Mitigating Truck Shadow Effects in Autonomous Vehicle Perception through Machine Learning algorithms

Autonomous vehicles (AVs) must operate safely and effectively in complex environments. Accurate object detection and navigation are vital to AV performance, especially as vehicles face challenges like shadows, occlusions, and varying lighting conditions. Improving AV perception systems using machine learning (ML) models can enhance detection and classification accuracy, ultimately enabling safer navigation. This study evaluates the performance of five machine learning classifiers for AV object detection and navigation, considering metrics like accuracy, precision, recall, and area under the receiver operating characteristic (ROC-AUC). The study focuses on identifying the most effective ML model for accurate object classification and navigation under challenging conditions, thus optimizing AV safety. To achieve this, we propose a structured ML framework to improve AV perception accuracy. A comprehensive ML pipeline was developed, encompassing data preprocessing, feature extraction, model training, and performance evaluation. Preliminary results show that Random Forest achieves high accuracy, while SVM demonstrates superior precision, suggesting a hybrid model could combine the strengths of both.