Rear-end collisions are frequent yet preventable crashes in the United States. Collision warning systems have the potential to prevent crashes and mitigate crash severity. However, their success depends on the algorithms used to trigger the warnings, and computational models of rear-end collision avoidance behaviors are critical for accurate calibration of warning algorithms. This dissertation addresses gaps in driver modeling research by developing models based on three different psychological perspectives. A driving simulator experiment was used to provide data of imminent collision with a stopped or decelerating lead vehicle in the presence or absence of a warning system. Three models were developed--two were based on the information-processing approach (static models) with different parameter associations, and one was based on the ecological approach and concepts of direct perception (dynamic model). The static model with independent stages considered parameters (reaction time, jerk, and deceleration) as independent; the static model with dependent stages used copula functions to construct trivariate distributions. Associations between variables suggests that assumptions of independence are invalid. Counterfactual analysis was used to perform benefits estimation, and results show that predicted benefits of the copula-based models and those of the static models with independent stages differ by 11%. The dynamic model used perceptual variables--visual angle and expansion rate--to model onset of braking (reaction time) and deceleration profiles. This dynamic model assumes that perception occurs in the light, i.e., ecological structures relevant to collision avoidance, such as visual looming of the lead vehicle, can be used to directly specify collision-avoidance actions. The dynamic models may represent drivers' braking responses more precisely, however, traditional statistical approaches cannot be used for the parameterization of such complex models. The Approximate Bayesian Computation technique was used to parameterize these models in this dissertation. Model parameters estimated with this technique indicate that combinations of perceptual variables generate dynamic collision-imminent deceleration profiles similar to those observed in the empirical data. Between-driver variances in deceleration were captured in the perceptual variable parameters. Taken together, the different models improve our understanding of the mechanisms governing drivers' rear-end collision avoidance and provide a basis for future behavioral modeling.

This book presents a general overview of the various factors that contribute to modelling human behaviour in automotive environments. This long-awaited volume, written by world experts in the field, presents state-of-the-art research and case studies. It will be invaluable reading for professional practitioners graduate students, researchers and alike.

This paper presents a modeling strategy of human driving behavior based on the controller switching model focusing on the driver's collision avoidance maneuver. The driving data are collected by using the three-dimensional (3-D) driving simulator based on the CAVE Automatic Virtual Environment (CAVE), which provides stereoscopic immersive virtual environment. In our modeling, the control scenario of the human driver, that is, the mapping from the driver's sensory information to the operation of the driver such as acceleration, braking, and steering, is expressed by Piecewise Polynomial (PWP) model. Since the PWP model includes both continuous behaviors given by polynomials and discrete logical conditions, it can be regarded as a class of Hybrid Dynamical System (HDS). The identification problem for the PWP model is formulated as the Mixed Integer Linear Programming (MILP) by transforming the switching conditions into binary variables. From the obtained results, it is found that the driver appropriately switches the "control law" according to the sensory information. In addition, the driving characteristics of the beginner driver and the expert driver are compared and discussed. These results enable us to capture not only the physical meaning of the driving skill but the decision-making aspect (switching conditions) in the driver's collision avoidance maneuver as well.

A road traffic participant is a person who directly participates in road traffic, such as vehicle drivers, passengers, pedestrians, or cyclists, however, traffic accidents cause numerous property losses, bodily injuries, and even deaths to them. To bring down the rate of traffic fatalities, the development of the intelligent vehicle is a much-valued technology nowadays. It is of great significance to the decision making and planning of a vehicle if the pedestrians' intentions and future trajectories, as well as those of surrounding vehicles, could be predicted, all in an effort to increase driving safety. Based on the image sequence collected by onboard monocular cameras, we use the Long Short-Term Memory (LSTM) based network with an enhanced attention mechanism to realize the intention and trajectory prediction of pedestrians and surrounding vehicles. However, although the fully automatic driving era still seems far away, human drivers are still a crucial part of the road‒driver‒vehicle system under current circumstances, even dealing with low levels of automatic driving vehicles. Considering that more than 90 percent of fatal traffic accidents were caused by human errors, thus it is meaningful to recognize the secondary task while driving, as well as the driving style recognition, to develop a more personalized advanced driver assistance system (ADAS) or intelligent vehicle. We use the graph convolutional networks for spatial feature reasoning and the LSTM networks with the attention mechanism for temporal motion feature learning within the image sequence to realize the driving secondary-task recognition. Moreover, aggressive drivers are more likely to be involved in traffic accidents, and the driving risk level of drivers could be affected by many potential factors, such as demographics and personality traits. Thus, we will focus on the driving style classification for the longitudinal car-following scenario. Also, based on the Structural Equation Model (SEM) and Strategic Highway Research Program 2 (SHRP 2) naturalistic driving database, the relationships among drivers' demographic characteristics, sensation seeking, risk perception, and risky driving behaviors are fully discussed. Results and conclusions from this short book are expected to offer potential guidance and benefits for promoting the development of intelligent vehicle technology and driving safety.

Advanced driver assistance systems (ADAS) are a subject of increasing interest as they are being implemented on production vehicles and also continue to be developed and researched. These systems need to work cooperatively with the human driver to increase vehicle driving safety and performance. Such a cooperation requires the ADAS to work with the specific driver with some knowledge of the human driver’s driving behavior. To aid such cooperation between human drivers and ADAS, driver models are necessary to replicate and predict human driving behaviors and distinguish among different drivers. This dissertation presents several lateral and longitudinal driver models developed based on human subject driving simulator experiments that are able to identify different driver behaviors through driver model parameter identification. The lateral driver model consists of a compensatory transfer function and an anticipatory component and is integrated with the design of the individual driver’s desired path. The longitudinal driver model works with the lateral driver model by using the same desired path parameters to model the driver’s velocity control based on the relative velocity and relative distance to the preceding vehicle. A feedforward component is added to the feedback longitudinal driver model by considering the driver’s ability to regulate his/her velocity based on the curvature of his/her desired path. This interconnection between the longitudinal and lateral driver models allows for fewer driver model parameters and an increased modeling accuracy. It has been shown that the proposed driver model can replicate individual driver’s steering wheel angle and velocity for a variety of highway maneuvers. The lateral driver model is capable of predicting the infrequent collision avoidance behavior of the driver from only the driver’s daily driving habits. This is important due to the fact that these collision avoidance maneuvers require high control skills from the driver and the ADAS intervention offers the most benefits, but they happen very infrequently so previous knowledge of driver behavior during these incidents cannot be assumed to be known. The contributions of this dissertation include 1) an anticipatory and compensatory lateral driver steering model capable of modeling a wide range of in-city and highway maneuvers at a variety of speeds, 2) the combination of the lateral driver model with the addition of defining an individual driver’s desired path which allows for increased modeling accuracy, 3) a predictive lateral driver model that can predict a driver’s collision avoidance steering wheel angle signal with no prior knowledge of the driver’s collision avoidance behavior, only data from every day, standard driving, 4) the addition of a longitudinal driver model that works with the existing lateral driver model by using the same desired path and is capable of replicating an individual driver’s standard highway and collision avoidance behavior, and 5) A feedforward longitudinal driver model based on regulating the driver’s velocity along his/her desired path is added to the existing feedback longitudinal driver model that together are capable of modeling an individual driver’s velocity for lane-changing and collision-avoidance maneuvers with less than 0.45 m/s (1 mph) average error.