Optimization of MIMO Unitary Space-Time Codes

摘要

One category of Space-Time codes are constellations of unitary matrix in parametric form. Optimization is essential for seeking parameters of constellation with the largest diversity product. In this research, we demonstrate that diversity product is not a good measure of constellation quality as widely adopted in communication community. Moreover, we show that good codes may not need to have full diversity. We propose better criteria for measuring the quality of constellations, which are also very amenable for optimization and particularly suitable for the gradient search method. Furthermore, we propose a new approach for signal constellation design. Instead of ambiguously discriminating low and high SNR, our techniques target the range of block error rate which is acceptable and not extremely small. Although the computational complexity of code designing can be formidable, we have developed techniques which significantly improve the efficiency. We obtain space-time codes which significantly outperform existing ones.