Warning Apps for Road Safety: A Technological and Economical Perspective for Autonomous Driving – The Warning Task in the Transition from Human Driver to Automated Driving

ABSTRACT

In the last decade, mobile warning apps and security services for drivers were developed, driven by the evolving market for smartphone-based applications. Now, in times of emerging autonomous vehicles, the value of warning detection in form of warning apps for road transport needs to be reappraised. In our study, we address the following questions: Does this kind of application have a technological and economical perspective for future mobility? How will the interaction between drivers, information systems (smartphone or onboard computer), and driving systems change? In this article, we bring together the aspects to consider in terms of technology development and business modeling amidst a changing human-computer interaction (HCI) toward autonomous driving. We provide a comprehensive picture of how technology-driven applications, such as warning detection apps, could develop their economic and technological impact within the transition of individual transport. By classifying the warning apps available in German-speaking regions it became apparent that apps exist for diverse areas of risk. Some of them are currently under development, others are already in use, but still with minor operational deficits. We found out that till today, the majority of warning services provides either an interim solution with a standalone mobile app, or a first test stage approach for prospective mobility solutions, only working on the basis of a car-to-car or car-to-pedestrian communication. Nevertheless, the effectiveness of warning apps was exemplarily proven by a survey about the Wildwarner, an app for wildlife-vehicle collision warnings. Looking ahead, warning-service-usage scenarios (1) without driving assistance, (2) with driving assistance and (3) fully autonomous driving are gone through, showing the still important but altered interactions between humans, warning information systems, and automated driving systems. From an economical perspective, already existing business models for warning services are identified and evaluated according to whether they could be transferred to business models for future mobility under changing ways of user interaction. The analysis of existing established warning apps indicates that more integrated approaches combining different warning types and delivering them as services either for autonomous vehicle navigation or human drivers ought to be supported for increasing road safety.

1. Introduction and research questions

Warning detection for the road traffic is not only provided by car-internal sensor systems but also by other origins, such as crowd-sourcing or official data in case of congestions or bad weather conditions, using real-time GPS traffic or satellite data. Thus, warning detection and services are often external data-driven services. Until now, several warning detections are provided to the users by mobile applications, and the number of such apps is increasing. In times of changing human-computer interaction (HCI) toward a higher level of driving automatization, the warning detection by internal-based as well as external-based sources should be given for a further use of current warning services. Therewith, the user interaction and therefore technical solutions have to be adjusted. Besides, also the business model of the service providers needs to be reassessed.

According to the European Commission (2017), 37% of autonomous car revenues are estimated to derive from safety product packages until 2022, which are suspected to be the largest item of autonomous car revenues. Besides automatic collision detection by internal-assisted driving features with sensors, warning and risk services calculated by service algorithms based on crowd- and real-time data may be the second pillar of this business. Already existing mobile applications including predictive and warning algorithms in the backend could provide the basis for the second pillar, provided that the car is connected online, and the country or region has a good internet coverage. In the times of autonomous driving, appropriate business models need to be defined for this new service type and should consider the interfaces between the three “actors” human driver, information system and automated driving system.

This article analyzes how already existing mobile apps for warning detection for German-speaking regions can contribute to the above-mentioned revenues and how the warnings can be better integrated. Analyses regarding the warning services’ aptitude in consideration to the transformation of the automotive sector toward autonomous driving are executed from the technological as well as the economical perspective. Thereby, we answer the following questions.

1. What are success factors for warning apps and how could they be transferred for warning detection in cases of autonomous driving?

2. Which kind of HCI is possible and desirable depending on the level of automation?

3. Which business models are applicable for warning apps on-road mobility at the moment, and which are suitable to enter the future market of autonomous driving?

The economic perspective will be discussed with regards to business models because the current market for warning applications is penetrated not only by established companies but also by several startups and young companies. These may need to adapt their revenue streams while changing from smartphone-based solutions to a stronger car-integrated service. This means that the discussion of user interaction and how users as humans or systems will receive information is strongly bound to the question of how the service can be maintained based on a business model and corresponding revenue streams.

