Application of integer quadratic programming in detection of high-dimensional wireless systems

High-dimensional wireless systems have recently generated a great deal of interest due to their ability to accommodate increasing demands for high transmission data rates with high communication reliability. Examples of such large-scale systems include single-input, single-output symbol spread OFDM system, large-scale single-user multi-input multi-output (MIMO) OFDM systems, and large-scale multiuser MIMO systems. In these systems, the number of symbols required to be jointly detected at the receiver is relatively large. The challenge with the practical realization of these systems is to design a detection scheme that provides high communication reliability with reasonable computational complexity, even as the number of simultaneously transmitted independent communication signals becomes very large. Most of the optimal or near-optimal detection techniques that have been proposed in the literature of relatively low-dimensional wireless systems, such as MIMO systems in which number of antennas is less than 10, become problematic for high-dimensional detection problems. That is, their performance degrades or the computational complexity becomes prohibitive, especially when higher-order QAM constellations are employed. In the first part of this thesis, we propose a near-optimal detection technique which offers a flexible trade-off between complexity and performance. The proposed technique formulates the detection problem in terms of Integer Quadratic Programming (IQP), which is then solved through a controlled Branch and Bound (BB) search tree algorithm. In addition to providing good performance, an important feature of this approach is that its computational complexity remains roughly the same even as we increase the constellation order from 4-QAM to 256-QAM. The performance of the proposed algorithm is investigated for both symbol spread OFDM systems and large-scale MIMO systems with both frequency selective and at fading channels. The second part of this work focuses on a reduced complexity version of IQP referred to as relaxed quadratic programming (QP). In particular, QP is used to reformulate two widely used detection schemes for MIMO OFDM: (1) Successive Interference Cancellation (SIC) and (2) Iterative Detecting and Decoding (IDD). First, SIC-based algorithms are derived via a QP formulation in contrast to using a linear MMSE detector at each stage. The resulting QP-SIC algorithms offer lower computational complexity than the SIC schemes that employ linear MMSE at each stage, especially when the dimension of the received signal vector is high. Three versions of QP-SIC are proposed based on various trade-offs between complexity and receiver performance; each of the three QP-SIC algorithms outperforms existing SIC techniques. Second, IDD-based algorithms are developed using a QP detector. We show how the soft information, in terms of the Log Likelihood Ratio (LLR), can be extracted from the QP detector. Further, the procedure for incorporating the a-priori information that is passed from the channel decoder to the QP detector is developed. Simulation results are presented demonstrating that the use of QP in IDD offers improved performance at the cost of a reasonable increase in complexity compared to linear detectors