A Survey on AI-Based Object Recognition and Autonomous Control for Smart Locomotives for Enhancing Railway Automation

• Version
• Download 1
• File Size 651.22 KB
• File Count 1
• Create Date 3 July, 2025
• Last Updated 3 July, 2025

Authors:Dr. Yogesh Kirange, Sudarshan Jadhav, Harshal Tawade, Kamini Rajput, Piyush Chaudhari, Pranav Shirsath

DOI :10.46335/IJIES.2025.10.9.1

Abstract – The rapid advancement of artificial intelligence (AI) has significantly transformed autonomous locomotive systems, enhancing their efficiency and safety. This study presents an AI-powered object detection framework for logistics-centric autonomous locomotives, leveraging the Robot Operating System (ROS) infrastructure. The research focuses on optimizing object recognition using a lightweight YOLOv4 Tiny model, ensuring high-speed inference while maintaining accuracy. The primary objective is to improve the locomotive's ability to detect and classify objects in real-time, reducing operational risks and enhancing automation reliability. The methodology involves training deep learning models on the Logistics Objects in Context (LOCO) dataset, followed by performance evaluation using precision metrics such as mean average precision (mAP) and intersection over union (IoU). Experimental results indicate a substantial improvement over conventional detection systems, with mAP reaching 46% and IoU achieving 50%. These advancements pave the way for further integration of AI-driven perception models in real-world logistics applications. Future research will focus on refining detection accuracy, integrating sensor fusion techniques, and implementing adaptive decision-making models. The proposed approach not only strengthens autonomous locomotive navigation but also contributes to the broader adoption of AI in railway automation, promoting safer and more efficient rail transport systems.