Skip to content

README.md

Latest commit

 

History

History
60 lines (42 loc) · 1.81 KB

README.md

File metadata and controls

60 lines (42 loc) · 1.81 KB
Raw

Docker Installation and Usage

The Docker workflow is the recommended way to run robo_orchard_sim.

Prerequisites

  • Linux with an NVIDIA driver compatible with Isaac Sim 5.1.0
  • Docker Engine and NVIDIA Container Toolkit
  • The simulation assets described in the root installation guide

Get the Image

docker pull horizonrobotics/robo_orchard_sim:ubuntu22.04-py3.11-cuda12.8-isaacsim5.1-isaac_lab-v2.3.2-v0.1

The image ships Ubuntu 22.04, Python 3.11, CUDA 12.8, PyTorch 2.7.0, Isaac Sim 5.1.0, Isaac Lab 2.3.2, and cuRobo 0.7.6, but not robo_orchard_sim itself.

Run the Container

export WORKSPACE=/absolute/path/to/your/workspace
bash docker/run_container.sh
docker exec -it robo_orchard_sim bash

WORKSPACE becomes the container HOME, and ${WORKSPACE}/.cache is reused as the Kit cache. Add your own -v for the asset directory, then set ORCHARD_ASSET and NV_ASSET_ROOT_DIR inside the container as described in the root installation guide.

If you write your own launch script, copy the docker run flags from run_container.sh — the Vulkan mounts, --cap-add=ALL, and NVIDIA_DRIVER_CAPABILITIES are required for Isaac Sim to render.

Install and Verify

Inside the container, from your clone of this repository:

cd robo_orchard_sim
make install-editable

Run With GUI

The container shares the host X11 socket, so prefix the command with the host DISPLAY and the window shows up on the host desktop:

DISPLAY=:{id} python3 your_isaac_example.py

Replace {id} with the host display number (echo $DISPLAY on the host). Launcher-based scripts need SimpleIsaacAppLauncher(headless=False, virtual_display=False).

Only GPU 0 can present to a display; runs on other GPUs must be headless.