DeepMind Control Suite
About
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• 2018
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Cartpole Swing Up | DeepMind Control Suite Cartpole standard (test) | Mean Final Performance862 | 4 | |
| Finger Spin | DeepMind Control Suite Finger standard (test) | Mean Final Performance985 | 4 | |
| Reacher Easy | DeepMind Control Suite Reacher standard (test) | Final Performance Score967 | 4 | |
| Cup Catch | DeepMind Control Suite Ball in Cup standard (test) | Mean Final Performance980 | 4 | |
| Walker Walk | DeepMind Control Suite Walker standard (test) | Mean Final Performance968 | 4 | |
| Cheetah Run | DeepMind Control Suite Cheetah standard (test) | Final Performance524 | 4 |
Showing 6 of 6 rows