Iterative Learning From Novice Human Demonstrations for Output Tracking

This paper studies human-guided iterative learning control for output-trajectory tracking tasks when the controller does not have direct access to the desired output and needs to infer it from human demonstrations. The main challenge is to develop such demonstration-based learning in situations where the novice human user might not be able to achieve precision tracking. The major contribution of this study is an inversion-based approach to infer the human intent and use it to design the input update for improving the output tracking. Additionally, conditions are developed for convergence of the proposed iterative output-tracking control for linear time-invariant systems. The proposed scheme is illustrated with a human-in-the-loop tracking experiment with nine human subjects. Results for each of the subjects show that the proposed human-guided learning scheme allows the machine to learn from the demonstrations and capture most of the human effort (about 90%). Moreover, the proposed approach achieves greater tracking precision (about 60% reduction in the average error) than what was achieved by the human alone.