DeepMind Control Suite

The DeepMind Control Suite is a benchmark environment used to evaluate reinforcement learning (RL) agents on a variety of continuous control tasks, primarily focusing on improving the robustness and generalization capabilities of these agents, especially when dealing with visual inputs. Current research emphasizes developing more efficient and robust RL algorithms, including model-based approaches like DreamerV3 and its variants (e.g., MuDreamer), and incorporating techniques like contrastive learning and self-supervised learning to improve representation learning from visual data. This research is significant because it pushes the boundaries of RL's ability to handle complex, real-world scenarios, ultimately contributing to the development of more adaptable and reliable AI agents for robotics and other applications.

Papers

November 28, 2022

Hypernetworks for Zero-shot Transfer in Reinforcement Learning
Sahand Rezaei-Shoshtari, Charlotte Morissette, Francois Robert Hogan, Gregory Dudek, David Meger
Reinforcement Learning Zero Shot Transfer Attention Hypernetworks DeepMind Control Suite Optimal Reinforcement Learning Meta RL

August 2, 2022

Implicit Two-Tower Policies
Yunfan Zhao, Qingkai Pan, Krzysztof Choromanski, Deepali Jain, Vikas Sindhwani
Action Space DeepMind Control Suite Implicit Policy

July 3, 2022

Stabilizing Off-Policy Deep Reinforcement Learning from Pixels
Edoardo Cetin, Philip J. Ball, Steve Roberts, Oya Celiktutan
Policy Reinforcement Learning Tetromino Pixel Convolutional Encoder DeepMind Control Suite Catastrophic Overfitting Policy Deep Reinforcement Learning Training Instability

June 30, 2022

Denoised MDPs: Learning World Models Better Than the World Itself
Tongzhou Wang, Simon S. Du, Antonio Torralba, Phillip Isola, Amy Zhang, Yuandong Tian
Reinforcement Learning World Model Policy Optimization Controllability Condition DeepMind Control Suite

June 25, 2022

Value-Consistent Representation Learning for Data-Efficient Reinforcement Learning
Yang Yue, Bingyi Kang, Zhongwen Xu, Gao Huang, Shuicheng Yan
Deep Reinforcement Learning Data Efficient Visual Representation Learning Action Prediction DeepMind Control Suite Consistent Representation Learning

May 20, 2022

Learning Task-relevant Representations for Generalization via Characteristic Functions of Reward Sequence Distributions
Rui Yang, Jie Wang, Zijie Geng, Mingxuan Ye, Shuiwang Ji, Bin Li, Feng Wu
Strong Generalization Visual Reinforcement Learning DeepMind Control Suite Reward Prediction Characteristic Function Reward Distribution Task Induced Representation

May 16, 2022

The Primacy Bias in Deep Reinforcement Learning
Evgenii Nikishin, Max Schwarzer, Pierluca D'Oro, Pierre-Luc Bacon, Aaron Courville
Deep Reinforcement Learning Deep Reinforcement Learning Agent DeepMind Control Suite Primacy Bias

April 21, 2022

Revisiting Gaussian mixture critics in off-policy reinforcement learning: a sample-based approach
Bobak Shahriari, Abbas Abdolmaleki, Arunkumar Byravan, Abe Friesen, Siqi Liu, Jost Tobias Springenberg, Nicolas Heess, Matt Hoffman, Martin Riedmiller
Policy Evaluation Policy Reinforcement Learning Gaussian Mixture Actor Critic Algorithm DIstribution Estimation DeepMind Control Suite Sample Based

February 21, 2022

Don't Touch What Matters: Task-Aware Lipschitz Data Augmentation for Visual Reinforcement Learning
Zhecheng Yuan, Guozheng Ma, Yao Mu, Bo Xia, Bo Yuan, Xueqian Wang, Ping Luo, Huazhe Xu
Data Augmentation Proper Issue Labeling Visual Reinforcement Learning DeepMind Control Suite Control Benchmark