Abstract: This study constructs a mixed-integer programming model for large logistics companies to address the pickup and delivery problems in complex urban road networks. The model comprehensively incorporates practical operational and distribution scenarios, including multiple transshipment points, multiple vehicle types, multiple trips, divisible demands, and time windows, with the objective of minimizing usage and travel costs. To validate the effectiveness of the model, the pr01 dataset is adopted in the study case. By establishing different test scenarios, the impact of various factors on model performance is investigated. Results indicate that the model considering all five factors outperforms those considering only four, with a significant reduction in costs. When the model is not sensitive to vehicle usage cost, the non-splitable demand strategy offers advantages in both economy and convenience. The factors of multiple transshipment points and multiple trips play a critical role in cost control, neglecting either can lead to a substantial increase in cost. Moreover, from the perspective of vehicle loading rate, the single-trip strategy demonstrates the best performance, contrasting sharply with conclusions drawn from the cost-based perspective. This comprehensive model provides strong support for operational planning and decision-making in large logistics enterprises, helping to reduce operational costs, enhance transportation efficiency, and improve customer satisfaction.