Artificial Intelligence in Simulations

Abstract

<p>Simulator systems are becoming increasingly popular within the automotive industry. Driving simulations are used to train new drivers, and to research and improve automobile related technologies. While recent technological advances have made simulators more affordable, they have also made simulators more complex. At McMaster's motion simulator laboratory, we wished to create a system to analyze how drivers behave when behind the wheel in a variety of situations. To accomodate this, we have created a complex, robust system capable of presenting the driver with customized scenarios, and measuring their reactions to those scenarios using a variety of standard psychology techniques, such as EEG. Our system utilizes a variety of software components such as scenario management, artificial intelligence, image generation, and content creation. This thesis describes the use of one software component, the Presagis AI.implant. The AI.implant is an artificial intelligence system used to control autonomous characters within the simulation, which the driver can interact with. The AI.implant has also been extended to allow for more possible experiment scenarios, and to improve the quality of the simulation. These extensions include: the addition of signal lights for vehicles, emergency vehicle characters, and improved character movement at intersections. This thesis demonstrates that an artificial intelligence is a useful component of a more complex simulation, in order to promote further research in this field.</p>