AI-based intelligent scheduling and cost optimization mechanism for logistics procurement supply chain in universities

Against the backdrop of increasingly fierce competition in global higher education, university logistics systems face unique challenges of multi-objectivity, strong seasonality, and uncertain demand. This paper aims to construct an intelligent scheduling and cost optimization mechanism that integrates multi-agent reinforcement learning and graph neural networks. By designing a multi-objective optimization function and four types of collaborative agents, the MADDPG algorithm and spatiotemporal graph convolutional network are employed for demand prediction, aiming to simultaneously optimize cost, service quality, and supply chain resilience. Experiments show that this model significantly reduces the total cost of ownership by 21.4% (from 2.175 million yuan in traditional methods to 1.326 million yuan), increases the order fulfillment rate to 96.2%, accelerates inventory turnover to 8.3 times per year, and ablation experiments verify that multi-agent collaboration is a key contributing factor (with a cost increase of 14.2% after removal). This AI solution effectively addresses the multi-objective trade-offs and spatiotemporal dependencies in university logistics. In the future, lightweight models and small-sample learning need to be explored to enhance practicality.