Predicting Environment Variables Using Accumulated Sensor Data
Publicerad
Författare
Typ
Examensarbete för masterexamen
Program
Modellbyggare
Tidskriftstitel
ISSN
Volymtitel
Utgivare
Sammanfattning
Environmental factors such as road roughness play a crucial role in operation of
heavy vehicles. In order to recommend appropriate specifications to customers,
manufacturers need to know which kind of road profile a vehicle will be operated
on. By recommending correct specifications, manufacturers not only improve life of
a vehicle but also trust of their customers.
Over the years, many methods have been proposed to find the roughness profile of
the roads. Our work focuses on how the road roughness can be classified using the
data obtained from the on-board sensors fitted in the vehicles and with the help of
HERE API which is a third party API that returns the roughness value information.
The data that we have used are the accelerometer readings, Global Positioning
System (GPS) and the roughness index values from HERE API for the classification.
These data are collected over different periods for different vehicles and stored in a
database. The main aim of this thesis is to build a Machine Learning model which
will be able to classify a specific route/road between two GPS points into three
main roughness categories - Good, Fair and Poor. This has been approached with
the Support Vector Machine model and Multi-Layer Perceptron model.
Beskrivning
Ämne/nyckelord
Machine Learning, accelerometer, GPS, on-board sensors, HERE API, Road roughness, Neural Network, MLP, SVM