MUMBAI, India, Aug. 12 -- Intellectual Property India has published a patent application (202611071052 A) filed by Noida Institute Of Engineering And Technology Niet on June 08, 2026, for Hardware-Secured Progressive Reinforcement Learning System For Autonomous Driving Decision Inference.
Inventors include Faizan Ahmad; and Meghali Das.
The application for the patent was published on August 07, 2026, under issue no. 32/2026.
Abstract: A hardware-secured progressive reinforcement learning system (100) for autonomous driving decision inference is disclosed. A supervised fine-tuning processor (106) cold-starts a multimodal language model on chain-of-thought annotated samples, a difficulty classification circuit (110) partitions training data into easy, medium, and hard subsets using threshold registers (146), and a Gaussian curriculum sampling controller (112) shifts batch composition from easy toward hard samples. An entropy computation circuit (116) quantifies disagreement among rollout trajectories and an advantage scaling circuit (118) modulates each policy update accordingly. Upon convergence, a policy weight commit controller (124) writes the parameters into a write-protected weight memory region (126) and asserts a commit signal that latches an interconnect firewall circuit (128) against direct memory access writes and causes a secure element (130) to increment a monotonic counter (132) and sign a manifest verified by a vehicle system-on-chip (134) before actuation is enabled.
Disclaimer: Curated by HT Syndication.
