dm_control: Software and Tasks for Continuous Control

By Yuval Tassa, Saran Tunyasuvunakool, Alistair Muldal, Yotam Doron, Siqi Liu, Steven Bohez, Josh Merel, Tom Erez, Timothy P. Lillicrap, Nicolas Manfred Otto Heess. 2020

526 citations recorded by the source corpus.

The dm_control software package is a collection of Python libraries and task suites for reinforcement learning agents in an articulated-body simulation. A MuJoCo wrapper provides convenient bindings to functions and data structures. The PyMJCF and Composer libraries enable procedural model manipulation and task authoring. The Control Suite is a fixed set of tasks with standardised structure, intended to serve as performance benchmarks. The Locomotion framework provides high-level abstractions and examples of locomotion tasks. A set of configurable manipulation tasks with a robot arm and snap-together bricks is also included. dm_control is publicly available at this https URL

Subfields: Artificial Intelligence, Control and Systems Engineering, Mechanical Engineering

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Control and Systems Engineering

In Mechanical Engineering

Citations in the corpus

robosuite: A Modular Simulation Framework and Benchmark for Robot Learning (2020, background): “In recent years, advances in physics-based simulations and graphics have led to a series of simulated platforms and toolkits [1, 14, 8, 2, 16] that have accelerated scientific progress on robotics and embodied AI.”
RL Unplugged: Benchmarks for Offline Reinforcement Learning (2020, methodology): “These tasks are made up of the corridor locomotion tasks involving the CMU Humanoid, for which prior efforts have either used motion capture data (Merel et al., 2019a,b) or training from scratch (Song et al., 2020).”
A Generalist Agent (2022, background): “On canonical DM Control Suite (Tassa et al., 2018), Gato achieves better than 50% of the expert score on 21 out of 30 tasks from state, and more than 80% for 18 tasks.”
Physically Embedded Planning Problems: New Challenges for Reinforcement Learning (2020, methodology): “The DeepMind locomotion suite and soccer environment (Liu et al., 2019; Tassa et al., 2020) includes both short-horizon, agility-focused problems as well as longer-horizon and multi-agent challenges.”
IKEA Furniture Assembly Environment for Long-Horizon Complex Manipulation Tasks (2019, background): “Simulated benchmarks [1]–[5] have played a key role in the recent advances in reinforcement learning (RL) and imitation learning (IL) for robotic manipulation.”
Instruction-driven history-aware policies for robotic manipulations (2022, background): “The emergence of robotic simulators, such as Gym [52], manipulaTHOR [53], dm control [54], Sapien [55], CausalWorld [56], and RL-Bench [19], also greatly accelerated the development of manipulation methods.”
The Primacy Bias in Deep Reinforcement Learning (2022, methodology): “We focus on two domains: discrete control, represented by the 26-task Atari 100k benchmark (Kaiser et al., 2019), and continuous control, represented by the DeepMind Control Suite (Tassa et al., 2020).”
Regret Minimization Experience Replay in Off-Policy Reinforcement Learning (2021, result): “In MuJoCo and DMC tasks, ReMERT outperforms baseline methods on four of six tasks and achieves comparable performance in the rest two tasks, i.e. HalfCheetah and Hopper, as shown in Fig.”
Introducing Symmetries to Black Box Meta Reinforcement Learning (2021, background): “In the CartPole environment the agent is rewarded for being as upright and centered as possible (Tassa et al. 2020).”
RL Unplugged: A Suite of Benchmarks for Ofﬂine Reinforcement Learning (2020, methodology): “These tasks are made up of the corridor locomotion tasks involving the CMU Humanoid [Tassa et al., 2020], for which prior efforts have either used motion capture data [Merel et al., 2019a,b] or training from scratch [Song et al., 2020].”
Learning Markov State Abstractions for Deep Reinforcement Learning (2021, methodology): “Next, we evaluate our approach in an online setting and train an agent on a collection of six imagebased, continuous control tasks from the DeepMind Control Suite (Tassa et al., 2020).”
From Motor Control to Team Play in Simulated Humanoid Football (2021, methodology): “It conforms with a standard environment interface (42) for RL environments, and it allows plug and play in the sense that different walker bodies can be used to play football, while the humanoid body used in this work can also be deployed in a range of other tasks (42).”
A Survey on Offline Reinforcement Learning: Taxonomy, Review, and Open Problems (2022, methodology): “The RL Unplugged benchmark consists of datasets from four different suites, including the DeepMind Control Suite [69], DeepMind Locomotion [70], Arcade Learning Environment [71] (ALE), and Real-World Reinforcement Learning Suite [72].”
Minigrid & Miniworld: Modular & Customizable Reinforcement Learning Environments for Goal-Oriented Tasks (2023, methodology): “This has been supported by suites of simulation environments such as OpenAI Gym [3] (now gymnasium ) and dm_control [27] that provide common benchmarks for comparing algorithms.”