Keywords

1 Introduction

Recently, collision avoidance technology (CAT) had emerged as a new way to reduce the number of car accidents. Numerous technologies have been introduced to develop a better solution for crash avoidance [1,2,3]. Breakthroughs in the development of CAT offer a promising future in vehicle safety and in saving human lives. Those technologies should be thoroughly evaluated in the context of driving to understand how CAT systems may influence driving performance and safety. However, the previous studies related to CAT evaluation were mainly focused on the false-positive warning rate, the lane change test, and the warning prevention test in driving simulators (or the closed-course test tracks). According to our previous study, drivers showed different visual scan paths and driving performance when they received different types of stimulus warning in a real driving condition [4]. For that reason, the need for testing the warnings in on-road environments is growing, and little research has been performed in open-road conditions to investigate the acceptance and the effectiveness of the collision avoidance systems.

There are also several factors that can influence driving behavior, such as personality, emotion, demographics, policy, spatial and temporal elements [5, 6], inadequate mental workload, distractions, stress, anxiety, and time pressure, which can affect driving performance [7, 8]. The combinations of these influential factors can affect the driver’s acceptance level and the effectiveness level of CAT systems in a vehicle. Hence, it is important to investigate whether or not implementing CAT could influence these factors negatively while driving an open track.

The frequent false alarms cause driver aversion to the warnings, sometimes high false alarm rate even can cause the increases in mental workload and distraction of driving [9]. According to the previous research did by Maltz and Shinar [10], drivers tend to ignore the false alarm and treat it as noise to avoid any unnecessary reaction. However, most of the previous research was conducted based on the driving simulator, so it is necessary to test the influence of collision warning in real driving, especially from the drivers’ perspective. The collision avoidance technologies have shown potential benefit in reducing vehicle accidents. However, their success has heavily relied on driver’s acceptance [11]. To measure user acceptance of CAT system, the technology acceptance model (TAM) [12] was used in this study. The model evaluates person’s attitude toward using the system (A) and the perceived usefulness (U). The model was applied in other research and tested how age, gender, and environment influence the acceptance and effectiveness of collision warning [13]. Compared to other models, such as AttracDiff, User Experience Questionnaire (UEQ), or meCUE, TAM is a simple and powerful model which focuses on user acceptance behavior on system-related technology.

The purpose of this study was to conduct a usability study of a collision avoidance system in public transportation. The benefits of implementing CAT in a public transportation system were revealed through this study. To conduct the usability study, five professional drivers from OATS tested the device for three weeks. The experiment was designed to monitor how the drivers performed differently in real-world driving environments after the instrument was installed. The tested CAT device sensed vehicles in the forward path and other lanes, as well as generated warnings in response to a collision threat. We hypothesized that the total number of alarms, including lane departure and forward collision warning, should be reduced significantly if there was an impact on driving habits caused by the warnings from CAT. Furthermore, this study was designed to conduct a user perception and opinion survey for the collision avoidance system, as we collected survey responses from the drivers. The survey was designed to gather their perception and feedback on the collision avoidance system. Based on the result from drivers’ survey responses and the warning reports, the acceptance and effectiveness of the collision avoidance system were evaluated.

2 Method

2.1 Apparatus

Five Ford E350 buses were used as testing vehicles for the experiment (see Fig. 1). The seating capacity is twelve passengers. These vehicles were operated by Older Adults Transportation Service (OATS) center in Columbia, Missouri. Since the drivers provide transportation service to customers, the driving path was totally based on the clients’ needs during the study without any prior design.

Fig. 1.
figure 1

Study vehicle

Full size image

The collision avoidance device typically consists of multiple sensors and provides the environmental information surrounding the vehicle. In this study, we selected one of the well-known aftermarket advanced driver assistance systems as a collision avoidance device. The device uses a combination of a vision sensor, camera, and an image processing board to detect signals. This combination allows it to better detect where the lines of roads are and to calculate the distance between vehicles. The device includes the forward collision warning (FCW) and the lane departure warning (LDW) features. The details of the collision avoidance device are below:

  • FCW (>15 mph): various levels of sound and visual warning. It could cover both high and low speed for the distance between front vehicles. The visual warning is represented by a red car icon on the display.

  • LDW (>37 mph): audio (a series of high-pitched beeping sound) and visual (flashing two white lane indication sign on display) alerts

  • Connected to a vehicle’s turning signal. This function is developed to avoid the LDW with turning signal since drivers intentionally change lanes.

