Driver Attention in Automatic Transmission Cars

A Study on the Control Authority Transition Characteristics by Driver Information

Poster Papers, CSCI-ISSC (SmartCities and Smart Mobility) HyunSuk Kim*, Woojin Kim, Jungsook Kim, DaeSub Yoon

Cognition & Transportation ICT Research Section Electronics and Telecommunications Research Institute 161 Gajeong-dong Yuseong-gu Daejeon, Rep. of Korea {hyskim, wjinkim, jungsook96, eyetracker}@etri.re.kr

Abstract—We observe and analyze the characteristics of control authority transition according to driver's age and gender in a highly automated vehicle driving environment in this paper. We made control authority transition scenarios and performed experiments using the vehicle simulator. We experimented five tasks such as No NDRT, Conversation, Drink, Texting and Movie. We measured the time it took the driver to take control when the takeover request (TOR) occurred and analyzed the data using an independent sample t-test. According to age and gender, TOR control transition times showed significant differences in some driver's tasks. Therefore, automated driving system (ADS) should be able to request transition to manual operation considering the driver’s age and gender. Drivers must be trained in rules and regulations to prevent accidents, even when driving an automated vehicle. In the future, in order to reduce the response and reaction time, research should continue on methods for pre-cue, methods for providing SA information, and methods for increasing readiness.

Keywords—Automated Driving, Take-over Request, Non- Driving Related Task, Control Authority

I. INTRODUCTION Recently, many researchers have been conducted to

produce automated vehicles in accordance with the development of sensor technology [1-4]. SAE J3016 suggests six stages of automated driving from level 0 (no automation) to level 5 (full automation) [5]. Fully automated driving is not possible in the level 3 vehicle, so it is necessary to be able to transfer driving control authority between the ADS and the driver. Since driver intervention is always required in situations where automated driving is difficult in Level 3 vehicle, there have been studies that measure the performance of the driver's manual driving re-engagement [6-13].

The driver can perform various non-driving related tasks (NDRT) in the automated driving section. In level 3 automated vehicle, drivers must start manual driving when a take-over request (TOR) occurs, which is a request for transfer of control authority. As humans get old, their cognitive and physical abilities decrease [14-15]. In studies [16], male drivers reported higher accident rates than female drivers. In order to secure control authority transfer, the transfer performance according to the driver's age or gender should be studied. In this paper, we conduct an experiment to find out whether the time performance required to transfer to manual driving is different according to the driver's age and gender characteristics, and perform statistical analysis.

II. EXPERIMENT DESIGN

A. Experiment Environment and Data Acquisition We experimented with a Jaguar S-type based simulator

vehicle. The simulator has all driver controls fully operational

and is housed within a constant temperature and humidity chamber (4.5m X 5m X 2.5m). The front road scene is displayed through a 100-inch projector, and the three rear projectors display the surrounding road scenes in the rearview mirror and side mirrors [13].

Fig. 1. Simulator experiment environment.

We have established a data acquisition system (DAQ) in which the necessary information can be collected from driver, vehicle, and environment during the experiment. The driver information includes the driver's democratic information, behavior information, physiological information, and driver's TOR response information. Vehicle information consists of longitudinal/lateral vehicle control information during automated or manual driving. The environmental information is composed of inter-vehicle distance, lane and outside environment information. The driver writes the questionnaire before and after the experiment, and the operator inputs it into the DAQ. Other information is automatically collected from simulators, sensors, and video cameras and stored in the DAQ.

B. Experimental Procedure When the participant arrives at the laboratory, he or she

listens to the explanation of the purpose of the experiment and writes a personal democratic questionnaire (driving experience, age, sex, etc.) and practices simulator driving. Participants conducted five NDRT experiments: No-NDRT, Conversation, Drink, Texting, and Movie. Participants repeat one experiment three times and fill out the questionnaire. In order to prevent data contamination, NDRT execution order of participants was determined considering counter-valence.

