Inverse Reinforcement Learning with Natural Language Goals

Abstract: Humans generally use natural language to communicate task requirements to each other. Ideally, natural language should also be usable for communicating goals to autonomous machines (e.g., robots) to minimize friction in task specification. However, understanding and mapping natural language goals to sequences of states and actions is challenging. Specifically, existing work along these lines has encountered difficulty in generalizing learned policies to new natural language goals and environments. In this paper, we propose a novel adversarial inverse reinforcement learning algorithm to learn a language-conditioned policy and reward function. To improve generalization of the learned policy and reward function, we use a variational goal generator to relabel trajectories and sample diverse goals during training. Our algorithm outperforms multiple baselines by a large margin on a vision-based natural language instruction following dataset (Room-2-Room), demonstrating a promising advance in enabling the use of natural language instructions in specifying agent goals.

06/12/2020

Ossama Ahmed, Frederik Träuble, Anirudh Goyal and
Alexander Neitz, Manuel Wuthrich, Yoshua Bengio, Bernhard Schoelkopf, Stefan Bauer

Machine Learning, Transfer, Adaptation, Multi-task Learning, Reinforcement Learning, Incremental Learning, Learning in Robotics

11:02

16/11/2020

Towards General and Autonomous Learning of Core Skills: A Case Study in Locomotion

Roland Hafner, Tim Hertweck, Philipp Kloeppner and
Michael Bloesch, Michael Neunert, Markus Wulfmeier, Saran Tunyasuvunakool, Nicolas Heess, Martin Riedmiller

Models for code, Differentiable program generator, Combinatorial optimization, Program obfuscation, Adversarial computer programs, Machine Learning (ML) for Programming Languages (PL)/Software Engineering (SE)

6:27

18/07/2021

Domain Adaption, Third-Person Imitation, Observational Imitation, Reinforcement Learning, Machine Learning, Mutual Information, Imitation Learning

4:51

06/12/2020

Learning Physical Common Sense as Knowledge Graph Completion via BERT Data Augmentation and Constrained Tucker Factorization

ACNMP: Skill Transfer and Task Extrapolation through Learning from Demonstration and Reinforcement Learning via Representation Sharing

Agent-based and Multi-agent Systems, Multi-agent Learning, Knowledge Representation Languages, Logics for Knowledge Representation, Reasoning about Knowledge and Belief

9:57