Published Online:June 2025
Product Name:The IUP Journal of Information Technology
Product Type:Article
Product Code:IJIT040625
DOI:10.71329/IUPJIT/2025.21.2.59-71
Author Name:Sumedha Dangi, Deepak Kumar and Vipin Khurana
Availability:YES
Subject/Domain:Engineering
Download Format:PDF
Pages:59-71
Self-driving vehicles must traverse different traffic situations and construct secure, coherent decisions and movements. They must detect diverse conditions such as windy weather, illumination, night and rainy and sunny days along the track. This paper takes a novel approach to autonomous driving systems for detecting faults and bugs, using deep learning. The efficient use of various optimizers helps in optimizing self-driving systems by identifying the errors using various estimation techniques such as mean absolute error (MAE), mean squared error (MSE) and R2 . A pilot net model with saliency maps for the detection of objects is implemented. Simulators are capable of making a replica of traffic scenes using real-world datasets that aid in training and evaluating the proposed algorithms. The framework proved its effectiveness in diverse realistic simulation scenarios, delivering an efficient model with both accuracy and reliability. The result showed optimized performance, as compared to traditional optimization methods, highlighting the capability of deep learning in enhancing the capacity of self-driving vehicles. Further validation was conducted using real-time datasets to assess the model’s robustness under different environmental conditions. The integration of saliency maps improved object detection accuracy, ensuring safer navigation. Comparative analysis with existing models demonstrated superior fault detection and adaptability. This paper underscores the potential of deep learning in advancing autonomous driving technology.
The advent of self-driving systems has the capability to transform vehicular transportation, with autonomous driving vehicles that offer problem solutions for road safety, scalability, and traffic management, and lessen human error.