Advanced Graph-Based Detection for Wireless Channels

Low-complexity graph-based algorithms provide a useful and powerful tool to limit the huge computational load typically encountered in the design of wireless communications receivers. This thesis focuses on the development of efficient detection techniques aimed at achieving an optimal trade-off between computational complexity and performance. Specifically, in the context of single-carrier time-varying channels affected by fading or strong phase noise, this research examines approximate message-passing schemes based on the expectation propagation framework. The primary objective is to develop a detection algorithm that, besides solving the complexity problem, operates effectively even with a reduced number of pilot symbols for the channel estimation process, thereby increasing the payload efficiency and improving adaptability to various communication standards. Moreover, current state-of-the-art algorithms rely on turbo iterations between the detector and decoder to refine estimates of both data and channel parameters. However, this approach introduces significant computational complexity, which is generally impractical, and the communication between the two units can sometimes be unfeasible. These limitations further motivate the research presented in this thesis. Additionally, this work addresses the detection problem for Orthogonal Time Frequency Space (OTFS)-based transmissions in challenging high-mobility conditions. It proposes a message-passing algorithm that leverages the particular structure of the equivalent channel matrix in the Doppler-delay domain, aiming to reach the performance benchmark with affordable complexity under realistic channel conditions. The performance of the proposed solutions is evaluated by assessing bit error rate and pragmatic capacity through MATLAB-based numerical simulations. The results demonstrate satisfactory performance, suggesting that the proposed approaches provide a viable low-complexity alternative to the existing methods. These findings offer practically implementable solutions in the receiver design for future research and applications.