arxiv-2609-08402 SRC-0222 · paper

Towards Embodied Air-Ground Cooperative Object Search: Benchmark, Dataset and Agentic Method

Boao Yu, Zimo Chen, Junreng Rao, Yue Hu, Zhengqiu Zhu, Yong Zhao

Created: 2026-09-08 · Ingested: 2026-09-19

Air-Ground Object Search (AGOS) in urban environments is a challenging embodied task, which requires an Unmanned Aerial Vehicle (UAV) and an Unmanned Ground Vehicle (UGV) to jointly search for and verify a specified target vehicle from multi-view visual references. To study this underexplored problem, we introduce AGOS-Bench, the first dedicated benchmark for evaluating whether general-purpose Vision-Language Models (VLMs) can integrate aerial discoveries and ground-level verification through UAV-UGV cooperation. We further provide AGOS-Dataset as the companion resource of exemplary trajectori

Referenced by

Claims in this catalog that draw on this source, and whether as support or counterpoint.

supportsIn simulated urban air-ground object search where a VLM drives both a UAV and a UGV, wrapping the model in a fixed search-handoff-verify protocol that leaves geometric projection and road-graph path planning to deterministic tools raises task success for backbones with adequate perception while cutting decision calls for all, but does not lift weak small open models, whose success can fall on harder splits.