List Sequential MIMO Detection: Noise Bias Term and Partial Path Augmentation

Near-capacity performance can be achieved in multiple-antenna systems by using a list sequential (LISS) detector for iterative equalization. Path augmentation is known to increase the performance of this detector, while a so-called bias term has been used to reduce its computational complexity, in applications other than MIMO detection. In this work, we extend the LISS MIMO detector [1] to incorporate different implementations of the bias term. We show that following the traditional approach of using an auxiliary stack for this purpose leads to a strong narrowing of the search tree, but also to a significant performance degradation, due to a reduced quality of the detector soft output. We therefore propose to use a so-called noise bias term instead, which can be implemented with negligible effort, significantly reduces the size of the search tree, but allows for almost retaining detector performance. Finally, we demonstrate that the extension of only a small fraction of the incomplete stack entries is sufficient to leverage the full gains of path augmentation, thus enabling further substantial savings in computational complexity.