Approximate Inference Low-Complexity MIMO Detection

Abstract: In the telecommunications field, the correct detection of a certain signal is of most interest as it improves the robustness of the system. With multiple-input multiple-output (MIMO) antenna systems a new problem arises, the increase in complexity of the detection algorithms with the number of antennas.This thesis consists of an overview over the classical and most recent derived from statistical learning detection algorithms.   Starting with the optimal approach, the Maximum Likelihood detector, whose solution minimizes the error rate, is formally presented along with its scalability problem.  To solve this, and enable the detection in systems of higher order, other sub-optimal receivers were developed along the years.  Zero-Forcing or MMSE detectors are two examples of these, and, even though they present an attracting simplicity they carry a poor performance.The development of statistical learning enabled a new class of algorithms such as expectation propagation (EP) or belief propagation (BP). The first iteratively refines an approximation to the true posterior by moment matching.  A generalization to the EP algorithm is the Power EP, and, in this work its performance for relevant MIMO scenarios was assessed.  Belief propagation is an algorithm that calculates the marginal of a joint distribution by passing messages over factor graphs from variable nodes to function nodes and vice-versa.   In the presence of a loop free graph, the result of this message passing technique yields the exact marginal distribution while in the presence of loops its result is an approximation to it.  The application of these two cases can be seen in what is known as the Gaussian Tree Approximation and the Loopy BP detectors,and, in this work their performance is assessed for relevant scenarios.