Abstract: During urban low-altitude UAV flights, communication outages frequently occur because conventional ground BS antenna layouts are suboptimal and existing statistical low-altitude channel models fail to accurately characterize complex multipath propagation. This calls for dedicated low-altitude coverage planning and parameter optimization, where—unlike ground BS engineering-parameter tuning—the aerial trajectory must also be optimized. We propose an alternating-optimization-based joint air–ground communication framework. A high-fidelity digital-twin environment is first built by coupling ray tracing with geographic data to extract scenario-specific channel impulse responses. The joint problem is then decomposed into two coupled subproblems: on the network side, a trust-region Bayesian optimization searches high-dimensional BS antenna parameters and transmit power; on the UAV side, 3D trajectory decision-making is cast as an MDP and solved by PPO. Under a TDMA access scheme, the framework reconstructs the low-altitude coverage distribution: versus trajectory-only optimization, average SINR gains ~5 dB, with noticeable suppression of co-channel interference and building shadowing. The approach improves communication continuity in low-altitude airspace and offers a scalable strategy for smart air–ground network deployment.