The collision avoidance device was installed in all testing vehicles. A sensor camera was located near the top center of the windshield. The display unit was located at the bottom left corner of the car. Figure 2 shows the location of the device display.

Fig. 2.
figure 2

Collision avoidance system setup

Full size image

2.2 Participants

Five professional drivers (4 males and 1 female) from OATS participated in the study. All participating subjects in the study had enough driving experience to be considered expert drivers. The mean age of male drivers was 70.25 (SD: 6.495), and the only one female driver is 39 years old. The average driving experience of male driver was 55.5 years (SD: 7.297), and the female driver has 22 years driving experience. All five participants had at least one year of working experience in this current position of OATS. They were all professional drivers and familiar with the driving path of daily working. Each driver had assigned vehicle; they drove the same vehicle every day for work.

2.3 Procedure

The experiment was conducted in three stages: (1) stealth stage, (2) training stage and (3) active stage. For the stealth stage, the collision avoidance device did not provide any audio and visual warnings to drivers. The participants were asked to drive their vehicles as they had in the past without any knowledge of the collision avoidance system. However, the device was turned on and kept working by itself. A black box was also installed to record the number of warnings from CAT device. When the device detected any threats in driving, those events were recorded as alerts in a database without notifying drivers. The stealth stage was set as the baseline of drivers to collect their normal driving data in daily working. The experiment was conducted for three weeks as the stealth stage. After that, the drivers learned how the collision avoidance system works. During the week of the training, the drivers learned how the warning works without hearing any warning from the device while driving. After the training, we checked the driver’s awareness regarding the collision warning. Finally, the audio and visual systems were turned on during the driving as the active stage. The data was collected for three weeks during the active stage as well to compare with data from stealth stage. After finishing the all three stages, we provided the questionnaires to drivers about the collision avoidance device based on they experienced and the feedbacks of using the CAT device.

3 Measures

3.1 Warning Report

After completing all three stages: stealth stage, training stage, and active stage, the total number of alarms for each day was gathered as the result of the warning report. This report helped us to see the effect of the collision avoidance system while driving. Comparing the warning data between the stealth stage and the active stage provided measurable outcomes of the effectiveness of the CAT device, such as normalized warning (how many warnings were generated per 100 miles and a total number of warnings per day).

3.2 Questionnaires

We designed and conducted a user perception and opinion survey for the collision avoidance system. This post-experimental survey contained 30 questions, including demographic information, driving experience, and a perceptive measures scale. The questions were reviewed and validated by a subject matter expert in the field of human environmental sciences. We visited the OATS center three times and handed out the questionnaires after the active stage. The drivers answered the questions and provided their feedback separately without any interactions between colleagues. All questions were related to the driver’s feelings about the device and experience of using it during the active stage, such as “How often did you meet heavy traffic with collision avoidance system?”, “Do you think the collision avoidance system helps you in reforming your driving habits?”, and “How often did you check the collision avoidance system monitoring display when there was a sound warning?”, to understand drivers’ overall driving behavior changes by collision warning and driving conditions. We also asked them to answer the questions about whether the CAT system provided a correct warning or not (e.g., “Do you think the collision avoidance system provides the accurate forward collision warning?”). The scale was from 1 to 7 (1 represents to “never”, 7 is “always”, medium level 4 as “sometimes”). Also, some questions were designed to analyze user’s acceptance and willingness of driving with the collision avoidance system (e.g., “Did you feel safe driving with this collision avoidance warning compared to driving without it?” and “Would you desire to drive with the collision avoidance warning?”). Questions also covered other items like “pleasant” and “annoying”. Unpleasant may happen when drivers think the system kept judging them even though the warning was helpful. Annoying is about the drivers’ negative feeling about the warning sound and sensitive level if they think the warning is the noise. The measuring scale was from point 1 to 7. The larger rating score meant a better user experience, like “very safe, very desirable”, while rated score 1 corresponded to “very unsafe, very undesirable.”

