Distributed Automation Systems (DAiSY) Laboratory 

AUTHOR: Hamza Bouanzoul

RELATOR: Stefania Santini

CO-TRELATOR: Gianluca Toscano

TUTOR: Aniello Mungiello

ABSTRACT: This thesis presents a comprehensive exploration into the performance of three Kalman filter variants—Extended Kalman Filter (EKF), Cubature Kalman Filter (CKF), and Adaptive Cubature
Kalman Filter—in the context of vehicle pose estimation during straight motion. The study leverages real-world data obtained from the Xsens MTi-680G inertial navigation system (INS) hardware, fixed in the XEV Yoyo car. The project is centred around a test scenario involving straight motion, providing a detailed analysis of the filters' performances under specific motion dynamics. The XEV Yoyo car, equipped with the MTi-680G, serves as the platform for evaluating the filters' performance in capturing the vehicle's pose. Performance metrics encompassed include accuracy in position and orientation estimation, computational efficiency, and the adaptability of the filters to the dynamic behaviour exhibited during straight-line motion. Results obtained from the real-world test provide insights into the strengths and limitations of the EKF, CKF, and Adaptive CKF in the context of straight-line motion scenario.

AUTHOR: Lorenzo Palma XU

RELATOR: Stefania Santini

TUTOR: Aniello Mungiello

ABSTRACT: An autonomous driving system is a complex combination of various components that can be defined as systems, in which the perception, decision-making and operation of the car is performed by electronic devices and machines instead of human drivers. In this thesis, we focused on the perception module. In particular, we carried out the design and validation of a VRTK sensor that combines information from cameras and GPS antennas with RTK technology. This makes it possible to achieve a localization with centimetre accuracy of the car. The validation of this sensor was carried out by real test on a scaled vehicle.  

AUTHOR: Diego Manganaro

RELATOR: Alberto Petrillo

CO-TRELATOR: Manuela Tufo

TUTOR: Aniello Mungiello

ABSTRACT: Within the context of self-driving vehicles, the realization of a control module requires signal processing from the surrounding environment, in particular we dealt with the one assigned to the interpretation of LiDAR signals.
The development was organized as follows:
In-depth study of LiDAR signal processing techniques;
Implementation, through the ROS2 framework, of LiDAR signal processing codes;
Validation in a virtual environment of the identified strategies;
Insight into the identified algorithms and their use;
Test phase on the scaled-down vehicle provided by Kineton.

AUTHOR: Davide Colucci 

RELATOR: Stefania Santini

TUTOR: Aniello Mungiello 

ABSTRACT: This dissertation presents the structure and the implementation of a multi-extended Kalman filter localization algorithm with fault detection capabilities. This algorithm has been developed for the driverless car of the Federico II racing team Unina Corse, that compete in the formula students’ events all over Europe.
The thesis will first introduce the odometry concept, principles and state of the art. Then proceed to introduce the context for which this algorithm has been developed: the formula student events, and competitions. After that the UniNa Corse team is briefly introduced, with its history and structure, then a broad overview of the car is given from the mechanical, electrical and software point of view where the components and the design choices are explained, with a more detailed explanation focused on the driverless algorithms structure.
Following this introduction, the odometry problem statement and the use cases contextualized in the formula student competitions will be presented, and only at this point will be explained the development of the structure of the localization algorithm that has followed the Model Based Control Design process (Appendix A2), starting with the analysis of the requirement and moving on with the different validation phases, finishing with the track test. In this section will also be given the overviews of the techniques used in the development process, such Extended Kalman Filter, Multi-Sensor Fusion, Chi-Square and drift detection algorithms.
In conclusion, ideas for further improvements of the algorithm and the testing phases will be discussed.