Internet of vehicles task offloading based on edge computing using double deep Q-network
摘要
With the rapid development of the internet of vehicles (IoV) and advancements in intelligence, IoV generates a large number of task requests and data. At the same time, the traditional cloud computing paradigm makes it difficult to meet real-time and reliability requirements. Under the trend of deep integration between edge computing and IoV, task offloading has emerged as a critical enabling technology. By offloading vehicular task requests to the network edge equipped with computational and storage resources, collaborative task processing can be achieved. This not only supports complex vehicular applications but also significantly enhances overall system performance. Therefore, this paper proposes an edge computing-based task offloading algorithm for IoV, enabling efficient offloading of vehicular task requests to network edges to improve task processing efficiency. Furthermore, a double deep Q-network (DDQN)-based agent model is deployed to optimize offloading decisions for edge nodes. The model comprehensively considers dynamic network states and edge resource distribution to compute a set of candidate edge nodes for task offloading. Simulation results demonstrate that compared with baselines, the proposed algorithm improves the task offloading success rate, offloading revenue-to-cost ratio, and offloading revenue by an average of