SBIR Phase I: Analog Neural Networks Enabled by Tunable Apertures

The broader impact of this Small Business Innovation Research (SBIR) Phase I project is in addressing the massive energy demands of modern artificial intelligence by developing a computing platform utilizing radio waves instead of traditional electrical circuits to perform complex calculations. This innovation enables advanced AI to operate on small, battery-powered devices such as medical monitors, autonomous safety sensors, and handheld communication tools. Furthermore, the project enhances scientific and technological understanding by demonstrating how physical wave interactions can be harnessed for high-speed data processing. This Small Business Innovation Research (SBIR) Phase I project addresses a unified radio-frequency (RF) accelerator that performs deep learning operations directly through electromagnetic wave interactions within a physical cavity. The primary research objective is to validate that a single hardware system can perform both feature extraction (convolution) and classification (matrix multiplication) using passive wave scattering. The research involves designing optimized metal cavities and tunable metasurfaces that use low-power diodes to "steer" waves and perform mathematical operations at the speed of light. To support this, the project will develop a physics-informed "digital twin” computer model that accurately predicts the behavior of the physical hardware for efficient training. This work establishes a foundation for a new class of energy-efficient analog hardware for diverse machine learning applications. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.