Keywords

1 Introduction

1.1 Adoption of ADAS

Road safety is believed to be mainly impaired by ‘human factor’ [1]. 90% of traffic accidents are caused by human failure, including fatigue, inattention, alcohol and etc.

Therefore, researchers, professionals and governments try to countermeasure ‘human factor’ with driving assistant systems. Effective ADAS can reduce 70% crash rate. In real world, according to the 2005–2008 U.S. GES Crash Records, ADAS technologies successfully reduced light vehicles’ crashes by 32.99% and heavy trucks’ crashes by 40.88% [2].

In 2003, European Union opened ‘eSafety’ program. In 2005, America started assessment of ‘Integrated Vehicle-Based Safety System’. In 1991, Japan started ‘Advanced Safety Vehicle Development Program’. These programs have facilitated ADAS adoption in major vehicle markets [3]. From policy perspective, China acts slow in this area. Till ‘Made in Chine 2025’ (2015), China officially encourages ADAS technologies’ development and adoption [4].

1.2 Understanding of ADAS Scope

Troppmann’s Model

Troppmann (2006) depicts ADAS technologies in Active-Safety dimensions [5]. In Active dimension, technologies are allocated to demonstrate the degree of machine control: they are just text reminders or can take over the vehicle control from its driver. In Safety dimension, technologies are arranged to show their Safety impact: they are crash-avoiding or making driving easier (Fig. 1).

Fig. 1.
figure 1

Troppmann’s Model of ADAS function categories

According to this framework, ADAS technologies can be categorized as:

Collision Mitigation (Active-Safety)

  • Collision Avoidance

  • Automatic Emergency Braking

  • Hazard Braking

  • Braking Preparation

Vehicle Control (Active-no-Safety).

  • Longitudinal Control

  • Lateral Control

  • Adaptive Cruise Control Stop &Go

  • Adaptive Cruise Control Stop &Roll

  • Adaptive Cruise Control

Passive Safety (Passive-Safety)

  • Traffic Member Recognition

  • Collision Warning

  • Pre-Crash

Driver Support (Passive-no-Safety)

  • Parking Assistance

  • Parking Aid

  • Blind Spot Detection

  • Night Vision

  • Lane Departure warning

NHTSA Model

America’s National Highway Traffic Safety Administration (NHTSA, 2020) describes ADAS as Driver Assistance Technologies on its website to public [6]. It has 4 categories: Forward Collision Prevention, Backing Up &Parking, Lane &Side Assist, Maintaining Safe Distance.

Forward Collision Prevention

  • Forward Collision Warning

  • Automatic Emergency Braking

  • Pedestrian Automatic Emergency Braking

  • Adaptive Lighting

Backing up &Parking

  • Rear Automatic Braking

  • Rear Video System or Backup Camera

  • Rear Cross Traffic Alert

Lane & Side Assist

  • Lane Departure Warning

  • Lane Keeping Assist

  • Blind Spot Detection

  • Lane Centering Assist

Maintaining Safe Distance

  • Traffic Jam Assist

  • Highway Pilot

  • Adaptive Cruise Control

ADAS is a collection of systems and subsystems that transform manual driving to autonomous driving. Though it has a relative clear main purpose, its sub-categories can be slightly various in different researches.

1.3 Challenge of ADAS Category Design

All category models above could be references for in-car HMI designers to structure ADAS manual/setting system. And we can learn the categorizing strategy from both models. In Troppmann’s Model, drivers need understand the vehicle’s active control level and safety impact. In NHTSA’s category, drivers should tell the technological configuration difference of ADAS functions.

However, all models above cannot directly represent Chinese drivers’ understanding of ADAS Category. Although China’s vehicle ownership grows rapidly, it is far behind major developed countries, like United States. Chinese drivers probably have less knowledge about ADAS. Also, China has a large population. People’s concern about ADAS may vary from other countries.

Based on these references and manufacturer information, we design a card sorting study to examine Chinese customers’ mental model of ADAS categories. In this study, we will compare drivers’ mental model with these 2 existed models to reveal ordinary drivers’ grouping strategy.

2 Method

2.1 Material

Based on collaborating manufacturer information, we prepare 18 ADAS terminology paper cards. Every card is about 5 cm × 4 cm with the terminology name and a brief explanation on it (Table 1).

Table 1. ADAS terminologies and brief explanations used in the study

2.2 Participants

14 private car drivers participate this study. Their driving experience range from 5 months to 8 years. All of them have ADAS function using experience (Table 2).