The cooperation between driver and vehicle is one major open question, especially with regards to warning detection. In this article, we explicitly distinguish between smartphone-based systems, onboard information systems and the combination of both. We will refer to both as information systems, in contrast to the automated driving systems. The automated driving systems are a combination of several control units, either supporting the human driver (driver assistance) or the fully autonomous driving. Two current transitions might have an impact, how warnings will be handled in the future. (1) The trend for more automation up to fully autonomous vehicles will shift the communication between a warning service and the driver to a machine to machine communication, thus warning service to automated driving system. (2) The integration of services with onboard information systems, smartphones and automated driving systems provides potential new interfaces between systems as well as between driver and systems.

The article is structured as follows: After the method description, we classify mobile warning apps for road safety in order to derive characteristics of the warning detection for further analysis, to check their level of integration and the market spread. After proving the sense of purpose, the warning services’ technical transformation to automated driving and possible changes in HCIs are pointed out. Apart from defining fundamental notions, the effect of warning detection is proven by results of a survey conducted about the Wildwarner App, a warning app for wildlife-vehicle collisions (Wildwarner, 2019) supported by further scientific literature. The Wildwarner App was chosen for this empirical study because it represents a dynamic detection warning application as standalone solution, which is in the early stage of market entry. Effectiveness and usability are two essential components for the future existence of the app. Afterward, the economic transformation is examined by selecting suitable business models from the 55+ St. Gallen Business Models (Gassmann et al., 2017), as these models are a pattern-based set of existing business models which can be recombined and adapted to the new conditions. The models are selected according to their applicability to warning apps and for the case of warning service for autonomous driving. Afterward, the results of the research are discussed, and guidelines are proposed.

2. Related work

Over 90% of road accidents are partly caused by human errors or inattentions (European Commission, 2003). Mobile applications and services that warn the driver in cases of risk are deployed to avoid human-driven collisions (European Commission, 2017). While the majority of literature warns for the three categories of distraction (manual, visual and cognitive distraction, see e.g., Quaresma & Gonçalves, 2014), and while the level of smartphone usage is still “alarmingly high” (Vollrath et al., 2016, p. 29), the contribution of apps to support drivers and especially the design of non-distracting human–computer interaction is still a topic of scientific discussion. Mainly highly integrated in-car apps are recommended (e.g., Chen et al., 2017) using the transparent windshield to display information and voice commands and gestural interaction, while Albert et al. (2016) indicate that there are even desirable apps: “collision warning, texting prevention (both no-typing and no-reading), voice control (both text-to-speech and commands), and Green Box (In Vehicle Data Recorder – IVDR)” (p. 54). Warning signals against congestion, weather, or obstacles of any kind (pedestrians, animals, vehicles) are provided not only by onboard information systems but in times of smartphones also by a diverse and growing number of mobile apps. As a consequence, the variety of warning systems, warnings and use of warning elements leads to a heterogeneous HCI, disregarding HCI patterns, and all together to the concepts of the cognitive dimensions. Now, with the upcoming of autonomous driving, considerations are taking place about how accidents can be reduced through the automatization and autonomy of the vehicle (Zlocki et al., 2014; Watzening & Horn, 2016) and whether the warning services and the underlying algorithms could be used for risk predictions (Pagany & Dorner, 2019). Whilst currently the issued warnings are addressing human drivers, they may in future be useful not only for drivers but also for the partially or fully automated driving and assistance systems. This will also result in changing roles and a distribution of tasks and information between drivers and systems (Zlocki et al., 2014). The paradigm of an HCI between driver and warning app should change to a computer-computer-human interaction, where a warning service interacts with the onboard information system as well as automated driving system. Warning detection consequently will not serve to cause an action by the driver, but provide information for the automated driving system and the on-board information system and as a maximum to provide the user of the vehicle and information.

