Coordinated Multi-Agent Reinforcement Learning (Masters Thesis)

Project Overview

This research focuses on developing a sophisticated Coordinated Multi-Agent Reinforcement Learning (CMARL) framework that enables autonomous vehicles to effectively coordinate with human drivers in complex mixed-traffic scenarios. The project utilizes the CARLA simulator for realistic traffic environment simulation and PyTorch for implementing deep reinforcement learning algorithms.

Technologies Used

Key Features

• Real-time coordination between autonomous and human-driven vehicles
• Advanced traffic scenario simulation using CARLA
• Deep Q-Network (DQN) implementation for decision making
• Safety-critical situation handling
• Performance metrics evaluation and analysis

Technical Challenges

• Handling unpredictable human driver behavior
• Ensuring safety in critical traffic scenarios
• Balancing efficiency and safety in coordination strategies
• Real-time processing requirements

Project Gallery