Year: 2025
Type: conference-paper
Abstract: Autonomous driving systems rely on accurate and real-time control decisions to ensure safe and efficient navigation. Among these, steering angle prediction is a critical task that directly impacts vehicle trajectory. Traditional rule-based systems often fall short in complex or dynamic environments, necessitating robust data-driven solutions. In this study, we implement and evaluate an end-to-end deep learning approach using the NVIDIA Convolutional Neural Network model for predicting steering angles from front-facing camera images. The dataset used includes simulated driving scenarios with corresponding telemetry, and extensive preprocessing steps such as image cropping, normalization, and data augmentation were applied to enhance generalization. The proposed model was benchmarked against EfficientNetB0, MobileNetV2, ResNet50, and a stacked ensemble using SVR. The NVIDIA CNN outperformed all baseline models, achieving a Mean Squared Error of 0.0118, Root Mean Squared Error of 0.1085, and an R² score of 0.7804, indicating high predictive accuracy and stability. These results highlight the model’s suitability for real-time deployment in autonomous systems.
Source: EPJ Web of Conferences
Authors: Yogesh Karemore, Swati Dronkar
Institution: Priyadarshini College of Engineering
Language: English
Subfield: Automotive Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: diamond
Funder: Nvidia
OpenAlex ID: W4411430420