Neuronové sítě a evoluční algoritmy pro videohry
Loading...
Date
2023-06-20
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Tato diplomová práce se zabývá řešením autonomního řidiče do hry TORCS. Pro řízení je použita neuronová síť optimalizovaná algoritmem NEAT. Byla provedena analýza konfigurace optimalizačních parametrů. Podle nejlepší nalezené konfigurace byly natrénovány dva modely (s důrazem na spolehlivost a s důrazem na rychlost). Modely byly otestovány na dostupných tratích v prostředí TORCS a to včetně tratí, na kterých nebyly trénovány. Modely se ukázaly být konkurenceschopné na většině tratí s modelem Berniw, který je implementovaný přímo v TORCS.
This thesis addresses the development of an autonomous driver for the TORCS game. A neural network optimized by the NEAT algorithm is utilized for driving. An analysis of optimization parameter configurations was conducted. According to the best found configuration, two models were trained (one focused on reliability and the other on speed). The models were tested on available tracks in the TORCS environment, including tracks that were not trained on. The models proved to be competitive on most tracks with the Berniw model, which is directly implemented in TORCS.
This thesis addresses the development of an autonomous driver for the TORCS game. A neural network optimized by the NEAT algorithm is utilized for driving. An analysis of optimization parameter configurations was conducted. According to the best found configuration, two models were trained (one focused on reliability and the other on speed). The models were tested on available tracks in the TORCS environment, including tracks that were not trained on. The models proved to be competitive on most tracks with the Berniw model, which is directly implemented in TORCS.
Description
Subject(s)
NEAT, TORCS, umělá inteligence, strojové učení, neuronové sítě, evoluční algoritmy, videohry