| Passenger vehicle automation is a promising technology with the potential to provide safety and convenience for millions of drivers across the world. While the current state of autonomous vehicle technology enables vehicles to drive safely under normal driving conditions, it fails to effectively mitigate collisions due to the errors of human drivers of other vehicles on the road. This research demonstrates the need for vehicles to have predictive abilities in order to anticipate and prevent dangerous situations caused by other vehicles. This dissertation proposes the idea of giving vehicles a theory of mind that allows them to better understand the mental states of other drivers and, consequently, allows them to predict the future behavior of other vehicles more accurately than prevailing methods.;The Cognitive Driving Framework is unique in that it explicitly models both the intent and the beliefs of other vehicles on the road. By acknowledging the fact that others' beliefs about the world may be incorrect, autonomous vehicles can estimate these potentially incorrect beliefs and predict when they are about to result in a dangerous situation. The Cognitive Driving Framework is a method for modeling and performing online inference on other vehicles' belief process. Using the online estimate of the vehicle's mental states, the Cognitive Driving Framework provides a method for predicting the future state of multi-vehicle systems and for determining the likelihood of a collision.;The impetus for the Cognitive Driving Framework is reducing the number of vehicle-vehicle automobile collisions and the catastrophic loss of life caused by them. It is the hope that this dissertation will spur further development and implementation of the method on both fully autonomous vehicles as well as on human-driven vehicles as an Advanced Driver Assistance System. Moreover, the algorithm is sufficiently general such that it can be applied to the prediction of many complicated behaviors performed by a wide variety of dynamic agents. |
