ML and DL Classifications of Route Conditions Using Accelerometers and Gyroscope Sensors

Ibrahim Khan, Zahid Ahmed

Abstract: The Modern Era has undergone vast transformations in terms of automation in a wide array of industrial applications. The Artificial Intelligence platform has revolutionized our daily lives with the advent of Intelligent systems. The usefulness of AI in Road Mapping and Route Classification is demonstrated in our study where an Intelligent Transport System (ITS) is proposed which enables monitoring and classification of road conditions by implementing Machine Learning (ML) and Deep Learning (DL) algorithms on data recorded by cellular accelerometer, gyroscope, and GPS sensors. Field Data was recorded in two different scenarios on different vehicles. The route mapping was performed by plotting latitude and longitudes on Google Earth. The labelling of different classes of road was done manually with correlation done via video camera recording. Road Terrain was classified into Bumps. Potholes, Rough and Smooth Roads. Six classical Supervised Machine Learning models (K Neighbors Classifier, Decision Tree Classifier, Random Forest Classifier, Support Vector Classifier, Gaussian Naive Bayes Model and Logistic Regression Model) were implemented. Furthermore, Ensembler classifier was used on all six classifiers. The selection of an Optimum Classification Model is done via Soft Voting Algorithm. Finally, K-Fold cross validation was performed to determine the accuracy of our trained model.