Automatic Joint Optimization of Iterative MIMO-OFDM Receiver Algorithms on a Meta Level

We present a method for automatically optimized selection and composition of algorithm components for iterative MIMO-OFDM receivers. Complexity is measured as clock cycles of the target processor cores (suitable for software defined radio based implementation). A receiver description language is used to specify the design space, and to enable enumeration of the design space using graph algorithms. Performance prediction of a composition of candidate algorithms uses a recently published method from stochastic convergence analysis of iterative processing. We illustrate the method in an example scenario, where the optimized receivers show both an extended operational SNR range compared to the standard receiver architecture, as well as significantly reduced complexity compared to iterative processing according to round-robin iteration scheduling using the same algorithm components.