Real-Time Online Low-Complexity Passenger Scheduling and Resource Optimization for Networked Multi-Airport Systems

The growing complexity of resource scheduling (e.g., coordinating airport slots, gates, and passengers) in multi-airport systems (MAS) poses significant challenges in real-time coordination, including high computational load and communication inefficiency. To tackle these issues, this work proposes a novel architecture for Networked Multi-Airport Systems (NMAS) and takes passenger scheduling as an entry point to optimize resource scheduling in MAS. First, we develop a multi-agent reinforcement learning method using fluid approximation and an approximate Bellman equation, establishing an efficient, low-complexity, and large-scale distributed framework for real-time passenger scheduling and resource optimization. Second, we design a directional communication protocol to coordinate communication-efficient dynamic scheduling. Finally, a novel SCA-Critic-only algorithm (successive convex approximation and Critic-only learning) is proposed within this framework. The effectiveness and superiority of the proposed scheme are verified using both synthetic and real-world airport operational datasets. Numerical results reveal that the proposed SCA-Critic-only algorithm can achieve significant performance gains over various state-of-the-art baselines.