Advertisement

Automotive Innovation

, Volume 1, Issue 1, pp 76–84 | Cite as

Evaluation of the Effectiveness of Awareness Messages for Road Traffic Hazards in Experimental Tests

  • Ai Takeda
  • Makoto Kondo
  • Tetsushi Mimuro
Article

Abstract

Advanced driver assistance systems, especially autonomous emergency braking and forward collision warnings, have become popular in Japan. To reduce the number of road traffic accidents, safety information should be provided to a driver earlier than avoidance or warning messages so as to avoid a risky situation. A series of actual running tests was conducted to evaluate the activation timing and effectiveness of awareness messages. Objective analysis showed that the drivers could avoid an obstacle with a sufficient safety margin thanks to any of the awareness messages. Subjective ratings showed that the best timing is 10 s before encountering the obstacle. The results of objective analysis are limited in the present paper, and further analyses are required.

Keywords

Active safety Advanced driver assistance systems Awareness message Onboard information system Traffic hazard 

1 Introduction

Many kinds of advanced driver assistance systems (ADASs) have been developed and deployed to reduce the number of road traffic accidents. The autonomous (advanced) emergency braking (AEB) systems with forward collision warning (FCW) function may be the most well-known ADASs in Japan. These ADASs were introduced to the market in 2003, in the name of the collision mitigation brake (CMB) systems at that time, and have become popular since. Members of the Euro New Car Assessment Program reported a 38% overall reduction in real-world, rear-end crashes for vehicles fitted with low-speed AEB when compared with a sample of equivalent vehicles [1]. The Insurance Institute for Highway Safety reported that FCW alone and FCW with AEB are effective in reducing the number of rear-end crashes, and FCW with AEB is effective in reducing the number of rear-end crashes resulting in injury, according to the crash experiences of drivers who purchased the optional technologies [2].

The basic idea of CMB came from the work of the Advanced Safety Vehicle (ASV) project commencing in 1991 [3]. The Ministry of Land, Infrastructure, Transport and Tourism (MLIT) of Japan established technical guidelines for CMB (i.e., damage mitigation braking systems) in 2003 [4]. Additionally, in UN/ECE/WP29, regulation R. 131 of Advanced Emergency Braking Systems for heavy commercial vehicles was established in 2012 [5]. Many countries including Japan have implemented the regulation as mandatory.

The MLIT technical guidelines essentially state that AEB should be activated no earlier than a time to collision (TTC) of 0.6 s in the case of passenger cars so as to prevent the driver becoming overly reliant on the system. Here, the TTC is defined as the intervehicle distance between the subject vehicle and the preceding vehicle divided by the relative speed of the two vehicles. Regulation R. 131 states that the emergency automatic brake should be activated no earlier than a TTC of 3.0 s in the case of heavy commercial vehicles. Each AEB system involves an FCW. Regulation R. 131 also requires that the first-stage warning should be presented more than 1.4 s before AEB activation, and the second stage warning should be presented more than 0.8 s before AEB activation.
Table 1

Four stages of ADAS functions

Four stages of ADAS functions

Definitions

Expected driver’s actions

Rendering and response times [17]

Information presentation

Informs the driver about an external possible hazard in the future

To update driver’s situation awareness state

3.7 s

Awareness message

Informs the driver about an external potential hazard within a short time in the future [18]

To prepare to avoid a potential hazard within a short time in the future [18]

3.2 s

Warning

Requests action be taken immediately to avoid an external hazard [18]

To respond accordingly with a corrective maneuver in a short time [18]

0.8 s

Collision avoidance

Takes an evasive action to avoid the just approaching collision

N/A

N/A

In the ASV project and other projects, it is sometimes explained that ADAS functions involve four or so stages, such as the stages of “information presentation,” “awareness message,” “warning” and “accident avoidance control” according to the situational urgency. Among them, consensus is building for the activation timings of warning and accident avoidance control, methods and effectiveness evaluation based on many studies on human factors. Basic requirements are written in CMB guidelines and the Advanced Emergency Braking Systems regulation mentioned before. There are also advanced research works on how to activate a warning or accident avoidance control earlier to realize safer conditions using additional information [6, 7, 8, 9, 10, 11].

