To address the coordination challenges in urban transportation and logistics systems, this study investigates a scheduling optimization method driven by digital twin technology. By constructing a digital twin framework that integrates multi-source data and high-fidelity models, dynamic perception and bidirectional interaction with the physical system are realized. This research proposes a collaborative decision-making mechanism based on agent-based modeling and reinforcement learning, and focuses on exploring its application in key scenarios such as dynamic warehouse reservation and emergency logistics channel guarantee. By effectively breaking down information barriers, it achieves system-level coordinated optimization of traffic flow and logistics resources, significantly improving system operation efficiency and emergency response capabilities. This provides an innovative theoretical framework and practical path for building integrated management of smart cities.