Joint Multiple Parameter Estimation and Channel Decoding for Physical-layer Network Coding and Multiuser Detection

This thesis investigates the joint multiple parameter estimation and channel decoding problem for physical-layer network coding (PNC) and multiuser detection (MUD) systems. Both of PNC and MUD can take advantages from the simultaneous transmissions by multiple users. However, the superimposition of multiple transmissions brings with it new challenges for signal processing. The first major challenge is the estimation of the multiple parameters at the receiver. The second major challenge is how to compensate for system impairments caused by these parameters. This thesis consists of two parts that tackle these challenges: The first part is related to PNC systems and the second part is related to MUD systems.;Part I:;The first part of this thesis addresses the problem of joint channel estimation and channel decoding in PNC systems. In PNC, multiple users transmit to a relay simultaneously. PNC channel decoding is different from conventional multiuser channel decoding: Specifically, the PNC relay aims to decode a network-coded message rather than the individual messages of the users. Although prior work has shown that PNC can significantly improve the throughput of a relay network, the improvement is predicated on the availability of accurate channel estimates. Channel estimation in PNC, however, can be particularly challenging because of 1) the overlapped signals of multiple users; 2) the correlations among data symbols induced by channel coding; and 3) time-varying channels. We combine the expectation-maximization (EM) algorithm and belief propagation (BP) algorithm on a unified factor-graph framework. In this framework, channel estimation is performed by an EM subgraph, and channel decoding is performed by a BP subgraph that models a virtual encoder matched to the target of PNC channel decoding. Iterative message passing between these two subgraphs allows the optimal solutions for both to be approached progressively. We present extensive simulation results demonstrating the superiority of our PNC receivers over other PNC receivers.;Part II:;The second part of this thesis investigates a channel-coded MUD system operated with orthogonal frequency division multiplexing (OFDM) and interleaved division multiple-access (IDMA). In general, there are many variations to MUD systems. Our choice of the combination of OFDM and IDMA is motivated by its ability to achieve multiuser diversity gain in frequency-selective multiple-access channels. However, to realize this potential advantage of OFDM-IDMA, we must first solve the frequency asynchrony problem induced by the multiple carrier frequency offsets (CFOs) of the signals of multiple users. This part of the thesis tackles the following two major challenges. The first, as in PNC systems, is the estimation of multiple channel parameters (e.g., CFOs and channel gains). A particular challenge is how to contain the estimation errors of the channel parameters of the multiple users, considering that the overall estimation errors may increase with the number of users because the estimations of their channel parameters are intertwined with each other. The second is how to compensate for the multiple CFOs. A particular difficulty is that, different from a single-user receiver for which there is only one CFO, it is not possible for our multiuser receiver to compensate for all the multiple CFOs simultaneously. To tackle the two challenges, we put forth a framework that solves the joint problem of multiuser channel-parameter estimation, CFO compensation, and channel decoding iteratively by employing the space alternating generalized expectation-maximization (SAGE) and expectation-conditional maximization (ECM) algorithms. Our study reveals that treating the data rather than the channel parameters as the hidden data in ECM will lead to better performance. We further show that Gaussian message passing is an effective complexity reducing technique. Simulations and real experiments based on software-defined radio (SDR) indicate that, compared with other approaches, our approach can achieve significant performance gains.;Overall, this thesis puts forth two frameworks (EM-BP for PNC, SAGE-ECM for MUD) to address the problem of multiple parameter estimation and channel decoding. We believe our frameworks are promising solutions for the signal processing challenges arising from the superimposition of multiple transmissions in multiuser systems.