Jul 2026· European Conference on Artificial Intelligence· pp. 1-6· 0 citations· 19 references
Abstract
Coverage map estimation is fundamental in planning wireless communication systems to enhance service quality and minimize operational costs. Although traditional deterministic propagation models offer high precision, their computational costs and processing times increase exponentially, especially in urban areas. In this study, a ResNet-based Conditional Variational Autoencoder (ResNet-CVAE) architecture is proposed for rapid and accurate coverage map generation in scenarios involving multiple diffractions. In order to generate dataset, a dynamic ray-tracing code integrating Geometric Optics (GO) and the Uniform Theory of Diffraction (UTD) was developed. By processing obstacle geometry and transmitter locations as both numerical and spatial condition information, the proposed deep learning model successfully captures abrupt signal level drops and physical shadowing effects behind obstacles. Experimental results demonstrate that the ResNet-CVAE model produces high accurate results with significantly lower computational overhead compared to traditional methods and adapts effectively to complex obstacle configurations. This approach offers significant potential for real-time analysis in network planning and base station placement processes.
Autonomous aerial vehicles have become increasingly important for data harvesting tasks in complex urban environments, where efficient area coverage, reliable data collection, and safe navigation are critical. Coverage path planning ensures that all regions of interest are visited with minimal redundancy, while data harvesting focuses on collecting data from distributed IoT sensor nodes under energy and safety constraints. In this work, we propose a Transformer-enhanced Double Deep Q-Network (TDDQN) framework that effectively integrates a Standard Transformer Encoder with a hierarchical global-local map representation for large-scale urban AAV navigation. This strategy combines a compressed global map with a local obstacle-aware map to enable scalable operation and temporal reasoning in complex urban landscapes. The proposed approach is evaluated through extensive simulations on Manhattan32 and Urban50 scenarios and compared against various existing models. Experimental results demonstrate that the proposed method consistently outperforms all baselines by achieving superior coverage efficiency, data collection performance, and landing success rates. Notably, in Urban50 environment, the proposed method improves the coverage ratio by 48.1% and the landing success rate by 14.3% over the baseline models. These results highlight the effectiveness of attention based architectures in enhancing AAV decision-making for coverage path planning and data harvesting tasks while maintaining stringent safety and energy requirements.
Radio maps (RMs) characterize the spatial distribution of wireless channel features and provide essential foundations for environment-aware wireless communication. Existing RM reconstruction methods mainly focus on pathloss prediction on fixed two-dimensional planes, while three-dimensional (3D) joint inference of pathloss, time of arrival (TOA), and direction of arrival (DOA) from sparse pathloss-only observations remains underexplored. In this letter, we propose HAMM-UNet, an end-to-end framework for joint reconstruction of 3D multi-modal RMs conditioned on sparse pathloss observations, building information, and base station locations. Through height-adaptive modulation and modality-specific decoding, the framework effectively captures propagation variations across different height layers and mitigates multi-modal reconstruction conflicts. Experimental results demonstrate that at a 10% sampling rate, the proposed method achieves optimal performance across all channel parameters, with a 69.4% reduction in pathloss root mean square error (RMSE) compared to the best baseline, and maintains robust performance even at sampling rates as low as 1%.
Bowen Zhu, Meng Wang, Pan Zhen et al.· IEEE Wireless Communications...· 0 citations
Deep learning based wireless channel prediction can benefit from full exploitation of propagation features, yet how to construct effective inputs for accurate channel prediction remains under-explored. This paper presents a deep learning-based approach that optimizes environmental input construction for accurate channel path loss prediction. A campus measurement campaign is conducted to obtain datasets, where satellite image and semantic layers are aligned at a unified environmental granularity. Using a fixed network backbone and identical training protocol, results show that channel prediction accuracy is sensitive to spatial coverage. Overly local patches are insufficient, while enlarging coverage to include propagation-relevant surroundings yields error reduction with diminishing returns. It is further found that orientation normalization, implemented by aligning the Tx-Rx direction to a fixed reference axis, improves performance by reducing geometric variability. Building on this spatial extent, Shapley-based attribution and ablation analysis indicate that buildings and roads have dominating impacts, vegetation offers complementary information, and location-related descriptors mainly contribute through interactions. The results validate that the input feature construction improves channel prediction accuracy and provides guidance for future intelligent channel prediction.
Zhicheng Qiu, Ruisi He, Bo Ai et al.· npj Wireless Technology· 0 citations
This paper presents a computationally efficient deep learning framework for accurate direction-of-arrival (DoA) estimation in portable radar applications. Leveraging a MobileNet architecture, the proposed model directly processes raw in-phase and quadrature-phase (IQ) data, enabling more effective learning of both spatial and temporal features. This direct input approach enhances DoA estimation accuracy, particularly under challenging conditions such as low signal-to-noise ratio (SNR) and limited snapshot scenarios. A unified training strategy is adopted for both single-source and multi-source target detection, ensuring consistency and robustness. Comprehensive simulation experiments demonstrate the proposed model’s competitive and robust performance across various conditions, including different SNR levels, closely spaced targets, and random off-grid angles. It also shows that our method achieves performance comparable to or better than recent deep learning approaches in several challenging scenarios, establishing its potential for resource-constrained environments where only low snapshot data are available. The proposed IQ-MobNet DoA estimation model achieves this competitive performance with substantially lower computational complexity, requiring only 0.24 million parameters and 0.42 million Floating Point Operations (FLOPs), representing a reduction of over 96% compared to the recent neural network models. To ensure practical applicability, the proposed IQ-MobNet framework is validated using real-world measured data, confirming its robustness beyond simulated environments.
Neeraja P. Kovilakam, Bindiya T. Sambasivan, Raghu C. Variyam· Electronics· 0 citations
This work proposes Physics-Unrolled Hybrid Neural Operator (PU-HNO), a three-stage cascade that predicts high-fidelity indoor radio maps from low-fidelity ray-tracing outputs and scene priors by progressively capturing reflection, diffraction, and scattering effects, rather than treating radio maps as generic images.
Rafid Umayer Murshed, Saif Ur Rahman, Mingyue Tang et al.· 0 citations
Accurate wireless channel prediction in urban environments is essential for network planning and optimization, but traditional ray tracing (RT) simulations are computationally expensive. This paper presents a deep learning approach that learns from sparse RT simulations to predict received power for unseen transmitter locations. We propose a spatial attention convolutional neural network with an encoder-decoder structure incorporating convolutional block attention modules to prioritize critical regions, such as propagation boundaries, complemented by a distance-aware loss emphasizing accuracy near transmitters and borders. Trained on 80 heatmaps from a $500 mathrm{m} \times 500 mathrm{m}$ urban area, each on a $\mathbf{3 4} \times \mathbf{3 4}$ receiver grid (1,156 positions), the model achieves RMSE of \~{}22 dB and MAE of \~{}12 dB, outperforming nearest-heatmap averaging by \~{}5 dB, with an $\mathbf{R}^{\mathbf{2}}$ of $\mathbf{0. 9 2}$. Results demonstrate effective capture of multipath, diffraction, and shadowing, offering a computationally efficient alternative to full RT for urban channel prediction.
Eran Greenberg, Itzik Klein· 2026 6th International Confe...· 0 citations