| PhD Seminar

Name of the Speaker: Ms. Vasumathi R (EE21D063)
Guide: Prof. Ramkrishna Pasumarthy
Venue: ESB-244 (Seminar Hall)
Date/Time: 23rd December 2024 (Monday), 3:00 PM
Title: Enhancing Autonomous Navigation Through Dynamic SLAM and V2X Communication. Abstract:

Abstract :

Developing autonomous driving algorithms requires rigorous testing in environments that closely emulate real-world conditions. Simulation environments, while instrumental in early-stage development, often fail to capture the dynamic complexities of real-world traffic, including diverse road conditions and environmental uncertainties. To address these challenges, we developed a novel 1:10 scaled traffic test track and integrated it with a one-tenth-scale electric vehicle platform, DEFT, developed at our lab to create a controlled yet realistic framework for evaluating autonomous functionalities for intelligent vehicles.

This work builds upon a localization framework incorporating FIR and Kalman filtering for accurate position estimation utilizing IMU and wheel encoder data in indoor environments. We then focused on the perception aspect, where we tailored a lane detection pipeline for autonomous navigation and validated it using DEFT and the scaled test track under challenging conditions, including lane occlusions and damages, demonstrating resilience to real-world imperfections—an ongoing challenge in the literature. Current efforts focus on advancing dynamic Simultaneous Localization and Mapping (SLAM) by integrating real-time sensors (stereo cameras, IMUs, and LiDAR) for segmenting static and dynamic objects and motion prediction algorithms to refine mapping accuracy. Vehicle-to-Everything (V2X) communication will enhance localization accuracy through infrastructure-based reference points, improving situational awareness and cooperative capabilities. By bridging the gap between simulation and full-scale testing, this work presents a scalable platform replicating real-world traffic scenarios, including curvy roads, intersections, on-off ramps, and parking spaces, to validate autonomous navigation algorithms for intelligent vehicles.