We propose to learn tasks directly from visual demonstrations by learning to predict the outcome of human and robot actions on an environment. We enable a robot to physically perform a human demonstrated task without knowledge of the thought processes or actions of the human, only their visually observable state transitions. We evaluate our approach on two table-top, object manipulation tasks and demonstrate generalisation to previously unseen states. Our approach reduces the priors required to implement a robot task learning system compared with the existing approaches of Learning from Demonstration, Reinforcement Learning and Inverse Reinforcement Learning.
@article{arxiv.1703.02658,
title = {What Would You Do? Acting by Learning to Predict},
author = {Adam Tow and Niko Sünderhauf and Sareh Shirazi and Michael Milford and Jürgen Leitner},
journal= {arXiv preprint arXiv:1703.02658},
year = {2017}
}
Comments
Submitted to International Conference on Intelligent Robots and Systems (IROS 2017)