Task-Oriented Communications for Edge-Assisted Multi-View Localization

Zhengru Fang; Huanhuan Lou; Senkang Hu; Yihang Tao; Zongdian Li; Yiqin Deng; Jingjing Wang; Yuguang Fang

cs.NISep 28, 2026

Task-Oriented Communications for Edge-Assisted Multi-View Localization

Authors: Zhengru Fang, Huanhuan Lou, Senkang Hu, Yihang Tao, Zongdian Li, Yiqin Deng, Jingjing Wang, Yuguang Fang

Organizations: Hong Kong JC STEM Lab of Smart City and the Department of Computer Science, City University of Hong Kong, Hong Kong · School of Cyber Science and Technology, Beihang University, Beijing, China

Abstract

Unmanned aerial vehicles (UAVs) and unmanned ground vehicles (UGVs) often lose satellite positioning in urban canyons, indoor facilities, and jammed or spoofed environments, making vision-based matching with geo-tagged databases important for absolute positioning. However, limited onboard computation and energy often require localization to be offloaded over wireless links with time-varying throughput. We present a network-adaptive task-oriented communication framework that jointly determines when to offload, which views and semantic rate to transmit, and which client to serve. The framework combines scalable orthogonality-regularized variational information bottleneck (O-VIB) encoding, value-of-information (VOI)-guided request control, and VOI-weighted Lyapunov scheduling. O-VIB supports importance-ordered latent prefixes and different view subsets, while edge assistance is requested only when its predicted reduction in localization risk exceeds the communication and service cost. Under a matched per-route traffic budget, VOI-guided control reduces mean and p95 route errors by 24.8% and 31.0% over budgeted periodic offloading on CARLA multi-view UAV data. In real-world indoor UAV and UGV experiments, the pipeline reduces mean position error by 28.0% and 14.4% over uncompressed all-view CLIP retrieval while cutting descriptor traffic by 98.6% and 98.2%, respectively. Under high congestion, value-aware shaping reduces edge-side p95 latency for the top-10% high-value requests by 76.2%, from 137.7 ms to 32.8 ms.

Figures & tables

Fig. 1: Motivating scenario and edge-assisted multi-view localization pipeline. Panel (a) illustrates localization drift and collision risk when GNSS is unavailable or unreliable. Panel (b) shows the proposed pipeline: the mobile client observes multiple views, selects a task-oriented semantic representation, transmits it over the client-edge link, and receives a corrected pose after edge decoding and database-assisted localization.

Feature	[ 16 ]	[ 10 ]	[ 30 , 31 ]	[ 6 , 7 , 5 ]	[ 40 ]	[ 9 ]	[ 24 , 47 , 25 ]	[ 1 ]	Proposed work
Task-oriented compression	✓	✓	$\triangle$	$\times$	$\triangle$	$\times$	$\times$	✓	✓
Variable-rate representation	$\triangle$	✓	✓	$\times$	$\triangle$	$\times$	$\times$	$\times$	✓
Request control	$\times$	$\times$	$\triangle$	$\times$	✓	$\triangle$	$\times$	$\times$	✓
View selection	$\times$	$\times$	$\triangle$	$\times$	$\times$	$\times$	$\times$	$\times$	✓
Coarse-to-fine retrieval	$\times$	$\times$	$\times$	$\triangle$	$\times$	$\triangle$	$\times$	$\times$	✓
Uncertainty feedback	$\times$	$\times$	$\times$	$\times$	✓	$\triangle$	$\times$	$\times$	✓

TABLE I: Feature coverage across representative works.

Fig. 2: System architecture of network-adaptive edge-assisted multi-view localization. The client maintains a local fast loop, requests edge localization only when the predicted value is high, and receives correction feedback after shared wireless and edge-scheduling stages.

Fig. 3: Scalable O-VIB encoder for view-rate adaptive semantic localization. View masks select the available camera subset, while nested latent prefixes provide low-, medium-, and high-rate semantic payloads for one deployed encoder.

Fig. 6: Self-collected multi-scene multi-view UAV data rendered in the CARLA simulator. The synchronized five-view RGB observations and logged simulator poses support retrieval, feature extraction, and model training.

Setting	Data path	Scenes	Frames/routes	Views	Ground truth	Main purpose
Single-scene split	Conference-version Town05	1 scene	4,994; 3,495/749/750	5	Simulator pose	Scalable compression and ablations
Raw multi-scene data	CARLA UAV	5 scenes	6,684 frames	5	Simulator pose	Raw samples and retrieval database
Sampled benchmark	CARLA subset	5 scenes	4,593 frames	5	Simulator pose	Coarse-to-fine retrieval and plots
Control workload	Town02 route replay	Held-out route	593 slots / 7 policies	Variable	Simulator pose	$g_{\psi}$ request and mode control
Retrieval queries	Offline query set	5 scenes	400 queries	5	Simulator pose	Scene coarse + tile-pruned fine search
Indoor UAV	Qualisys capture	4.1 m $\times$ 4.0 m	76 ref. / 58 query	5 sequential	6-DoF mocap	Real-scene compact localization

TABLE II: Datasets and evaluation configurations used in the study.

Method	Models	Storage	Rates	Latency @ 8 KB/s
Fixed-rate O-VIB-32	1	2.8 MB	1	50.4 ms
Fixed-rate O-VIB-128	1	2.9 MB	1	100.6 ms
Fixed-rate bank $\{32,128\}$	2	5.7 MB	2	50.4–100.6 ms
Scalable O-VIB $k=35$	1	13.9 MB	128	52.0 ms
Scalable O-VIB $k=16$	1	13.9 MB	128	42.0 ms

TABLE III: Deployment overhead of fixed-rate and scalable O-VIB configurations under the Jetson Orin NX latency profile.

