Not-so-large MIMO signal detection based on damped QR-decomposed belief propagation
Tanabe Shogo, Akihide David Shigyo, Koji Ishibashi · International Symposium on Information Theory and its Applications · 2016
In this paper, signal detection based on QR-decomposed belief propagation (QR-BP) in combination with message damping for not-so-large multi-input multi-output (MIMO) systems where the number of antennas is around ten. Bit error rate (BER) performance of belief propagation (BP) based detection cannot approach to that of maximum likelihood detection (MLD) since factor graphs defined by typical MIMO channels are fully-connected, namely heavily loopy. Although it is known that QR-BP can achieve near-optimal performance by transforming factor graphs into those with less edges, exponential computational complexity is still necessary. Hence, we investigate complexity reduction of QR-BP applying partial marginalization (PM) which enables to achieve a good tradeoff between computational complexity and performance. Furthermore, we introduce a message damping to QR-PM-BP to further reduce the complexity. Numerical results confirm that QR-PM-BP with message damping can achieve near-optimal performance with feasible complexity.