Automated driving is a revolutionary technology that is fundamentally changing worldwide transportation networks. An essential component of autonomous vehicle operation is path planning, which guarantees secure and effective navigation across intricate and ever-changing surroundings. This work presents an innovate neural network design that integrates deep Q-learning with policy gradients, therefore improving the current deep reinforcement learning paradigm. This methodology allows the self-driving vehicle to acquire knowledge and adjust to intricate situations using huge quantities of data, strict adherence to demanding global criteria. To verify the efficacy of our suggested approach, we performed simulations using both realistic and benchmark scenarios. The findings illustrate that our methodology greatly improves the vehicle's performance in difficult circumstances while upholding a commendable degree of safety. An analysis was conducted on important parameters like route completion time, energy usage, and adaptation to changing environmental circumstances. This approach demonstrates superior performance compared to both conventional and modern autonomous driving methods, therefore highlighting its capacity to propel the field of route planning in autonomous systems forward. The present study makes a valuable contribution to the advancement of autonomous driving technology by the introduction of an advanced deep learning framework that enhances path planning and decision-making capabilities.
Supervised Deep Learning Methodology for Autonomous Vehicle Routing
22.11.2024
331810 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch