Offloading optimization of UAV-assisted mobile edge computing based on DQN
FENG Yixiong
XIONG Dan
JIN Kebing
WU Xuanyu
HONG Zhaoxi
TAN Jianrong
Abstract：[Objective]In mobile edge computing(MEC)systems in dynamic environments,traditional task offloading strategies generally have problems such as inflexible scheduling,weak adaptability to environmental changes,and limited delay control capabilities,making it difficult to meet the processing requirements of delay-sensitive tasks.To this end,this paper proposed a MEC offloading optimization method that integrated unmanned aerial vehicle(UAV)-assisted mechanisms to improve the system's service quality and task response efficiency.[Methods]Considering the dynamic user distribution and frequent link state fluctuations in UAV-MEC scenarios,this paper jointly modeled task offloading,user scheduling,and UAV trajectory control as a Markov decision process(MDP),and used the deep Q-network(DQN)framework to learn approximate optimal strategies.In state modeling,factors such as UAV energy consumption constraints,user task attributes,and timeliness requirements were fully considered,with action space discretization implemented to adapt to the DQN architecture.The reward function introduced delay loss and timeout penalty mechanisms to guide the agent in adaptively learning effective offloading strategies.[Results]The simulation results show that the proposed method is superior to the benchmark strategies such as full local computing and full edge offloading in terms of cumulative rewards,average task processing delay,and the number of task timeout penalties,showing good strategy convergence and environmental adaptability,especially when the communication link fluctuates or computing resources are limited.[Conclusions]The proposed DQN-based UAV-assisted edge computing joint optimization strategy can significantly improve the system's processing efficiency and scheduling performance for time-sensitive tasks in a dynamic and complex environment,providing a feasible method path and theoretical support for the design and optimization of high-mobility mobile edge computing systems.
Keywords：UAV-assisted computingtask offloadingmobile edge computingtask schedulingdeep Q-network
Publication Date：2025-07-25
Online Publishing Date：2025-09-18(First online date of this platform, not the publication date of the document)
Pages：8( 409-416 )