UrbanDrivePathway: A Decision-Making Framework for Navigating Urban Autonomous Vehicles in Complex Traffic Systems

The rapid advancement of autonomous vehicle (AV) technology necessitates robust decision-making frameworks for navigating complex urban traffic environments. This study introduces UrbanDrivePathway, a novel decision-making framework that leverages deep learning, reinforcement learning, and real-time traffic analytics to enhance AV performance. The framework integrates EfficientDet for object detection (97.69% accuracy), Graph Neural Networks (GNN) for trajectory prediction (97.42% accuracy), and Monte Carlo Tree Search (MCTS) for decision-making (97.38% accuracy). Model Predictive Control (MPC) is employed for path planning, achieving 97.61% accuracy, while an Eco-Optimized Driving Strategy enhances fuel efficiency with 97.69% accuracy in energy consumption analysis. Evaluated on diverse real-world and synthetic datasets, the proposed framework outperforms conventional models in collision avoidance, intersection handling, and energy optimization. Comparative analysis with existing methods highlights superior accuracy, reduced decision latency, and enhanced safety compliance. The UrbanDrivePathway framework ensures reliable navigation in dense urban settings, addressing challenges <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">re</sup>lated to real-time adaptability and dynamic interactions. Future enhancements will focus on multi-agent collaboration, adversarial scenario training, and advanced explainability techniques. These improvements will further optimize AV decision-making in unpredictable urban environments.