IoV mobile edge computing task offloading based on MADDPG algorithm

Abstract

The Internet of Vehicles has become a crucial component of contemporary transportation as a significant subset of the Internet of Things. The demand for periphery computing is increasing as vehicle intelligence and connectivity continues to advance. However, the task unloading of onboard edge computing encounters several obstacles, including limited computing power, communication delay, etc. This paper proposes a task discharge scheme for Internet of Vehicles edge computing based on the MADDPG algorithm to address these issues. The scheme employs a multi-agent reinforcement learning algorithm to accomplish cooperation and communication between vehicles and optimizes the task allocation strategy to improve the efficiency and performance of onboard edge computing. Simulation results indicate that, in comparison to other algorithms, this algorithm can significantly reduce the system's overall execution latency and possesses strong adaptability.