ML-MAS-FRAMEWORK

Improving Self-Driving Performance using ML-MAS Framework

• 📖 Table of Contents
• 📍 Overview
• 📦 Features
• 🪧 Project Demo
• 📂 Repository Structure
• ⚙️ Modules
• 🚀 Getting Started
  • 🔧 Installation
  • 🤖 Running ML-MAS-Framework
  • 🧪 Results
• 🤝 Contributing
• 📄 License
• 👏 Acknowledgments

This project investigates a neuro-symbolic AI framework using ML and MAS, with the aim to enhance the driving capabilities of autonomous vehicles. The ML component enables the vehicle to learn patterns and behaviours from vast amounts of data, while the MAS component provides rational decision-making and adaptive behaviour in response to complex driving scenarios.

Previous approaches to autonomous driving have predominantly focused on purely ML-based solutions or rule-based systems. ML models, such as deep neural networks, have shown remarkable capabilities in perception tasks like object recognition and lane detection. However, they often lack explicit reasoning and interpretability, making them prone to uncertainty and limited adaptability in dynamic environments. On the otherhand, rule-based systems can provide rational behaviour but struggle tohandle the complexity and variability encountered in real-world driving scenarios.

By incorporating Neuro-symbolic AI with ML and MAS, this project takes a hybrid approach that combines the learning capabilities of ML models with the rational behaviour and adaptability of agent-based systems. This integration aims to overcome the limitations of purely ML or rule-based approaches by leveraging the strengths of both paradigms. The project strives to demonstrate that a balanced approach, combining learning-based models with rational decision-making agents, can yield superior results in autonomous driving.

The project builds upon the previous work of Hilal Al Shukairi, a former MSc AI student, who developed the ML-MAS framework and showcased improved driving scores.

└── ML-MAS-Framework/
    ├── MLMAS_Project/
    │   ├── MLMAS_Framework/
    │   ├── ML_Models/
    │   ├── environment.yml
    │   ├── leaderboard/
    │   ├── results/
    │   ├── run_MLMAS_LAV_eval.bat
    │   └── scenario_runner/
    ├── PSEEB_Annex_A_Ethics_Checklist.pdf
    ├── Project Plan and Abstract.pdf
    ├── Project Report.pdf
    └── setup.bat

Dependencies

Please ensure you have the following dependencies installed on your system:

CARLA 0.9.10.1 and its Additional Maps

LAV Model weights

Anaconda and Java installed

1. Clone the ML-MAS-Framework repository:

git clone https://github.com/Abhijith14/ML-MAS-Framework

2. Change to the project directory:

cd ML-MAS-Framework

3. Install the dependencies:

setup.bat

This should setup both carla and lav model weights.

4. Setup the conda environment

cd MLMAS_Project
conda env create -f environment.yml

cd *TO CARLA FODLER*
Carla.exe

cd ..
cd MLMAS_Project
run_MLMAS_LAV_eval.bat

Recordings for each route can be found at:

└── ML-MAS-Framework/
    └── MLMAS_Project/
       └── results/
           └── records/

And inside results folder, the metrics for each route can also be found.

Contributions are always welcome! Please follow these steps:

1. Fork the project repository. This creates a copy of the project on your account that you can modify without affecting the original project.
2. Clone the forked repository to your local machine using a Git client like Git or GitHub Desktop.
3. Create a new branch with a descriptive name (e.g., new-feature-branch or bugfix-issue-123).

git checkout -b new-feature-branch

4. Make changes to the project's codebase.
5. Commit your changes to your local branch with a clear commit message that explains the changes you've made.

git commit -m 'Implemented new feature.'

6. Push your changes to your forked repository on GitHub using the following command

git push origin new-feature-branch

7. Create a new pull request to the original project repository. In the pull request, describe the changes you've made and why they're necessary. The project maintainers will review your changes and provide feedback or merge them into the master branch.

This project is licensed under the GPL-3.0 License. See the GPL-3.0 file for additional info.

I would like to express my deepest gratitude and appreciation to all those who have contributed to the successful completion of this project on integrating a Neuro-symbolic AI framework using Machine Learning (ML) and Multi-Agent Systems (MAS) for enhancing the driving capabilities of self-driving cars.

First and foremost, I would like to thank my project supervisor, Dr. Rafael C. Cardoso, for his guidance, support, and invaluable insights throughout the duration of this project. His expertise and dedication have been instrumental in shaping the direction and scope of our work.

I am immensely grateful to Hilal Al Shukairi, the former MSc AI student, whose previous work on the ML-MAS framework laid the foundation for my project. His efforts and innovative ideas have significantly contributed to my understanding and implementation of Neuro-symbolic AI in autonomous driving.

I extend my thanks to the faculty and staff of University of Aberdeen, whose resources and facilities have provided me with a conducive environment for conducting my research. Their commitment to fostering academic excellence and pushing the boundaries of knowledge has been a constant source of motivation

This readme file is created by eli64s/README-AI

↑ Return