Driverless Cars Enhancing Self Driving Prediction with Advanced Deep Learning Techniques

Abstract

The advancement of driverless cars relies on accurate steering angle prediction to ensure safe and efficient navigation. This work tries to create a deep learning models that minimizes prediction errors while maintaining computational efficiency. The primary objective is to enhance steering control by leveraging advanced neural network architectures. Deep learning models—including EfficientNetV2, EfficientNetV2B3, Xception, and VGG19—were trained on a driving scenario dataset to attain this. Evaluation of the models started with three key performance criteria: loss, mean absolute error (MAE), and mean squared error—MSE. Every model was tuned to maximize generalization and feature extraction therefore guaranteeing strong performance in practical settings. With the lowest error rates (Loss: 0.0110, MAE: 0.0798, MSE: 0.0110), Xception showed great performance among the models tested, therefore demonstrating great accuracy in steering angle predictions. The outcomes underline how much deep learning improves the precision of autonomous car control systems. This work effectively meets its goal by using a very effective model that raises dependability and forecast accuracy. Future research will investigate methods of real-time deployment and further optimization strategies to improve autonomous car decision-making capacity.