Deep-Reinforcement-Learning-Based Distributed Computation Offloading in Vehicular Edge Computing Networks

Vehicular edge computing has emerged as a promising paradigm by offloading computation-intensive latency-sensitive tasks to mobile-edge (MEC) servers. However, it is difficult provide users with excellent Quality-of-Service (QoS) relying only on these server resources. Therefore, in this article, we propose formulate the computation policy based deep reinforcement learning (DRL) vehicle-assisted vehicular network (VAEN) where idle resources of vehicles are deemed Specifically, each task represented directed acyclic graph (DAG) and offloaded nodes according our proposed subtask scheduling priority algorithm. Further, formalize problem under constraints candidate service vehicle models, which aims minimize long-term system cost, including delay energy consumption. To end, distributed algorithm multiagent DRL (DCOM), an improved actor–critic (IACN) devised extract features, joint mechanism prioritized experience replay adaptive $n$ -step (JMPA) enhance efficiency. The numerical simulations demonstrate that, VAEN scenario, DCOM achieves significant decrements latency consumption compared other advanced benchmark algorithms.