Flatland: a Lightweight First-Person 2-D Environment for Reinforcement Learning

Abstract : Flatland is a simple, lightweight environment for fast prototyping and testing of reinforcement learning agents. It is of lower complexity compared to similar 3D platforms (e.g. Deep-Mind Lab or VizDoom), but emulates physical properties of the real world, such as continuity, multi-modal partially-observable states with first-person view and coherent physics. We propose to use it as an intermediary benchmark for problems related to Lifelong Learning. Flatland is highly customizable and offers a wide range of task difficulty to extensively evaluate the properties of artificial agents. We experiment with three reinforcement learning baseline agents and show that they can rapidly solve a navigation task in Flatland. A video of an agent acting in Flatland is available here: https://youtu.be/I5y6Y2ZypdA.
Document type :
Conference papers
Complete list of metadatas

Cited literature [34 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-01951945
Contributor : David Filliat <>
Submitted on : Tuesday, December 11, 2018 - 5:27:01 PM
Last modification on : Wednesday, July 3, 2019 - 10:48:05 AM
Long-term archiving on : Tuesday, March 12, 2019 - 3:49:38 PM

File

1809.00510 (1).pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01951945, version 1

Citation

Hugo Caselles-Dupré, Louis Annabi, Oksana Hagen, Michael Garcia-Ortiz, David Filliat. Flatland: a Lightweight First-Person 2-D Environment for Reinforcement Learning. Workshop on Continual Unsupervised Sensorimotor Learning, ICDL-EpiRob 2018, Sep 2018, Tokyo, Japan. ⟨hal-01951945⟩

Share

Metrics

Record views

40

Files downloads

132