Abstract: Aiming to the extreme environmental challenges of network disconnection, power failure and road blockage in complex confined spaces such as long railway tunnels under sudden disasters, as well as the inherent defects of traditional wireless ad-hoc networks in long-hop chain topologies—including single-point gateway failure, severe hidden terminal collisions, high multi-service delays and sharp throughput degradation—this paper proposes an air-ground cooperative ad-hoc network architecture featuring air-ground collaborative networking and ground-based multi-hop access as well as a cross-layer scheduling mechanism. First, wireless fading models and deployment curvature constraint models applicable to tunnel confined spaces are established to quantitatively characterize signal transmission attenuation and node deployment limitations. Second, a distributed gateway reconstruction algorithm based on multi-attribute utility maximization is designed, and a backoff scale factor is introduced to alleviate congestion on control channels and improve reconstruction efficiency. Third, a physical-layer and medium access control (MAC) layer scheduling mechanism oriented to heterogeneous service priorities is proposed to suppress hidden terminal interference, and an asymmetric priority backoff algorithm based on the Markov steady-state model is constructed to optimize transmission scheduling. Simulation and field tests show that the average topology reconstruction time is controlled within 8.2 s; under a 9-hop cascaded long-chain topology constrained by ground-based multi-hop bottlenecks, the stable end-to-end available bandwidth reaches 5.0 Mbps and the average single-hop delay is 6.7 ms. The proposed mechanism effectively overcomes delay deterioration and bandwidth attenuation in long-chain confined-space topologies, realizes deterministic scheduling for multi-source heterogeneous services, and provides reliable theoretical and equipment-technical support for high-survivability, high-reliability air-ground cooperative emergency communications in tunnel disaster rescue scenarios.