research Kit. Human participants completed training, retention, and transfer phases across different rail shapes and motion scaling ratios. In parallel, a linear-quadratic-Gaussian (LQG) controller performed the same task by combining state estimation with optimal feedback control. Humans and the model showed qualitatively similar patterns of improvement with practice and generalization across complexity. These results suggest that a simple control theory model can capture features of human teleoperation learning, providing a mechanistic model to predict how motor skill transfers and how robotic interface settings influence motor learning.