A COMPREHENSIVE MICROSCOPIC AUTONOMOUS DRIVER MODEL FOR USE IN HIGH- FIDELITY DRIVING SIMULATION ENVIRONMENTS

Abstract

Driving simulation is an invaluable tool for conducting driving- related research. In any virtual driving environment, autonomous traffic is necessary to complete the driver's experience. The higher the fidelity of a driving simulator, the more stringent are the requirements for the realistic appearance of autonomous traffic. Unfortunately, an increase in the fidelity, or the complexity, of the traffic simulation model does not always enhance the appearance of realism. In our experience with driving simulation applications, the number of behaviors exhibited by the autonomous driver models is the biggest factor in how observers rate the realism of the autonomous traffic, even when compared with a smaller number of high-fidelity or validated behaviors. Based on this feedback, we have developed a comprehensive driver model that focuses on breadth of behaviors and is founded on a software architecture that supports growth by providing a framework that allows progressive addition and improvement of behaviors