DeepMind Control Suite

In plain words: A set of simulated robot-control tasks built on a physics engine, all sharing the same setup and clear scoring so learning agents can be compared fairly and the tasks easily changed. It runs in Python and includes results from several learning algorithms.

Abstract

The DeepMind Control Suite is a set of continuous control tasks with a standardised structure and interpretable rewards, intended to serve as performance benchmarks for reinforcement learning agents. The tasks are written in Python and powered by the MuJoCo physics engine, making them easy to use and modify. We include benchmarks for several learning algorithms. The Control Suite is publicly available at https://www.github.com/deepmind/dm_control . A video summary of all tasks is available at http://youtu.be/rAai4QzcYbs .

Yuval Tassa, Yotam Doron, Alistair Muldal, Tom Erez, Yazhe Li, Diego de Las Casas, David Budden, Abbas Abdolmaleki, Josh Merel, Andrew Lefrancq, Timothy Lillicrap, Martin Riedmiller
arXiv:1801.00690 · cs.AI · submitted Jan 2, 2018
abstract · pdf · html · 24 pages, 7 figures, 2 tables