Skip to content

UniEnv

UniEnv overview

UniEnv is a robotics-oriented interface layer for environments, worlds, data pipelines, and backend-aware transformations.

It is built to let the same codebase span:

  • simulation environments and real-world control loops
  • stateful and functional environment styles
  • structured observations and flattened tensor views
  • online interaction and offline dataset workflows
  • multiple array backends through a shared compute abstraction

Why UniEnv

UniEnv gives you one vocabulary for the parts that tend to fragment robotics codebases:

  • Environment APIs for reset, step, render, and typed spaces
  • World composition for building environments from reusable nodes
  • Wrappers and transformations for action scaling, flattening, frame stacking, backend conversion, and recording
  • Data interfaces for replay buffers, storage backends, and dataset adapters

The result is a package that can sit between your simulator, robot stack, dataset format, and learning code without forcing each layer to invent its own data model.

Start Here

  • :material-rocket-launch-outline: Getting Started

    Install UniEnv, learn the package layout, and see the smallest useful workflow.

    Open the guide

  • :material-shape-outline: Core Concepts

    Learn how environments, spaces, backends, worlds, and nodes fit together.

    Browse the concepts

  • :material-book-open-page-variant-outline: Guides

    Go straight to wrappers, replay buffers, storages, and dataset integrations.

    Browse the guides

  • :material-code-braces: API Reference

    Inspect the generated module reference for unienv_interface and unienv_data.

    Open the API reference

Package Map

Area What it covers Key modules
Interface layer Spaces, environments, wrappers, transformations unienv_interface.space, unienv_interface.env_base, unienv_interface.wrapper, unienv_interface.transformations
Composition layer Worlds, nodes, and environment composition unienv_interface.world
Data layer Batches, replay buffers, storage, samplers unienv_data.base, unienv_data.replay_buffer, unienv_data.storages, unienv_data.samplers
Integrations External dataset and training ecosystem adapters unienv_data.integrations

Reading Order

If you are new to the project, a good path is:

  1. Read Getting Started.
  2. Read Environments and Spaces and Backends.
  3. Read World Composition if you are composing complex simulators or robot systems.
  4. Read Wrappers and Transformations and Replay Buffers and Storage for day-to-day usage.