Rotating Non-Uniform and High-Dimensional Constellations Using Geodesic Flow on Lie Groups (1309.5247v1)

Abstract: We use a numerical algorithm on the Lie group of rotation matrices to obtain rotated constellations for Rayleigh fading channels. Our approach minimizes the union bound for the pairwise error probability to produce rotations optimized for a given signal-to-noise ratio. This approach circumvents explicit parametrization of rotation matrices, which has previously prevented robust numerical methods from being applied to constellation rotation. Our algorithm is applicable to arbitrary finite constellations in arbitrary dimensions, and one can thus apply our method to non-uniform constellations, which are of interest for practical concerns due to their ability to increase BICM capacity. We show how our rotations can improve the codeword error performance of non-uniform constellations, and we also apply our method to reproduce and improve rotations given by ideal lattices in cyclotomic fields.