U6G XL-MIMO Radiomap Prediction: Multi-Config Dataset and Beam Map Approach

Xiaojie Li; Yu Han; Zhizheng Lu; Shi Jin; Chao-Kai Wen

eess.SPMar 6, 2026

U6G XL-MIMO Radiomap Prediction: Multi-Config Dataset and Beam Map Approach

Authors: Xiaojie Li, Yu Han, Zhizheng Lu, Shi Jin, Chao-Kai Wen

Organizations: School of Information Science and Engineering, Southeast University, Nanjing 210096, China · Institute of Communications Engineering, National Sun Yat-sen University, Kaohsiung 80424, Taiwan

Abstract

The upper 6 GHz (U6G) band with XL-MIMO is a key enabler for sixth-generation wireless systems, yet intelligent radiomap prediction for such systems remains challenging. Existing datasets support only small-scale arrays (up to 8x8) with predominantly isotropic antennas, far from the 1024-element directional arrays envisioned for 6G. Moreover, current methods encode array configurations as scalar parameters, forcing neural networks to extrapolate array-specific radiation patterns, which fails when predicting radiomaps for configurations absent from training data. To jointly address data scarcity and generalization limitations, this paper advances XL-MIMO radiomap prediction from three aspects. To overcome data limitations, we construct the first XL-MIMO radiomap dataset containing 78400 radiomaps across 800 urban scenes, five frequency bands (1.8-6.7 GHz), and nine array configurations up to 32x32 uniform planar arrays with directional elements. To enable systematic evaluation, we establish a comprehensive benchmark framework covering practical scenarios from coverage estimation without field measurements to generalization across unseen configurations and environments. To enable generalization to arbitrary beam configurations without retraining, we propose the beam map, a physics-informed spatial feature that analytically computes array-specific coverage patterns. By decoupling deterministic array radiation from data learned multipath propagation, beam maps shift generalization from neural network extrapolation to physics-based computation. Integrating beam maps into existing architectures reduces mean absolute error by up to 60.0% when generalizing to unseen configurations and up to 50.5% when transferring to unseen environments. The complete dataset and code are publicly available at https://lxj321.github.io/MulticonfigRadiomapDataset/.

Figures & tables

Dataset	Area (m 2 )	Scenes	Multi-band	MIMO	Antenna	Height Map
Ours	$1280\times 1280$	800	✓	$32\times 32$ UPA	directional	✓
DeepMIMO v4.0b [ 17 ]	N/A	152	✓	$8\times 8$ UPA	isotropic	✓
BeamCKM [ 18 ]	$256\times 256$	100	$\times$	$8$ ULA	isotropic	✓
UrbanRadio3D [ 19 ]	$256\times 256$	701	$\times$	$\times$	isotropic	✓
CKMImageNet [ 20 , 21 ]	$1280\times 1280$	42	$\times$	$\times$	isotropic	✓
SpectrumNet [ 22 ]	$1280\times 1280$	764	✓	$\times$	isotropic	$\times$

TABLE I: Comparison with Existing Radiomap Datasets

Fig. 1: System model of XL-MIMO radiomap prediction, where the radiomap characterizes spatial signal strength distribution across a $K\times K$ grid observation plane with color indicating received power level (yellow: high, dark blue: low).

Fig. 2: XL-MIMO radiomap dataset generation pipeline comprising geographical acquisition, 3D modeling, ray-tracing simulation, and SSB beamforming stages.

Band	Antenna (UPA)	Beams	Sweeping	Interval
6.7 GHz	$32\times 32$	64	$[-32^{\circ},31^{\circ}]$	$1^{\circ}$
	$16\times 16$	16	$[-28^{\circ},24.5^{\circ}]$	$3.5^{\circ}$
	$8\times 8$	8	$[-28^{\circ},21^{\circ}]$	$7^{\circ}$
	$16\times 32$ , $8\times 16$	1	$0^{\circ}$	–
4.9 GHz	$8\times 16$ , $8\times 8$	1	$0^{\circ}$	–
3.5 GHz	$8\times 8$	1	$0^{\circ}$	–

TABLE II: Multi-Dimensional Configuration Parameter Space

Fig. 3: Geographic diversity of the proposed dataset: (a) sparse urban (0.1% building coverage), (b) mixed urban (16.5%), and (c) dense urban (51.7%) scenes.

