Steering Angle Prediction of Autonomous Vehicle Using Machine Learning

Intelligent Transportation Systems (ITS) encompass a variety of subsystems and technologies, such as basic car navigation management systems, traffic signal control systems, road actor classifications, lidar systems, and autonomous control systems. One crucial aspect of autonomous control systems is steering angle prediction, which is essential for keeping the vehicle within road lines a challenging task. Steering angle prediction has garnered significant attention from researchers, manufacturers, and insurance companies due to its importance in controlling autonomous vehicles (AVs). Various Deep Learning (DL) architectures have been employed to predict AV steering angles in different contexts. However, these methods often involve a large number of parameters and complex algorithms, requiring substantial computational resources and incurring communication overhead and privacy concerns due to the sharing of raw training data. This makes them unsuitable for widespread deployment within ITS. This paper proposes a federated learning-based approach for training a convolutional neural network (CNN) model with significantly fewer parameters. Experimental results demonstrate that our approach outperforms state-of-the-art methods, achieving an average training mean square error of 0.0274 using Sully Chen’s dataset.