The Equivalence of Semidefinite Relaxation MIMO Detectors for Higher-Order QAM

In multi-input multi-output (MIMO) detection, semidefinite relaxation (SDR) has been shown to be an efficient high-performance approach. For BPSK and QPSK, it has been found that SDR can provide near-optimal bit error probability performance. This has stimulated a number of recent research endeavors that aim to apply SDR to the high-order QAM cases. These independently developed SDRs are different in concept, structure and complexity, and presently no serious analysis has been given to compare these methods. This paper analyzes the relationship of three such SDR methods, namely the polynomial-inspired SDR (PI-SDR) by Wiesel et al., the bound-constrained SDR (BC-SDR) by Sidiropoulos and Luo, and the virtually-antipodal SDR (VA-SDR) by Mao et al. Rather unexpectedly, we prove that the three SDRs are equivalent in the following sense: The three SDRs yield the same optimal objective values, and their optimal solutions have strong correspondences. Specifically, we establish this solution equivalence between BC-SDR and VA-SDR for any 4(q)-QAM constellations, and that between BC-SDR and PI-SDR for 16-QAM and 64-QAM. Moreover, the equivalence result holds for any channel, problem size, and signal-to-noise ratio. Our theoretical findings are confirmed by simulations, where the three SDRs offer identical symbol error probabilities. Additional simulation results are also provided to demonstrate the effectiveness of SDR compared to some other MIMO detectors, in terms of complexity and symbol error performance.