Fig. 4: Frequency-dependent coverage characteristics for an $8\times 8$ UPA: (a)–(d) radiomaps across four carrier frequencies (2.6–6.7 GHz), and (e) coverage ratio at three RSRP thresholds.

Fig. 5: Impact of array architecture on coverage patterns at 6.7 GHz: (a)–(c) radiomaps for $8\times 8$ , $16\times 16$ , and $32\times 32$ UPA configurations; (d) peak-to-mean power ratio quantifying beam concentration; and (e) top-5% power concentration measuring spatial energy focusing.

Fig. 6: Beam steering characteristics for a $32\times 32$ UPA at 6.7 GHz across five azimuth angles ( $-32^{\circ}$ to $+31^{\circ}$ ).

Task	Input Features			Data Split	Evaluation Goal
Task	$\mathbf{X}_{\text{env}}$	$\mathbf{X}_{\text{array}}$	$\mathbf{X}_{\text{sample}}$	Data Split	Evaluation Goal
Task 1 (Blind)	✓	✓	$\times$	Random	In-distribution
Task 2 (Sparse)	$\circ$	$\circ$	✓	Random	In-distribution
Task 3a (Cross-Config)	$\circ$	$\circ$	$\circ$	Config-disjoint	Extrapolation
Task 3b (Cross-Env)	$\circ$	$\circ$	$\circ$	Scene-disjoint	Extrapolation
✓: required; $\circ$ : optional; $\times$ : not used

TABLE III: Task Taxonomy for XL-MIMO Radiomap Prediction

Fig. 7: Beam map-enhanced radiomap prediction framework. Left: physics-based beam map computation. Middle: task-dependent feature concatenation. Right: integration with neural network architectures.

Task	Method	MAE (dB)	RMSE (dB)	$\Delta$ MAE
Task 1 (Blind)	RadioUNet	12.98	16.98	–
	RadioUNet+Beam	7.69	11.61	$\downarrow$ 40.8%
	RME-GAN	12.94	17.05	–
	RME-GAN+Beam	6.27	10.53	$\downarrow$ 51.5%
Task 2 (Sparse)	RadioUNet	8.73	12.03	–
	RadioUNet+Beam	5.87	8.86	$\downarrow$ 32.8%

TABLE IV: Radiomap Prediction Performance: Blind Prediction (Task 1) vs. Sparse Reconstruction at 5% Sampling (Task 2)

Scenario	Method	MAE (dB)	RMSE (dB)	$\Delta$ MAE
Cross-Config (Task 3a)	RadioUNet	19.91	25.55	–
	RadioUNet+Beam	7.96	12.06	$\downarrow$ 60.0%
	RME-GAN	13.27	17.41	–
	RME-GAN+Beam	6.24	10.38	$\downarrow$ 53.0%
Cross-Env (Task 3b)	RadioUNet	13.16	17.23	–
	RadioUNet+Beam	7.85	12.12	$\downarrow$ 40.4%

TABLE V: Cross-Distribution Generalization Performance (Task 3)

Fig. 8: Qualitative comparison of radiomap prediction methods on a representative example. Top: ground-truth and predictions. Bottom: building heights and absolute error distributions.

Fig. 9: Cross-distribution generalization comparison. Top: Task 3a (cross-configuration). Bottom: Task 3b (cross-environment). Each row: ground truth, baseline prediction, and beam map-enhanced prediction.

Input Configuration	MAE (dB)	RMSE (dB)
Environment Only	13.52	17.41
Env + Config Encoding	12.98	16.98
Env + Beam Map	7.69	11.61

TABLE VI: Ablation Study on Input Feature Configurations (Task 1, RadioUNet)

Radio Propagation Modeling

Radio Propagation Modeling predicts how electromagnetic signals travel through environments for wireless planning, channel estimation, and sensing.

