ROMar 8, 2017

What Would You Do? Acting by Learning to Predict

arXiv:1703.02658v12 citations
Originality Incremental advance
AI Analysis

This work reduces priors needed for robot task learning compared to existing methods like Learning from Demonstration, addressing a domain-specific challenge in robotics.

The authors tackled the problem of enabling robots to learn tasks directly from visual demonstrations by predicting action outcomes, achieving generalization to unseen states in table-top object manipulation tasks.

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.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes