The advancement of autonomous driving technology raises significant safety concerns, necessitating careful attention and robust safety measures. Challenges include ensuring the reliability of sensors, perception systems, and decision-making processes. To address these challenges, research proposes integrating Large Language Models (LLMs) into autonomous vehicle software, particularly focusing on enhancing safety through contextual understanding. LLMs can improve decision-making and safety assessment by incorporating linguistic and contextual knowledge, aiming to handle complex driving scenarios. Translating LLM decisions into warnings, leveraging human-like reasoning to predict hazards and understand other road users' intentions, thereby enhancing awareness from environment during driving.
cd safety_evalution_with_llm
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
cd safety_evalution_with_llm
python auto_safety_framework.py output_video1.mp4
We can do this for output_video2.mp4 and output_video3.mp4. These videos are created from KITTIMOTS dataset.
The expected output is a string. The example for output_video1.mp4 is:
Exercise caution and maintain awareness of cyclists while driving