Fig. 9: Reproduction of the conference-version single-scene O-VIB experiments. Panel (a) shows ARD-controlled active dimensions and error. Panels (b) and (c) report the fixed-

k=16

sweep, and panel (d) reports Jetson-profile compute cost.

Fig. 11: Calibration and evaluation of VOI-Control. Panel (a) shows the train-only context/action median residual. F, FS, H4, and A5 denote front, front-plus-sides, four-horizontal, and all-five views. C and R denote $k=16$ and $k=32$ . Panel (b) shows tuning-seed budget search, and panel (c) shows held-out policies with the oracle action.

High congestion: $N=30$ 4 KB/s weak link
Scheduler	E2E p95	Edge p95	Stale	Cost
MaxWeight	198 ms	142.4 ms	0.022 m	1.87
Base-DPP	193 ms	137.7 ms	0.022 m	1.82
VOI-Lyap.	88.5 ms	32.8 ms	0.012 m	1.12

TABLE IV: High-congestion scheduler comparison for top-10% high-value requests.

Fig. 15: Real-world UAV/UGV testbed and edge-assisted localization pipeline. The platforms transmit visual inputs over Wi-Fi, while Qualisys provides pose ground truth.

Plat.	Representation	KB	Mean	p90	R@0.5
			(m)	(m)
UAV	CLIP	10.000	0.407	0.758	86.2%
UAV	O-VIB-8	0.051	0.286	0.458	89.7%
UAV	O-VIB-32	0.145	0.293	0.447	91.4%
UAV	O-VIB-128	0.520	0.274	0.459	93.1%
UGV	CLIP	8.000	0.346	0.396	98.6%

TABLE V: Held-out real-scene localization for representative O-VIB prefixes.

Explore similar work

PiLoT v2: Pixel-to-Orthogonal Map Alignment for Free-view UAV Geo-localization
Jun 30, 2026Xinyi Liu, Xiaoya Cheng, Rouwan Wu +4Drone-View Geo-Localization Global Navigation Satellite System
National University of Defense Technology
Joint Optimization of Trajectory Control, Resource Allocation, and Task Offloading for Multi-UAV-Assisted IoV
May 6, 2026Maoxin Ji, Qiong Wu, Pingyi Fan +4Multi-Uav Power Allocation
School of Internet of Things Engineering, Jiangnan University, Wuxi 214122, China · School of Information Engineering, Jiangxi Provincial Key Laboratory of Advanced Signal Processing and Intelligent Communications, Nanchang University, Nanchang 330031, China · Department of Electronic Engineering, State Key laboratory of Space Network and Communications, Beijing National Research Center for Information Science and Technology, Tsinghua University, Beijing 100084, China +4
A Distributed Multi-UGV Exploration Framework With Loop-Aware Planning and Descriptor-Aided Localization in Resource-Limited Environments
Jun 9, 2026Zhiwei Li, Haiou Liu, Xijun Zhao +3Multi-Uav Exploration
School of Mechanical Engineering, Beijing Institute of Technology, Beijing 100081, China · China North Artificial Intelligence & Innovation Research Institute, Collective Intelligence & Collaboration Laboratory (CIC), Beijing 100081, China

Jun 30, 2026, cs.CVJ/K move · Enter open · S save

PiLoT v2: Pixel-to-Orthogonal Map Alignment for Free-view UAV Geo-localization

Xinyi Liu, Xiaoya Cheng, Rouwan Wu, Zhaochen Wang+3

National University of Defense Technology

Drone-View Geo-Localization Global Navigation Satellite System Image Registration Pilot

Open paper AskPDF arXiv

Real-time, drift-free UAV geo-localization is essential for autonomous missions in GNSS-denied environments. The pioneering system, PiLoT, achieves high precision via Neural Pixel-to-3D Registration, aligning UAV video streams with a single rendered reference view from 3D meshes. However, its reliance on heavy 3D meshes incurs massive storage overheads, complex map acquisition, and significant computational rendering costs, severely hindering deployment on embedded platforms. To address these bottlenecks, we propose PiLoT v2, a lightweight yet robust evolution that shifts the paradigm to direct pixel-to-orthogonal map registration for free-view UAV geo-localization. By leveraging True Digital Orthophoto Maps (TDOMs) and Digital Surface Models (DSMs) as the reference substrate, PiLoT v2 replaces GPU-intensive 3D rendering with a highly efficient, CPU-friendly map cropping operation. To bridge the severe geometric discrepancy between these 2.5D orthogonal crops and free-view oblique UAV imagery, we train a cross-view feature registration network using a novel, large-scale geometrically annotated dataset. Furthermore, we integrate onboard sensor prior--specifically gravity direction and single-point laser rang--directly into the pose optimization manifold to enhance robustness against cross-view visual degradation. Experimental results demonstrate that PiLoT v2 achieves performance comparable to, or even exceeding, its Pixel-to-3D predecessor, while offering drastically lower storage and computational costs.

Abstract

Explore similar work

PiLoT v2: Pixel-to-Orthogonal Map Alignment for Free-view UAV Geo-localization

Joint Optimization of Trajectory Control, Resource Allocation, and Task Offloading for Multi-UAV-Assisted IoV

A Distributed Multi-UGV Exploration Framework With Loop-Aware Planning and Descriptor-Aided Localization in Resource-Limited Environments