Changelog

Changelog#

Version 1.5.2#

  • Bug Fixes:

    • Fixed domain randomization compatibility for mujoco!=3.1.1.

    • Corrected dexterous hand mounting for Panda and pot-with-handles geometry for rectangular pots.

    • Fixed Mink IK with delta input, duplicated inertial properties, free-joint issues, and Baxter mesh problems.

    • Resolved mobile base XML and controller logic issues, binding utils on Windows, and EGL error reporting.

    • Hot-fixed Task.generate_id_mappings() to remove invalid EEF targets.

  • Features / Enhancements:

    • Add env seeding for determinism.

    • Added support for multiple cameras across renderers and observation pipelines.

    • Enabled instance segmentation to include arena bodies.

    • Added DualSense controller support and improved SpaceMouse auto-detection.

    • Introduced scalable arenas and MuJoCo objects via set_scale.

    • Allowed base type specification and manipulator mounts for mobile manipulation envs.

    • Added new robots and hands (xArm7 revised, Inspire/Fourier grasp qpos).

    • Extended whole-body IK with input_ref_frame=base and skip_wbc_action.

    • Added sensors, shell inertia for Sawyer meshes, and joint position observations.

  • Documentation Updates:

    • Updated v1.5 docs, fixed broken links, improved dark-mode visibility, and refined demo/device documentation.

    • Added USD requirements and SpaceMouse dependencies to docs and extras.

    • Improved demos page and clarified base type comments.

  • CI / Release / Tooling:

    • Added and refined PyPI publishing workflows.

    • Improved build–docs synchronization and pinned Mink version.

  • Miscellaneous:

    • Added render modalities and updated demo scripts to include all manipulators by default.

    • Refactored controller configs and composite controller logic.

    • Improved data collection wrapper resets and environment seeding.

Version 1.5.1#

  • Bug Fixes:

    • Fixed segmentation demo, part controller demo, demo renderer, joint and actuator issues, equality removal, and orientation mismatch in Fourier hands.

  • Documentation Updates:

    • Updated basicusage.md, overview page, demo docs, teleop usage info, devices, and related docs.

    • Added a CI doc update workflow and .nojekyll fix.

    • Simplified composite controller keywords and updated robots section and task images.

  • Features/Enhancements:

    • Added GymWrapper support for both gym and gymnasium with dict_obs.

    • Updated DexMimicGen assets and added a default whole-body Mink IK config.

    • Improved CI to check all tests and print video save paths in demo recordings.

    • Add GR1 and spot robot to demo_random_actions.py script.

  • Miscellaneous:

    • Added troubleshooting for SpaceMouse failures and terminated mjviewer on resets.

    • Adjusted OSC position fixes and updated part controller JSONs.

Version 1.5.0#

The 1.5 release of Robosuite introduces significant advancements to extend flexibility and realism in robotic simulations. Key highlights include support for diverse robot embodiments (e.g., humanoids), custom robot compositions, composite controllers (such as whole-body controllers), expanded teleoperation devices, and photorealistic rendering capabilities.

New Features#

  • Diverse Robot Embodiments: Support for complex robots, including humanoids, allowing exploration of advanced manipulation and mobility tasks. Please see robosuite_models for extra robosuite-compatible robot models.

  • Custom Robot Composition: Users can now build custom robots from modular components, offering extensive configuration options.

  • Composite Controllers: New controller abstraction includes whole-body controllers, and the ability to control robots with composed body parts, arms, and grippers.

  • Additional Teleoperation Devices: Expanded compatibility with teleoperation tools like drag-and-drop in the MuJoCo viewer and Apple Vision Pro.

  • Photorealistic Rendering: Integration of NVIDIA Isaac Sim for enhanced, real-time photorealistic visuals, bringing simulations closer to real-world fidelity.

Improvements#

  • Updated Documentation: New tutorials and expanded documentation on utilizing advanced controllers, teleoperation, and rendering options.

  • Simulation speed improvement: By default we set the lite_physics flag to True to skip redundant calls to env.sim.step()

Migration#

  • Composite controller refactoring: please see example of usage

Contents