The work of this thesis aims at automatically recognize road irregularities (potholes, bumps, etc.) with the aid of a variety of low-cost sensors mounted in soft-mobility means (i.e., electric kick-scooters and electric bikes) and machine learning techniques. The work can be divided into three main steps: data acquisition, signal processing and training and performance assessment. In the first part, low-cost sensors (sonars, cheap cameras, accelerometers) are mounted in the vehicle and connected to an Arduino MKR1010 Wi-Fi board that sends the signals to another board connected to the computer through a Wi-Fi connection. Several runs with different road conditions have been taken into account and the signals have been labeled accordingly. The dataset collected during the first step is preprocessed, filtered and then given in input to a variety of machine learning classifiers in order to train the model to automatically recognize different road irregularities. Once the classifiers are trained, the model is tested in a real testbed and the performances of all the algorithms taken into account evaluated, finally selecting the most suitable algorithm.

The work of this thesis aims at automatically recognize road irregularities (potholes, bumps, etc.) with the aid of a variety of low-cost sensors mounted in soft-mobility means (i.e., electric kick-scooters and electric bikes) and machine learning techniques. The work can be divided into three main steps: data acquisition, signal processing and training and performance assessment. In the first part, low-cost sensors (sonars, cheap cameras, accelerometers) are mounted in the vehicle and connected to an Arduino MKR1010 Wi-Fi board that sends the signals to another board connected to the computer through a Wi-Fi connection. Several runs with different road conditions have been taken into account and the signals have been labeled accordingly. The dataset collected during the first step is preprocessed, filtered and then given in input to a variety of machine learning classifiers in order to train the model to automatically recognize different road irregularities. Once the classifiers are trained, the model is tested in a real testbed and the performances of all the algorithms taken into account evaluated, finally selecting the most suitable algorithm.

Recognition of Road Irregularities by means of Low-Cost Sensors and Machine Learning Techniques

PASTI, MATTIA
2022/2023

Abstract

The work of this thesis aims at automatically recognize road irregularities (potholes, bumps, etc.) with the aid of a variety of low-cost sensors mounted in soft-mobility means (i.e., electric kick-scooters and electric bikes) and machine learning techniques. The work can be divided into three main steps: data acquisition, signal processing and training and performance assessment. In the first part, low-cost sensors (sonars, cheap cameras, accelerometers) are mounted in the vehicle and connected to an Arduino MKR1010 Wi-Fi board that sends the signals to another board connected to the computer through a Wi-Fi connection. Several runs with different road conditions have been taken into account and the signals have been labeled accordingly. The dataset collected during the first step is preprocessed, filtered and then given in input to a variety of machine learning classifiers in order to train the model to automatically recognize different road irregularities. Once the classifiers are trained, the model is tested in a real testbed and the performances of all the algorithms taken into account evaluated, finally selecting the most suitable algorithm.
2022
Recognition of Road Irregularities by means of Low-Cost Sensors and Machine Learning Techniques
The work of this thesis aims at automatically recognize road irregularities (potholes, bumps, etc.) with the aid of a variety of low-cost sensors mounted in soft-mobility means (i.e., electric kick-scooters and electric bikes) and machine learning techniques. The work can be divided into three main steps: data acquisition, signal processing and training and performance assessment. In the first part, low-cost sensors (sonars, cheap cameras, accelerometers) are mounted in the vehicle and connected to an Arduino MKR1010 Wi-Fi board that sends the signals to another board connected to the computer through a Wi-Fi connection. Several runs with different road conditions have been taken into account and the signals have been labeled accordingly. The dataset collected during the first step is preprocessed, filtered and then given in input to a variety of machine learning classifiers in order to train the model to automatically recognize different road irregularities. Once the classifiers are trained, the model is tested in a real testbed and the performances of all the algorithms taken into account evaluated, finally selecting the most suitable algorithm.
Machine Learning
Signal Processing
Road Irregularities
Low-Cost Sensors
File in questo prodotto:
File Dimensione Formato  
Pasti_Mattia.pdf

accesso aperto

Dimensione 21.04 MB
Formato Adobe PDF
21.04 MB Adobe PDF Visualizza/Apri

The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/20.500.12608/60408