dm-control

What it is and what it does

dm_control is Google DeepMind's infrastructure for physics-based simulation and reinforcement learning, built on the MuJoCo physics engine. It provides Python bindings to MuJoCo, a curated suite of RL environments (cartpole, humanoid, walker, etc.), an interactive viewer for debugging, and higher-level tools like MJCF model composition and a component-based environment builder for creating custom control tasks.

The package is designed for researchers and practitioners building continuous control agents. It handles the complexity of physics simulation, rendering, and environment management, letting you focus on algorithm development. The suite component offers standardized benchmarks; the viewer provides real-time visualization; and the composer and mjcf modules support building custom multi-agent scenarios like soccer tasks. It requires modern Python (3.9+) and system OpenGL libraries for rendering.

Use it for:

• Benchmark reinforcement learning algorithms on standardized continuous control tasks like humanoid locomotion or manipulation.
• Develop and test custom physics-based control environments by composing reusable components with the composer library.
• Visualize and debug agent behavior in real-time using the interactive viewer during development.
• Build multi-agent scenarios such as competitive or cooperative tasks using the soccer module and environment composition tools.
• Prototype robot control policies in simulation before deployment, using MuJoCo's accurate physics and efficient computation.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

dm_control provides Python bindings to the MuJoCo physics engine and a suite of reinforcement learning environments for continuous control tasks, along with tools for composing and modifying physics-based simulations.

Yes. dm_control is actively maintained, has no known vulnerabilities, installs with low friction, and is widely used in RL research. Install it if you need a physics simulation framework for continuous control research or prototyping. Be aware that editable mode installation is not supported and you must have an OpenGL backend available on your system.

Install

dm-control on PyPI

pip install dm-control

uv add dm-control

poetry add dm-control

Installing dm-control

Before you install

Low friction install with a wheel distribution available. Package is actively maintained with a recent release (17 days old) and an active repository. Note: cannot be installed in editable mode due to legacy auto-generated components; standard pip install required.

License in practice

Apache-2.0 permissive license allows commercial and private use with minimal restrictions, making it suitable for research and production applications.

Quickstart

pip install dm_control

import dm_control
from dm_control import suite

env = suite.load(domain_name='cartpole', task_name='balance')
action_spec = env.action_spec()
observation = env.reset()

Requires at least one OpenGL rendering backend (GLFW, EGL, or OSMesa) installed on the system; GLFW requires a windowing system and cannot be used on headless machines. On Linux: libglfw3, libglew2.0, or libgl1-mesa-glx + libosmesa6 depending on backend choice.

Verify before relying

• Whether the 15 runtime dependencies (including mujoco, numpy, scipy, lxml, pyopengl, glfw) are all required for basic usage or if some are optional for specific components.
• Performance characteristics and scalability limits for large-scale multi-agent simulations beyond the soccer tasks mentioned.
• Compatibility with specific MuJoCo versions beyond what the mujoco package dependency specifies.

Package facts

Evidence: dm_control-1.0.44-py3-none-any.whl

Keywords: machine, learning, control, physics, MuJoCo, AI

Tags

More Artificial Intelligence packages

Further reading