The corresponding molecule library number of winning neurons
for Pair 1852-1847 was 23. Figure 6 represents these two followers’ mean acceleration responses associated with the eight common winning neurons. As shown in the figure, the two followers (VINs 1794 and 1852) had different mean acceleration magnitudes for the same winning neuron. In two of the winning neurons, the signs of the accelerations are opposite. Overall, VIN 1794 has higher magnitudes of acceleration while VIN 1852 has heavier deceleration. The differences may be caused by the followers’ driving habits. Figure 6 Differences in mean response between two followers. 5.4. Intradriver Heterogeneity Another pair of passenger cars (Pair 350-346) in test data set I was selected to illustrate that, even if the same driver is presented with similar stimuli, his/her response may be inconsistent. This pair of vehicles has 50.5 seconds of vehicle-following observations, resulting in 101 vectors at 0.5 second intervals. Figure 7 plots the follower’s acceleration profiles over the duration of observation. The vertical color coded bars represent the winning
neurons identified by the SOM. The Δt in t + Δt in the horizontal axis is to account for the time lag when the stimulus occurs at time t. Five neurons are highlighted here as they have sufficient winning frequencies for subsequent analysis. Figure 7 Acceleration profile of selected vehicle pair and winning neurons. Figure 8 shows the distributions of VIN 350′s responses in three of the five winning neurons identified in Figure 7. According to Figure 4, on average, drivers decelerate in neurons (x = 10, y = 0), (x = 10, y = 1), and (x = 10, y = 3). It appears that, on average, VIN 350 has the same deceleration signs at neurons (x = 10, y = 0) and (x = 10, y = 3) which is consistent with the driver population in the training and test
data sets. However, the driver of VIN 350 has, on average, acceleration response at neuron (x = 10, y = 1) (see Figure 8(b)) while the average response in the data sets is deceleration. As plotted in Figure 8, when faced with similar inputs Brefeldin_A belonging to the same winning neuron, the driver of VIN 350 had varied responses. This evidence suggests that the same driver responded inconsistently when the stimulating factors are considered analogous. Figure 8 Distribution of response by VIN 350. 5.5. Inter-Vehicle-Type Heterogeneity In this subsection, the distribution of responses among the vectors in test data sets I and II was compared. Test data set I consisted of data from “car following car” scenarios while test data set II consisted of “car following truck” scenarios. For each stimulus at neuron (x, y), a two-tail paired t-test was conducted to see if the difference between the mean responses is significant.