Iot-Based Road Abnormality Detection Using Ai/Ml Techniques

Abstract

The condition of the road network plays a vital role in affecting rolling resistance, driving comfort, and road safety. Consequently, it is imperative to regularly and meticulously inspect the road infrastructure to identify any areas of damage or potential hazards. Based on car sensors and supervised machine learning categorization, techniques have been developed to thoroughly and automatically digitise the road infrastructure and evaluate the condition of the road. One classifier cannot be used to other cars since different types of vehicles have varied suspension systems with unique reaction capabilities. The process of gathering training data for each unique vehicle and classifier often takes a long time. To solve this issue, a dataset of inertial sensors data for Road Surface Type Classification (RSTC) from Kaggle has been used. The findings demonstrate that the best algorithm for reliably transmitting road surface data from a single vehicle to the cloud using IoT is the long-short term memory (LSt-TMem), which is based on the performance of the Support Vector Machine (Sup-Vt-Mac) and long-short term memory (LSt-TMem). Additionally, this approach provides the opportunity to combine the output and data on the road network from many aberrant locations, allowing for a more accurate and reliable forecast of the ground truth.