Isaac Sim Automator

The Isaac Sim Automator tool automates the deployment of Isaac Sim to public clouds, enabling efficient and scalable simulations.

Installation

Installing Docker

Ensure Docker is installed on your system for container management. For installation guidance, see Docker Installation.

Obtaining NGC API Key

Obtain an NGC API Key to download Docker images from NVIDIA NGC. This can be done at NGC API Key Setup.

Building the Container

To build the Automator container, use the following command in your project root directory:

./build

This builds and tags the Isaac Sim Automator container as 'isa'.

Usage

Running Automator Commands

You can run the Automator commands in two ways:

Enter the automator container and run commands inside:

./run
./somecommand

Or, prepend the command with ./run:

./run ./somecommand <parameters>

Examples include:

./run ./deploy-aws
./run ./destroy my-deployment

Deploying Isaac Sim

Choose the appropriate cloud provider (AWS, GCP, Azure, Alibaba Cloud) and follow the provided steps to deploy Isaac Sim using the automator container.

K-Scale Sim Library

The K-Scale Sim Library is built atop Isaac Gym for simulating Stompy, providing interfaces for defined tasks like walking and getting up.

Getting Started with K-Scale Sim Library

Initial Setup

First, clone the K-Scale Sim Library repository and set up the environment:

git clone https://github.com/kscalelabs/sim.git
cd sim
conda create --name kscale-sim-library python=3.8.19
conda activate kscale-sim-library
make install-dev

Installing Dependencies

After setting up the base environment, download and install necessary third-party packages:

wget https://developer.nvidia.com/isaac-gym/IsaacGym_Preview_4_Package.tar.gz
tar -xvf IsaacGym_Preview_4_Package.tar.gz
conda env config vars set ISAACGYM_PATH=`pwd`/isaacgym
conda deactivate
conda activate kscale-sim-library
make install-third-party-external

Running Experiments

Setting Up Experiments

Download and prepare the Stompy model for experiments:

wget https://media.kscale.dev/stompy.tar.gz
tar -xzvf stompy.tar.gz
python sim/scripts/create_fixed_torso.py
export MODEL_DIR=stompy

Training and Evaluation

For leg-specific tasks:

python sim/humanoid_gym/train.py --task=legs_ppo --num_envs=4096 --headless

For full-body tasks:

python sim/humanoid_gym/train.py --task=stompy_ppo --num_envs=4096 --headless

Evaluate the models on CPU:

python sim/humanoid_gym/play.py --task=legs_ppo --sim_device=cpu

Troubleshooting

Common issues and solutions for setting up and running Isaac Gym and K-Scale simulations.

git submodule update --init --recursive
export LD_LIBRARY_PATH=PATH_TO_YOUR_ENV/lib:$LD_LIBRARY_PATH
sudo apt-get install vulkan1

Isaac Sim Automator

Contents