Optimizing task offloading and resource allocation in latency-constrained vehicular edge computing

Optimizing task offloading and resource allocation in latency-constrained vehicular edge computing

연구 분야: Networking

학회: Cluster Computing

To satisfy the increasing demand of vehicular computing tasks, offloading tasks to edge computing nodes in vehicular edge computing (VEC) scenarios has attracted much attention. However, existing studies have neglected the high mobility of vehicles, the variability of edge server computing resources, and the volatility of wireless network conditions. To this end, this paper proposes a combinatorial optimization algorithm that aims to jointly optimize the task offloading strategy and resource allocation by simultaneously considering the time-varying channel conditions and edge server resources to achieve the goal of minimizing the task execution delay. First, considering the challenge of obtaining real-time channel state information (CSI) in complex VEC systems, we propose a distributed deep deterministic policy gradient (D4PG)-based task offloading algorithm, which predicts future CSI based on historical data and determines the optimal task offloading decision. Second, after determining the optimal task offloading strategy using the D4PG algorithm, the optimization problem can be transformed into a resource allocation problem and the dynamic resource allocation scheme can be implemented using convex optimization theory. Simulation results show that our method outperforms the baseline method in reducing task execution delay.