Effect of distraction task on driving performance of experienced taxi drivers

The GDS-300s driving simulator (Gridspace Inc., Korea) provides front, left, and right visual information via 3 32-inch liquid crystal display monitors. The driving simulator is similar to the ‘Click’ model from Hyundai Motors, having driving devices (steering wheel, accelerating pedal, brake pedal, parking brake, direction indicator lever, emergency lamp lever, windshield wiper lever, headlight lever, gear lever, and safety belt lamp) and displays (direction indicator lamp, speedometer, revolutions per minute meter, temperature gauge, fuel gauge, and several warning lights). Motor-driven power steering, which is a motor control type simulator, was used as the steering wheel.

The protocol for the research project was approved by the Institutional Review Committee of Konkuk University (KUH116062) where the work was undertaken, and the protocol conforms to the principles of the Declaration of Helsinki and its contemporary amendments. The overall purpose and protocol for the experiments were explained to the participants before their enrollment in the study, and experiments were conducted after written consent informed was obtained.

Twelve male taxi drivers (age: 56.3 ± 4.4 y, with 28.4 ± 6.4 y taxi-driving experience) and 14 female taxi drivers (age: 55.5 ± 3.5 y, with 19.4 ± 5.0 y taxi-driving experience) participated in this study. Before starting the experiment, each participant was familiarized with the driving simulator through a practice driving session. The experimental design was controlled for external factors, such as the use of tobacco, alcohol, or coffee, which could possibly influence driving performance and physiological signals.

Participants were instructed to drive without changing lanes, while maintaining a constant 30 m headway gap from the car ahead on a 3-lane straight road in a vacant downtown area. The front car maintained a speed of 90 km/h. The headway distance was displayed on the screen of the simulator in real time. The experimental paradigm is shown in Figure 1.

Figure 1

After resting in the driver’s seat for 3 min (rest phase), participants drove for 2 min while maintaining a constant gap (30 m) and speed (90 km/h). Participants were driving only for the first 1 min (control phase). For an additional 1 min, they were instructed to drive only or perform a task (STM or SN) while driving (task phase). For a task that involves following the car in front, if the task takes too long, the driver may feel bored and may not be able to focus on the experiment. Therefore, as suggested by preliminary experiments, the driving time was restricted to 2 min so that most subjects did not feel bored. In the case of STM, subjects were instructed to use their own mobile phones and put the phone in the most convenient place they could find within 30 cm of the steering wheel (Figure 2).

Figure 2

In the case of SN, participants were instructed to use the navigation system provided by the experimenter who placed a car-mounted-type navigation on the dashboard. In particular, when participants were not accustomed to the navigation, they were instructed to familiarize themselves with and practice use of the navigation system in advance. All subjects were right handed and they were requested to conduct the tasks with their right hand. All subjects participated in 3 types of experiments: driving only, driving + STM, and driving + SN. During the experiments, the experimenter counted the number of crashes with the car in front. The order of experiments was counterbalanced. In detail, after completing one experiment, participants took a 30-min rest to account for possible simulator sickness before moving on to the next experiment. The experiment time was approximately 1.5 h for each participant.

The activation level of the sympathetic nervous system was measured using SCL. An MP100 data acquisition system and AcqKnowledge version 3.8.1 (Biopac Systems) were used to measure the SCL from the index and middle fingers on the left hand. The sampling rate of the physiological data was 500 samples/s.

The relative change of SCL was calculated, which was obtained by dividing the difference between the SCL in the task phase (SCL_Task) and the SCL in the control phase (SCL_Control) by the SCL _Control (Eq. 1).

Relative change of SCL (%)=SCLTask−<!-- - -->SCLControlSCLControl×100$$\begin{array}{} \displaystyle \text{Relative change of SCL }\big(\text %\big) = \frac{{\rm SCL_{Task} - SCL_{Control}}}{{\rm SCL_{Control}}} \\\qquad\qquad\qquad\qquad\qquad\qquad\quad\,\,\times 100 \end{array} $$(1)

The driving simulator supplied information on the distance from the car ahead as well as the vehicle’s speed and location. Using the information regarding the distance from the car ahead, the average following distance was calculated.

The speed deviation was calculated by using the vehicle speed and obtained by dividing the difference between the intended design speed (90 km/h) and operating speed by the design speed (Eq. 2).

