OpenArmLab

Isaac Lab extension for the OpenArm manipulator, providing base uni-/bimanual reaching tasks and a Gymnasium entrypoint for simulation/RL experiments in Isaac Sim.

Features

• OpenArm robot setup and scenes on top of NVIDIA Isaac Lab.
• Base uni- and bimanual reaching task configs (observations, rewards, curriculum).
• Sample launcher that wraps tasks as Gymnasium envs for quick rollouts.

Prerequisites

• NVIDIA GPU drivers + CUDA (container base: CUDA 12.8).
• Isaac Sim with Isaac Lab assets (pulled via isaaclab[isaacsim,all]==2.3.1).
• Python 3.11, managed by uv (curl -fsSL https://astral.sh/uv/install.sh | sh).
• Optional: Docker with docker compose (see below).

Quickstart (local, uv)

uv sync --extra dev           # install deps and dev tools
uv run python scripts/01_create_env.py --task <TASK_NAME> --num_envs 1 --headless
# or activate venv: source .venv/bin/activate

Replace <TASK_NAME> with a registered Isaac Lab task id; add --robot_name if your task uses a different robot key.

Using Docker

Build and launch the Compose services (needs NVIDIA runtime):

OPENARMLAB_PATH=/workspace/openarmlab WEB_PORT=6080 docker compose up -d --build
docker compose run --rm openarmlab bash   # enter the app container
cd /workspace/openarmlab
uv sync --extra dev
uv run python scripts/01_create_env.py --task <TASK_NAME> --num_envs 1 --headless

desktop service exposes a web VNC at WEB_PORT for GUI runs; add/remove --headless as needed.

Development

• Lint: uv run ruff check . (config in pyproject.toml).
• Tests: add/execute with uv run pytest as needed.
• For dependency changes, edit pyproject.toml (uv resolves/locks via uv sync).