RevEAL: Reliability vs Energy Optimization for Autonomous Vehicles Using Large Language Models

As autonomous vehicles continue to gain traction, the need for highly accurate and energy-efficient systems to enhance safety and performance becomes increasingly critical. Effectively managing the tradeoff between energy consumption and reliability in these systems requires the ability to predict various operational conditions. With the rapid advancements in Large Language Models and models like ChatGPT, new opportunities for improving predictions in autonomous vehicle operations have emerged. This paper proposes RevEAL, which utilizes Large Language Models as map reader co-drivers to predict essential parameters for optimizing the energy-reliability balance during AV operations. Experimental results demonstrate that RevEAL achieved up to 67% driving accuracy and a 53.4% reduction in total energy consumption, depending on the operating scenario. Additionally, RevEAL reduced power consumption by 33% compared to selected baseline configurations, highlighting its strength in maintaining a practical balance between navigation performance and energy efficiency. These findings underscore the potential of RevEAL to contribute to the development of more adaptive and resource-aware autonomous driving systems.