One-Shot Reinforcement Learning for Robot Navigation with Interactive Replay

By Jake Bruce, Niko Sünderhauf, Piotr Wojciech Mirowski, Raia Hadsell, Michael Milford. 2017

63 citations recorded by the source corpus.

Recently, model-free reinforcement learning algorithms have been shown to solve challenging problems by learning from extensive interaction with the environment. A significant issue with transferring this success to the robotics domain is that interaction with the real world is costly, but training on limited experience is prone to overfitting. We present a method for learning to navigate, to a fixed goal and in a known environment, on a mobile robot. The robot leverages an interactive world model built from a single traversal of the environment, a pre-trained visual feature encoder, and stochastic environmental augmentation, to demonstrate successful zero-shot transfer under real-world environmental variations without fine-tuning.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

A Survey of Deep Network Solutions for Learning Control in Robotics: From Reinforcement to Imitation (2016, neutral): “Bruce et al. (2017) propose a method that enables zero-shot transfer for a mobile robot to learn to navigate to a ﬁxed goal in an environment with variations unseen during the training.”
Learning Deployable Navigation Policies at Kilometer Scale from a Single Traversal (2018, methodology): “These modiﬁcations are simple, but surprisingly effective and computationally efﬁcient ways of providing unlimited variability to the observations of the agent, and stochastic augmentations like these have been shown to be crucial for achieving generalisation beyond the training data [24].”
Learning to Navigate in Cities Without a Map (2018, methodology): “Photographic environments have been used to train agents on short navigation problem in indoor scenes with limited scale [9, 1, 7, 35].”
Reinforced Imitation: Sample Efficient Deep Reinforcement Learning for Mapless Navigation by Leveraging Prior Demonstrations (2018, background): “[23] use interactive experience replay to learn how to navigate in a known environment to a fixed goal from images by traversing it only once.”
Transfer Learning for Related Reinforcement Learning Tasks via Image-to-Image Translation (2018, background): “…learned disentangled representation of the image (Higgins et al., 2017), analogies between sets of instructions (Oh et al., 2017), interactive replay (Bruce et al., 2017) while training and learn general policies by training on multiple tasks in parallel (Espeholt et al., 2018; Sohn et al., 2018).”
Visual Representations for Semantic Target Driven Navigation (2018, background): “More recently, the success of data driven machine learning strategies for a variety of control and perception problems opens new avenues for overcoming the limitations of previous approaches [4], [5], [6], [7], [8].”