3.3 User Acceptance Model

In the questionnaires, some questions were designed to calculate the user acceptance level of the collision avoidance system. Son et al. [13] studied drivers’ acceptance model of the advanced driving assistance system in terms of ‘safe,’ ‘desirable,’ ‘pleasant’, and ‘comfort.’ In the user acceptance model, ‘Safe’ means the level of feeling safe while using the system. ‘Desirable’ is defined as the rate of driver’s interest to use the system after the test. ‘Pleasant’ reflects the driver’s preference level of this technology. ‘Comfort’ reflects the level of tranquility about the technology. Among these four subjective experiences of feeling, ‘safe’ and ‘desirable’ represent positive responses, while ‘unpleasant’ and ‘annoying’ indicate negative feedback. The Technology Acceptance Model (TAM) was also used to quantify the computer-related technology acceptance behavior (A) of the user [12]. It is based on perceived usefulness (U) and perceived ease of use (EOU).

$$ A = U + EOU $$
(1)

The usefulness (U) is based on ‘safe’ and ‘desirable’ rating scale from the participant’s response.

$$ U = (S_{safe} + S_{desirable} )/2 $$
(2)

Where \( {\text{S}}_{\text{safe}} \) is the subjective score of safety rating. Sdesirable is the subjective score of desirable rating from the survey results.

The ease of use (EOU) term is calculated by using the subjective rating of ‘pleasant’ and ‘comfort’ in the questionnaires:

$$ EOU = \left( {S_{pleasant} + S_{comfort} } \right)/2 $$
(3)

Where Splesant is the subjective score of pleasant rating. Scomfort is the subjective score of comfort rating.

Finally, the user acceptance of CAT is measured by the mean of the usefulness (U) and the ease of use (EOU).

$$ A = \left( {\frac{U + EQU}{{2*C_{Ratingscale} }}} \right)*100\% $$
(4)

Where CRatingscale is the subjective rating scale (from 1 to 7).

4 Results

4.1 Warning Data Analysis

Table 1 shows the normalized number of weekly warnings. Because of a failure in a data collection device in vehicle #4, the warning data from four testing vehicles (#1, #2, #3, and #5) were used in this analysis. According to the result, we could see the decrease in the number of FCW when the data from stealth stage was compared to the active stage [F(1,17) = 7.934, P = 0.012]. It means that FCW from the device significantly influenced the driving behavior related to a forward vehicle collision. As time goes by, drivers tried to adjust their driving behavior to avoid warning after experiencing with CAT warning for a period of time. Also, there was a weak significant difference on LDW [F(1,17) = 4.78, P = 0.044]. According to the individual warning report three subjects showed significant difference on FCW between stealth mode and active mode (subject 1 [F(1,24) = 24.55, P < 0.001], subject 2 [F(1,6) = 10.63, P = 0.004], and subject 5 [F(1,22) = 12.5, P = 0.002]). For LDW, only one subject showed a significant difference between stealth mode and active mode (F(1,24) = 16.03, P = 0.001).

Table 1. Comparison of normalized weekly warning
Full size table

4.2 Questionnaires

The survey results (see details in Table 2) showed that the average condition of heavy traffic driving was 5.4 out of 7, with around 50% on the highway while 30% urban driving. The path familiarity rating was 5.4 out of 7. Both results indicated that the drivers had experienced medium-high level of workload during the experiment. Regarding the attention paid to the collision avoidance system, the warnings influenced their visual attention, and the drivers checked the device visual display after they acknowledged alarms. Based on the driver’s experience during the experiment, they experienced that both warnings had a similar accuracy level [F(1,4) = 2.08, P = 0.187]. The accuracy of FCW and LDW from the user’ perspective rating were 4.8 and 5.8 out of 7, respectively.

Table 2. Survey results
Full size table

4.3 Questionnaires

Table 3 shows the average score of the user acceptance level of the device. According to the five drivers’ responses, their acceptance level of the collision avoidance system was about 69.28 out of 100. Also, there was no significant difference on user acceptance level between FCW and LDW [F(1,4) = 0.05, P = 0.824].

Table 3. Acceptance level (Scale 0–100%)
Full size table

5 Discussion and Conclusion

This study investigated the acceptance and effectiveness of a collision avoidance device in public transportation. The objective of the study was to estimate the potential benefits of implementing collision avoidance technology (CAT) to public transportation. Five professional drivers from OATS participated and tested the collision avoidance device for several weeks.

