Learning-based approaches for wireless PHY layers from the perspective of conventional machine learning to foundation models: A comprehensive survey

The future of wireless communication requires low latency, ultra-reliable connectivity, and the ability to manage a large number of IoT devices in real time. Achieving these demands for quality of service (QoS) can be addressed through machine learning (ML), deep learning (DL), and foundation models integrated into wireless systems and devices. Foundation models, in particular, show promise for overcoming the limitations of conventional approaches, and improving system performance was observed to be between 9.63 % and 12.80 %, with the possibility of exceeding this range under certain conditions. This review explores how ML, DL, and foundation models can be applied at the physical (PHY) layer in wireless communications. It covers various learning algorithms such as deep, recurrent, and feedforward neural networks, explaining their design, training methods, and challenges. Key applications include channel estimation, constellation design, signal detection, and optimizing signal modulation schemes to achieve better spectral efficiency and noise resilience. The paper also discusses how the ML, DL, and foundation models can enhance MIMO systems with improved detection performance. In addition, it highlights the challenges and opportunities in adopting these models in different communication domains, including trade-offs between accuracy, complexity, and generalization. Conventional ML performs better in scenarios with small datasets, low computational complexity, and tasks requiring high interpretability, whereas DL approaches tend to outperform traditional methods in large-scale, high-dimensional wireless problems such as CSI prediction, interference classification, and spectrum sensing. This survey offers valuable insights into the evolving landscape of intelligent communication systems, guiding practitioners in implementing learning-aided strategies for the next generation of wireless technology.