Table 2. Participants information
Fig. 2.
figure 2

Picture examples of the study conducting process

2.3 Procedures

It is an individual open card-sorting research design, each participant sorts cards individually. There is no number limit on group quantities and participants could name their sorted groups by their own language. In this study, each participant trial takes about a half hour in conduct.

Step 1 - Understand 18 ADAS terminologies.

18 ADAS-related terminologies and their brief explanations were printed on separate paper cards. A participant has enough time to read these cards and ask questions until he/she is well prepared to do the card sorting.

Since all participants have ADAS feature using experience, with brief explanation along with the terminology, they can understand these 18 ADAS terminologies easily. No one ask for further terminological explanation, but some inquiry about the technological feasibility in real products. For example, some participants show interests on Traffic Sign Recognition technological feasibility.

Step 2 - Sort 18 ADAS terminologies into groups.

Participants are told to sort these cards into groups that could show the relationship of cards in their mental model. Participants are encouraged to do a rough categorization first, and then refine it on the second round.

Step 3 Name the groups.

Participants are told to name their card groups as what they would want to see in their own car. They were given blank cards to write down those names and put them on the top of related card stacks. They can organize these ADAS terminologies in their own language, not limited to existed models (Fig. 2).

3 Result

3.1 Category Distribution

To analyze card sorting result, 2 researchers merged similar group names into 7 groups based on participant interview: Driving Assist, Collision Prevention, Parking Assist, Safety Setting, General, Information, and Start-Stop Assist, showing in Table 3.

The percentage represents the ratio of participants who group a terminology under this category. The background color goes deeper as the number increases.

12 out of 18 cards can be clearly grouped, being put under the same group by more than (or equals) 50% participants. 6 terminologies that cannot be clearly grouped are Auto Hold, Blind Spot Detection, Traffic Sign Recognition, Pedestrian Detection, Automatic Post-Collision Braking, and Rear Cross Traffic Alert. If we set 40% as the group reference, Auto Hold can be grouped into Driving Assist, and Pedestrian Detection, Rear Cross Traffic Alert into Collision Prevention. 3 out of 7 group get cards clearly followed. They are Driving Assist, Collision Prevention, and Parking Assist.

3.2 Model Comparison

In Table 4, we compare mental model extracted in this study with Troppmann’s Model and NHTSA Model. We can find that participants tend to organize terminologies by driving status: driving, colliding, and parking.

Unlike the Active concern in Troppmann’s model, participants are not so aware of ADAS’s vehicle control extent. Also, participants would not differentiate assists from which sides of the vehicle. Therefore, some category scope differs from what in NHTSA Model. For example, Lane Keeping Assist is categorized into Lane &Side Assist in NHTSA Model, while it’s labeled as Driving Assist with Cruise Control in this study (Maintaining Safe Distance in NHTSA).

Table 3. ADAS category distribution in this study
Table 4. ADAS category model comparison

4 Discussion

4.1 Inspiration for in-Car HMI Design

This study reveals normal drivers’ mental model of ADAS functions that in-car HMI designer can exploit. Optimizing the manual/setting information architecture based on the study model, drivers can have a higher probability to find the target function items easily.

The results demonstrate that drivers are incline to group these functions according to driving status. They are not so aware of vehicle’s active control level and which direction side the function is tackling.

This phenomenon suggests liability between ADAS and drivers is ambiguous in driver’s mental model. In current situation, all liabilities are on drivers’ side. So, they do not argue for machine’s liability. However, once cars become more autonomous, technology’s liability may be need to clearly informed to drivers and considered by lawmakers.

This phenomenon also suggests drivers are not familiar with technology configuration. No matter collision detection in the front or in the back, they tend to group all these function as ‘Collision Prevention’. Once they get more technology familiarity, they may incline to group as NHTSA Model.

4.2 Limitation of Card Sorting

The card sorting method makes it more reasonable to retrieve information from the structure when users browse information, but it is difficult to solve personalized settings, such as shortening the information level of common settings. Designers should consider shortcuts, i.e. ‘my collection’, to empower user personalize their in-car HMI, reducing the search cost of users.

4.3 Limitation of Sample Representation

This study suggests Chinese drivers using driving status as a clue to group ADAS technologies. However, whether all ordinary drivers tend to use this strategy need to test in other country and culture as well. It is possible in other culture, with more technological knowledge, people will agree to Troppmann’s Model or NHTSA Model. And we can dig deep into the influence of clear vehicle active control level, safety impact and technological configuration.