An Integrated Approach for Autonomous Vehicle Urban Route Choice through Path Optimization and Machine Learning

The rapid emergence of Autonomous Vehicles (AVs) brings significant benefits, including improved urban transportation and greater logistics automation. However, current route planning technologies are not fully compatible with the needs of AVs, especially in developing cities and under-construction roads. To address these challenges, we proposed an integrated approach that combines dynamic path planning with Q-learning techniques to enhance urban route selection for AVs. Our framework uses Q-learning to optimize AV routing in real-time, adapting to dynamic traffic, safety factors, and efficiency goals. This Integrated Approach is capable of responding to varying road congestion levels and infrastructure changes, providing the vehicle with continuous feedback from its environment to make real-time decisions. Key cost factors, such as distance, time, energy consumption, and safety, are incorporated into the route planning. Additionally, human desirability factors like scenery, noise, and smell are considered to ensure a more holistic approach. The framework was implemented on urban roads in Shiraz city, showcasing the effectiveness of integrating these technologies for improved AV navigation.

Keywords:

Autonomous Vehicles , Route Planning , Dynamic path planning , Q-learning , Urban Transportation