Real-time Motion-Based Obstacle Avoidance on the NVIDIA Isaac Sim Platform using CuRobo library

⚠️ This example is intended solely for research purposes and not for commercial use.

This is a example of hard real-time control protocol for the development of AI applications. This demonstration showcases a hard real-time control protocol for the development of AI applications. It illustrates how motion commands rapidly generated by CuRobo within Isaac Sim can be executed in high-frequency real-time cycles on a Doosan robot, enabling fast and responsive control.

System Architecture

Prerequisites

• 1. Docker is required for isaac container example.
  Docker Installation for Ubuntu

• 2. Install for use in Isaac Sim https://curobo.org/get_started/1_install_instructions.html#install-for-use-in-isaac-sim

• 3. Install Doosan Robot Ros2 package https://github.com/DoosanRobotics/doosan-robot2

---

Installation Steps

1. Install Isaac Sim + CuRobo in Docker

After building the emulator, use the provided script to build and launch the Isaac Sim + CuRobo container.

cd ~/doosan_curobo/docker
./build_and_start_docker.sh

2. Install Real-time Kernel for communication between Isaac Sim and doosanrobot (Another PC)

This script connects the simulated robot to the real-time API using the Doosan control framework under hard real-time conditions. We recommend enabling the PREEMPT_RT kernel via Ubuntu Pro.

How to Run

1. (GPU PC) Access the container and run the example script:

docker exec -it automatica2025_container_isaac_sim_4.2.0 bash
omni_python automove.py

2. (Real-time Kernel PC) Inside the sim2dr directory, run the real-time API script:

cd sim2dr
./API_DRFL_ISAAC.sh