DeepMind Control Suite

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 .

24 pages, 7 figures, 2 tables

References in corpus (1)

• Reproducibility of Benchmarked Deep Reinforcement Learning Tasks for Continuous Control

Cited by in corpus (26)

• Challenges of Real-World Reinforcement Learning
• Benchmarking Model-Based Reinforcement Learning
• Latent ODEs for Irregularly-Sampled Time Series
• MinAtar: An Atari-Inspired Testbed for Thorough and Reproducible Reinforcement Learning Experiments
• MushroomRL: Simplifying Reinforcement Learning Research
• Investigating Generalisation in Continuous Deep Reinforcement Learning
• Neural Graph Evolution: Towards Efficient Automatic Robot Design
• Rewriting History with Inverse RL: Hindsight Inference for Policy Improvement
• Value constrained model-free continuous control
• Q-Learning in enormous action spaces via amortized approximate maximization
• Continuous-Discrete Reinforcement Learning for Hybrid Control in Robotics
• TF-Replicator: Distributed Machine Learning for Researchers
• Scheduled Intrinsic Drive: A Hierarchical Take on Intrinsically Motivated Exploration
• Motion Perception in Reinforcement Learning with Dynamic Objects
• PlaNet of the Bayesians: Reconsidering and Improving Deep Planning Network by Incorporating Bayesian Inference
• Acceleration of Actor-Critic Deep Reinforcement Learning for Visual Grasping in Clutter by State Representation Learning Based on Disentanglement of a Raw Input Image
• Learning a Multi-Modal Policy via Imitating Demonstrations with Mixed Behaviors
• Sample-based Distributional Policy Gradient
• Blue River Controls: A toolkit for Reinforcement Learning Control Systems on Hardware
• Model Primitive Hierarchical Lifelong Reinforcement Learning
• An Actor-Critic-Attention Mechanism for Deep Reinforcement Learning in Multi-view Environments
• Arena: a toolkit for Multi-Agent Reinforcement Learning
• Continual Reinforcement Learning with Diversity Exploration and Adversarial Self-Correction
• IMPACT: Importance Weighted Asynchronous Architectures with Clipped Target Networks
• Human-like Planning for Reaching in Cluttered Environments
• Accelerated Target Updates for Q-learning