Skip to content

Any plan for lekiwi? #50

Closed
Closed

Description

@mscwu

Is there any plan for supporting lerobot mobile manipulator derivatives, such as lekiwi?

Activity

  1. EverNorif commented on Sep 29, 2025

    @EverNorif
    Collaborator

    Yes, of course. We plan to support more lerobot-related devices in the future, such as lekiwi.

  2. vpraise00 commented on Dec 10, 2025

    @vpraise00

    https://github.com/vpraise00/lekisaac
    I’m working on making lekisaac function as a submodule of leisaac.
    Would it be possible for you to integrate this repository in the future? I plan to continue developing it.

  3. EverNorif commented on Dec 10, 2025

    @EverNorif
    Collaborator

    @vpraise00 Thank you!

    Actually, I’m working on the integration of lekiwi, and some parts are already finished. Maybe this week or next week it can be fully released.

    I think integrating lekiwi directly into leisaac would be better, which can make the usage logic unified. Perhaps after the lekiwi integration commit is submitted, we can discuss and see which parts can be improved.

  4. EverNorif commented on Dec 19, 2025

    @EverNorif
    Collaborator

    Hi there, we’ve just added teleoperation support for lekiwi. Feel free to give it a try.

  5. vpraise00 commented on Dec 19, 2025

    @vpraise00

    Have you solved the omni-wheel problem?

    I am trying to address the omni-wheel problem by adding meshes for the rollers on each wheel. However, modeling each individual roller results in a very high computational cost.

    If you have already solved this problem, could you please share the methods you used?

  6. EverNorif commented on Dec 19, 2025

    @EverNorif
    Collaborator

    @vpraise00 We adopted an approach to bypass explicit wheel modeling. As you mentioned, modeling each small roller on the wheels would be very difficult. Instead, we added three extra dummy joints to control motion along the x-axis, motion along the y-axis, and rotation around the robot’s own axis. The corresponding USD can be found here.

    This approach improves the teleoperation experience. However, in this case, the base motion is no longer driven by the wheels, which somewhat compromises physical realism.

  7. neso613 commented on Jan 14, 2026

    @neso613

    Hi
    I do not finf LeKiwi env files for teleopeartions, could some one help or share the link

  8. EverNorif commented on Jan 15, 2026

    @EverNorif
    Collaborator

    @neso613 We have uploaded the lightwheel_loft scene to a new link. You can find the download link here.