See all 60 papers

Sep 1, 2026

BeamRMX: Radiation-Pattern-Driven Learning for Generalizable Beam Radio Map Prediction and Beam Management

The evolution toward sixth-generation (6G) wireless networks is driving larger antenna arrays and highly directional multi-beam transmission, making accurate knowledge of beam-dependent spatial coverage important for beam management and environment-aware network operation. Radio maps (RMs) provide such a representation, yet conventional RM prediction assumes omnidirectional or transmitter-level radiation. In beamformed multiple-input multiple-output (MIMO) systems, one propagation scene instead gives rise to many configuration-dependent beam radio maps (BeamRMs), creating challenges in beam representation and generalization. Existing methods either condition prediction on beam descriptors or use beam maps as auxiliary inputs to generic architectures. We propose BeamRMX, which, to the best of our knowledge, is the first dedicated framework to treat the spatial radiation pattern as the primary BeamRM query and learn how scene geometry transforms it into the received power field. XBase learns multiscale interactions between the radiation query and scene geometry, while an optional Evidence Adapter uses a few cross-configuration BeamRMs from the same scene. Matched-domain and zero-shot experiments show consistent gains over deterministic and diffusion baselines, including mean absolute error reductions of 26.1% on unseen scenes and 47.8% on an unseen configuration. Cross-configuration evidence further improves reconstruction and intra-sector beam refinement.

Yue Zhang, Xiucheng Wang, Wenshuo Chen +1

State Key Laboratory of ISN and the School of Telecommunications Engineering, Xidian University, Xi’an 710071, China · The Hong Kong University of Science and Technology (Guangzhou), Guangzhou 511453, China

6G Networks Radio Propagation Modeling

Ask PDF

Sep 10, 2026

Rethinking Radiomap Blind Prediction with Limited Environment and Configuration Representations

Radiomap blind prediction infers radiomaps from observable representations of the propagation environment and base station (BS) configuration without field measurements. These representations are inherently incomplete and cannot uniquely determine the target radiomap. Under squared loss, we identify the conditional-mean radiomap as the population-optimal deterministic target and decompose domain risk into target-approximation error and irreducible uncertainty. The train-test risk gap motivates propagation priors as cross-domain guidance, although their partial or simplified forms may bias the attainable predictor. We therefore propose RadioDecomp, which treats a prior-guided predictor as a correctable base and uses deterministic residual refinement to learn its remaining predictable discrepancy. We instantiate RadioDecomp as RadioLSR (LoS-Shadow-Residual). Experiments under cross-configuration and cross-environment settings show that RadioLSR is especially effective for cross-configuration generalization and provides overall gains over a controlled monolithic counterpart under cross-environment generalization.

Xiaojie Li, Yu Han, Han Fang +3

School of Information Science and Engineering, Southeast University, Nanjing 210096, China · National Mobile Communications Research Laboratory, Southeast University, Nanjing 210096, China · School of Cyber Science and Engineering, Southeast University, Nanjing 210096, China +2

Residual Learning Cross-Domain Generalization Radio Propagation Modeling

Ask PDF

Mar 17, 2026

Structure-Aware Multimodal LLM Framework for Trustworthy Near-Field Beam Prediction

In near-field extremely large-scale multiple-input multiple-output (XL-MIMO) systems, spherical wavefront propagation expands the traditional beam codebook into the joint angular-distance domain, rendering conventional beam training prohibitively inefficient, especially in complex 3-dimensional (3D) low-altitude environments. Furthermore, since near-field beam variations are deeply coupled not only with user positions but also with the physical surroundings, precise beam alignment demands profound environmental understanding capabilities. To address this, we propose a large language model (LLM)-driven multimodal framework that fuses historical GPS data, RGB image, LiDAR data, and strategically designed task-specific textual prompts. By utilizing the powerful emergent reasoning and generalization capabilities of the LLM, our approach learns complex spatial dynamics to achieve superior environmental comprehension...

Mengyuan Li, Qianfan Lu, Jiachen Tian +5

School of Information Science and Engineering, Southeast University, Nanjing 210096, China · Institute of Communications Engineering, National Sun Yat-sen University, Kaohsiung 804, Taiwan

Wireless Communications

Ask PDF

Abstract

Radio Propagation Modeling

Explore similar work

BeamRMX: Radiation-Pattern-Driven Learning for Generalizable Beam Radio Map Prediction and Beam Management

Rethinking Radiomap Blind Prediction with Limited Environment and Configuration Representations

Structure-Aware Multimodal LLM Framework for Trustworthy Near-Field Beam Prediction