Speed deviation (%)=Design speed - Operating speedDesign speed×100$$\begin{array}{} \displaystyle \text{Speed deviation } \big(\text %\big) = \frac{\text{Design speed - Operating speed}}{\text{Design speed}} \\ \qquad\qquad\qquad\qquad\qquad\times 100 \end{array} $$(2)

Using the information on distance from the car ahead, the APCV, which represents vertical control of the car, was calculated. The APCV is obtained by dividing the standard deviation (AP_STD) of the distance from the car ahead by the mean value (AP_Mean) (Eq. 3). The APCV, as an indicator showing the variability of the front and rear movements of a car, is one of the control variables influenced by the driving performance.

APCV=APSTDAPMean$$\begin{array}{} \displaystyle \text{APCV}=\frac{{\rm AP_{STD}}} {{\rm AP_{Mean}}} \end{array} $$(3)

Using the location data of the car from the driving simulator, the MLCV, which describes the lateral controllability of the car, was calculated. MLCV was defined by Eq. 4, the standard deviation of the positional coordinates of the medial–lateral movement (ML_STD) of the car divided by the mean value (ML_Mean). This is an index explaining the variation in the left and right car movements in the center of the driving lane and is one of several control variables influenced by driving performance.

MLCV=MLSTDMLMean$$\begin{array}{} \displaystyle \text{MLCV}=\frac{{\rm ML_{STD}}} {{\rm ML_{Mean}}} \end{array} $$(4)

Relative change of SCL was calculated for 40 s, excluding the first 10 s and final 10 s in the control phase and the task phase that included driving only, driving + STM, and driving + SN. Average following distance, speed deviation, APCV, and MLCV were calculated for 40 s, excluding the first 10 s and final 10 s in the task phase including driving only, driving + STM, and driving + SN (Figure 1).

Concerning the factors relative change of SCL, average following distance, speed deviation, APCV, and MLCV, a repeated one-way analysis of variance, which considers an independent variable condition (driving only, driving + STM, and driving + SN) was conducted (SPSS, version 12.0k).

As seen from Table 1, significant differences were found in the relative change of SCL between driving only, driving + STM, and driving + SN (F = 11.8, P < 0.001). Bonferroni post hoc tests showed (Figure 3) that the relative change of SCL of driving + SN (P < 0.001) and driving + STM (P < 0.01) increased significantly, compared with that of driving only.

Table 1

Mean and standard deviation according to conditions (driving only, driving + sending text message [STM], driving + searching a navigation device [SN]) in the relative change of skin conductance level (SCL), average following distance, speed deviation, anterior–posterior coefficient of variation (APCV), and medial–lateral coefficient of variation (MLCV)

Figure 3

As seen from Table 1, there were significant differences in the average following distance between driving only, driving + STM, and driving + SN (F = 28.4, P < 0.001). Bonferroni post hoc tests showed (Figure 4) that the average following distance of driving + SN increased significantly, compared with that of driving only (P < 0.001) and compared with that of driving + STM (P < 0.05). The average following distance of driving + STM increased significantly, compared with that of driving only (P < 0.001).

Figure 4

As seen from Table 1, significant differences were found in the speed deviation between driving only, driving + STM, and driving + SN (F = 25.5, P < 0.001). Bonferroni post hoc tests showed (Figure 5) that the speed deviation of driving + SN (P < 0.001) and driving + STM (P < 0.001) increased significantly, compared with that of driving only.

Figure 5

As seen from Table 1, there were significant differences found in the APCV between driving only, driving + STM, and driving + SN (F = 16.7, P < 0.001). Bonferroni post hoc tests showed (Figure 6) that the APCV of driving + SN increased significantly compared to that of driving only (P < 0.001) and that of driving + STM (P < 0.05). The APCV of driving + STM increased significantly compared with that of driving only (P < 0.01). There were no crashes with the car in front during the experiments.

Figure 6

As seen from Table 1, significant differences were found in the MLCV between driving only, driving + STM, and driving + SN (F = 42.4, P < 0.001). Bonferroni post hoc tests showed (Figure 7) that the MLCV of driving + SN increased significantly, compared with that of driving only (P < 0.001) and that of driving + STM (P < 0.05). The MLCV of driving + STM increased significantly compared with that of driving only (P < 0.01).

Figure 7