Designing Interfaces for Automated Driving Systems

Essay & Workshop 2017

Figure 1: The Mercedes-Benz F 015 gives an idea of how the automotive future might look like (image retrieved from PCWorld, 2015).

Abstract

As driving automation advances, appropriate user interface design becomes increasingly important. It plays a key role in the usability and trustworthiness of driver assistance systems and automated driving systems. Thus, interface design determines whether systems get actually used by drivers or not. This essay gives an overview about which aspects of automation and interfaces need to be considered when designing automated driving systems. Therefore, state-of-the-art literature is reviewed to determine influencing factors evolving from interface components, automation problems & challenges and human psychology. As a result, design considerations are discussed in the context of automated driving featuring ergonomic and aesthetic aspects. Furthermore, general design principles were used to design and evaluate interface designs at an early stage. They can be considered as an appropriate method for the design and evaluation of automated driving systems within a human-centered design process.

1 Introduction

In the last decade public attention on automated driving increased significantly. This correlates with the growing number of available advanced driver assistance systems in modern cars and autonomous vehicles hitting the road, as well as with the rise of obtrusive advertising for these systems. Concept cars giving a glimpse of how the automotive future might look like (see Figure 1). OEMs compete for pole positions in the field of automated driving and invest huge amounts of money in research and development aiming to be ready for releasing driverless cars by 2020 (Muoio, 2016). Future mobility and infrastructure concepts are being discussed alongside ethical questions and legal consequences pointing out the large impact automated driving will have on our daily life. However, there’s also still a huge number of issues and open questions concerning the interaction between humans and automated systems that haven’t been solved yet. As a lot of them are related with human behavior and psychology the design of appropriate interfaces becomes increasingly important. It’s a necessity to have interfaces capable of integrating an effective and efficient relationship between the system and the user on a trustful basis, especially when system parts get automated that are important in high-risk situations like for example driving in a city. This essay gives an overview about what needs to be considered when designing automated driving systems, which include SAE levels 3, 4 and 5 (SAE International, 2016). In a first step, basic aspects of interfaces get clearly outlined and set in an automotive context while taking an exemplary look at technological innovations and resulting consequences for the design and automation process. Afterwards, problems and challenges of automation are discussed along with psychological aspects of human behavior, that need to be kept in mind when designing for automation. Therefore, a literature survey has been conducted. As a result of that, this essay takes a look on ergonomic and aesthetic aspects of interfaces that need to be addressed. It investigates a user-centered design process within the automotive context and discusses how design principles and heuristics can be used to evaluated automotive interfaces

2 Interfaces in Cars

An interface is defined as “A point where two systems, subjects, organizations, etc. meet and interact” (Oxford Dictionary, 2017). In the automotive context this means basically something which establishes the interaction between a human driver and a car system. Bubb, Bengler, Grünen, & Vollrath (2015) describe it as the point where the car and the driver exchange information. They split interfaces up into displays and control elements.

2.1 Displays

Displays are technical components, that transport information about a certain entity of the environment to the driver (Bubb et al., 2015). The information can be perceived via all human senses, but in the automotive context optical, acoustic and haptic displays become especially relevant as they feature appropriate ways to transport information from the car to the driver (Bubb et al., 2015). As technology advances we see more and more enhanced display techniques become available. For example, head-up displays (HUDs) feature less distractive ways to supply the user with the right information at the right time while keeping him focused on the driving task. However, with technology like this further challenges evolve. Israel (2012) describes HUDs as contact-analog displays merging pictorial situation-analog displays with the real world to an augmented reality. While he generally focused on scaling down the amount of physical space HUDs need in order to make it ready for standard use, he also states, that there are no universal design principles yet available. As a result of that, Israel (2012) recommends using iterative methods checking the design again and again for the right interpretation of visual signals and the usage of design principles. This can basically be applied to all sort of interface design and will be discussed more detailed within chapter 4.

