ROAD SURFACE CLASSIFICATION USING SUPERVISED MACHINE LEARNING MODEL AND INERTIAL DATA

Abstract

Recent advancements in transportation and technology highlight the need
for accurate road surface classification in autonomous vehicles and navigation
systems. Traditional methods using cameras and Lidar sensors face limitations in
real-time performance and cost. Our research suggests that inertial sensors,
coupled with machine learning, can achieve over 90% accuracy in real-time
classification of asphalt, dirt, and cobblestone surfaces, offering a low-cost and
readily available alternative. This paper proposes evaluating the classification
performance using fundamental machine learning models on publicly available
datasets for these three road surface types.