III. DATA ANALYSIS AND RESULTS

A. Participants and Data Preprocessing We recruited 36 persons through external announcements

considering driving experience, gender difference, and age difference. We provided $ 27 to participants in compensation for the experiment. In our experiment, pre-cue was provided with "tti-tti" sound and TOR informed by voice saying, "Please drive manually." When the TOR occurred, the driver operated the paddle shift and proceeded to the manual driving.

1562

2019 International Conference on Computational Science and Computational Intelligence (CSCI)

978-1-7281-5584-5/19/$31.00 ©2019 IEEE DOI 10.1109/CSCI49370.2019.00297

Authorized licensed use limited to: Florida Institute of Technology. Downloaded on September 15,2021 at 19:04:05 UTC from IEEE Xplore. Restrictions apply.

As shown in figure 2, the reaction time for the TOR was measured by the difference between the paddle-shift operation time and the TOR time. Since the drivers performed all 5 NDRTs and performed the same NDRT three times each, we obtained 15 TOR reaction times for one driver and collected 540 data for 36 drivers. We calculated the mean of the TOR response times for each task and determined that the outlier was three times larger than the standard deviation. Then we removed outliers and used 518 data for analysis.

Fig. 2. Transition time calculattion after TOR.

B. Analysis and Results We classified that middle-aged drivers were 40~59 years

old and young-aged were 20~39 years old. We performed independent sample t-test to know the difference between the TOR transition time of middle-aged and young drivers according to NDRT (Table 1). In the five types of NDRT experiments, when comparing the response time for TOR, middle-aged drivers were slower than younger drivers. Independent t-test results showed significant differences (p <0.05). In particular, when TOR occurs in during conversations, drinking or movie work, middle-aged drivers respond on average 0.5 seconds slower than younger drivers.

TABLE I. INDEPENDENT T-TEST RESULTS BY AGE

a. Middle-aged, b.Young-aged We performed independent sample t-test to know the

difference between the TOR transition time of female and male drivers according to NDRT (Table 2). There was a significant difference in TOR response time between female drivers and male drivers in conversation and drink experiments (p <0.05). In the conversation experiment, the female driver responded 0.5 seconds faster than the male driver, and the drink experiment showed that the female driver responded 0.7 seconds faster than the male driver.

TABLE II. INDEPENDENT T-TEST RESULTS BY GENDER

a.Female, b.Male

IV. CONCLUSION We performed Tor performance experiments considering

the various NDRTs that the driver can perform freely during automated driving. Conversation and drink experiments

showed that the age or gender of the driver influences the time to switch to manual driving. In the movie experiment, only the driver's age affected the TOR response time.

Therefore, ADS should be able to request transition to manual operation considering the driver’s age and gender. Drivers must be trained in rules and regulations to prevent accidents, even when driving an automated vehicle. In order to further reduce the reaction time, research should continue on methods for pre-cue, methods for providing SA information, and methods for increasing readiness.

ACKNOWLEDGMENT This research was supported by a grant (18TLRP-

B131486-02) from Transportation and Logistics R&D Program funded by Ministry of Land, Infrastructure and Transport of Korean government. The authors acknowledged Youngdal Oh ([email protected]) of the Korea Institute of Automotive Technology (KATECH) for their assistance in the experiment.

REFERENCES [1] M. W. Park, S. W. Lee, and W. Y. Han, "Development of Steering

Control System for Autonomous Vehicle Using Geometry-Based Path Tracking Algorithm," ETRI Journal, Volume 37, Number 3, June 2015.

[2] S. Y. Noh, B. J. Park, K. H. An, Y. B. Koo, and W. Y. Han, "Co-Pilot Agent for Vehicle/Driver Cooperative and Autonomous Driving," ETRI Journal, Volume 37, Number 5, October 2015.

[3] D. Asljung, J. Nilsson, and J. Fredriksson, "Using Extreme Value Theory for Vehicle Level Safety Validation and Implications for Autonomous Vehicles," IEEE Transactions on Intelligent Vehicles, vol. 2, no. 4, 2017.

