h12_adaptive_policy

Humanoid policy adaptive to end-effector payloads.

Installation

Both uv and conda flows are supported. uv is the recommended path: it resolves the full stack (RMA policy deps + the h12_ros2_controller and unitree_sdk2_python submodules) from a single pyproject.toml and creates a local .venv.

Option A: uv (recommended)

1. Install uv (e.g. curl -LsSf https://astral.sh/uv/install.sh | sh).

2. Clone this repo and initialize its submodules:

   git clone https://github.com/correlllab/h12_adaptive_policy.git
   cd h12_adaptive_policy
   git submodule update --init --recursive

3. Let uv create the environment and install every dependency (both the h12_ros2_controller and unitree_sdk2_python submodules are installed as editable path dependencies):

   uv sync

   This creates .venv/ pinned to Python 3.10. All subsequent commands should be invoked via uv run ..., which activates the venv automatically.

4. Smoke-test the install:

   uv run python -c "import mujoco, torch, pinocchio, h12_ros2_controller; print('ok')"

5. Run the Mujoco RMA deployment from the repo root:

   uv run python h12_adaptive_policy/deploy/mujoco_deploy_h12_rma.py

See docs/mujoco_frame_controller_howto.md for running the Mujoco sim together with the h12_ros2_controller FrameController stack.

Option B: conda (legacy)

• Initialize submodules:

  git submodule update --init --recursive

• The Unitree Python SDK is vendored as a submodule under submodules/unitree_sdk2_python (cloned by the git submodule update above).

• Install python dependencies from environment.yml:

  conda env create -f environment.yml
  conda activate adaptive_env
  pip install -e submodules/unitree_sdk2_python
  pip install -e submodules/h12_ros2_controller

Files

• data/ contains saved data such as model checkpoint and evaluation results.
• figures contains generated figures.
• h12_adaptive_policy/: contains source code.
  • deploy/ contains scripts to deploy the policy and run experiments.
  • example/ contains example scripts.
  • plot/ contains scripts to plot figures.
  • RMA/ contains implementation of the RMA modules.
• submodules/: contains external dependencies.

Usage

• Run scripts from the root directory python h12_adaptive_policy/deploy/eval_rma_hand_sweep_6d.py.