MUMBAI, India, Aug. 17 -- Intellectual Property India has published a patent application (202641094644 A) filed by Vardhaman College Of Engineering on August 05, 2026, for Autonomous Vehicle Navigation Using Hybrid Slam And Deep Learning Models.
Inventors include Dr. Shaik Imam Saheb; Mr. Naveedkumarreddy K; Mr. V Shashivanth; Mr. Vankudoth Vinod Kumar; Mr. Uppala Nihar; and Ms. Bogolu Rupa.
The application for the patent was published on August 14, 2026, under issue no. 33/2026.
Abstract: Autonomous Vehicle Navigation Using Hybrid SLAM and Deep Learning Models is the proposed invention. The present invention provides a navigation system for autonomous vehicles that utilises Hybrid Simultaneous Localisation and Mapping (Hybrid SLAM) and deep learning models to obtain accurate localisation, intelligent perception of the environment, semantic mapping, and real-time path planning. The system integrates data from cameras, LiDAR, radar, Global Navigation Satellite Systems (GNSS), inertial measurement units, wheel encoders and ultrasonic sensors in a multi-sensor fusion framework. A transformer-based deep learning perception engine performs semantic scene segmentation to detect road infrastructure, lane markings, traffic signs, pedestrians, vehicles and dynamic obstacles. Hybrid SLAM produces a high quality, continuously-updated map with good localisation, even in GPS denied environments, from visual, LiDAR, inertial and odometry data. Loop closure detection and graph optimisation are to reduce the drift of localisation and thus improve the consistency of the long-term map. The semantic map, the predicted obstacle trajectories, traffic conditions, and vehicle dynamics are used by a navigation engine based on deep reinforcement learning to compute safe, energy-efficient, and collision-free routes in real-time. The sensors give feedback at all times and depending on the environment, the car can steer, brake and control speed accordingly.
Disclaimer: Curated by HT Syndication.
