Towards a more accurate agent-based multi-lane highway simulation

The next several years will see an acceleration of the adoption of intelligent driving aid devices. Studying the impact of such devices on the overall traffic performance and safety requires highly realistic microscopic traffic simulation models, which account not only for the overall traffic flow, but also for the details and variability of the individual driver’s behavior. In this paper, we first describe a series of improvements on current state of the art multilane highway driving models. The general theme of these improvements is to make the models more agent-like, by considering the specific goals and limitations of the individual drivers and vehicles. We are also performing a more detailed simulation of some of the critical steps in multi-lane highway driving, such as the process of merging at highway entrances and changing lanes. We apply our model to a real world example of the busy commuter highway 408, which crosses Orlando. Through a series of experiments, we investigate the effects of individual driver behavior on the traffic flow.