A positive effect of smartphone applications as warning systems on driving behavior has already been proven by several studies (Albert et al., 2016; Botzer et al., 2017; Cardamone et al., 2016). The automation of driving, ranging from assistance systems to envisaged fully autonomous vehicles will also result in changing roles and a distribution of tasks and information between drivers and systems. Biondi et al. (2019) have already provided an understanding of human-vehicle cooperation in automated driving, even detecting changes in responsibility and authority between passengers and also automated systems. They state that driving will be transferred to a more transportation-as-a-service system while demands on the driver’s attention will be reduced. The main control will be situated within the driving system itself. Nevertheless, a passengers’ situational awareness (trust in technology) and the possibility to always intervene will still be required (“human-computer symbiosis”).

3. Method

3.1. Warning detection with user interaction classification according to driving automatization levels

To explore the concepts of the recently evolved warning services for drivers and to analyze their current market situation, we used a combination of quantitative and qualitative methods. First, we searched for smartphone warning applications covering road transport in the Google Play Store. Secondly, we searched for scientific projects in Scopus using the term “mobile AND app AND warn* AND road OR traffic” in the title, abstract, and keywords. The aim of the following classification was not only to analyze the current market situation of warning solutions but also to estimate future market potentials when driving assistance and autonomous driving will be pushed forward. Furthermore, we wanted to understand the underlying concepts for user interaction within the different applications to gather a better insight into the different technological streams. We used the level of integration, market penetration (number of downloads and market spread) as well as the type of warning as selection criteria. Based on different automation levels, we developed a classification system for user interaction concepts.

3.2. Case study “Wildwarner” – effectiveness and potentials of a mobile warning service

In a second step, we applied these concepts to a young product, a smartphone app to warn drivers about the risk of wildlife-vehicle collisions (WVCs). The app Wildwarner (Translation: Wildlife warning system; Wildwarner, 2019) has been developed since 2016. In its beginnings, the system provided an audible and visual warning signal for the risk of WVCs based on statistical analyses of historical accidents. Later on, a messaging system to communicate with the local police and responsible local hunters was integrated, as well as a portal for hunters to manage their hunting area and report street segments at risk in combination with relevant periods of the day and year. The app is currently used by more than 50,000+ users (available for Android and Apple iOS), mainly in Germany, but partially also in Austria and Switzerland. About 500 users are registered users and provide the basis for our survey. We investigated the effectiveness and usage of warning detection, exemplarily for the Wildwarner App, aiming at understanding the users’ perception of the app. A questionnaire was sent out to 542 registered app users between 22 May and 11 June 2019. In total, the participants had to answer 19 questions about their app usage, driving behavior and the warning frequency of the app.

3.3. Business models for warning services

As a last step, we analyzed existing business models of warning apps and compared them to the 55+ Business Models of the St. Gallen Business School (Gassmann et al., 2017). In a dynamic market, where the change in user interaction also has strong impacts on the business model under which the service providing company is running, new business models might be required due to this alteration. General patterns such as the 55+ business models of the St. Gallen Business Model Navigator (Gassmann et al., 2017) are established and could provide help for running a business. We investigated a combination of models to depict which models are already established in the domain and which changes in models should take place. We analyzed which changes from a manually driving to the paradigm of (fully) automated driving will consequently result in a new business model variation for road warning apps. On the basis of the classified warning apps and the 55+ patterns of St. Gallen Business Model Navigator, we identified patterns for warning services in the future and derived concepts of how to direct the business model in alignment with the technical development.

4. Outcomes

4.1. Classification of warning apps

We identified and categorized different warning types regarding individual mobility. Not only warnings against collisions or lane departure are included but also warnings against traffic jams, weather phenomena, and obstacles, hence, warnings which affect driving or walking within the traffic system. For any kind of potentially risky object for driving, we selected an exemplary warning application. Our focus was thereby set on mobile applications usable in Germany.

