Practical Robot Learning from Demonstrations using Deep End-to-End Training
This addresses the need for practical robot learning in human environments, though it is incremental as it applies existing deep learning methods to a specific robotic task.
The paper tackled the problem of enabling robots to learn behaviors efficiently by training a ResNet variant to map camera images to end-effector velocities using human demonstrations via joystick, achieving task learning in under an hour with as little as 16 minutes of demonstrations.
Robots need to learn behaviors in intuitive and practical ways for widespread deployment in human environments. To learn a robot behavior end-to-end, we train a variant of the ResNet that maps eye-in-hand camera images to end-effector velocities. In our setup, a human teacher demonstrates the task via joystick. We show that a simple servoing task can be learned in less than an hour including data collection, model training and deployment time. Moreover, 16 minutes of demonstrations were enough for the robot to learn the task.