Improving Kinesthetic Teaching of Fine Tasks Using a Teaching Agent

In this study, programming by demonstration (PbD) teaching agent was proposed. PbD of fine movements proves to be difficult since sub-millimetre positioning accuracy is required. The agent presented in this study provided users with the visual enhancement feature in order to reduce the positioning error of demonstrations. Furthermore, an optimization of Gaussian mixture models (GMM) components positioning is presented, which allows for better detection and learning of relevant sub-millimetre movements within a larger movement.

Automated summary

A programming by demonstration teaching agent is proposed in this study, which utilizes visual enhancement feature and optimization of Gaussian mixture models (GMM) components positioning to achieve accurate detection and learning of sub-millimetre movements.