Intelligent and Decentralized Resource Allocation in Vehicular Edge Computing Networks

With the rise of intelligent transportation systems and the increasing diversity of vehicular applications, such as safety-related features, parking navigation, and multimedia applications, vehicular edge computing has garnered significant attention. However, managing task offloading efficiently to meet the demands of various tasks remains a fundamental research challenge due to the workload dynamics at multi-access edge computing (MEC) and the unpredictable arrival of tasks. To tackle these challenges, this work proposes a task offloading algorithm for a dynamic vehicular network based on task priority. We introduce a new resource allocation problem to ensure critical tasks meet their response time requirements. The algorithm utilizes Multivariate Long Short-Term Memory (LSTM) to develop an intelligent workload prediction for each MEC node. Additionally, we employ distributed deep reinforcement learning to enhance the efficiency and accuracy of the proactive resource allocation algorithm. Extensive numerical analysis and results demonstrate that our proposed algorithm can significantly increase the ratio of accepted critical tasks. Overall, our task offloading algorithm can effectively manage resources and meet the demands of various tasks in a dynamic vehicular network.