This paper presents a human-centered computing design approach for passenger experience optimization in smart airports. Based on WiFi traces, security check queue logs, self-service check-in records, indoor positioning streams, and 186,420 groups of passenger-service events and interactive feedback from the three terminal areas, a unified multi-source service computing framework is constructed. Passenger states are encoded by timing diagram features, context embedding, and service touch point representation, and the service decision module generates adaptive guidance, window diversion, and interactive prompts. The feedback-driven collaboration layer updates the service policy based on real-time satisfaction signals and waiting time changes. Experimental results show that the proposed method achieves 91.8%experience score, 89.6%response consistency and 0.914 F1 value, which is better than BiLSTM, GCN and Transformer benchmarks, and maintains high response stability and consistency in multi-period scenarios. The average reasoning delay of this method is maintained at 84 ms, which shows strong applicability of human-centered service computing in a large terminal operating environment.
