強(qiáng)化圖注意力網(wǎng)絡(luò)模型在選址路徑問題中的應(yīng)用

打開文本圖片集

中圖分類號(hào)：TP391 文獻(xiàn)標(biāo)志碼：ADOI： 10.12475/aie.20250307

本文引用格式：，，.等.強(qiáng)化圖注意力網(wǎng)絡(luò)模型在選址路徑問題中的應(yīng)用[J].自動(dòng)化與信息工程，2025，46（3）：43-51.HUANG Shuo， ZHANG Xuexi， XIE Xingwang， et al. Application of reinforced graph atention network model forlocation routing problem[J]. Automation & Information Engineering，2025，46（3）：43-51.

文章編號(hào)：1674-2605（2025）03-0007-09開放獲取

Application of Reinforced Graph Attention Network Model for Location Routing Problem

HUANG Shuo ZHANG XuexiXIE XingWang ZHANG Tao （Guangdong University of Technology， Guangzhou 51ooo6， China）

Abstract：Toaddress the limitationof traditional atention network modelsinefectivelypreserving graph structural informationforLocationRoutingProblems，thispaperproposesareinforced graphatentionnetworkmodel.First，whileretaining graph structuralinformation，theencoderextracts node informationfromthe graph structurethroughatention mechanisms toobtain high-dimensional featurerepresentationsofnodesandglobal graph feature information.Then，thedecoderutilzes GatedRecurrent Units to efectivelycapture temporaldependencies innodesequencesandacquirescompletesolutions viastep-by-stepdecodng. Finaly，anauxiliaryValueNetwork isintroduced toevaluatethevalueofachaction，guiding policyupdatestoenhanctraiing efficiency.Experimentalresultsdemonstratethatthisreinforcedgraphatentionnetworkmodelcanrapidlyobtainhigh-quality solutions for LRP.

Keywords： location routing problem; reinforcement learing; graph atention network; value network; graph structure

0 引言

選址路徑問題（location routing problem，LRP）是一個(gè)組合優(yōu)化問題，旨在同時(shí)優(yōu)化設(shè)施位置與車輛運(yùn)輸路徑，以最小化總成本，廣泛應(yīng)用于物流、供應(yīng)鏈管理等領(lǐng)域。（剩余10980字）