44 but does require sufficient real traffic and action data to capture the underlying relationship between states and actions. Therefore, ANN models estimate actions on the basis of the real state–action mapping of natural behavior. Data from the naturalistic driving database of the Naturalistic Truck Driving Study (NTDS) (1), collected by the Virginia Tech Transportation Institute and Blanco et al. (2), are used to find the real causalities and responses of truck drivers in car-following situ- ations. In this paper, first the process of calibrating the well-known Gazis–Herman–Rothery (GHR) car-following model for an individual driver is described. Then, a back-propagation (BP) neural network is constructed to train agents that represent individual drivers. Both methods use the same naturalistic data set. CAR-FOLLOWING MODELS Brief Review Many car-following models have been developed in the past 50 years to represent longitudinal driver behavior, including safety distance models, collision avoidance models, psycho-physical action point models, and models based on fuzzy logic (3). Most well-known car-following models have been embedded in microsimulation software, such as the Pipes model in CORSIM (4), the Gipps model in AIMSUN (5), the Fritzsche model in Paramics (6 ), and the Wiedemann model in VISSIM (7 ). Car-following models assume that the following vehicle reacts according to observed stimulus from its leader according to predefined rules. The models mentioned above require specific defined functions to relate stimuli that the following vehicle observes to the reaction it takes. For example, the GHR model uses speed difference and space headway as stimuli to determine acceleration of the following vehicle. The Wiedemann model divides headway and speed difference space into several driving regimes with predefined thresholds, where the following vehicle reacts differently each regime. The Wiedemann model uses the differences between actual and desired following distances as a stimulus in the closing-in regime, a calibrated accel- eration in the following regime, and a desired speed as the driving objective in the free-driving regime (7 ). The Gipps model uses vehicle dynamics as constraints and derives acceleration of the following vehicle from estimated deceleration of the leading vehicle (5). Driver Behavior Simulation The calibration of a car-following model is an important process to represent driver behavior and simulate vehicle trajectory. Calibration parameters are considered to be driver dependent and to remain Simulation of Driver Behavior with Agent-Based Back-Propagation Neural Network Linsen Chong, Montasir M. Abbas, and Alejandra Medina Two microscopic simulation methods are compared for driver behavior: the Gazis–Herman–Rothery (GHR) car-following model and a proposed agent-based neural network model. To analyze individual driver charac- teristics, a back-propagation neural network is trained with car-following episodes from the data of one driver in the naturalistic driving database to establish action rules for a neural agent driver to follow under perceived traffic conditions during car-following episodes. The GHR car-following model is calibrated with the same data set, using a genetic algorithm. The car-following episodes are carefully extracted and selected for model calibration and training as well as validation of the calibration rules. Performances of the two models are compared, with the results showing that at less than 10-Hz data resolution the neural agent approach out- performs the GHR model significantly and captures individual driver behavior with 95% accuracy in driving trajectory. The simulation of driver actions in traffic is an important part of modeling microscopic driver behavior. A driver action indicates driver behavior in terms of causalities and responses to traffic flow. Microscopic car-following models provide many powerful simula- tion tools to study individual driver behavior, interactions between leading and following vehicles, and cumulative macroscopic traffic phenomena. The performance of car-following models relies on the parameters of individual drivers that can represent unique driving behavior. Parameter calibration becomes a necessary process before car-following models can be applied to a simulation environment. Driver actions in car-following models are defined by predefined rules. These rules are mostly interpreted by relating the traffic state that a driver observes to the response or action that the driver takes. Different car-following models consider different criteria as causal- ities that stimulate a driver’s reactions. However, in reality, these rules specified by car-following models might not capture natural driving behavior because of the complexity and instability of the human decision-making process. In the proposed approach, instead of using predefined driving rules from car-following models, a reactive-structure artificial neural network (ANN) is used to relate traffic states to driver actions. An ANN does not require a function to connect traffic states to actions L. Chong, 301-D Patton Hall, and M. M. Abbas, 301-A Patton Hall, Charles Via, Jr., Department of Civil and Environmental Engineering; and A. Medina, Virginia Tech Transportation Institute, 3500 Transportation Research Plaza; Virginia Polytech- nic Institute and State University, Blacksburg, VA 24061. Corresponding author: M. M. Abbas, abbas@vt.edu. Transportation Research Record: Journal of the Transportation Research Board, No. 2249, Transportation Research Board of the National Academies, Washington, D.C., 2011, pp. 44–51. DOI: 10.3141/2249-07