In contrast, there are few studies on the timing or effectiveness evaluation of information presentation and the awareness message [12, 13], although they are widely implemented in various onboard information systems and ADASs. It might be difficult to find common rules that make a preferable information presentation because there are many situational factors involved [14, 15, 16]. Meanwhile, there are possibilities of establishing rules that provide a better awareness message. The present paper thus focuses on how to provide an effective awareness message by conducting actual running tests.

2 Consideration of an Effective Awareness Message

Table 1 summarizes the four stages of ADAS functions with their definitions, expected driver’s actions, and rendering and response times. In the table, rendering and response times are taken from ASV guidelines [17], and the definitions and expected actions of the awareness message and warning are referred to from the International Organization for Standardization (ISO) [18]. Figure 1 illustrates the ISO timeline, which includes hazard detection, hazard notification, driver reaction and vehicle state change, together with the rendering and response time of the ASV. Although the ISO defines two intervals before and after “hazard notification” at the end of “information rendering,” ISO also states that before rendering is finished, driver reaction may overlap with it.

Japanese road users frequently come across road signs 200, 100 and 50 m ahead of roadworks. In-car information systems can be designed to provide a similar gradual sequence of information. As a result of the European Union project SAFESPOT, Diederichs et al. [19] reported the effectiveness of presenting a series of comfort, safety and critical warnings to a driver using an audiovisual human machine interface. Here, comfort, safety and critical warnings could correspond to the three stages of the information presentation, awareness message and warning.
Fig. 1

Rendering and response time, and hazard notification, which definitions are from [17, 18]

In human factor guidelines published by the Ministry of Infrastructure and the Environment and the Dutch program Connecting Mobility [16], timely presentation requirements are described as follows.
  1. 1.

    Information should be presented in a timely manner, not too late or too early.

     
  2. 2.

    Information should be presented preferably about 36 s before the point of action or 200 m before the first road sign.

     
  3. 3.

    Information should be presented minimally 9 s before the point of action.

     
We suppose that the information presented 36 s before the point of action is similar to information presentation while that presented 9 s before is similar to the awareness message.

Hirose et al. [20] reported better activation timings for the “information,” “caution (awareness message)” and “warning” of an ADAS with vehicle-to-vehicle communication at an intersection obtained in driving simulator experiments. In the case of braking before an intersection, the preferable activation times are 5.1, 4.0 and 2.0 s, respectively. They are introduced by adding safety margins to the rendering and response times in Table 1.

Liu et al. [12] executed actual running tests with two cars. The preceding car stopped randomly on a blind curve, and the subject car followed at an initial distance of 500 m. The driver provided with obstacle information passed the stopped car at a safer speed. When no obstacle information was provided, 32.3% of drivers did not brake, 40.4% braked far from the preceding vehicle and 27.3% braked close to the preceding vehicle. These percentages improved to 39.7, 44.2 and 16.1% when obstacle information was provided.

3 Test Protocol

Studies on awareness messages are still limited in scope. To contribute to this research field, the present study conducted a series of actual running tests with an obstacle located on a blind curve. To obtain an appropriate driver evaluation of the timing of the awareness message, it is important to execute tests in a practical environment that is similar to real-world conditions and not to impair the driver’s feeling of speed or sense of distance. Meanwhile, dangerous situations cannot be reproduced using actual vehicles. The ADAS with virtual infrastructure for vehicle communication provides an awareness message for the obstacle on the curve. The objective of the tests is to evaluate the activation timing and effectiveness of “awareness messages.”

The course was a two-lane road at our university campus and consisted of a straight section with length of about 300 m and a blind curve with radius of about 60 m. Figure 2 shows the layout, which is an arrangement of the figure in Annex A of ISO 18682 [18]. Two road cones were set at Point A or B on the curve to symbolize roadworks (Fig. 3). These two points were randomly selected so that the driver hardly expected the obstacle location. The ISO uses the notation that \(T_\mathrm{ns}\) is the time until processing by external hazard detection and notification systems, \(T_\mathrm{rh}\) is the driver reaction time, and \(T_\mathrm{rv}\) is the time during which the state of vehicle changes. \(T_\mathrm{os}\) is the time from hazard notification until avoidance. Like Hirose et al. [20], we studied the suitable activation timing of an awareness message, which almost corresponds to \(T_\mathrm{os}\). As shown in Fig. 1, Point 2 is defined as the end of information rendering by the ISO, but we define Point 2 as the beginning of information rendering.
Fig. 2

Course layout and definitions of time intervals. Reproduced with permission from [18]

Fig. 3

Test vehicle avoiding road cones

Table 2

Test pattern order and subjective rating form for a driver (sample)

Test no.