Through the scientific Scopus research, 13 publications (five journal articles, eight conference papers) were identified. Six articles are related to the topic of warning detection for road safety using mobile apps. These articles deal with safety-aware navigation apps for pedestrian protection (Hwang et al., 2014), with a driver warning system for the android system (Ke et al., 2016), with collision warning and avoidance (Ghantous et al., 2017), with the warning app SafeStreet (Singh et al., 2018), with speed prediction in curves using mobile smartphones (Chu et al., 2018), and with an augmented proactive perception for driving hazards (Yang et al., 2019).

None of the detected applications are available in public app stores in Germany and, consequently, they are not included in the further analyses. The remaining seven publications deal with the distantly related topic of the impact of warning detection apps on driver behavior (Botzer et al., 2017; Payre & Diels, 2020), and the following foreign topics: service provider communication (Chen et al., 2017), hotspot detection via smartphones (Cuenca et al., 2018), an app for safely texting-while-walking in indoor environments (Huang, 2017), network forensics analysis of iOS social networking and messaging apps (Bhatt et al., 2018) and the design of driver safety UI (Lee & Lee, 2018). These projects are not further analyzed.

In the German Android app store, 18 different providers of warning apps can be identified. Some of them offer more than one warning application, e.g., the ADAC apps (ADAC, 2019). Comparing the selected warning services, the factors of Table 1 could be identified, classifying the apps into the following four upper groups: (1) Warning services integrated in navigation apps, (2) Standalone warning apps, (3) Apps with image recognition and (4) Others. We further categorized the selected applications according to the object the application is warning of, the type of warning signal (visual, auditory), the market scope (spatial spread and number of installations), and the maturity status (1 = publicly available/integrated in navigation system, 2 = publicly available/used as standalone app, 3 = prototype) (Table 1).

Table 1. Different types of warning apps and their classification

		Warnings against						Signal		Scope
Group	App service name	Collision (headway distance)	Lane departure	Speed/police control*	Traffic jam, building sites	Other obsta-cles **	Weather and natural disasters	Visual	Audi-tory	Spatial coverage	Number of install.(only Android)	Maturity 1 = in navigation, 2 = in use, 3 = prototype
1. Warning services integrated in navigation apps	ADAC apps (collective)				X			X	X	Inter-continental	4,000,000+	1/2****
	Google Maps				X			X	X	Global	5,000,000,000+	1
	HERE WeGo				X			X	X	Global	10,000,000+	1
	INRIX Traffic				X			X	X	Global	1 000 000+
	TomTom GO Mobile			X	X			X	X	Global	10,000,000+	1
	Sygic GPS Navigation & Maps			X	X			X	X	Global	50,000,000+	1
	Via Michelin			X	X		X	X	X	Global	5 000 000+	1
	Waze			X	X			X	X	Global	10,000,000+	1
2. Standalone warning apps	Blitzer.de			X	X			X	X	Inter-national	5,000,000+	2
	Carmunis			X				X	X	Global	50,000+	2
	PedSafe					X		X		Not limited	1,000+	2
	Wild-warner					X		X	X	Inter-national	10,000+	2
3. Image recognition	CarVi	X	X					X	X	Not limited***	1,000+	2
	iOnRoad	X						X	X	Not limited	500,000+	2
	Wavyn	X	X	X				X	X	Not limited	1,000+	2
4. Others	KAT-WARN						X	X	X	Germany	1,000 000+	2
	NINA				X		X	X	X	Germany	1,000,000+	2
	X2Safe					X		X	X	-	-	3

* Speed/police control: speed control does not just mean the warning against speed cameras next to the street, but also warnings in case the speed limit is surpassed