[4] H. Marzbani, H. Khayyam, C. N. TO , D. V. Quoc, and R. N. Jazar, "Autonomous Vehicles: Autodriver Algorithm and Vehicle Dynamics," IEEE Transactions on Vehicular Technology, vol. 68, no. 4, 2019.

[5] SAE J3016, “Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles,” September 2016.

[6] J. Radlmayr, C. Gold, L. Lorenz, M. Farid, and K. Bengler, “How traffic situations and non-driving related tasks affect the take-over quality in highly automated driving,” in the Human Factors and Ergonomics Society Annual Meeting, vol. 58, 2014.

[7] C. Gold, M. Körber, D. Lechner, and K. Bengler, “Taking over control from highly automated vehicles in complex traffic situations: the role of traffic density,” Human factors, vol. 58, no. 4, pp. 642–652, 2016.

[8] K. Zeeb, A. Buchner, and M. Schrauf, “Is take-over time all that matters? the impact of visual-cognitive load on driver take-over quality after conditionally automated driving,” Accident Analysis & Prevention, 2016.

[9] A. Eriksson and N. A. Stanton, “Takeover Time in Highly Automated Vehicles: Noncritical Transitions to and From Manual Control,” Human Factors, Vol 59, pp. 689-705, 2017.

[10] J. S. Kim, H. S. Kim, W. J. Kim and D. S. Yoon, "Take-over performance analysis depending on the drivers’ non-driving secondary tasks in automated vehicles," International Conference on ICT Convergence (ICTC) 2018.

[11] H. S. Kim, W. J. Kim, J. S. Kim, S. J. Lee and D. S Yoon, “An Analysis of Transition Characteristics of Driver Control Authority according to NDRT Type,” International Symposium on Future Active Safety Technology toward Zero Accidents (FAST-zero) 2019.

[12] H. S. Kim, W. J. Kim, J. S. Kim, S. J. Lee and D. S Yoon, “A study on the Effects of Providing Situation Awareness Information for the Control Authority Transition of Automated Vehicle,” ICTC2019.

[13] W. J. Kim, J. S. Kim, H. S. Kim, S. J. Lee and D. S Yoon, “A Study on the Driver’s Response Performance according to Modality of Planned TOR in Automated Driving,” ICTC2019.

[14] Daniel L. Murman, “The Impact of Age on Cognition”, Seminars in Hearing, Vol. 36, pp. 111–121, 2015.

[15] H. S. Kim, D. S. Yoon, H. S. Shin, C. H. Park, “Driving Characteristics Analysis of Young and Middle-aged Drivers”, ICTC 2016.

[16] A. Bener and D. Crundall, Role of gender and driver behaviour in road traffic crashes, International Journal of Crashworthiness, vol. 13, 2008.

NDRT No-NDRT Talk Drink Texting Movie

Age Ma Yb M Y M Y M Y M Y

N 54 52 49 53 52 54 54 52 48 50

Avg.(s) 2.73 2.67 2.98 2.49 3.35 2.85 3.89 3.49 4.17 3.54

t-test t=0.345 p=0.731 t=2.201 p=0.030

t=2.293 p=0.024

t=1.550 p=0.124

t=2.242 p=0.027

NDRT No-NDRT Talk Drink Texting Movie

Age F M F M F M F M F M

N 53 53 53 49 52 54 52 54 48 50

Avg.(s) 2.57 2.84 2.49 2.98 2.76 3.42 3.60 3.79 3.88 3.81

t-test t=-1.601 p=0.112 t=-2.193 p=0.031*

t=-3.078 p=0.003*

t=-0.742 p=0.460

t=0.235 p=0.815

1563

Authorized licensed use limited to: Florida Institute of Technology. Downloaded on September 15,2021 at 19:04:05 UTC from IEEE Xplore. Restrictions apply.