MUMBAI, India, June 26 -- Intellectual Property India has published a patent application (202641071293 A) filed by Vellore Institute Of Technology on June 08, 2026, for Federated Generative Reconstruction System For Robust Autonomous Vehicle Lane Detection.
Inventors include K Nivitha; and Karthik K.
The application for the patent was published on June 19, 2026, under issue no. 25/2026.
Abstract: Federated Generative Reconstruction System for Robust Autonomous Vehicle Lane Detection The present invention relates to a federated generative reconstruction system for robust lane detection in an autonomous vehicle environment. The system employs a vehicle-mounted camera module to capture road scene images and a CNN-based lane detection module to identify lane features from the captured images. A confidence estimation module evaluates the reliability of detected lane structures, while a threshold evaluation engine selectively activates a generative reconstruction engine when the detected lane features exhibit low confidence due to faded, interrupted, degraded, or partially occluded lane markings. The generative reconstruction engine reconstructs missing lane structures to generate an improved lane representation. A feature fusion engine combines reconstructed lane structures with detected lane features to generate a unified lane map for vehicle navigation and lane guidance operations. The invention further supports federated learning through collaborative model training across multiple vehicle nodes while retaining raw driving data locally to preserve privacy.
Disclaimer: Curated by HT Syndication.
