The distribution of emergency perishable materials after a disaster, such as an earthquake, is an essential part of emergency resource dispatching. However, the traditional single-period distribution model can hardly solve this problem because of incomplete demand information for emergency perishable materials in affected sites. Therefore, for such problems we firstly construct a multi-period vehicle path distribution optimization model with the dual objectives of minimizing the cost penalty of distribution delay and the total corruption during delivery, and minimizing the total amount of demand that is not met, by applying the interval boundary and most likely value weighting method to make uncertain demand clear. Then, we formulate the differential evolutionary whale optimization algorithm (DE-WOA) combing the differential evolutionary algorithm with the whale algorithm to solve the constructed model, which is an up-and-coming algorithm for solving this type of problem. Finally, to validate the feasibility and practicality of the proposed model and the novel algorithm, a comparison between the proposed model and the standard whale optimization algorithm is performed on a numerical instance. The result indicates the proposed model converges faster and the overall optimization effect is improved by 23%, which further verifies that the improved whale optimization algorithm has better performance.

As the logistics industry modernizes, living standards improve, and consumption patterns shift, the demand for fresh food continues to grow, making cold chain logistics for perishable goods a critical component in ensuring food quality and safety. However, the presence of both soft and hard time windows among demand nodes can complicate the single-network distribution of perishable goods. In response to these challenges, this paper proposes an optimization model for multi-distribution center perishable goods delivery, considering both one-echelon and two-echelon network joint distributions. The model aims to minimize total costs, including transportation, fixed, refrigeration, goods damage, and penalty costs, while measuring customer satisfaction by the start time of service at each demand node. A two-stage heuristic algorithm is designed to solve the model. In the first stage, an initial solution is constructed using a greedy approach based on the principles of the k-medoids clustering algorithm, which considers both spatial and temporal distances. In the second stage, the initial routing solution is optimized using a linear programming approach from the Ortools solver combined with an Improved Adaptive Large Neighborhood Search (IALNS) algorithm. The effectiveness of the proposed model and algorithm is validated through a case study analysis. The results demonstrate that the initial solutions obtained through the k-medoids clustering algorithm based on spatio-temporal distance improved the overall cost optimization by 1.85% and 4.74% compared to the other two algorithms. Among the three two-stage heuristic algorithms, the Ortools-IALNS proposed here showed enhancements in the overall cost optimization over the IALNS, with improvements of 3.24%, 1.12%, and 0.41%, respectively. The two-stage heuristic algorithm designed in this study also converged faster than the other two heuristic algorithms, with overall optimization improvements of 1.55% and 1.28%, further validating the superior performance of the proposed heuristic algorithm.

Network-scale traffic speed prediction plays a central role in signal timing, routing, traffic control and congestion management. However, existing methods face challenges in capturing complex spatiotemporal dependencies among road segments, and in achieving computational efficiency in large-scale urban networks. To address these challenges, we propose an enhanced Dynamic Spatiotemporal Residual Network (DST-ResNet) framework for network-scale traffic speed prediction. First, the method employs a multi-scale grid partitioning strategy to segment urban road networks at varying levels of granularity, enabling precise predictions at both local and global scales. Second, a dynamic convolution mechanism is introduced to integrate real-time traffic data with static road network features, allowing the model to adaptively capture the dynamic and heterogeneous characteristics of urban traffic. Then, a spatiotemporal attention mechanism is incorporated to capture sequential dependencies in traffic patterns, improving the accuracy and robustness of long-term traffic speed prediction. Last, we conduct experiments based on GPS trajectory data from taxis in Xi’an city, China. The results demonstrate that DST-ResNet outperforms five baseline models. The model maintains stable performance across five spatial resolutions, and a mask-aware retraining strategy reduces the average MSE by 49.65% under missing observations. Ablation studies confirm the contributions of dynamic convolution and attention mechanisms, while residual units mitigate the vanishing gradient problem, facilitating deeper network training.