Test pattern

Danger

Timing

Effective

Bother same

Expected

1

A-

4

   

1

2

A20

1

\(-\) 1

1

2

0

3

A15S

1

0

1

2

0

4

A5

3

1

\(-\) 2

2

2

5

A10

1

0

2

0

0

6

B10

1

\(-\) 1

1

2

0

7

A-

3

   

0

8

B15

1

\(-\) 2

1

0

0

9

A15

1

\(-\) 1

1

0

0

10

A15S

3

\(-\) 1

\(-\) 1

2

0

11

B20

3

\(-\) 2

0

2

0

12

A5

3

1

\(-\) 1

0

2

At the beginning of a series of tests involving a driver, the facilitator explained the test outline and the driver drove the course in the test vehicle for the sake of familiarity. In each test run, the facilitator took the passenger seat of the test vehicle and directed the driver to start driving and to then cruise at around 40 km/h, which is the regulation speed for two-lane roads in Japanese cities. Each driver was requested to drive normally as they do daily. The driver drove while feeling the potential risk of encountering oncoming cars and pedestrians, although other participants were controlled not to enter the scene.

The awareness message M5, “5 s ahead, roadworks!” was announced at Point 2, which is 55.6 m before Point A. Here, 55.6 m is the distance travelled at a speed of 40 km/h for 5 s. The distance between Points A and B was also 55.6 m. We prepared the messages M5, M10, M15 and M20, which, respectively, correspond to distances of 5, 10, 15 and 20 s. This time range was selected by referring to our previous test results and to previous research [17, 18, 19, 20]. Our previous test results suggest that distances corresponding to 10 or 15 s may be optimal. Although the interval is a little bit longer as 5 s, there may be still the high possibility that there will be no significant difference because of uncertainty in the subjective evaluation of the driver. Meanwhile, if the interval becomes shorter, the number of tests becomes enormous.

Each awareness message was given by a female voice and lasted about 3 s. Test patterns A5, A10, A15, A20, B10, B15 and B20 were created by combining the different messages and Points A and B. Pattern B5 was not used because a driver can see the Point B obstacle at the 5 s distance. Table 2 gives an example of the series of 12 test patterns randomly created for a driver. In the series, there are two test patterns denoted A- and two test patterns denoted A15S. Here, A- indicates that the obstacle is set at Point A and there is no awareness message. The order of presenting A- patterns is fixed (Nos. 1 and 7). A15S indicates the A15 pattern with a subtask for which results are not included in this paper. After each test run, the facilitator asks the driver to make subjective ratings (see the columns in Table 2 and “Appendix B”) and comments. After the series of tests, the facilitator provides the driver a questionnaire (“Appendix C”).

The test was permitted by the Akita Prefectural University Research Ethics Review Committee Regarding Human Subjects (No. 15-09). Nine male drivers and three female drivers provided written informed consent and drove the test vehicle as research participants. They were aged between 22 and 65 years (average age of 37.3 years), had possessed a driving license for between 1 and 45 years (average possession of 17.7 years), and had average annual mileage between 300 and 30,000 km (average annual mileage of 14,900 km) (see “Appendix A”). Although there were only 12 participants, they had wide ranging ages and driving experience, and it is thought that meaningful conclusions can be drawn from the results of the study.