2.2 Control elements

Control elements are technical equipment that help to convey information from a human to a machine using extremities (e.g. fingers or legs) or – as nowadays more and more popular – sensory based information processing (Bubb et al., 2015). Bubb et al. (2015) state furthermore, that the latter technique uses sensors like cameras or microphones to detect patterns, e.g. for speech or gestures recognition. Usage of these technologies increases rapidly with the availability of such systems, similar to the rise of gesturebased interaction on touchscreens. Availability and popularity of speech recognition systems like Siri or Alexa increases as well (Goode, 2016). A major reason for this is, that they offer people a more natural and, in some cases, an easier way of interacting with things, e.g. when both hands are already needed for another task. In the context of automotive, designers should always keep in mind what the main task in the respective situation is and decide based on that knowledge which control elements would be appropriate to use and which wouldn’t. While the driver might for example find it more convenient to use gesture-based interactions on touchscreens, instead of navigating through lists with a push and rotary switch, it might also be more distracting. This might not be a problem while driving autonomously, but it increases reaction time, what could have impacts on take-over time as well.

3 Designing for Automation

3.1 Problems and challenges

Automation at its core tries to replace the human operator with an automatic system, but the attempt of achieving that leads controversially to an increased importance of the human operator as a back-up role supervising, maintaining and adjusting the system, even within highly automated systems (Bainbridge, 1983). Bainbridge (1983) describes this as one of the “ironies of automation” and concludes “the more advanced a control system is, so the more crucial may be the contribution of the human operator” (p. 775). As an obvious result of that, one can say, that even highly automated systems are still human-machine systems (Bibby, Margulies, Rijnsdorp, Withers, & Makarow, 1975). Norman (1989) declares that the problems related with automation emerge out of inappropriate system design and thus blames in particular bad feedback and interaction design. After his opinion, it’s necessary that the automation process “should either be made less intelligent or more so“ (p. 1), but not stay at an inappropriate, intermediate level. Thus, problems and challenges of automation are acquainted since the early 1970s, they are still relevant today. The exact same problem as previously mentioned is nowadays discussed when talking about automated driving with a regard to the SAE levels of automation (SAE International, 2016). Some researchers and OEMs, e.g. Ford, therefore suggest to skip SAE level 3 in order to overcome associated problems (Davies, 2017). In level 3 the human driver would need to overtake control in certain situations the system isn’t able to handle. This results in the requirement that the driver has to continuously monitor the system in order to be capable of reacting in time. Molloy & Parasuraman (1996) carried out several exemplary user tests determining inefficiency of automation monitoring in contrast to manual control tasks. Thus, tasks where humans are only responsible for monitoring the automated system, become more susceptible to errors. To avoid that, appropriate design is needed (Norman, 1989). Adapting Norman’s (1989) conclusions to the context of automated driving this means, that the car system should always accept that errors either raised by the system itself, the driver or the environment can occur at any time. Norman (1989) therefore suggests that the driver should be kept in the loop with the ability to interact effectively with the system, while appropriate feedback, e.g. about the automation status, should be persistent available and the system “should allow for the worst of situations” (p. 1).