** Obstacles: wildlife, pedestrians, cyclist

*** No limitation: usage is independent of spread

**** Not every ADAC app is integrated in navigation, some are standalone

Nearly half of the warning solutions are integrated in navigation systems, mostly warning of traffic congestions or serving as danger alarms. Except PedSafe, all warning Apps use more than one type of warning, visual and audible. A majority uses real-time GPS tracking data or official report data, e.g., provided by traffic control systems. ADAC apps (ADAC, 2019), Google Maps (Google, 2019), Here WeGo (HERE, 2019), TomTom Go mobile (TomTom, 2019), INRIX Traffic (INRIX, 2019), Sygic GPS Navigation & Maps (Sygic, 2019), ViaMichelin Navigation (ViaMichelin, 2019), and Waze (Waze, 2019) can be assigned to this group (1). The apps of group (2) are characterized as standalone services. They are not combined with other warnings or integrated in a navigation system like group (1). These standalone products deal with warnings of wildlife-vehicle collisions (Wildwarner (Wildwarner, 2019)), speed cameras (Blitzer.de (EIFRIG MEDIA, 2019), Carmunis (Carmunis, 2019)) or other road users (PedSafe (PedSafe, 2019)). Applications of group (3), namely iOnRoad (Vegasrome, 2019), Wavyn (Wavyn, 2019) and CarVi (CarVi, 2019) are based on computer vision, comparable to driving automation techniques (video inspection, sensors for distance or driving control). Nevertheless, these app solutions are not connected to driving automation systems, but HCI is still necessary if the app identifies a risk. Group (4) consists of the applications KATWARN (Fraunhofer-Gesellschaft, 2019), NINA (BBK, 2019) and X2Safe (ZF, 2016). The first two warning-apps warn in case of emergencies, such as extreme weather conditions, and receive warning data from official authorities such as the German Weather Service (DWD). The latter app X2Safe (ZF, 2016) warns pedestrians as well as drivers against the presence of another road user by the use of car-to-car or car-to-pedestrian communication.

By comparing the number of app functions to their respective market spread (see Figure 1), it is obvious that warning services which are integrated in larger systems such as onboard and navigation systems, have a larger regional spread and higher app download numbers than standalone warning apps. The latter seem to be less popular because their use does not have such a large impact (several advantages according to the number of functionalities) as a complete and integrated package for navigation, warnings or other help services would have. Especially Sygic GPS Navigation, TomTom Go mobile, Waze, Google Maps and Here WeGo provide more than one functionality and are globally available, resulting in the most successful warning apps with regard to their market share.

Figure 1. Categorization of warning apps according to market spread and functionality – bubble size correlates with download numbers

The prototype X2Safe needs a full car-to-car communication and can therefore not be publicly used yet. Hence, the market potential regarding download numbers is currently restricted. Nearly all applications under consideration are implemented by commercial providers, except for NINA and KATWARN which were developed based on a governmental intent. Besides the integrated functionalities, the data origin is another success factor for traffic warning detection. In all apps with high market penetration, real-time data are the basis for the algorithm to warn against congestions or obstacles. The data are indirectly crowd-based, delivered by tracking-app and navigation-app users (location, speed, etc.). In some cases, such as in the ADAC apps, floating car data are delivered from partnering systems. In Sygic GPS Navigation & Maps, the smartphone’s integrated camera provides additional information. Also all apps from group (3) use camera and sensor information. Blitzer.de, Wildwarner, KATWARN and NINA use data from official sources, such as collision data from the police or weather information, while PedSafe does not access external data sources, but is based solely on the smartphone’s sensors. X2Safe is also based on crowd-data in combination with real-time calculations, comparable to group (1) of warning services.¹

4.2. User interaction classification according to driving automatization levels

Besides the integration of sensors and camera systems, services for safe driving will also be included in automated driving systems. For current warning services, the way to integrate them into the technical car system is an open question. Until now, drivers are the information receivers and, consequently, the acting part of the system; but is this HCI still mandatory and useful for automated driving? In the following, we discuss which warnings are still suitable for automated driving and which actors can be warning providers, receivers, and controllers. We value all classified warning apps (Table 1) by using three scenarios, which we defined according to the international standard levels of driving automation (SAE, 2018). In these three scenarios (Table 2), the differences of reactions on warnings (1) without driving assistance, (2) with driving assistance, and (3) fully autonomous driving are demonstrated. The scenarios are separated according to Biondi et al. (2019), depending on their level of HCI. It is especially discussed, if and how much drivers are responsible to react to warning detections, to interact with and to supervise the vehicle’s driving system.

