Reinforcement Learning via Corrective Feedback

interactive imitation learning research

About

This is my bachelor thesis project, where I researched interactive imitation learning techniques. Combined with human corrective feedback, the robots rapidly learn to perform various control tasks in simulation.

This work was jointly (and mainly) supervised by Prof. Youling Yu at the College of Electronics and Information Engineering and was rated as Outstanding Bachelor Thesis at Tongji University, Shanghai.

husky_simulation

husky_simulation

Thesis

The full-text thesis can be found here.

Note that this is a translated version just for demonstration purposes, which is not reviewed.

Supplementary Video