Razvoj metrik za ocenjevanje udobnosti vožnje v avtonomnih vozilih

Valič, David (author), Sodnik, Jaka (mentor)

Avtonomna vožnja je zelo relevantna tema, ki postaja ključna za prihodnost transporta. Poleg varnosti v avtonomnem vozilu je zelo pomemben tudi občutek udobja. Da lahko sistem za avtonomno vožnjo razume, kaj je udobno in kaj ne, so potrebne metrike za ocenjevanje udobja vožnje. Magistrska naloga se posveča razvoju metrik, ki lahko za poljubno vožnjo ocenijo njeno udobje. Naloga je sestavljena iz štirih delov. V prvem delu je predstavljen razvoj mobilne aplikacije za zajem podatkov premikanja avtomobila med vožnjo in istočasno tudi psihofizične lastnosti potnika ter potnikova ocena opravljene vožnje. Drugi del opisuje uporabniško študijo, v kateri je bila aplikacija preizkušena s potniki v pravem vozilu. Skupaj je bilo opravljenih 27 različnih voženj s šestimi potniki. Tretji del predstavlja analizo in obdelavo podatkov za nadaljnjo uporabo pri razvoju metrik. Zadnji del naloge predstavi nekaj različnih poskusov razvoja metrik za ocenjevanje udobja vožnje. Kot najboljša izmed razvitih metrik se je izkazala metrika, ki določi oceno vožnje na podlagi pravil za različne pospeške. Metrike na podlagi strojnega učenja se niso obnesle najbolje zaradi majhne količine zajetih podatkov.

• Language: Slovenian
• Keywords: mobilna aplikacija, zajem podatkov, udobje vožnje, metrike udobja vožnje
• Work type: Master's thesis
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2024
• PID: 20.500.12556/RUL-164915
• Publication date in RUL: 15.11.2024

Secondary language

• Language: English
• Title: Development of metrics for assessing driving comfort in autonomous vehicles

Abstract

Autonomous driving is a very relevant topic and is becoming crucial for the future of transportation. Besides the safety of autonomous vehicles, the perception of comfort is also a very important topic. Metrics for evaluating driving comfort are necessary to help an autonomous driving system understand what comfortable driving is. The master's thesis focuses on developing metrics that can assess the driving comfort of any given ride. The thesis is divided into four parts. The first part presents the development of a mobile application for capturing vehicle movement data during a ride, as well as the psychophysiological characteristics of the passenger and the comfort score of the ride. In the second part, the developed application is used to collect data from rides with passengers. In total, data for 27 different rides with six passengers were obtained. The third part includes the analysis and preprocessing of data from the completed rides. The final part focuses on developing different metrics for assessing driving comfort. The most successful of the developed metrics was estimating the driving comfort based on rules related to different accelerations. Metrics based on machine learning did not perform well due to the small amount of collected data.

• Keywords: mobile app, data collection, driving comfort, driving comfort metrics