4 Measuring Devices of the Test Vehicle

The test system included a 2.4-l minivan as the test vehicle (Figs. 3, 4). As illustrated in Fig. 4, the laptop computer in the vehicle acquires signals of the vehicle speed, accelerator pedal stroke, three-axis acceleration, three-axis angular velocity, relative positions of forward obstacles provided by a laser sensor, driver’s heart rate and video signal of a pedal camera. The sampling time is 20 ms except for the laser sensor, heart rate and pedal camera. The drive recorder acquires not only the video signal of the front and driver face cameras, but also the stop lamp signal (ON/OFF) and turn signals (left and right) at the video rate. A tablet displays information, but was not working in the tests.
Fig. 4

Measuring devices of the test vehicle

Fig. 5

Typical time history data for the A15 pattern (participant I)

5 Test Results

Typical time history data of a participant for the A15 pattern are shown in Fig. 5 after processing with an 11-point moving average. In the figure, the accelerator pedal stroke takes a value of 0–100%. The vehicle initially travelled at around 40 km/h. At 7 s, the M15 awareness message commenced. At 14 s, the driver released the accelerator pedal to decelerate the vehicle and take the curve. At 16 s, the vehicle entered the curve with steady lateral acceleration of 1.6 m/s\(^2\). The vehicle passed the road cones at a speed of 35 km/h at 22 s and then returned to its own lane while accelerating. In almost all cases, the vehicle was smoothly handled and avoided the obstacle without stopping, and hard braking was scarcely observed. Figure 6 illustrates an example of the vehicle avoidance trajectory on a Google map obtained using a Global Positioning System (GPS) device. Two road cone icons were appended to the picture to explain the avoidance trajectory.
Fig. 6

Example of the vehicle avoidance trajectory on a Google map recorded by a GPS device with two road cones illustrated

Fig. 7

Comparison of peak deceleration without expectation, with expectation, and for different timings of the awareness message

While the drivers did not expect the obstacle to appear in test No. 1, they were conscious of the possibility of an obstacle appearing in subsequent tests even if they did not receive explicit information. The difference may be seen in the driver’s avoidance control. Figure 7 compares the peak decelerations before the obstacle among A- without expectation (No. 1), A- with expectation (No. 7), A5, A10, A15 and A20 for the 12 participants. The mean values for Nos. 1 and 2 are 1.87 and 1.19 m/s\(^{2}\), respectively, and this pair has significant difference (*\(p= 0.03 < 0.05\)). No significant difference is seen among the mean values of No. 7, A5, A10, A15 and A20. The series B5, B10, B15 and B20 has a similar tendency, but with lower decelerations because Point B is located at the back of the curve and the vehicle speed decreases along the curve. Regardless of the point, the drivers avoided the obstacle with sufficient safety margin thanks to the expectation or awareness message. It is thus difficult to find differences among the objective responses. However, there may be possibilities of differences in handling operations or accelerator pedal release timings, and further analyses are required.

Hereafter, instead of conducting objective response analysis, the subjective ratings given after each run are analyzed; items are summarized in “Appendix B”.

Figures 8, 9, 10 and 11 show the subjective evaluations made by the 12 drivers using five-level scales by test pattern. These evaluations were made immediately after each run. The question for Fig. 8 is “Q11: Did you sense danger when you encountered the obstacle?” Although no remarkable feature is seen in series B, A5 is evaluated to be slightly more dangerous than A10, A15 and A20. Figure 9 shows the evaluation result of awareness message timings. In series A and B, the relations between timing and evaluation score are common and a timing of 10 s is considered best. There is a significant difference (**\(p = 0.006 < 0.01\)) between A10 and A15.
Fig. 8

Subjective evaluation for “Q11: Did you sense danger when you encountered the obstacle?”

Fig. 9

Subjective evaluation for “Q12: How about the awareness message timing?”

The result shown in Fig. 9 is supported by the comparison of message effectiveness in Fig. 10. Again, 10 s is the best option although the result is not so clear. It is considered that the characteristics of A5 and B5 are similar to those of “warning” and help a driver avoid the obstacle, but the driver passes the obstacle not so smoothly and with a smaller safety margin than in the case of other earlier awareness messages. In contrast, because A20 and B20 are too early for avoidance, drivers require another awareness message having a more appropriate timing. Such driver requests are presented in the next section.
Fig. 10

Subjective evaluation for “Q13: Was the awareness message effective?”

The question for Fig. 11 is “Q14: Was the awareness message bothersome for you?” The results are the same as for the results of the reverse expression of effectiveness in Fig. 10.
Fig. 11

Subjective evaluation for “Q14: Was the awareness message bothersome for you?”

