AirGroundVLN: A Large-Scale Benchmark for Goal-Oriented Air-Ground Collaborative Vision-and-Language Navigation

Zhenxuan Zeng; Qingle Wu; Wei Suo; Maojia Wu; Bairong Zhang; Hangzheng Yu; Peng Wang

cs.ROOct 7, 2026

AirGroundVLN: A Large-Scale Benchmark for Goal-Oriented Air-Ground Collaborative Vision-and-Language Navigation

Authors: Zhenxuan Zeng, Qingle Wu, Wei Suo, Maojia Wu, Bairong Zhang, Hangzheng Yu, Peng Wang

Organizations: School of Computer Science, Northwestern Polytechnical University, China.

Abstract

Goal-oriented Vision-and-Language Navigation (VLN) requires agents to locate and reach targets described in natural language without prescribed routes. Air--ground collaboration is valuable for tasks requiring both wide-area search and fine-grained localization. However, systematic study of goal-oriented air--ground collaborative VLN remains limited by the lack of large-scale, diverse benchmarks and two core challenges: 1) substantial differences between aerial and ground views, together with useful observations becoming unavailable as navigation proceeds, make it difficult to maintain spatially consistent context across platforms and over time; and 2) asymmetric spatial observability makes ground perception locally detailed but spatially limited and aerial perception broad but locally coarse, limiting the reliability of single-platform planning. To address these limitations, we introduce AirGroundVLN, a benchmark containing 10,281 navigation episodes and 955 target instances across 19 Unreal Engine environments, with seen/unseen splits and an aerial-visibility protocol for systematic evaluation. Alongside the benchmark, we propose AG-CoNAV, a trainable reference framework comprising two key components: Spatiotemporally Anchored Collaborative Memory (SACM) and Aerial-Guided Regional-to-Local Planning (AGRLP). SACM maintains and retrieves spatially consistent historical context across aerial and ground observations. Meanwhile, AGRLP combines regional aerial guidance with fine-grained ground navigation. Extensive experiments demonstrate the effectiveness of AG-CoNAV and establish AirGroundVLN as a comprehensive benchmark for future exploration.

Figures & tables

Figure 1: Overview of AirGroundVLN . The benchmark contains 10,281 navigation episodes, 955 target instances, and 19 diverse scenes for goal-oriented air–ground collaborative VLN. It supports evaluation across aerial-visible/aerial-occluded target conditions and seen/unseen environments. AG-CoNAV is provided as a trainable reference framework for this setting.

Figure 2: Overview of AirGroundVLN data collection and an illustrative episode . The collection process includes target and environment description, start-location selection, goal-oriented instruction generation, and dense air–ground waypoint generation.

Benchmark	Ground	Air	Simulator	$\mathbf{N_{\mathrm{traj}}}$	$\mathbf{L_{\mathrm{traj}}^{\mathrm{ground}}}$	$\mathbf{L_{\mathrm{traj}}^{\mathrm{aerial}}}$	$\mathbf{L_{\mathrm{instr}}}$	$\mathbf{N_{\mathrm{vocab}}}$
R2R ( Anderson et al., 2018b )	$\surd$	$\times$	Matterport3D	7,189	10.0	$\times$	29	3.1K
REVERIE ( Qi et al., 2020 )	$\surd$	$\times$	Matterport3D	7,234	10.0	$\times$	18	1.6K
RxR ( Ku et al., 2020 )	$\surd$	$\times$	Matterport3D	13,992	14.9	$\times$	129	7.0K
VLN-CE ( Krantz et al., 2020 )	$\surd$	$\times$	Habitat	4,475	11.1	$\times$	–	4.3K
AerialVLN ( Liu et al., 2023 )	$\times$	$\surd$	AirSim	8,446	$\times$	661.8	83	4.5K
TravelUAV ( Wang et al., 2025 )	$\times$	$\surd$	OpenUAV	12,149	$\times$	255.0	104	10.8K

Table 1: Comparison with representative VLN benchmarks.

