20.500.12556/DKUM-80629 Uporaba metod zastopanja znanja za preučevanje naprednih sistemov za pomoč voznikom in avtonomne vožnje masterʹs thesis Knowledge representation as a means to assess advanced driver assistance systems and autonomous driving Področje avtonomne vožnje je eno izmed najbolj raziskovanih tem v avtomobilski industriji. Varnost avtonomnih vozil v vsakdanjih situacijah na cesti predstavlja enega izmed glavnih izzivov, zaradi vrste različnih situacij do katerih pride v resničnem svetu. Avtonomna vozila bi morala prevoziti več sto miljard kilometrov, da bi lahko potrdili njihovo zanesljivost pri odločanju v primeru nevarnosti. Zato ima virtualno testiranje scenarijev, kjer lahko simuliramo poljubne situacije, pomembno vlogo pri validaciji in preverjanju delovanja avtonomnih vozil. Tako virtualno testiranje predstavlja nujen postopek v razvoju naprednih sistemov avtonomne vožnje. Vendar je ustvarjanje raznolikih scenarijev velikokrat okoren in zamuden postopek. Zaradi tega je zaželjena večkratna in ponovna uporaba podatkov iz realnega sveta. Ker ontologije predstavljajo večkrat uporabno zastopanje informacij, so idealen kandidat za ustvarjanje raznolikih scenarijev. V nalogi predstavimo avtomatiziran postopek za razbiranje informacij iz podatkovne baze za namene večkratne in ponovne uporabe le teh v ontologijah. Ključnega pomena za razvoj ontologije je dobro razumevanje strukture podatkovne baze, zato se velik del naloge osredotoča na analizo le te. Podatke z baze je potrebno s pomočjo programske kode razbrati in interpretirati ter šele nato uskladiti ontologijo z bazo. Pri tem razvijemo dve nove metodi, ki temeljita na geometričnih algoritmih. Na podlagi le-teh, raziščemo bazo in zbrane podatke očiščimo s z uporabo statistične analize ter opredelimo v ontologijo. Rezultat naše naloge je ontologija zapolnjena z osnovnimi koncepti, ki so definirani na podlagi podatkov zbranih s podatkovne baze. Autonomous driving is one of the most promising and challenging research topics in the automobile industry. Testing the safety of autonomous vehicles in everyday situations on the road is a major challenge due to large variety of circumstances that occur in the real world. Autonomous vehicles would have to be driven hundreds of billions of kilometres to demonstrate their reliability in terms of fatality and injuries. Due to this, virtual testing is essential in the development phase in which, we can experiment on arbitrary scenarios. Virtual testing is an essential part in developing advanced driver assistance system functions, but it can be challenging to generate new assessment test cases. That is why reusability of real-world data for test case generation is a crucial. Since ontologies introduce a sharable and reusable knowledge representation, they present an ideal candidate for solving this task. This thesis presents an automated procedure for information acquisition from the database in order to align the ontology with derived knowledge. Understanding the basic structure of the dataset is paramount in order to correctly develop the ontology. Therefore, a substantial part of the thesis focuses on analysing the dataset. Data has to be acquired and interpreted first and only then can we coordinate it with the ontology. For this purpose, we have developed three new methods based on geometrical algorithms with the help of which we have analysed the dataset and cleaned the acquired data through the methods of statistical analysis. The findings of this thesis present an ontology aligned with the basic concepts that are defined based on data collected from the dataset. ontologije avtonomna vožnja avtomatizacija Waymo Open Dataset Ontologies autonomous driving automation Waymo Open Dataset true true false [G. Ovsenjak] Slovenski jezik Angleški jezik Magistrsko delo/naloga 2021-09-23 14:51:29 2021-12-20 08:38:51 2023-04-12 15:09:20 0000-00-00 00:00:00 2021 0 Maribor 0 Maribor IX f., 61 str. 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 531.8(043.2) 89438211 MAG_Ovsenjak_Gregor_2021.pdf MAG_Ovsenjak_Gregor_2021.pdf 1 A95937FC8EBE5D8C631436966B59C77C 306488d35310987f02cd2d8cf78caada789679f947842a12bd4c664d6da08402 ecfd4535-26bb-11ec-aa8d-00155d000105 20.500.12556/dkum/b1423aaf-e67c-4841-86f5-29d0395debd6 https://dk.um.si/Dokument.php?lang=slv&id=154844 Fakulteta za naravoslovje in matematiko 0 0 0