[Paper Review] Layered Orthogonal Lattice Detector for Two Transmit Antenna Communications

Massimiliano Siti, Michael P. Fitz|ArXiv.org|Aug 12, 2005

Advanced Wireless Communication Techniques11 references17 citations

TL;DR

This paper proposes the Layered Orthogonal Lattice Detector (LORD), a novel MIMO detection algorithm for two transmit antennas that achieves maximum-likelihood (ML) performance with significantly reduced complexity compared to exhaustive search. By leveraging a layered, orthogonal lattice structure and a deterministic, parallelizable search, LORD enables optimal hard-decision detection and exact max-log bit log-likelihood ratio (LLR) computation at low complexity, making it suitable for real-time, high-data-rate systems with efficient hardware implementation.

ABSTRACT

A novel detector for multiple-input multiple-output (MIMO) communications is presented. The algorithm belongs to the class of the lattice detectors, i.e. it finds a reduced complexity solution to the problem of finding the closest vector to the received observations. The algorithm achieves optimal maximum-likelihood (ML) performance in case of two transmit antennas, at the same time keeping a complexity much lower than the exhaustive search-based ML detection technique. Also, differently from the state-of-art lattice detector (namely sphere decoder), the proposed algorithm is suitable for a highly parallel hardware architecture and for a reliable bit soft-output information generation, thus making it a promising option for real-time high-data rate transmission.

Motivation & Objective

• To address the high computational complexity of maximum-likelihood (ML) detection in MIMO systems with two transmit antennas.
• To overcome the limitations of existing lattice detectors like the sphere decoder, which are inherently serial and sensitive to parameters.
• To develop a detection algorithm that enables reliable, low-complexity soft-output generation for iterative decoding in coded MIMO systems.
• To design a detector compatible with highly parallel hardware architectures for real-time, high-throughput applications.

Proposed method

• Formulates the MIMO detection problem in a real-valued lattice domain using a transformed channel matrix, enabling orthogonal decomposition.
• Applies a preprocessing step based on QR-like decomposition to transform the channel matrix into a form that allows layered detection.
• Performs a layered lattice search over $2S$ sequences (where $S$ is the constellation size), instead of the $S^4$ sequences of exhaustive ML search.
• Uses a deterministic, parallelizable search strategy that avoids the variable search volume of sphere decoding.
• Derives exact max-log bit LLRs by comparing minimum metrics over subsets of constellation points corresponding to bit values 0 and 1.
• Reorders the symbol vector and reuses the same detection structure to compute LLRs for all bits, enabling parallel soft-output computation.

Experimental results

Research questions

• RQ1 Can a lattice-based MIMO detector achieve ML performance with complexity significantly below exhaustive search for two transmit antennas?
• RQ2 Can the detector support reliable, low-complexity soft-output generation for iterative decoding in coded systems?
• RQ3 Is the detection process amenable to a highly parallel hardware implementation, unlike existing sphere decoders?
• RQ4 Can the algorithm maintain a fixed, predictable number of lattice points to search, avoiding parameter sensitivity?
• RQ5 Does the layered orthogonal structure enable exact max-log LLR computation without approximation?

Key findings

• LORD achieves optimal maximum-likelihood performance in hard-output demodulation for two transmit antennas, matching the performance of exhaustive ML detection.
• The algorithm requires searching only $2S$ sequences, reducing complexity from $O(S^4)$ for exhaustive ML to $O(S)$, where $S$ is the constellation size.
• LORD enables exact max-log bit LLR computation with a complexity linear in the number of transmit antennas, suitable for iterative decoding.
• The detector supports fully parallel hardware implementation, unlike the inherently serial sphere decoder.
• Simulation results confirm that LORD achieves near-ML performance in both uncoded 64QAM and BICM-coded systems with $L_t = 2$, $L_r = 2$, over Rayleigh fading channels.
• In BICM systems, LORD provides optimal soft-output LLRs with low complexity, enabling near-capacity performance in coded MIMO transmission.