Exact and Approximated Factor Graph Methods for MIMO Detection

Factor graphs (FG) are a tool to derive efficient algorithms for complex systems by graphically displaying the structure of the mathematical model. The construction of the graph and the choice of the messages which are passed over its edges significantly determine the derived algorithm. In our work we show the application of a FG to a MIMO detector and the derivation of three different algorithms by applying different message passing methods, such as an exact, an exact relaxed and a particle algorithm. Further on, we discuss the bit error performance depending on message passing methods and the computational complexity depending on message representation and graph construction.