ARTIFICIAL INTELLIGENCE FOR SIMULATED GRAND PRIX RACE TRACK CAR RACING

V diplomski nalogi učimo strategijo umetne inteligence voziti avtomobil po simulirani dirkalni progi Grand Prix. Učenje izvedemo s samoprilagodljivim algoritmom diferencialne evolucije. Avtomobil je simuliran v računalniškem simulatorju TORCS, kjer smo v pogon igre vstavili že obstoječo arhitekturo odjemalec/strežnik za komunikacijo med simulatorjem in našim programom za strategijo vožnje. Ta arhitektura je bila predlagana na tekmovanju 2011 Simulated Car Racing Championship. Cilj tekmovanja je čim bolje naučiti strategijo vožnje avtomobila po dirkalni progi.In this diploma work we teach an artificial intelligence strategy to drive a racing car on a simulated Grand Prix track. Teaching is done using self-adaptive differential evolution algorithm. This car was simulated in the game TORCS, in which we integrated already existing client/server architecture. This architecture was proposed for the 2011 Simulated Car Racing Championship. The aim of this competition is to learn a strategy in the best possible way for simulated grand prix car racing