Generative Adversarial Receiver (GAR)

We propose to achieve a quantum leap in RF performance by applying the technology from recent generative AI/ML advances. In the past 2 years, we have seen staggering innovation in generative AI technologies like ChatGPT and Dall-E. Our goal in this program is to leverage these ground-breaking innovations in generative AI/ML to increase RF communications distance and provide more robustness against interference and jamming. Specifically, we will deploy Generative Adversarial Network (GAN) deep learning methods to pull RF signals out of noise using a technique we call Generative Adversarial Receiver (GAR). We will demonstrate the effectiveness of these GAN denoising techniques in small scale experiments to start. Then, the project will culminate in two concrete outcomes: 1) Implementation and over-the-air fielding in our Zylinium Unilateral Radio (ZUR) platform. 2) Productionalization of GAR as an FPGA IP core that can be reused in a wide variety of DoD and commercial RF systems. If successful, this technology will not only be valuable in DoD applications, but will be a huge boon in commercial communication applications as well.