{"title":"Collaborative computation offloading and trajectory planning in locally observable multi-UAV MEC networks","authors":"Yuan He , Xie Jun , Yaqun Liu , Xijian Luo","doi":"10.1016/j.comnet.2025.111554","DOIUrl":null,"url":null,"abstract":"<div><div>Unmanned Aerial Vehicles (UAVs) can be deployed as aerial edge servers to provide ground users with mobile edge computing (MEC) services in a collaborative manner. In this paper, we investigate the motion and computation decision-making problem of the multi-UAV-assisted MEC in a locally observed environment. Firstly, a multi-UAV-assisted MEC network model is proposed taking into account the collaborative computation among UAVs and the task’s queuing delay. Secondly, we propose a reinforcement learning algorithm based on Graph Attention Networks (GAT) named GatPPO to implement UAV control based on local information. The local networks observed by UAVs are abstracted into heterogeneous and homogeneous graphs. Then, we extract and aggregate the graphs’ features with GAT to alleviate the problem of inconsistent input dimensions caused by the number of neighbors and users changing. In addition, two actor-critic networks are designed in each UAV agent for the trajectory planning and computation decisions respectively to solve the problem of asynchronous actions selection due to different frequencies. The numerical simulation results show that compared with the benchmark algorithms, GatPPO reduces the computation delay by about 10 %–30 % and improves user satisfaction by about 113 % at most when a single UAV has limited computing resources.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"270 ","pages":"Article 111554"},"PeriodicalIF":4.4000,"publicationDate":"2025-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Computer Networks","FirstCategoryId":"94","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S1389128625005213","RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"COMPUTER SCIENCE, HARDWARE & ARCHITECTURE","Score":null,"Total":0}
引用次数: 0
Abstract
Unmanned Aerial Vehicles (UAVs) can be deployed as aerial edge servers to provide ground users with mobile edge computing (MEC) services in a collaborative manner. In this paper, we investigate the motion and computation decision-making problem of the multi-UAV-assisted MEC in a locally observed environment. Firstly, a multi-UAV-assisted MEC network model is proposed taking into account the collaborative computation among UAVs and the task’s queuing delay. Secondly, we propose a reinforcement learning algorithm based on Graph Attention Networks (GAT) named GatPPO to implement UAV control based on local information. The local networks observed by UAVs are abstracted into heterogeneous and homogeneous graphs. Then, we extract and aggregate the graphs’ features with GAT to alleviate the problem of inconsistent input dimensions caused by the number of neighbors and users changing. In addition, two actor-critic networks are designed in each UAV agent for the trajectory planning and computation decisions respectively to solve the problem of asynchronous actions selection due to different frequencies. The numerical simulation results show that compared with the benchmark algorithms, GatPPO reduces the computation delay by about 10 %–30 % and improves user satisfaction by about 113 % at most when a single UAV has limited computing resources.
期刊介绍:
Computer Networks is an international, archival journal providing a publication vehicle for complete coverage of all topics of interest to those involved in the computer communications networking area. The audience includes researchers, managers and operators of networks as well as designers and implementors. The Editorial Board will consider any material for publication that is of interest to those groups.