Topological Game Self-Play Scheduling Framework for eVTOL Vertiport Low-Altitude Landing and Takeoff

This preprint proposes an AlphaZero-inspired topological game self-play scheduling framework to resolve core scheduling bottlenecks of high-density eVTOL vertiport urban air mobility (UAM). A modular 2.5D low-altitude airspace topological primitive system is defined to construct generable vertiport simulation environments, transforming multi-eVTOL takeoff/landing slot competition into a temporal occupancy combinatorial multi-agent cooperative-competitive game. The framework integrates Topology-aware Mixture-of-Experts (TopoMoE) graph neural networks for multi-scale airspace feature encoding and temporal discount improved MCTS-UCT for long-horizon sequential slot reasoning. A hierarchical multi-objective reward function balances flight safety, pad throughput, delay and power consumption, and teacher-student policy distillation generates lightweight airborne decision modules supporting distributed independent vehicle scheduling without central UTM solvers. Different from separated arrival/departure scheduling models, this work realizes unified modeling of landing and takeoff tasks under one graph topology and MCTS pipeline. It eliminates reliance on manual scheduling rules and global centralized optimization, featuring interpretable multi-step temporal planning, modular airspace extension, communication-fault-tolerant distributed coordination, self-emergent cooperative flight yielding behaviors and integrated arrival-departure decision-making. Staged simulation verification schemes, four core scientific assumptions and five-phase experimental roadmaps are provided for quantitative benchmarking against FCFS heuristics and MILP solvers. The research provides a novel distributed self-play scheduling technical path for dense urban vertiport operation and a general topological game paradigm for low-altitude time-resource scheduling problems.