3.2 Human behavior

Concerning human usage of technology Parasuraman and Riley (1997) state that automation can be used, misused, disused or abused. The following section gives an overview what these aspects mean in the context of automated driving. As a good comprehension of this can on the one hand lead to a better understanding of problems that might occur, when developing or using automated systems, and therefore on the other hand improve one’s ability to design and evaluate such systems. Parasuraman and Riley (1997) describe use as „the voluntary activation of automation by the user“ (p. 230), which can be influenced by a person’s trust in the automated system, the experienced workload and the risk in the respective situation. In a modern car an example for a good usable interface could be the activation of cruise control via an easy available lever behind the steering wheel featuring low risk of failure and a low amount of cognitive resources needed. Misuse happens when the user over-relies on automation resulting in neglectful monitoring duties (Parasuraman & Riley, 1997). This can lead to serious incidents, like last year’s fatal accident of a Tesla Model S crashing into a crossing tractor trailer while being operated in Autopilot mode (National Highway Traffic Safety Administration, 2017). In contrast to that disuse appears when users don’t trust the system or don’t trust it anymore, e.g. because of false alarms, and therefore don’t use it or start to refuse using it, e.g. by ignoring system alarms (Parasuraman & Riley, 1997). Consequently, trust is a critical factor of automated systems. Lee & See (2004) declare that an appropriate level of trust is needed and emphasize the increasing importance of trust in consumer products, like automotive automation, regarding on the one hand user acceptance and on other hand safety and performance implications. Abuse is probably the most abstract aspect of automation. Parasuraman and Riley (1997) describe it as “the automation of functions […] without due regard for the consequences for human performance“ (p. 230). This can lead to inadequate system feedback confusing the operator, who’s therefore no longer able to handle the situation (Parasuraman & Riley, 1997) or, as mentioned above, to monitor processes adequately. Parasuraman and Riley (1997) state that increasing automation consequently might reduce the risk emanating from the operator’s behavior, but might also make it eventually more vulnerable to design errors. This means in turn, that humans are needed as a control authority supervising the system and scanning it for eventual occurring errors. This phenomena correlates with what Norman (1989) describes as “The Problem of Automation” and Bainbridge (1983) as the “Ironies of Automation”. Focussing back on automtive, Bubb et al. (2015) are declaring that both the driver and the car have a particular idea about each other. They state, that the driver needs to have a certain (anticipated) knowledge of how to use the car, whereas the car does have a concept of the driver’s characteristics deposited which is determined by the designer and thus vulnerable to errors, as the driver probably doesn’t use the system as originally intended by the designer. To prevent abuse Parasuraman and Riley (1997) are proposing similar steps as Norman (1989) does when he talks about appropriate design considerations. They also conclude that it is necessary to keep the operator in the loop, what would also feature safety benefits. If the system doesn’t cooperate with the operator and incapacitates him instead, abuse might also evoke Misuse and Disuse and hence lead to further problems.

4 Design Considerations

4.1 Ergonomic and aesthetic aspects

Besides requirements related to the nature of automation, there are also general ergonomic and aesthetic aspects that need to be considered when designing interfaces for (automated) cars. Schmid & Maier (2012) split the cockpit of a car in three areas distinguished by the way the driver can percept information and the way he’s able to physically interact with his environment, which is determined by his anthropometrical measurements. They state that the driver can use his central field of view and two-hand interactions in the primary area, whereas in the secondary area he uses the peripheral field of view and single-hand interactions. Elements within the tertiary area are outside of the driver’s field of view and are only accessible within an extended one-hand area (Schmid & Maier, 2012). Such a division can be used to decide about an element’s physical position based on its importance and its interaction modalities. This obviously leads to the conclusion, that the most important interface components should be situated in the primary area. Nowadays there are lots of different manufacturerspecific operating logics available, consequently leading to inconsistent interfaces (Schmid & Maier, 2012). Because of that, Schmid & Maier (2012) proposed that all manufacturers should use the same basic ergonomic concept with limited, but still enough room for aesthetic adjustments to recognize the respective brand. They furthermore claim, that such an overall concept would not only have practical benefits, but would also reduce operating errors and thus be beneficial towards driving safety. While Schmid & Maier (2012) give some good points, it might not be that realistic and also not necessary that all manufacturers commit on one basic concept. Though, interface design shouldn’t be dependent on brand-specific guidelines, that restrict ergonomic quality. Instead, designers should be able to ensure that drivers can use the system in an effective, efficient and satisfying way, what is also described by the ISO definition of usability (DIN, 2011). Driver assistance systems should feature transparent system behavior, offer control logics that are easy to use and to learn, meet the user expectations and clearly communicate system boundaries (Winner, Hakuli, & Wolf, 2012). This applies not only to systems used in manual driving, but to all levels of automation, where the user is still able to intervene or take over control of the system.

4.2 Human-centered design process

To achieve a high usability, it is essential to have a deep understanding of the context and the user with its individual properties and characteristics and to use this knowledge to design and develop the system (DIN, 2011). The ISO 9241-210 consequently proposes an iterative usercentered design (UCD) process (DIN, 2011). In this process, the solution gets designed and evaluated on the basis of user requirements which get determined by the context of use. A user-centered design process does hence also match the recommendation of Israel (2012) regarding the design of HUDs. Furthermore, UCD supports Lee & See’s (2004) implications to accomplish a trustable automation by integrating context and capabilities of the system and individual user characteristics. A major aspect of UCD is the evaluation of created solutions based on previously determined user requirements. Evaluations lead to further iterations until the solution meets the defined requirements.