N_{\mathrm{traj}}

denotes the number of navigation episodes (paired air–ground trajectories for AirGroundVLN), while

L_{\mathrm{traj}}

and

L_{\mathrm{instr}}

denote average trajectory length (m) and instruction length (words), respectively. “–” indicates statistics not officially reported by the corresponding benchmark.

Figure 3: Statistics of AirGroundVLN . (a) Distribution of UAV and UGV reference trajectory lengths. (b) Distribution of navigation instruction lengths. (c) Distribution of target instances across six major semantic categories.

Figure 4: Overview of the AG-CoNAV framework . Given goal-oriented instructions, aerial–ground observations, and a UAV-maintained fused BEV map, SACM maintains and retrieves spatially consistent context across views and time, while AGRLP coordinates regional aerial guidance with fine-grained ground execution.

Training-Free Methods
Method	Seen					Unseen
	SR $\uparrow$	OSR $\uparrow$	SPL $\uparrow$	NE UGV $\downarrow$	NE UAV $\downarrow$	SR $\uparrow$	OSR $\uparrow$	SPL $\uparrow$	NE UGV $\downarrow$	NE UAV $\downarrow$
Random	0.52	0.52	0.34	79.66	–	0.47	0.59	0.23	72.43	–
Air-Ground (Qwen3.5-2B)	4.05	4.26	2.38	81.08	100.85	1.30	1.77	0.89	69.60	92.12
Air-Ground (Qwen3.5-4B)	13.10	15.59	9.54	63.33	68.47	10.14	11.56	6.75	54.96	62.58
AGC-VLN (Qwen3.5-4B)	13.41	16.63	8.42	65.75	63.99	10.50	12.38	4.95	54.37	51.16

Table 2: Comparison with baselines on AirGroundVLN. Results are reported across seen and unseen environments and under different aerial visibility conditions.

SACM	AGRLP	Seen					Unseen
SACM	AGRLP	SR $\uparrow$	OSR $\uparrow$	SPL $\uparrow$	NE UGV $\downarrow$	NE UAV $\downarrow$	SR $\uparrow$	OSR $\uparrow$	SPL $\uparrow$	NE UGV $\downarrow$	NE UAV $\downarrow$
✗	✗	25.68	27.65	22.63	48.59	55.21	17.10	19.93	15.61	50.02	49.30
✗	✓	26.30	29.42	24.03	46.71	54.49	17.92	20.64	16.18	48.59	48.76
✓	✗	31.39	33.47	27.79	41.93	45.62	19.22	22.29	17.56	45.76	45.13
✓	✓	32.64	35.65	29.85	39.72	44.81	20.05	24.06	18.46	44.43	44.04

Table 3: Ablation of SACM and AGRLP components in AG-CoNAV on seen/unseen environments.

Temporal Prior	Spatial Prior	Seen					Unseen
Temporal Prior	Spatial Prior	SR $\uparrow$	OSR $\uparrow$	SPL $\uparrow$	NE UGV $\downarrow$	NE UAV $\downarrow$	SR $\uparrow$	OSR $\uparrow$	SPL $\uparrow$	NE UGV $\downarrow$	NE UAV $\downarrow$
✓	✗	29.83	31.91	26.75	43.92	47.03	18.99	21.82	16.92	47.14	46.19
✗	✓	31.19	34.10	28.56	41.37	45.92	19.69	23.11	17.53	45.61	44.86
✓	✓	32.64	35.65	29.85	39.72	44.81	20.05	24.06	18.46	44.43	44.04

Table 4: Detailed ablation of the temporal and dispersion-aware spatial priors in SACM on seen and unseen environments. AGRLP is enabled for all variants.

Appendix figures & tables10 assets

Supplementary material from the paper’s appendix.

Appendix

Figure 5: Prompt for target instance description.

Figure 6: Prompt for environmental context description.

Figure 7: Prompt for goal-oriented instruction synthesis.

Figure 8: Representative aerial-visible and aerial-occluded targets.

Figure 9: Additional statistics of AirGroundVLN. (a) Target name word cloud. (b) Visual attribute word cloud.

Figure 10: Air–ground waypoint generation prompt.

Figure 11: Target verification prompt.

