企业信息化背景下电商到家平台跨站点协同履约的双目标优化模型与进化求解

Bi-Objective Optimization Model and Evolutionary Solution for Cross-Depot Collaborative Fulfillment in E-Commerce On-Demand Platforms under Enterprise Informatization

摘要: 本地生活到家服务平台在订单需求的时空波动与站点供给异质性作用下，容易出现供需失衡；此外，随着多人协作类订单占比的增加，履约过程呈现出技能互补和时间耦合特征。若无法在预约时间窗内实现多人员的同步到达与同步启动，将导致等待、返工和投诉等体验损失。针对跨站点运力共享治理和协同订单同步约束难以在同一框架下进行统筹和优化的问题，本文在多站点网络中引入了受控的跨站点运力共享机制，构建了统一的派单与路径排程模型，适用于普通订单和协同订单。为此，本文提出了以服务体验损失最小化和履约成本及系统稳定性损失最小化为核心的双目标优化框架。体验目标主要从协同启动偏差、迟到成本项以及严重不同步或大幅迟到的额外成本项三个维度刻画服务质量；成本与稳定目标在行驶、服务与加班成本的基础上，引入工时差异成本项来反映负载均衡与人员稳定性；同时，对触发跨站点协同的订单设置额外成本项，体现协调成本和治理风险。本文采用NSGA-II算法进行求解，并与MOEA/D和SPEA2在统一编码/解码和目标评估框架下进行公平对比。基于扩展Solomon VRPTW的多站点协同履约算例体系和不同协同强度设置的实验结果表明：随着协同订单比例的上升，可行解域显著压缩，求解难度加大。在多数算例及关键指标(HV、IGD、GD等)上，NSGA-II表现出更优的前沿覆盖和收敛性，且统计检验结果表明其差异具有显著性。研究结果为平台在成本与体验权衡下的治理参数配置与跨站点协同调度策略提供了重要的决策支持。

Abstract: Local life on-demand service platforms are prone to supply-demand imbalances due to the spatiotemporal fluctuations in order demand and the heterogeneity of depot supply capacities. Additionally, as the proportion of multi-person collaborative orders increases, the fulfillment process exhibits characteristics of skill complementarity and temporal coupling. If simultaneous arrival and simultaneous start of multiple personnel cannot be achieved within the appointment time window, it will lead to experience losses such as waiting, rework, and complaints. To address the challenge of balancing cross-depot capacity sharing governance and collaborative order synchronization constraints within a single framework, this paper introduces a controlled cross-depot capacity sharing mechanism within a multi-depot network, and constructs a unified dispatching and routing scheduling model applicable to both standard and collaborative orders. Accordingly, this paper proposes a bi-objective optimization framework focused on minimizing service experience loss and minimizing fulfillment costs and system stability loss. The experience objective characterizes service quality from three dimensions: synchronization start deviation, lateness cost items, and additional cost items for severe asynchrony or significant delays. The cost and stability objective, based on travel, service, and overtime costs, introduces a labor-hour difference cost item to reflect load balancing and workforce stability. Additionally, an extra cost item is set for orders triggering cross-depot collaboration to reflect coordination costs and governance risks. The NSGA-II algorithm is employed for solving this problem, and a fair comparison is made with MOEA/D and SPEA2 under a unified encoding/decoding and objective evaluation framework. Experimental results based on an extended Solomon VRPTW multi-depot collaborative fulfillment instance system and various collaboration intensity settings show that as the proportion of collaborative orders increases, the feasible region significantly compresses, and the solving difficulty increases. NSGA-II demonstrates better frontier coverage and convergence in most instances and key metrics (such as HV, IGD, GD), with statistical tests confirming the significance of these differences. The research results provide important decision-making support for platform governance parameter configuration and cross-depot collaborative scheduling strategies under the trade-off between cost and experience.