Professor, imec - IDLab Ghent - Ghent University (Belgium)

Title of talk: Wireless Physical Layer Foundation Models 

Artificial Intelligence (AI) plays a crucial role in the evolving landscape of wireless communications, addressing challenges that traditional approaches cannot solve. This talk discusses the evolution of wireless AI, emphasizing the transition from isolated, task-specific models to more generalized and adaptable AI models, inspired by the recent success of large language models (LLMs). To overcome the limitations of task-specific AI strategies in wireless networks, Wireless Physical Layer Foundation Models are proposed. The concept of Wireless Physical layer Foundation Models is to create generic models trained on wireless data (e.g., IQ signals, RSSI, network KPIs) that can be applied to a variety of tasks such as interference detection, activity detection, power allocation, channel estimation, and more. To realize this vision, several key challenges must be addressed, such as identifying effective pre-training tasks, supporting embedded time-series data, and enabling human-understandable interaction. Furthermore, it is essential for Wireless Physical Layer Foundation Models to interact with LLMs, which can assist in extracting meta-data (such as classifications, semantic description of the wireless network conditions, sensing applications, human behavior, etc.) from these models. This integration with LLMs can lead to continuous optimization of wireless networks.