基于分层多智能体的轨道车辆装配协同 调度优化研究

打开文本图片集

中图分类号：TP399 文献标识码：A DOI：10.7535/hbkd. 2026yx02004

Abstract：Todeal with the scheduling problem inrail vehicleassembly，where assemblylinetask alocation iscomplexand carbodycomponentsrequirefrequentcrosstationtransfersrelyingontrolleys，thisstudyproposedanend-toendhierarchical multi-agentdepeinforcementlearningframeworkforshedulingoptimizationFirstlytheallcationofassemblytasksacross multipleassemblylines was modeledasasequentialdecisionproblem.Thehigh-levelagent encodedtheassemblytaskandline features using a Transformer and generated line assgnment strategies with a Pointer Network. Secondly，the lower-level agents coordinated the selectionofoperations，stationasignments，anddollyscheduling，andused Graph Atention Networks to extractrelationalfeatures from heterogeneous nodes.Finaly，multiplecomparison experiments wereconducted to validate theefectivenessof the proposedmethod.Theresultsshowthatthemethodachievesoptimal scheduling across different instance scales. The coordination of low-level agent strategies achieves an average maximum makespan gap of 11.36% ，which outperforms the 15.00% achieved by the graph isomorphism network method，and the method provides high-quality scheduling withcomputation eficiencysignificantly higher than theLateAcceptance Hill Climbing algorithm.Theproposed hierarchical collaborative schedulingframework achievesunified modelingandcoordinatedoptimizationof assemblytask assignmentand multi-esourcescheduling，providinganeficientandadaptableintellgentoptimizationapproachforailvehicleassemblysheduling.

KeyWords：computer aided manufacturing；rail vehicle asembly；deepreinforcement learning；multi-agent；hierarchical collaborative scheduling

轨道车辆装配系统的生产组织形式与离散制造系统高度契合[1]，其生产流程具有工艺链条长、装配工位数量多以及推进调度复杂等特点。（剩余17921字）