QoS-aware task offloading and resource allocation optimization in vehicular edge computing networks via MADDPG

Abstract

To meet the stringent requirements for real-time performance of computing tasks on the Internet of Vehicles (IoV) scenario, the Mobile Edge Computing (MEC) technique is introduced to combine edge servers with vehicles that have storage and computing resources, thereby reducing latency. However, successfully accessing the channel and completing the offloading computation within the specified time remains a big challenge in scenarios with multitasking vehicles and multibase stations equipped with multiple channels. To address this problem, we propose the Multi-Agent Deep Deterministic Policy Gradient-based Offloading and Resource Allocation (MADDPG-RA) algorithm. First, a sub-optimal offloading strategy is determined using the MADDPG algorithm to minimize the sum of system latency and energy consumption. Specifically, this strategy determines whether each vehicle should perform local or offload computation, which MEC server to choose, and which channel to access if offloaded. Based on the above, a closed form expression for the optimal computational resource allocation for MEC is derived using Lagrange multipliers. The simulation results demonstrate that the proposed MADDPG-RA algorithm can effectively reduce the total system latency and energy consumption compared to the existing algorithms.