Description:
Reference #: 1757
The University of South Carolina is offering licensing opportunities for ML-Based Zero-Constellation Design for BMOCZ.
Background:
Binary modulation on conjugate-reciprocal zeros (BMOCZ) is a modulation scheme, which works by encoding message bits onto the zeros of a polynomial and letting the coefficients of this polynomial modulate a carrier. Within the last decade, non-coherent communication schemes have gained traction for their ability to scale with the dramatic increase in wireless connectivity anticipated with the emergence of the Internet-of-Things (IoT). Binary modulation on conjugate-reciprocal zeros (BMOCZ) is a novel non-coherent communication scheme for short-packets that has many advantages over existing non-coherent communication strategies. Therefore, it is of paramount importance to optimize both the performance and flexibility of BMOCZ.
Invention Description:
This invention presents two frameworks to design zero-constellations for BMOCZ schemes. The first method provides a method to optimize constellation parameters for a DiZeT decoder by using machine learning methods. The second method introduces an AI-based decoder and jointly optimizes the zero-constellation and decoder. Although trained under AWGN, the proposed MLP decoder is shown to directly generalize to flat-fading channels. Furthermore, numerical simulations demonstrate superior error rate performance compared to canonical BMOCZ constellations, where the MLP decoder achieves significant gains at the cost of higher complexity.
Potential Applications:
The technology has potential applications in cellular systems, AI-native radios, military applications, sensor networks, aviation, autonomous vehicles, robotics, and more
Advantages and Benefits:
This invention improves the reliability and practicality of BMOCZ-based communication systems.