Innovative navigation artificial intelligence for motor racing games

Abstract

Motor racing games are pushing the boundaries of realism and player experience. Artificial Intelligence (AI) allows developers to create believable opponents. By getting their AI to follow a racing line that is similar to that taken by real racing drivers, developers are able to create a sense that the AI racers are trained drivers. This paper identifies two methods used in the field: the sector based system and the sensor based system. The sector based approach offers two or more predetermined lines for the AI to follow, with added logic allowing the AI to judge when to switch between lines. The sensor method is able to guide AI vehicles around tracks with sensors, offering more possible behaviours and lines. After implementation, the strengths and weaknesses of both methods are realised. The planning and development of a hybrid system was based on these findings. The resulting system is able to produce a more believable line for the AI. With the setting up process of a race track the sector method taking a long time, exploration into tool development is conducted to reduce the process. The subsequent tool reduced the time needed to set up a track, providing results similar to the old method.

Date of Award	May 2016
Original language	English
Awarding Institution	University of Bedfordshire
Supervisor	Rob Manton (Supervisor) & Hong Qing Yu (Second supervisor)