Executive Summary : | The global internet protocol (IP) is expected to experience a 55% annual increase in data traffic between 2020 and 2030 due to the rapid development of IoE, AI, and smart homes. The 5G and beyond (5GB) network is required for higher spectrum efficiency, energy efficiency, and mobility due to the rapid growth of mobile data services and smartphone usage. 5GB wireless network technologies, such as orthogonal frequency division multiplexing (OFDM) and Orthogonal Time Frequency Space (OTFS), are proposed to support high mobility scenarios. OFDM offers benefits such as reduced inter-symbol interference, high spectral efficiency, and ease of implementation. OTFS modulation converts rapidly time-varying fading channels into almost invariant channels, simplifying equalizer design and reducing channel estimation overhead. However, these systems are sensitive to synchronization parameters, such as symbol timing offset (STO) and carrier frequency offset (CFO), which significantly deteriorate performance. Joint estimation of STO and CFO is crucial for these systems. Conventional statistical methods have limitations, but deep learning methods can alleviate these limitations. Deep learning models can be optimized for real-time systems with all impairments and non-linearities, allowing for the joint estimation of STO and CFO for IRS-assisted OFDM/OTFS system. This data-driven scheme can handle large data and satisfy 5GB system requirements. A real-time implementation will be carried out on a reconfigurable radio frequency front-end testbed. |