They found that intra-driver variability rather than interdriver variability accounts for a large part of the calibration errors. Siuhi and Kaseko appear to be the first to use the Next Generation SIMulation (NGSIM) vehicle trajectory data set to analyze vehicle-following behavior Afatinib EGFR inhibitor [6]. They calibrated the GHR model (without the Δt term in the follower’s velocity) using the data collected at the U.S. 101 Freeway in Los Angeles. They showed the distributions of Δt during acceleration and deceleration, with deceleration having a smaller mean Δt value. The same study also analyzed the distributions of m and k values and recommended

different sets of m and k values for acceleration and deceleration, respectively, even for the same drivers. The different Δt, m, and k values in acceleration and deceleration lead to the so-called asymmetric vehicle-following phenomenon.

Siuhi [5] affirmed that different Δt, m, and k values are necessary to also account for vehicle types of the leader and the follower. Wang et al. studied interdriver and intradriver heterogeneities using vehicle trajectory data collected at the A2 Motorway in Utrecht, the Netherlands [7]. They calibrated the Helly model, Gipps model, and Intelligent Driver model. They found that, for the majority of the drivers, (i) the Δt for deceleration was smaller than that for acceleration; (ii) when the same vehicle-following model was fitted to the data, the fitted parameter values for acceleration and deceleration conditions were different; and (iii) the best fitted model took different forms in acceleration and in deceleration. Ossen and Hoogendoorn presented the results of five vehicle-following models which were calibrated against vehicle trajectory data collected at the A2 Motorway in Utrecht and the A15 Motorway

in Rotterdam, the Netherlands [8]. They compared the models when a car was following a car and when a car was following a truck. Among the findings were (i) different vehicle-following models best fitted different passenger cars; (ii) truck tended to be driven in a relatively lower speed variance compared to passenger cars; and (iii) the desired headways are lower when a car was following a car compared to a car following a truck. Their findings showed interdriver heterogeneity between passenger cars and well as the heterogeneity depending on the leader’s vehicle type. The above recent studies Cilengitide have shown that heterogeneities in vehicle-following behavior exist (i) for the same follower during acceleration and deceleration; (ii) for the same follower, when the leaders are of different vehicle types; (iii) between different followers, even when the leader-follower pairs are of the same vehicle combination. 2.2. Self-Organizing Feature Map The SOM, introduced by Kohonen [20], is motivated by the self-organization characteristics of the human cerebral cortex.