A Relax-and-Round Approach to Complex Lattice Basis Reduction

Abstract

We propose a relax-and-round approach combined with a greedy search strategy for performing complex lattice basis reduction. Taking an optimization perspective, we introduce a relaxed version of the problem that, while still nonconvex, has an easily identifiable family of solutions. We construct a subset of such solutions by performing a greedy search and applying a projection operator (element-wise rounding) to enforce the original constraint. We show that, for lattice basis reduction, such a family of solutions to the relaxed problem is the set of unitary matrices multiplied by a real, positive constant and propose a search strategy based on modifying the complex eigenvalues. We apply our algorithm to lattice-reduction aided multiple-input multiple-output (MIMO) detection and show a considerable performance gain compared to state of the art algorithms. We perform a complexity analysis to show that the proposed algorithm has polynomial complexity.