Chapter 6 Impact of Secondary Tasks on Driving Performance
6.3 Pre-Analysis
6.3.1 Data Exploring Analysis
Before the main analysis, data were explored for the normality, homogeneity and existing outliers. Most of the driving performance measurements followed the requirements of the ANOVA Test, therefore most analyses were conducted using the ANOVA Test. When a required assumption was violated, a Non-parameter Test would be applied instead. Some outliers were observed in the DEA processes. They were then either excluded from the analysis, or remained, depending on whether the outliers were legitimate or not (see Section 5.4.1). Similar to what was described in Chapter 5, for skewed and un-symmetric parameters, or the ones with outliers, non-linear transformations were conducted for these measurements, and the new variables were named after the original ones, with “Trans_” as prefix. For example, the transformed MNHW was named as Trans_MNHW. Again, for a better explanation of the data, the results for transformed parameters were only reported when the conclusion was different from the original parameters. It was found that the headway data from one of the test runs by Subject # 7 were remarkably higher than those collected in other tests, and appeared as outliers,
see Figure 6.2 as the Box Plot for MNHW when drivers performing different types of tasks. This dataset was then scrutinised by examining the distance headway from the video data. The length of the road markings, vehicle type and size in the image were used as reference to estimate the range of the headway. It was finally decided that these outliers were due to technical issues, and should be excluded from further analysis. Other parameters derived from data of the same test run were double-checked by the same method.
Figure 6.2 Outliers in Mean Headway (MNHW)
6.3.2 Impact of Leading and Tailing Vehicle Types
Previous research has suggested that factors such as types of leading vehicles can cause changes in driving performance (Shehab, 2007). The effects of leading and tailing vehicle types (LVT and TVT respectively) were therefore investigated before the main analysis, to rule out the impact of these factors. Information of leading vehicle and tailing vehicle types were included in the driving performance database, where the leading and tailing vehicle types were
differentiated from the video images (collected by the IV during the experiment) according to their size and appearance, and the information of the adjacent vehicle condition (including the leading and tailing vehicle types, and the starting/ending time of each following-episode) were recorded manually (see Chapter 4). As mentioned in Chapter 4, four vehicle types were considered in this research for both LVT and TVT: passenger car (represented as vehicle type “1” in the following analysis), heavy goods vehicle (vehicle type “2” in the following
description), van (vehicle type “3”), and no leading or following vehicle within distance which could impact the driver (recorded as “0”).
6.3.2.1 Impact of Leading and Tailing Vehicle Types in Car-Following
In the Car-Following scenario, all data were collected when there was a vehicle ahead (i.e. with a leading vehicle), therefore, the LVT investigated were only 1 (Passenger Car as leading vehicle), 2 (Lorry) or 3 (Van).
ANOVA tests showed that LVT had no effect on almost any driving performance measurements, apart from MinTHW (F (2, 997) = 10.224, p < 0.001). Tukey HSD Post Hoc Tests suggested that the impact on these parameters mainly existed when LVT = “3” (i.e. Van as leading vehicle) comparing with other two groups. The effects of LVT on MinTHW are demonstrated in Table 6.2. In the Car-Following scenario, there were 164 cases when data collected with a passenger car as a leading vehicle (i.e. LVT = 1), 761 cases of a lorry as leading vehicle, and 75 cases of vans as leading vehicles. As cases of following a van ahead was also the minority in the dataset, the cases with van-following were excluded for the analysis in Car-Following.
Table 6.2 Effects of LVT on MinTHW in Car-Following
LVT Mean N SD
Passenger Car 1.20 164 0.28 Lorry 1.19 761 0.31 Van * 1.36 75 0.25 Note: * denotes the p-value < 0.05.
The effect of TVT was investigated in 4 groups: 1: no tailing vehicle; 1: Passenger Car; 2: Lorry; and 3: Van. No effect of TVT has been found for the current dataset. Therefore, in the
Car-Following Scenario, cases of all TVT were included in the main analysis.
6.3.2.2 Impact of Leading and Tailing Vehicle Types in Free-Driving
In the Free-Driving Scenario, the purpose of investigating the effect of LVT and TVT was to avoid the effect of leading and tailing vehicle types on the speed choice and other driving behaviour. It was found that LVT had no effect on driving performance measures in any tasking condition (p < 0.05). The TVT had impacts on drivers steering behaviour. When there was no task or task demand was relatively low (i.e. performing auditory tasks), the effects mainly reflected on the MNSP and SDSP, while when performing visual tasks, the effect were more noteworthy on steering behaviour (p < 0.05). Tukey HSD Post Hoc Tests suggested that the impact on these parameters mainly existed when LVT = “2”, i.e. when there was a lorry behind as a tailing vehicle, significantly fewer steering wheel adjustments and lower steering entropy were observed. Again, as there were only 27 cases with a lorry as tailing vehicle (out of 839
cases in total), the cases with lorry as TVT were excluded for the analysis in Free-Driving. The effects of TVT on steering behaviour are demonstrated in Table 6.3.
Table 6.3 Effects of LVT on steering behaviour in Car-Following No Tailing Vehicle Passenger Car Lorry * Van TVT
Mean N SD Mean N SD Mean N SD Mean N SD
RR1.5 34.49 483 15.15 37.28 266 16.56 28.21 27 14.90 32.90 63 13.87 SE 0.69 483 0.10 0.70 266 0.10 0.64 27 0.14 0.68 63 0.09
Note: * denotes the p-value < 0.05.
6.3.2.3 Summary on the Impact of Leading and Tailing Vehicle Types
The effects of leading and tailing vehicle types (LVT and TVT respectively) on driving
performance were investigated, to avoid potential interaction with the impact of secondary tasks
and driver characteristics. In the Car-Following scenario, TVT did not affect driving behaviour, but LVT had significant effects on drivers - following a van ahead caused longer minimum time headway than for other leading vehicle types. In the Free-Driving Scenario, there were no LVT effects, but TVT did cause some steering behaviour changes - drivers made significantly less steering wheel adjustment per minute when there was a lorry behind. Therefore, these cases (where performance was significantly affected by the effects of LVT and TVT) were excluded from the main analysis. In total, 75 cases of following a van were excluded from the Car- Following Scenario, and 27 cases of tailing a lorry excluded from the Free-Driving Scenario, which represents less than 7.5% and 3% of the total respectively.