Improved Multiuser Detection in Code-division Multiple Access Communications

In Code Division Multiple Access (CDMA) communications, Multiuser Detection (MUD), that reduces the multi-access interference (MAI) and that solves the near-far problem, has been widely studied for over 15 years. Since optimal multiuser detection is generally NP hard, many sub-optimal algorithms that provide reliable performance and ensure polynomial complexity have been proposed. However, there is still a large gap between the performance of the sub-optimal detectors and that of the optimal detector. Due to the advances in the hardware computational speeds, advanced MUD algorithms that achieve near-optimal performance , while maintaining high computational efficiency, are of special interest to both researchers and industry. The main objective of this research is to improve the existing multiuser detectors and propose new advanced near-optimal and optimal detectors. In our research, we improve the performance of the Group Decision Feedback (GDF) detector by finding the optimal user partitioning and ordering. We solve the time labeling issue in asynchronous CDMA and improve the performance of the DF detector by finding the optimal time labeling and user ordering. Based on the user ordering and Branch-and-Bound search, we improve the computational efficiency of the optimal detector by proposing a fast optimal algorithm that significantly reduces the average computational cost. In addition, we also develop a new multiuser detection algorithm based on the idea of Probabilistic Data Association (PDA) from target tracking. The PDA detector achieves near-optimal performance in both synchronous and asynchronous systems with O(K 3) complexity where K is the number of users. The situation of overloaded system in both synchronous and asynchronous cases are also studied. The soft-output feature of the PDA method makes it extremely flexible and easy to extend to multiuser detection problems in a wide variety of communication settings.