Abstract
Generalized Spatial Modulation (GenSM) and Generalized Quadrature Spatial Modulation (GenQSM) are promising techniques for enhancing spectral and energy efficiency in MIMO systems. By utilizing antenna indices to carry information, these methods reduce the number of required radio frequency (RF) chains. However, the performance of these systems is heavily dependent on the design of the transmit vector set. This paper proposes two signal shaping approaches—Optimization-Based Signal Shaping (OBSS) and Codebook-Based Signal Shaping (CBSS)—aimed at maximizing the minimum Euclidean distance (MMED) between transmit vectors. While OBSS provides optimal performance through recursive size optimization and complex entry optimization, CBSS offers a practical, low-complexity alternative for large-scale systems.
Introduction
In the evolution of wireless communications, Generalized Spatial Modulation (GenSM) and Generalized Quadrature Spatial Modulation (GenQSM) have emerged as efficient alternatives to traditional MIMO. These techniques activate a subset of transmit antennas, known as a Transmit Antenna Combination (TAC), to convey information through both the selected indices and the transmitted symbols. A critical challenge lies in designing the optimal set of transmit vectors to minimize error rates under various channel conditions.
The Challenge of Joint Optimization
Traditionally, the "spatial" component (antenna selection) and the "signal" component (QAM/PSK symbols) in GenSM/GenQSM have been treated as independent design problems. This decoupled approach is inherently suboptimal. To achieve the lowest Symbol Error Rate (SER), the system must maximize the Minimum Euclidean Distance (MMED) across all possible transmit vectors. This requires a joint optimization where the signal constellations are tailored to the specific antenna combinations being used.
The optimization is computationally intensive. For a typical system, the number of potential distributions of signal points across TACs is vast, leading to a non-convex, large-scale problem that scales poorly with system size.
Proposed Solutions: OBSS and CBSS
The researchers introduced a unified framework addressing three levels of Channel State Information at the Transmitter (CSIT): no CSIT, statistical CSIT, and perfect CSIT. Two primary methodologies were developed:
- Optimization-Based Signal Shaping (OBSS): This approach serves as the performance benchmark. It employs a recursive design to determine the optimal number of signal points for each TAC and uses Quadratically Constrained Quadratic Programming (QCQP) to optimize the coordinates of these points. Although it achieves the theoretical performance limit, its high computational complexity makes it less suitable for real-time adaptive systems.
- Codebook-Based Signal Shaping (CBSS): Designed for practical implementation, CBSS selects the best transmit vectors from a discrete pre-defined codebook (e.g., 16-QAM). By using a progressive selection algorithm to maximize a "constellation figure of merit," CBSS significantly reduces the computational burden while maintaining performance close to OBSS.
Key Results and Performance
Simulation results in transmit-correlated Rayleigh channel environments demonstrate the efficacy of these signal shaping methods. A key finding is that joint signal shaping can effectively mitigate the performance degradation caused by high antenna correlation.
As shown in the performance curves, the OBSS approach provides substantial reliability gains. In a GenSM system with high antenna correlation ($\delta = 0.3$), OBSS outperforms standard BPSK-based spatial modulation by approximately 8 dB at an SER of $10^{-2}$. Furthermore, when perfect CSIT is available, the proposed methods exceed the performance of existing precoding-aided benchmarks by 1 to 4 dB. The CBSS approach, despite its lower complexity, tracks the performance of OBSS closely, making it a viable candidate for large-scale hardware deployments.
Conclusion
This research highlights the importance of joint signal and spatial optimization in next-generation MIMO systems. By maximizing the minimum Euclidean distance through OBSS and CBSS, GenSM and GenQSM systems can achieve significantly higher reliability and efficiency. While OBSS defines the performance ceiling, CBSS provides a scalable path forward for practical, high-capacity wireless networks. Future work may explore the integration of machine learning to further accelerate the optimization process and the inclusion of hardware-specific constraints like peak-to-average power ratio (PAPR).
Summary updated