4.3 Evaluation methods & design principles

Concerning the evaluation of design solutions, literature describes several approaches, e.g. usability inspection methods (Nielsen, 1994) or driving simulator studies (Bubb et al., 2015). One pretty common and straightforward approach is to compare the design with a set of design principles. Nielsen (1994) describes this as heuristic evaluation since he proposed to use ten heuristic principles for the evaluation, which he describes as “broad rules of thumb” (Nielsen, 1995). Heuristic evaluation is a usability inspection method, that can be used to rate everything from early concepts to already implemented systems. The method itself isn’t limited to Nielsen’s heuristics. They can easily be replaced with another set of principles, e.g. the dialogue principles specified in the ISO 9241-110 (DIN, 2008) or the more fundamental principles by Norman (2013). Norman’s (2013) set consist of 7 principles (affordances, signifiers, constraints, mappings, feedback and conceptual model) that describe general aspects and requirements of a product. Norman (2013) emphasizes that conceptual model might be the most important principle as it “provides true understanding” (p. 10). Regarding the usage of design principles for the evaluation of an interface, the designer should figure out in advance on what aspects of the design he wants to focus and thus decide for an appropriate set.

4.4 Experimental application of design principles

As part of a workshop we conducted in the seminar “Human Factors of Automated and Cooperative Driving” at the Technical University of Munich we used Norman’s (2013) principles to create and evaluate interface concepts for highly automated driving systems. The objective here was to motivate participants not to limit their ideas and concepts just to displays. Thus, they were guided to take a more holistic approach and to think for instance of element transitions, e.g. a retractable steering wheel. Therefore, we decided to use Norman’s (2013) fundamental design principles. The set worked quite well while serving on the one hand as a starting point and guidance during the actual design process and on the other hand as an evaluation basis. Participants worked in groups of four to five people, visualized their ideas using a given car interior template (see Figure 2) and evaluated the solutions of other groups with a heuristic evaluation using Norman’s (2013) principles to describe both advantages and issues that might occur.

Figure 2: Car interior design template used in the seminar workshop.

5 Conclusions

Designing interfaces for automated driving systems is complex. Reviewed literature showed that not only aesthetic and ergonomic aspects need to be considered, but also knowledge of human psychology & behavior along with an understanding of phenomena that occur due to the automation of processes is required. Thus, it is critical to understand users, context and system limitations to design appropriate human-automation interactions. An iterative human-centered design approach features an adequate process to include these aspects and therefore an appropriate basis for developing suitable systems. Since, there are no general design principles and requirements specified yet for designing interfaces of automated driving systems, it can be recommended, to use established general methods. Existing design principles offer an adequate approach for both designing and evaluating such interfaces.

Acknowledgments

This essay was created as part of the seminar “Human Factors of Automated and Cooperative Driving” at the Technical University of Munich. Seminar sessions have been split up in groups of two to three students. Thanks to my team mates Ann-Kathrin Stadler and Marvin Loch for the excellent collaboration, as well as to our tutor Sebastiaan Petermeijer for his great support.