  9. neso613 commented on Jan 15, 2026

    @neso613

    Thank you @EverNorif for sharing.
    I can see on this link - Kitchen env which has many USD assets.
    also when I run command - python scripts/environments/list_envs.py given on link
    -------------------------------------------------------------------------------------+
    | S. No. | Task Name | Entry Point | Config |
    +--------+---------------------------------------------+---------------------------------------------------------------------------------------+-------------------------------------------------------------------------------------------------+
    | 1 | LeIsaac-SO101-AssembleHamburger-v0 | isaaclab.envs:ManagerBasedRLEnv | leisaac.tasks.assemble_hamburger.assemble_hamhurger_env_cfg:AssembleHamburgerEnvCfg |
    | 2 | LeIsaac-SO101-AssembleHamburger-BiArm-v0 | isaaclab.envs:ManagerBasedRLEnv | leisaac.tasks.assemble_hamburger.assemble_hamhurger_bi_arm_env_cfg:AssembleHamburgerBiArmEnvCfg |
    | 3 | LeIsaac-SO101-CleanToyTable-v0 | isaaclab.envs:ManagerBasedRLEnv | leisaac.tasks.clean_toy_table.clean_toy_table_env_cfg:CleanToyTableEnvCfg |
    | 4 | LeIsaac-SO101-CleanToyTable-BiArm-v0 | isaaclab.envs:ManagerBasedRLEnv | leisaac.tasks.clean_toy_table.clean_toy_table_bi_arm_env_cfg:CleanToyTableBiArmEnvCfg |
    | 5 | LeIsaac-SO101-CleanToyTable-BiArm-Direct-v0 | leisaac.tasks.clean_toy_table.direct.clean_toy_table_bi_arm_env:CleanToyTableBiArmEnv | leisaac.tasks.clean_toy_table.direct.clean_toy_table_bi_arm_env:CleanToyTableBiArmEnvCfg |
    | 6 | LeIsaac-LeKiwi-CleanupTrash-v0 | isaaclab.envs:ManagerBasedRLEnv | leisaac.tasks.cleanup_trash.cleanup_trash_env_cfg:CleanupTrashEnvCfg |
    | 7 | LeIsaac-SO101-FoldCloth-BiArm-v0 | isaaclab.envs:ManagerBasedRLEnv | leisaac.tasks.fold_cloth.fold_cloth_bi_arm_env_cfg:FoldClothBiArmEnvCfg |
    | 8 | LeIsaac-SO101-FoldCloth-BiArm-Direct-v0 | leisaac.tasks.fold_cloth.direct.fold_cloth_bi_arm_env:FoldClothBiArmEnv | leisaac.tasks.fold_cloth.direct.fold_cloth_bi_arm_env:FoldClothBiArmEnvCfg |
    | 9 | LeIsaac-SO101-LiftCube-v0 | isaaclab.envs:ManagerBasedRLEnv | leisaac.tasks.lift_cube.lift_cube_env_cfg:LiftCubeEnvCfg |
    | 10 | LeIsaac-SO101-LiftCube-DigitalTwin-v0 | leisaac.enhance.envs:ManagerBasedRLDigitalTwinEnv | leisaac.tasks.lift_cube.lift_cube_env_cfg:LiftCubeDigitalTwinEnvCfg |
    | 11 | LeIsaac-SO101-LiftCube-Mimic-v0 | leisaac.enhance.envs:ManagerBasedRLLeIsaacMimicEnv | leisaac.tasks.lift_cube.lift_cube_mimic_env_cfg:LiftCubeMimicEnvCfg |
    | 12 | LeIsaac-SO101-LiftCube-Direct-v0 | leisaac.tasks.lift_cube.direct.lift_cube_env:LiftCubeEnv | leisaac.tasks.lift_cube.direct.lift_cube_env:LiftCubeEnvCfg |
    | 13 | LeIsaac-SO101-PickOrange-v0 | isaaclab.envs:ManagerBasedRLEnv | leisaac.tasks.pick_orange.pick_orange_env_cfg:PickOrangeEnvCfg |
    | 14 | LeIsaac-SO101-PickOrange-Mimic-v0 | leisaac.enhance.envs:ManagerBasedRLLeIsaacMimicEnv | leisaac.tasks.pick_orange.pick_orange_mimic_env_cfg:PickOrangeMimicEnvCfg |
    | 15 | LeIsaac-SO101-PickOrange-Direct-v0 | leisaac.tasks.pick_orange.direct.pick_orange_env:PickOrangeEnv | leisaac.tasks.pick_orange.direct.pick_orange_env:PickOrangeEnvCfg |
    +--------+---------------------------------------------+---------------------------------------------------------------------------------------+-------------------------------------------------------------------------------------------------+

    Task 1 is LeIsaac-SO101-AssembleHamburger-v0, but I do not see env for this.
    Error in 'pxrInternal_v0_24__pxrReserved__::UsdStage::Open' at line 1059 in file C:\g\201060795\USD\pxr\usd\usd\stage.cpp : 'Failed to open layer @C:\Users\Admin01\Desktop\Physical AI\leisaac\assets\scenes\kitchen_with_hamburger\scene.usd@'

    can you pls guide here- how to proceed to train on this task?

  10. EverNorif commented on Jan 16, 2026

    @EverNorif
    Collaborator

    @neso613 We haven’t uploaded this scene yet, see also here. However, we are currently organizing this task and will release it very soon.

  11. yanyoufu commented on Jul 8, 2026

    @yanyoufu

    @neso613 @EverNorif Hello, the current download link for the "lightwheel_loft" scene is no longer available (it jumps to simready.com and can not find the assets event i login in for searching). Could you please inform me how to download it now? Thank you.

    @neso613 We have uploaded the lightwheel_loft scene to a new link. You can find the download link here.

  12. EverNorif commented on Jul 8, 2026

    @EverNorif
    Collaborator

    @yanyoufu please try this download link

  13. yanyoufu commented on Jul 8, 2026

    @yanyoufu

    @yanyoufu please try this download link

    thanks a lot.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions