AGC-VLN: Air-Ground VLN via Shared Bird's-Eye Maps
Loading video
Loading videoAGC-VLN is the first training-free baseline for air-ground collaborative vision-and-language navigation. The drone renders the ground robot's reported pose and the VLM-anchored target as CAR and GOAL markers with distance labels onto a shared bird's-eye map; the ground robot plans a road-following path from that global view and the UAV flies 3D-SPF. No learned communication protocol and no cross-agent training: 77.0% joint success over 100 closed-loop CARLA-Air Town10HD episodes, lifting the ground robot from 12% alone to 75%.
VLNAir-Ground CollaborationUAV-UGVNavigationTraining-FreeCARLAEmbodied AIAerial-ground collaborationVision-language navigation
Category: research
Author: @heetezition
Date: 2026-09-07T00:00:00
Duration: 119.304s
Reference: https://arxiv.org/abs/2609.03483
