AEGIS: Human Attention-based Explainable Guidance for Intelligent Vehicle Systems

[ACM SIGCHI 2025]
Official repository for the paper: AEGIS: Human Attention-based Explainable Guidance for Intelligent Vehicle Systems, presented at ACM SIGCHI 2025.

---

⚙️ Create the Environment

Make sure you have Anaconda or Miniconda installed.

Then, run the following command in your terminal:

conda env create -f environment.yml

or

conda create -n aegis python==3.7
conda activate aegis
pip install -r requirements.txt

What do AEGIS provide? Please click to view the video ↓↓↓

🚗 Running the CARLA Simulator

# Create and navigate to the CARLA directory
mkdir carla
cd carla

# Download and extract CARLA 0.9.14
wget https://carla-releases.s3.us-east-005.backblazeb2.com/Linux/CARLA_0.9.14.tar.gz
tar -xvf CARLA_0.9.14.tar.gz

# Launch the simulator
./CarlaUE4.sh --world-port=2000

🧠 Training AEGIS for the Car-Following Scenario

python train_car_following.py --simulator_port 2000

Training AEGIS for the Left-Turn Scenario

python train_left_turn.py --simulator_port 2000

Evaluate AEGIS for the Car-Following Scenario

python eval_car_following_save.py

You should be able to visualize the machine attention using this evalutation script.

📄 Paper

• Zhuang Z, Lu CY, Wang YK, Chang YC, Thomas Do, Lin CT. "AEGIS: Human Attention-based Explainable Guidance for Intelligent Vehicle Systems". ACM CHI Conference on Human Factors in Computing Systems, 2025.

You can read our paper on arXiv here:
AEGIS: Human Attention-based Explainable Guidance for Intelligent Vehicle Systems (arXiv:2504.05950)