The objective of this paper is to use the SOM to study the STAT Signaling Pathway heterogeneities of vehicle-following behavior. We use a trained SOM to show that when presented with similar stimuli (i) different car drivers respond with different magnitudes of acceleration when following cars; that is, car drivers have interdriver heterogeneity; (ii) the same car driver responds with different magnitudes of acceleration when following the same car; that is, the same driver has intradriver heterogeneity; and (iii) car drivers respond with different magnitudes of acceleration when the leaders are of different vehicle types. We called this phenomenon inter-vehicle-type heterogeneity.

In additional to proposing the SOM as a nonparametric vehicle-following model, the findings of interdriver heterogeneity, intradriver heterogeneity, and inter-vehicle-type heterogeneity serve as complements to limited earlier studies. After this introduction, the next section of this paper reviews the vehicle-following models and SOM. This is followed by a description of the data used in this research. The next section presents the SOM training. Subsequently, we present the results of using the trained SOM to analyze the interdriver, intradriver and inter-vehicle-type heterogeneities. The

findings are summarized towards the end of this paper. 2. Literature Review 2.1. Vehicle-Following Models A vehicle-following model is an equation (or a set of equations) that describes the movement of a driver-vehicle in response to the dynamics of the driver-vehicle immediately ahead,

when both vehicles are traveling in the same direction in the same lane. As a fundamental building block of microscopic traffic simulation, the realism of a vehicle-following model improves the accuracy of the simulation outcome, which in turn enables better transportation decision making. The historical development of vehicle-following models from 1958 to 1999 has been summarized in [1]. Many vehicle-following models have been proposed, tested, and used in microscopic simulation models over the years [2]. Brefeldin_A The deterministic model proposed by Gazis, Herman, and Rothery [3], often known simply as the GHR model, is one of the earliest and the most well-known models. The GHR model, also known as the General Motors (GM) model, takes the following form: x¨ft+Δt=λfx˙ft+Δtmx˙lt−x˙ftxlt−xftk, (1) where x¨ft is the acceleration of the follower f at time t; x˙ft is the velocity of the follower f at time t; x˙lt is the velocity of the leader l at time t; xf(t) is the position of the follower f at time t; xl(t) is the position of the leader l at time t; λf is the follower’s sensitivity constant; Δt is the time lag in the follower’s response; and m and k are calibration constants. The GHR model equates the follower’s response to the follower’s sensitivity multiplied by the external stimulus.