MUMBAI, India, April 17 -- Intellectual Property India has published a patent application (202531129514 A) filed by Asansol Engineering College, Asansol, West Bengal, on Dec. 20, 2025, for 'hybrid analog-digital ai accelerator for low-energy edge devices.'
Inventor(s) include Dr. Ashmi Chakraborty; Mr. Sujoy Kumar Basu; and Dr. Lipika Mandal.
The application for the patent was published on April 17, under issue no. 16/2026.
According to the abstract released by the Intellectual Property India: "The present invention provides a hybrid analog-digital artificial intelligence accelerator designed to achieve ultra-low-power and high-efficiency machine learning inference in edge devices. The system integrates an analog crossbar computing engine, based on memristor or resistive memory devices, to perform massively parallel multiply-accumulate operations, with a digital control and activation unit responsible for nonlinear processing, quantization correction, and precise computation. A novel analog-digital synchronization module continuously monitors analog device drift, noise, and non-linearities, and dynamically adjusts sampling parameters using predictive machine learning techniques. A calibration and error-correction unit periodically tunes the analog conductance states to maintain long-term accuracy, while an energy-aware mode controller adaptively allocates workload between analog and digital subsystems based on available power and required inference precision. By combining high-density analog computation with robust digital processing, the invention significantly reduces inference energy consumption-by up to 70-90%-while enabling reliable, real-time neural network execution suitable for IoT nodes, wearable devices, autonomous sensors, robotics platforms, and other resource-constrained embedded systems."
Disclaimer: Curated by HT Syndication.
