Isaac Sim Navigation Demo

🔗 Resources

	Resource	Link
📦	Dataset	Scenes · Meta · SceneGraph
🛠️	Code	Generaion pipeline

---

Demo Video

This demo allows you to test the navigation performance for 10 episodes each of fine_grained and coarse_grained tasks.

Prerequisites

1. Download the Isaac Simulator.

We use Isaac Sim 4.5.0. Download it from offical page, which is a zip file. Simply unzip it to a folder (Let's say ISAACSIM_ROOT.).

2. Create and set up a conda environment for the simulation.

conda create -n goodnav python=3.10
conda activate goodnav
conda install pytorch==2.5.1 torchvision==0.20.1 torchaudio==2.5.1  pytorch-cuda=11.8 -c pytorch -c nvidia
pip install pandas
pip install scipy==1.10.1
cd ISAACSIM_ROOT
source setup_conda_env.sh

3. Tested environment

We test the demo on Ubuntu 24 with RTX4090 (Driver Version: 570.195.03 CUDA Version: 12.8 ).

Setup

Download the necessary scene files and code:

git clone https://github.com/william13077/IAmGoodNavigator
cd IAmGoodNavigator
bash download.sh

How to Run

To run a specific demo episode, use the following command:

python demo.py --task <task_type> --index <episode_index> --work_dir path_store_result

• --task: Choose either fine or coarse.
• --index: Choose an index from 0 to 9.

Example:

python demo.py --task fine --index 0 --work_dir ./myresults

The script will load the specific scene USD file and data based on the provided task and index.

Interaction & Controls

Interface

• Instruction Panel: Once the simulation starts, you will see an instruction panel on the left side of the screen displaying the task descriptions.
• Camera Setup: To see the robot's perspective:
  1. Click the Perspective button at the top (slightly to the left of the center).
  2. Select Cameras.
  3. Select FloatingCamera.

Controls

• W: Move forward
• S: Move backward
• A: Turn left
• D: Turn right

Completion

Once you have reached the goal or completed the task, press the Enter key to view your performance evaluation.