6 Questionnaire Results

After the series of running tests, the facilitator provided each driver the questionnaire shown in “Appendix C”. There were not many negative evaluations of the test management, but five drivers checked “Q234: Because there were many runs, it was difficult to make an evaluation for each one.” The understanding of the awareness message was good. Nine drivers checked “Q241: I correctly understood the purpose of awareness messages and received the tests.” while only one driver checked “Q242: I am not sure that I received the tests dedicated for awareness messages.” and “Q243: I could not understand the meaning of awareness messages.” The above results reveal that the tests were conducted appropriately.

Figures 12 and 13 show the main results of the effectiveness and activation timing of the awareness message. All drivers answered affirmatively in regard to effectiveness, and 11 drivers checked the box that the message was “very effective” or “extremely effective.” Eight drivers considered 10 s to be the best activation timing. The one driver who selected 20 s was the driver who checked Q242 and Q243.
Fig. 12

Questionnaire result for “Q21: Was the awareness message function effective?”

Fig. 13

Questionnaire result for “Q22: Which awareness message timing was the best?”

Q251–255 are related to the expectation of the awareness message function. Nine drivers checked “Q251: An awareness message is more important than a warning or collision avoidance.” In contrast, no one checked “Q252: There is no need for an awareness message function if my car is equipped with a warning function or collision avoidance function.” Eight drivers checked “Q253: It may be helpful if my car is equipped with an awareness message function.” Four drivers checked “Q254: It would be better if there is also a display.” Six drivers checked “Q255: It would be better if there are awareness messages on multiple occasions.” Here plural presentations mean for example that the first awareness message is presented at 20 s and the second message is presented at 10 s.
Table 3

Attributes of research participants

ID

Gender

Age

Duration of driving license (year)

Annual mileage (km)

Cumulative mileage (km)

A

Male

65

45

9000

400,000

B

Male

22

1

300

300

C

Male

23

5

25,000

150,000

D

Male

23

3

20,000

30,000

F

Male

22

4

5000

16,000

G

Male

63

41

6000

55,000

H

Male

41

23

25,000

400,000

I

Female

37

17

6000

60,000

J

Male

63

45

30,000

350,000

K

Female

24

3

2000

6000

L

Female

23

3

20,000

100,000

M

Male

42

23

30,000

500,000

 

Average

37.3

17.7

14,858

172,275

 

Min

22

1

300

300

 

Max

65

45

30,000

500,000

7 Conclusions

A series of actual running tests was conducted to evaluate the activation timing and effectiveness of an awareness message. Test vehicle drivers drove a course with an obstacle located on a blind curve at an initial speed of 40 km/h. Awareness messages were provided with various timings before the obstacle was encountered. The obstacle was moved or sometimes removed so that the driver did not know in advance the position of the obstacle.

There was no significant difference in the peak deceleration among various timings of messages. It is considered that the drivers could avoid the obstacle with a sufficient safety margin thanks to any of the awareness messages.

According to the subjective rating, a timing of 10 s before encountering the obstacle is optimal in the present situation. It is considered that the 5-s awareness message is more similar to a warning having a smaller safety margin than other earlier awareness messages. The 20-s awareness messages were too early for avoidance, such that messages should be repeated at 10 s.

We only used 12 drivers in the present work, and therefore, only a minimal number of results had significant differences. In future work, by increasing the number of drivers, we would like to get more meaningful results and improve reliability.

The present study only objectively analyzed peak decelerations. Further analyses are required to address possibilities of differences in handling avoidance, the accelerator pedal release timing and the heart rate.

Notes

Acknowledgements

Part of this work was supported by JSPS KAKENHI Grant Number 26350455. Advice and comments given by Mr. Kenji Kimura helped in the construction of the onboard system. We express our gratitude to faculty members and laboratory students for participating in the tests.