The results showed that there was a significant difference in the total number of weekly warnings between the stealth stage and the active stage. It means that the collision avoidance system was accepted by drivers and appeared to have positive effects on their driving behavior. However, the impact of LDW was less significant than that of FCW. Three out of five drivers showed a significant reduction in the number of FCWs. On the contrary, for LDW, only one driver demonstrated a major improvement in his driving behavior. One of the possible explanations for this difference is dissimilar driving behavior regarding forward collision and lane departure. According to the research done by Gasper et al. [14], drivers commonly responded with a combination of braking and steering when they detected threats of forward collision or lane departure. They found that the drivers chose different maneuvers to avoid the potential hazard, depending on their detection time of the threat. In the present study, the detection time of forward collision and lane departure were different, and it was based on the detection algorithms of the device. It means that the sensitivity level of the algorithms could influence the effectiveness of the collision avoidance device.

For the acceptance level, the overall acceptance was about 69%. Compared to the previous study [13], there is no significant difference in the acceptance level between public transportation vehicles and passenger vehicles. However, the user acceptance level for LDW from the driver in a public transportation vehicle was higher than the passenger vehicle (FCW: 71.7%, LDW: 69.5%).

Besides, according to the drivers’ feedback, three drivers (subjects 1, 3, and 4) reported positive feedback for the lane departure warning. Also, two drivers (subjects 1 and 3) felt confident about the forward collision warning. Subjects 1 and 3 (one male and one female) recommended using the collision avoidance system with public transportation. They said that the device helped them to maintain lane position and to be aware of measuring the distance from a leading vehicle. Although it took time to become familiar with both LDW and FCW, both drivers had no problems with using the device after they got used to it. Many drivers reported that potentially safer after installing the collision warning device since it can correct driving habits in some degree. Some drivers said they used turn signal more frequent with lane departure warning exist. The user acceptance level of subjects 1 and 3 were 80% and 78%, respectively. However, subjects 2 and 4 showed negative feedback with using the collision avoidance system. Subject 2 said that the warnings distracted and annoyed him due to a lot of false warnings. His user acceptance level was 20%. He stated that he would prefer to use the device in a personal car rather than a public transportation system because passengers get nervous due to the audible alarms. Also, when he drove in a city with heavy traffic congestion, he was often frustrated by the many forward collision warnings generated because of the close distance with front cars, which are hard to avoid in heavy traffic. Subject 4 also had several negative comments about the device. He felt the FCW function should be improved further because the device was not beneficial in larger cities. Subject 5 agreed to keep the collision avoidance system in a vehicle.

This study showed that a CAT system appeared to have positive effects on their driving behavior. The overall effectiveness of a CAT system was about 75%, and the acceptance was about 69%. In short, 75% of drivers (3 out of 4) showed significant differences in driving behavior after they used a CAT device. 60% of drivers reported positive feedback for the lane departure warning. 40% of drivers (2 out of 5) felt confident about the forward collision warning. The OATS drivers reported 69.28% of acceptance level of using the CAT device. All our findings support that there are potential benefits to implementing CAT in a public transportation system. However, it is vital to understand how drivers set their threshold to accept the warning from collision avoidance features. It can be concluded that it is essential to study how a driver uses a CAT device to avoid unexpected negative impacts and provide appropriate safety parameters. Encouragement should be provided to administrators to install the CAT systems in public transportation vehicles as the results of the current study show satisfactory effectiveness and acceptance levels.

In conclusion, CAT systems are beneficial to prevent vehicle accidents. However, these technologies should be evaluated, in the context of driving, to understand how the CAT devices influence driving performance and safety in real driving conditions.

There are several identified limitations in this study. First, this study considered only a limited age group of drivers. According to the research done by Son et al. [13], an advanced driver assistance system is more efficient for older drivers over age 70. They also reported that the acceptance of collision warning might differ by drivers’ age and gender on a wider variety of vehicles. Future research should consider recruiting the entire age group to see if the acceptance and effectiveness changed in different age groups. Second, although we could estimate the benefits of using CAT in a public transportation vehicle, it was impossible to see the reduction rate of vehicle crashes during the experiment. Therefore, future research should study the effectiveness of CAT in a public transportation system for a longer duration (>6 months) to observe the changes in the number of vehicle crashes. Moreover, for this study, we did not include any physiological measures to assess drivers’ response to collision warning in a real driving environment. In the future, we plan to introduce some physiological data analysis, such as eye tracking data, electromyography (EMG) data to reflect the collision warning impact on driving behavior.