Research Projects

The emerging reconfigurable intelligent surfaces (RIS)-aided integrated sensing and communication (ISAC) provides a unified framework that successfully integrates the communication and radar sensing tasks. An RIS-aided ISAC systems improve spectral efficiency, reduce hardware costs, and support environment-aware wireless systems. In addition, incorporating RIS in the ISAC system is forecast to provide better sensing and communication performance. A major issue is the difficulty in the dynamic environment and the increased resource demand of RIS-aided ISAC network. Scalability and adaptability are also critical problems, as the number of RIS element phase configurations can grow exponentially. Various machine learning (ML)-based models have been investigated in the literature to address these challenges. The working principle of ML-based models is limited by the requirement for large datasets and labeled data. To address this problem, a large wireless model (LWM) can support the complex joint design, optimization, and real-time adaptation required in RIS-aided ISAC networks by learning unified, task-agnostic representations of the wireless environment. An LWM is a pre-trained foundation model with a self-supervised way to learn universal features and embeddings. So, the key idea is to train these downstream models, such as RIS-assisted network or RIS-assisted ISAC model, based on the embeddings of the foundation model, which is already trained with a large number of real case data sets instead of training themself directly on the raw channel data.