References

  1. 1.
    Fildes, B., Keall, M., Bos, N., et al.: Effectiveness of low speed autonomous emergency braking in real-world rear-end crashes. Accid. Anal. Prev. 81, 24–29 (2015)CrossRefGoogle Scholar
  2. 2.
    Cicchino, J.B.: Effectiveness of Forward Collision Warning Systems With and Without Autonomous Emergency Braking in Reducing Police-Reported Crash Rates. Insurance Institute for Highway Safety, Arlington (2016)Google Scholar
  3. 3.
    Brochure on “Achievements of Promotion in ASV phase2”. http://www.mlit.go.jp/jidosha/anzen/01asv/resourse/data/asv2development.pdf (2001)
  4. 4.
    Grover, C., Knight, I., Okoro, F., et al.: Automated emergency braking systems: technical requirements, costs and benefits, PPR 227, TRL. https://circabc.europa.eu/sd/a/3ab87fdc-5715-4733-af50-c3608034ca56/report_aebs_en.pdf (2008). Accessed 2 April 2017
  5. 5.
    UN Regulation No. 131, EU Regulation No. 347. Advanced emergency braking systems (2012)Google Scholar
  6. 6.
    Toyota News Release: Toyota enhances pre-crash safety system with eye monitor (2008)Google Scholar
  7. 7.
    Jamson, A.H., Lai, F.C.H., Carsten, O.M.J.: Potential benefits of an adaptive forward collision warning system. Transp. Res. Part C Emerg. Technol. 16(4), 471–484 (2008)CrossRefGoogle Scholar
  8. 8.
    Park, M., Lee, H., Lee, S., et al.: Development of communication based collision warning systems. In: AVEC ’12 (2012)Google Scholar
  9. 9.
    Marumo, Y., Tanaka, K., Suzuki, H.: Assistance system to predict driving behavior considering information on pre-preceding vehicle. In: FAST-Zero, 20134665 (2013)Google Scholar
  10. 10.
    Machida, T., Mimuro, T., Takanashi, H.: Effectiveness evaluation of adaptive forward collision warning using actual vehicles. In: AVEC’14 (2014)Google Scholar
  11. 11.
    Yan, X., Zhang, Y., Ma, L.: The influence of in-vehicle speech warning timing on drivers’ collision avoidance performance at signalized intersections. Transp. Res. Part C Emerg. Technol. 51, 231–242 (2015)CrossRefGoogle Scholar
  12. 12.
    Liu, H., Wei, H., Yao, Z., et al.: Effects of collision avoidance system on driving patterns in curve road conflicts. Soc. Behav. Sci. 96, 2945–2952 (2013)CrossRefGoogle Scholar
  13. 13.
    Research on the influence of information presentation for safety driving support on driving behaviour. Japan Safe Driving Center (in Japanese) (2009). https://www.jsdc.or.jp/library/research/tabid/123/Default.aspx. Accessed 2 April 2017
  14. 14.
    Oppenheim, I., Shinar, D.: A context-sensitive model of driving behaviour and its implications for in-vehicle safety systems. Cognit. Technol. Work 14, 261–281 (2012)CrossRefGoogle Scholar
  15. 15.
    Shinar, D.: Traffic Safety and Human Behaviour. Elsevier, Amsterdam (2007)CrossRefGoogle Scholar
  16. 16.
    Kroon, E.C.M., Martens, M.H., Brookhuis, K.A., et al.: Human factor guidelines for the design of safe in-car traffic information services. Ministerie van Infrastructuur en Milieu (2014)Google Scholar
  17. 17.
    Ministry of Land, Infrastructure, Transport and Tourism of Japan: The guideline of driving assist system by communication. (in Japanese) (2013). http://www.soumu.go.jp/main_content/000288894.pdf. Accessed 2 April 2017
  18. 18.
    Intelligent transport systems—external hazard detection and notification systems—basic requirements [M]. ISO 18682:2016Google Scholar
  19. 19.
    Diederichs, J.P.F., Marberger, C., Hinder, V.: Iterative design and assessment of an audio visual warning concept for Car2x communication systems. In: 17th ITS World Congress (2010)Google Scholar
  20. 20.
    Hirose, T., Ohtsuka, Y., Gokan, M.: Activation timing of a collision avoidance system with V2V communication. SAE Paper 2017-01-0039 (2017)Google Scholar

Copyright information

© Society of Automotive Engineers of China (SAE-China) 2018

Authors and Affiliations

  1. 1.Graduate School of Systems Science and TechnologyAkita Prefectural UniversityYurihonjoJapan
  2. 2.Akita Prefectural UniversityYurihonjoJapan

Personalised recommendations