Q-Learning Based Computation Offloading with Minimizing the Cost in Vehicle Edge Computing

The advancement of autonomous driving and vehicle networking technologies is continuously spawning new vehicle applications, improving traffic safety and the overall driving experience is a key goal. Yet, this is challenged by the limited computing and storage capacities available on-board. certain computation-heavy tasks miss their deadlines, necessitating the use of edge computing. Vehicle edge (VEC) serves as a potent solution for enhancing vehicular computing capabilities. This paper presents a model for task offloading in vehicular edge computing (VEC).Our approach leverages idle resources in vehicles to support edge computing offloading within a heterogeneous network. To reduce system expenses, encompassing both latency and energy, we integrate reinforcement learning (RL) with deep learning (DL) for enhanced convergence efficiency. An advanced Q-learning algorithm is proposed to address both task offloading and processing challenges, enabling optimal offloading decisions. Simulation outcomes indicate that the enhanced Q-learning algorithm successfully lowers overall system costs and boosts the service quality in Vehicle Edge Computing (VEC).