RISKS REDUCING THROUGH INTELLIGENT HEADLIGHT MANAGEMENT: OPTIMIZING Q-LEARNING FOR ELECTRIC VEHICLES

This paper proposes an intelligent headlight management system for Electric vehicles (EVs) based on an adaptive Q-learning framework that considers enhancing safety and reducing risks. This includes formulating a Q-learning strategy for real-time control of headlights operating in modes suitable for the current conditions and vehicle operations. Evaluation of the performance of the adaptive Q-learning system is presented in this study in terms of safety metrics such as visibility distance and energy efficiency indicators such as power consumption through comprehensive simulations across various turning scenarios. These results show significant improvements compared to traditional systems with fixed beam patterns and rules-based control systems. This approach proves effective and expresses the research prospects of enhancing the safety of night-time driving, reducing risks, minimizing energy usage, and improving the overall performance of the approach with traditional routing methods, demonstrating its superior performance in various scenarios. This paper not only contributes to the optimization of last-mile delivery using shipping drones but also highlights the potential of reinforcement learning techniques, such as deep Q-learning, in addressing complex routing challenges in dynamic, real-world environments in smart logistics. Ultimately, further exploration into the utilization of reinforcement learning for complex optimization issues across various domains is recommended.