Figure 12: Bird’s-eye-view trajectory comparisons on AirGroundVLN. Joint UAV and UGV trajectories from four methods (Qwen3.5-2B, GPT-5.5, UGV + Global BEV Map , and AG-CoNAV) are visualized across three representative episodes alongside ground-truth (GT) paths.

Figure 13: Qualitative navigation trajectory visualization on AirGroundVLN. Navigation trajectories toward language-specified targets are visualized in seen environments with visible aerial observations.

Figure 14: Qualitative navigation trajectory visualization on AirGroundVLN. Navigation trajectories toward language-specified targets are visualized in unseen environments with occluded aerial observations.

Explore similar work

Air-Ground Collaborative Vision-and-Language Navigation via Shared Bird's-Eye Maps
Sep 3, 2026Shuning Zhang, Liang Li, Yunheng Wang +3Aerial Vision-Language Navigation Cooperative Air Combat
Robotics and Autonomous Systems Thrust, Systems Hub, The Hong Kong University of Science and Technology (Guangzhou), China · School of Communications and Information Engineering, Nanjing University of Posts and Telecommunications, China
CoNav-UAV: Cooperative Dual-Altitude Aerial Navigation via Stackelberg Learning
Aug 3, 2026Junru Song, Wenhao Zhang, Yang Yang +7Aerial Vision-Language Navigation Multi-Uav
1Shanghai Jiao Tong University · 2Intelligent Game and Decision Laboratory · 3Renmin University of China +1
From Region Arrival to Instance-Level Grounding in Vision-and-Language Navigation
Jul 4, 2026Xiangyu Shi, Ruoxi Yang, Wei Tao +3Vision-Language Navigation Contextual Grounding
Australian Institute for Machine Learning, Adelaide University · Ruyi Dynamics · Fudan University +1

Sep 3, 2026, cs.ROJ/K move · Enter open · S save

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

Shuning Zhang, Liang Li, Yunheng Wang, Tao Wang+2

Robotics and Autonomous Systems Thrust, Systems Hub, The Hong Kong University of Science and Technology (Guangzhou), China · School of Communications and Information Engineering, Nanjing University of Posts and Telecommunications, China

Aerial Vision-Language Navigation Cooperative Air Combat Bird'S-Eye View Navigation Single-Agent Baselines

Open paper AskPDF arXiv

Air-ground collaborative Vision-and-Language Navigation (VLN) pairs an unmanned aerial vehicle (UAV) with a global bird's-eye view and an unmanned ground vehicle (UGV) with a local first-person view, yet the setting remains largely unexplored: existing training-free methods solve single-agent tasks but offer no collaboration mechanism, and a recent CARLA-Air evaluation found no stable cooperative behavior across five state-of-the-art VLA models; naive semantic communication or bidirectional coupling even degrades performance. We establish AGC-VLN (Air-Ground Collaborative VLN), the first training-free baseline for air-ground collaborative VLN. The key insight is that training-free methods decompose navigation into VLM-based semantic reasoning and deterministic geometric execution, exposing a collaboration interface: the UAV's global view, over which it renders the UGV's reported pose and the VLM-anchored target as CAR/GOAL markers with distance labels, yielding a shared bird's-eye map. From this map, the UGV acquires global spatial context its first-person view cannot provide, plans a road-following path with a frozen VLM, and executes it under closed-loop control; in parallel, the UAV runs 3D-SPF, a spatial-search upgrade of SPF that localizes the target in the downward view and flies toward it. On 100 closed-loop episodes in CARLA-Air's Town10HD scene, AGC-VLN reaches a 77.0% joint success rate, a collaboration gain of +27.0% over the weaker individual agent (the UAV, 50.0%), and exceeds the strongest published single-agent baseline (Travel UAV, 53.0%) by 24.0 points, stemming from the complementarity of the UAV's global view and the UGV's road-following execution. Project page: https://github.com/ZSN2024/AGC-VLN.

Abstract

Explore similar work

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

CoNav-UAV: Cooperative Dual-Altitude Aerial Navigation via Stackelberg Learning

From Region Arrival to Instance-Level Grounding in Vision-and-Language Navigation