References

Bainbridge, L. (1983). Ironies of automation. Automatica, 19(6), 775–779. http://doi.org/10.1016/0005-1098(83)90046-8
Bibby, K. S., Margulies, F., Rijnsdorp, J. E., Withers, R. M. J., & Makarow, I. M. (1975). Man’s role in control systems. 6th IFAC Congress Boston.
Bubb, H., Bengler, K., Grünen, R. E., & Vollrath, M. (Eds.). (2015). Automobilergonomie. Automobilergonomie. Wiesbaden: Springer Vieweg. http://doi.org/10.1007/978-3-642-13505-7
Davies, A. (2017). The very human problem blocking the path to selfdriving cars. Retrieved 22 March 2017, from https://www.wired.com/2017/01/human-problem-blocking-pathself- driving-cars/
DIN. (2008). DIN EN ISO 9241-110:2008-09 Ergonomics Of Human- System Interaction – Part 110: Dialogue Principles (ISO 9241- 110:2006) English Version Of DIN EN ISO 9241-110:2008-09.
DIN. (2011). DIN EN ISO 9241-210:2011-01 Ergonomie der Mensch- System-Interaktion – Teil 210: Prozess zur Gestaltung gebrauchstauglicher interaktiver Systeme (ISO 9241-210:2010); Deutsche Fassung EN ISO 9241-210:2010.
Goode, L. (2016). Hey Siri, who’s better: you or Alexa? Retrieved 20 March 2017, from http://www.theverge.com/2016/4/28/11463336/apple-siri-vsamazon- alexa-which-is-better-versus Israel, B. (2012). Potenziale eines kontaktanalogen Head-up Displays für den Serieneinsatz. Dissertation an Der Technischen Universität München.
Lee, J. D., & See, K. a. (2004). Trust in automation: designing for appropriate reliance. Human Factors, 46(1), 50–80. http://doi.org/10.1518/hfes.46.1.50.30392
Molloy, R., & Parasuraman, R. (1996). Monitoring an Automated System for a Single Failure: Vigilance and Task Complexity Effects. Human Factors: The Journal of the Human Factors and Ergonomics Society, 38(2), 311–322. http://doi.org/10.1177/001872089606380211
Muoio, D. (2016). These 19 companies are racing to put driverless cars on the road by 2020. Retrieved 24 March 2017, from http://www.businessinsider.de/companies-making-driverlesscars- by-2020-2016-8?op=1
National Highway Traffic Safety Administration. (2017). ODI Resume on Investigation: PE 16-007.
Nielsen, J. (1994). Heuristic Evaluation. In J. Nielsen & R. L. Mack (Eds.), Usability Inspection Methods (pp. 25–62). New York: John Wiley & Sons. http://doi.org/10.1089/tmj.2010.0114
Nielsen, J. (1994). Usability inspection methods. Conference Companion on Human Factors in Computing Systems - CHI ’94, 25(1), 413–414. http://doi.org/10.1145/259963.260531
Nielsen, J. (1995). 10 Usability Heuristics for User Interface Design. Retrieved 10 March 2017, from https://www.nngroup.com/articles/ten-usability-heuristics/
Norman, D. A. (1989). The ‘Problem’ of Automation: Inappropriate Feedback and Interaction, Not ‘Overatuomation’. San Diego.
Norman, D. A. (2013). The Design of Everyday Things: Revised and expanded edition. New York: Basic Books. http://doi.org/10.1002/hfm.20127
Oxford Dictionary. (2017). Definition of interface in English. Retrieved 20 March 2017, from https://en.oxforddictionaries.com/definition/interface
Parasuraman, R., & Riley, V. (1997). Humans and Automation: Use, Misuse, Disuse, Abuse. Human Factors: The Journal of the Human Factors and Ergonomics Society, 39(2), 230–253. http://doi.org/10.1518/001872097778543886
PCWorld. (2015). Mercedes-Benz F 015. Retrieved 29 March 2017, from http://www.pcworld.com/article/2865478/mercedes-benzs-f015- concept-is-a-self-driving-hydrogen-powered-living-room.html
SAE International. (2016). J3016 Surface Vehicle Recommended Practice, Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles.
Schmid, M., & Maier, T. (2012). Ergonomie versus Ästhetik? Die funktionale und formale Integration im Fahrzeugcockpit. Zeitschrift Für Arbeitswissenschaft, 66(2–3), 129–139.
Winner, H., Hakuli, S., & Wolf, G. (Eds.). (2012). Handbuch Fahrerassistenzsysteme. Handbuch Fahrerassistenzsysteme (2nd ed.). Wiesbaden: Springer Fachmedien. http://doi.org/10.1007/978- 3-8348-8619-4

My tasks

Research
Presentation
Workshop
Essay