In this article, a new cross-docking approach with two characteristics, namely paired-doors and preemption, is developed based on the sustainable development goals (SDG12) paradigm that expresses targets to decrease the food losses throughout the supply chain. For every receiving door, there is a corresponding shipping door in front of it. The packages can only be transferred from a receiving door to its shipping counterpart. This new cross-dock is suitable for distributing the perishable products due to the products’ time-sensitive feature. The proposed cross-docking system, which is easier to implement considering the lower automation level required, is compared with the conventional approach to evaluate the new characteristics. Moreover, a genetic algorithm and grey wolf optimizer are proposed to solve the model for the large-sized problem instances. The results show that the proposed cross-docking approach can transfer the products faster than the traditional approach. Also, a predictive model is built by a machine learning algorithm based on the mathematical modeling outputs to predict the makespan by analyzing the factors such as the total product number, number of inbound and outbound vehicles, and the number of paired doors. The results showed that the predictive model could predict the makespan (with an average of 92.8%).

在本文中，基于可持续发展目标（SDG12）范式开发了一种具有两个特征的新交叉配送方法，即配对门和抢占，该范式表达了减少整个供应链中粮食损失的目标。对于每个收货门，其前面都有一个相应的发货门。包裹只能从接收门转移到其运输对应门。由于易腐烂产品具有时间敏感性，这种新型交叉配送中心适合配送这些产品。考虑到所需的自动化水平较低，所提出的越库配送系统更容易实施，与传统方法进行比较以评估新特性。而且，提出了遗传算法和灰狼优化器来解决大型问题实例的模型。结果表明，所提出的越库配送方法可以比传统方法更快地转移产品。同时，根据数学建模输出，通过机器学习算法建立预测模型，通过分析产品总数、进出车辆数量、配对门数量等因素来预测完工时间。结果表明，预测模型可以预测完工时间（平均为92.8%）。基于数学建模输出，通过机器学习算法建立预测模型，通过分析产品总数、进出车辆数、配对门数等因素来预测完工时间。结果表明，预测模型可以预测完工时间（平均为92.8%）。基于数学建模输出，通过机器学习算法建立预测模型，通过分析产品总数、进出车辆数、配对门数等因素来预测完工时间。结果表明，预测模型可以预测完工时间（平均为92.8%）。