MUMBAI, India, June 24 -- Intellectual Property India has published a patent application (202441097049 A) filed by Geetanjali Udgirkar; Swara Pawanekar; and Cmrit on December 09, 2024, for Processor Architecture For Differentiable Autonomous Driving.
Inventors include Geetanjali Udgirkar; and Swara Pawanekar.
The application for the patent was published on June 12, 2026, under issue no. 24/2026.
Abstract: The proposed processor architecture for differentiable autonomous driving incorporates specialized hardware for real-time neural network computation, gradient-based optimization, and sensor fusion. It is designed to handle the complex, real-time decision-making required for autonomous vehicles while optimizing for energy efficiency and safety. This architecture represents a scalable and efficient solution for developing autonomous driving systems capable of learning and adapting in real- time. Differentiable autonomous driving refers to using machine learning models that arc fully differentiable, meaning they can be trained using gradient-based optimization techniques such as backpropagation. This approach allows autonomous driving systems to fine-tune their decision- making processes in real-time, by directly minimizing the error or loss between predicted actions (steering, acceleration, etc.) and desired outcomes (staying in lane, avoiding collisions, etc.).
Disclaimer: Curated by HT Syndication.
