We develop a novel mobility-aware transformer-driven tiered structure (MASSFormer) based co-operative spectrum sensing method that effectively models the spatio-temporal dynamics of user movements. Unlike existing methods, our method considers a dynamic scenario involving mobile primary users (PUs) and secondary users (SUs) and addresses the complexities introduced by user mobility. The transformer architecture utilizes an attention mechanism, allowing the proposed method to model the temporal dynamics of user mobility by effectively capturing long-range dependencies. The proposed method first computes tokens from the sequence of covariance matrices (CMs) for each SU. It processes them in parallel using the SU-transformer to learn the spatio-temporal features at SU-level. Subsequently, the collaborative transformer learns the group-level PU state from all SU-level feature representations. The main goal of predicting the PU states at each SU-level and group-level is to improve detection performance even more. The proposed method is tested under imperfect reporting channel scenarios to show robustness. The efficacy of our method is validated with simulation results that demonstrate its higher performance compared to existing methods in terms of detection probability Pd, sensing error, and classification accuracy (CA).

Intelligent reflecting surfaces (IRS) technology has demonstrated considerable potential in enhancing wireless communication by improving signal quality and extending coverage. However, IRS-assisted systems face unique issues in channel estimation caused by their passive nature and the complexity of the channel environment. Deep learning-driven methods provide powerful tools to address complexities such as non-linearities and the high dimensionality inherent in these systems. This paper offers an extensive survey of existing channel estimation techniques in IRS-assisted systems, laying a foundation for future research. To achieve this, a comprehensive literature search was conducted across eight reputable databases and search engines, including IEEE Xplore, Google Scholar, and Scopus etc. After applying rigorous inclusion criteria, 57 key articles were identified as highly relevant, forming the basis of this review. The survey covers traditional methods, such as least squares (LS), minimum mean squared error (MMSE), and linear MMSE (LMMSE), and contrasts them with advanced approaches, including matrix decomposition, compressed sensing, and deep learning techniques. The survey then systematically categorizes the selected studies into three groups: discriminative (supervised learning), generative (unsupervised learning), and hybrid learning. This study reveals that convolutional neural networks (CNNs) are well-suited for resource-constrained or real-time applications, while transformers provide excellent adaptability and accuracy, albeit with higher computational demands. The survey concludes with insights into future research directions, emphasizing the need for improved estimation efficiency and robustness in next-generation wireless systems.