Reduced-complexity sphere decoding with dimension-dependent sphere radius design

Abstract

Abstract-A modified sphere decoding (SD) scheme is proposed for multiple-input multiple-output (MIMO) communication systems in this paper. The conventional SD goes from the lower dimension to the higher dimension to examine whether a lattice point lies inside the sphere of some radius, which remains fixed for all dimensions. Since the sphere radius directly affects the search range and thus the complexity, it is an important parameter to design. The proposed scheme employs a set of dimension-dependent sphere radii, which performs more aggressive search-space reduction in low dimensions. The proposed method is shown by computer simulation to offer substantial complexity benefits with little symbol error rate (SER) performance loss (0.5 dB), compared to the optimal maximum likelihood (ML) decoding. The contribution of this paper includes the complexity advantage yielded by the proposed scheme as well as the introduction of a systematic approach to sphere radius design and control.

Best-First Tree Search with Probabilistic Node Ordering for MIMO Detection: Generalization and Performance-Complexity Tradeoff

“…For simplicity, in this paper we show results for ∞-norm only. demonstrated for different signal-to-noise ratios (SNRs) 7 for a 4 × 4 system with 64-QAM in Fig. 4.…”

“…8 Among schemes with variable latency, DSPS has the smallest average latency as well as the least amount of variability, both at a given operating SNR and across different SNRs. As a result, DSPS needs to maintain the smallest number of I/O buffers which could limit the detection throughput and pose an extra overhead in 7 The SNR is defined as E[∥Hx∥ 2 ]/E[∥v∥ 2 ]. 8 While the -best detector has a fixed iteration count independent of , the time required for each iteration to complete is proportional to [11].…”

“…Since the path metric of nodes from different layers constitute different partial sums in the likelihood metric according to (7), the path metric does not objectively represent the "cost" of nodes from the detection point of view. In the following, we present the proposed modified algorithm adopting a new cost function that leverages the statistical properties of the path metric to achieve more efficient node maintenance and tree exploration.…”