Training-Free Air-Ground Collaborative Vision-and-Language Navigation via Bird's-Eye Maps

The AGC-VLN framework enables collaboration between UAVs with global views and UGVs with local views using shared bird's-eye maps. It decomposes navigation into VLM-based semantic reasoning and deterministic geometric execution to avoid the performance degradation seen in naive bidirectional coupling.