In the field of wireless communication, the increasing number of devices makes limited spectrum resources more scarce and accelerates the complexity of the electromagnetic environment, posing a serious threat to the sustainability of the industry’s development. Therefore, new effective technical methods are needed to mine and analyze the activity rules of spectrum resources to reduce the risk of frequency conflict. This paper introduces the idea of graphs and proposes a spectrum resource analysis and prediction architecture based on big data. In this architecture, a spatial correlation model of spectrum activities is constructed through feature extraction. In addition, based on this correlation model, a depth learning network based on graph convolution is designed, which uses the prior information of spatial activity to achieve the efficient prediction of spectrum resources. Numerical experiments were carried out on two datasets with different spatial scales. Compared with the best baseline model, the prediction error is reduced by 8.3% on the small-scale dataset and 11.7% on the large-scale dataset. This shows that the proposed method is applicable to different spatial scales and has more obvious advantages in complex scenes with large spatial scales. It can effectively use the results of spatial domain analysis to improve the prediction accuracy of spectrum resources.
18 July 2025 in IEEE Transactions on Mobile Computing
Radio map estimation (RME) is a crucial method for analyzing spectrum space utilization and network coverage, serving as an essential tool for the mobile communication. However, physical constraints, security, privacy, and other issues often render some areas inaccessible, resulting in extremely sparse and unevenly distributed measurement data. To address these challenges, we propose a multimodal feature fusion graph convolution network (MFFGCN). The model incorporates a dual-encoder architecture with an adaptive multi-feature fusion module to exploit environmental information and learn the shadowing effects of radio-signal propagation. We then convert the coarse estimation into regional feature patches and construct a graph over these patches. A graph neural network aggregates contextual information among them, thereby alleviating the impact of uneven spatial sampling. Extensive experiments on open datasets demonstrate that our method achieves state-of-the-art performance, effectively reducing the effects of uneven sampling.
24 February 2026 in IEEE Transactions on Mobile Computing
Urban unmanned aerial vehicle (UAV) navigation operates under tight bandwidth, compute, and latency budgets. Interference and building blockage cause rapid link fluctuations, making interference-aware path planning on grid signal-to-interference-plus-noise ratio (SINR) maps essential for reliable communication. In this paper, we formulate the problem as a joint update–navigation decision: map uncertainty triggers on-demand partial refreshes that are co-optimized with motion under a bandwidth budget. We introduce UT-Grid, an uncertainty-triggered grid-update framework, that refreshes conditionally triggered upon necessity to reduce overall map-update traffic, and MoE-D3QN, a Dueling Double DQN with sparse Mixture-of-Experts (Top-1 routing) that activates a single expert per step, cutting per-decision active parameters and FLOPs while matching or surpassing comparable dense D3QN planners. In an urban simulation with multi-source interference, the framework outperforms static-map, periodic-refresh, and dense D3QN baselines, increasing reaching probability and path efficiency while markedly reducing communication overhead.
22 April 2026 in IEEE Transactions on Consumer Electronics
Accurate spatial priors of received power spectral density (PSD) are important for communication-aware urban sensing and learning-based resource management. Sampling-free radio map (RM) construction avoids costly in-area measurements by directly constructing a full PSD-based radio map from environmental priors. However, mainstream diffusion-based methods incur high inference latency and lack an explicit coarse-to-fine generation path. To address these issues, we propose RadioVAR, a visual autoregressive (VAR) progressive framework for sampling-free RM construction. RadioVAR replaces iterative denoising with a feed-forward next-scale token prediction process, which substantially reduces inference computation and latency. RadioVAR first represents radio maps in a discrete multi-scale latent space using a frozen tokenizer, and performs coarse-to-fine next-scale token prediction conditioned on multi-channel priors (e.g., building layout, Tx location, and vehicle occupancy maps). We further introduce dynamic residual-token supervision to construct state-dependent token targets from the residual-to-go, improving training–inference alignment. Extensive experiments on public datasets demonstrate that RadioVAR improves reconstruction quality over diffusion-based baselines while significantly reducing inference latency, thereby offering a better accuracy–efficiency trade-off for edge deployments.