On the decoding order of MIMO maximum-likelihood sphere decoder: linear and non-linear receivers

The paper investigates the use of sphere decoding on multiple-input multiple-output (MIMO) maximum-likelihood (ML) detection for complexity reduction. The problem is important for both multiuser detection and space-time decoding. In particular, we focus on the design of the decoding order which, if designed properly, can miniaturize the decoding hypersphere faster, leading to much complexity reduction. Our proposal is to sort the lattice points, within the intervals defined by the radius of the hypersphere, according to their distance from a given reference signal point, and search the coordinates closest to the chosen reference. A rather surprising result is that by defining the reference point as the soft-output signal point of a conventional receiver, such as zero-forcing (ZF), minimum mean-square-error (MMSE), or interference cancellation (IC), we can naturally combine any known receivers into a sphere decoder for ML detection and greatly reduce the decoding complexity without compromising the performance.