Table 2. Scenario partitioning depending on the HCI

Scenario	(SAE, 2018) Levels	Human-Computer Interaction (HCI)
I	0	No technical interaction between warning service and driving system; But completely manually with 100% HCI
II	1, 2, 3	Level of driver assistance with some interactions between warning service and driving system; the warning service indicates a warning to the vehicle, the vehicle reacts with steering, acceleration/deceleration; HCI required
III	4, 5	Highly and fully autonomous driving; human interaction is no longer required, but possible, e.g., if a person wants to change conditions or activate/deactivate any warnings

First, the ADAC apps were evaluated whether and if yes, how they can be applicable for all scenarios (see evaluation overview in Table 3). The ADAC services provide congestion warnings and an alarm system, e.g., for wrong-way drivers. Using real-time floating car data (FCD) in combination with official police reports, drivers can be warned against danger on the road and suggestions for alternative routes in case of traffic jams are made. Thus, the apps cannot only be applied for warning the driver (Scenario I), but can also be helpful in Scenario II and III. Driving assistance (II) or the autonomous vehicle (III) may reduce speed, for instance, to avoid high speeds in case of a traffic jam ahead. In Scenario II, the driver may still interact directly with the warning detection and with the driving system. In case the ADAC services send out a warning signal, the driver could instruct the system to navigate via an alternative route (information system-human communication). Not only the driver but also the driving assistance would interact with the warning system (information system-automated driving system communication). While the driving assistance reacts automatically, for example, with speed reduction, the driver still decides on his or her own an alternative route. In a higher level of autonomy (Scenario III), the information delivered from ADAC’s warning algorithm could take over the function of alternative route decisions for the driver (completely information system–automated driving system communication). HCI is neither obligatory with the warning information system, nor with the driving system.

Table 3. Evaluation of scenarios II and III for the classified warning apps

Group	Name	Driver Assistance System	Autonomous Driving
Warning apps integrated in navigation	ADAC apps (collectively)	X	X
	Google Maps	X	X
	HERE WeGo	X	X
	INRIX Traffic	X	X
	TomTom GO Mobile	X	X
	Sygic GPS Navigation & Maps	X	X
	ViaMichelin	X	X
	Waze	X	X
Standalone warning apps	Blitzer.de	X
	Carmunis	X
	PedSafe
	Wildwarner	X	X
Apps with image recognition	CarVi	X
	iOnRoad	X
	Wavyn	X
Others	KATWARN	X	X
	NINA	X	X
	X2Safe		X

In the case of Google Maps, Here WeGo and INRIX Traffic, Sygic GPS Navigation, Waze, TomTom GO Mobile and ViaMichelin Navigation, the drivers receive warnings during the navigation process. They all have the potential for application in scenario II and III (for driving assistance and for autonomous driving). Based on the applied FCD from tracking the app users, best or fastest routes are calculated. In Scenario II, the drivers are still able to decide which route they would like to choose, while the app just provides support to navigate and gives information about the current traffic situation (information system-human communication). This potential can be conveyed to the level of autonomous driving (Scenario III). With transferring the alternative route calculation into the car, the app’s services, visualized on a mobile system, will become needless, but the algorithms behind can be used for best route calculations without the driver’s involvement. Nonetheless, vehicle passengers should have the choice to select between different routing options (fastest or most beautiful route) depending on their requirements but the tasks will be executed by the car (mobility as a service; information system-automated driving system communication).

The Blitzer.de app is applicable to scenarios I and II, warning drivers when passing a speed control (I) or stopping the driving assistance such as cruise control (II). In contrast to the apps above, for autonomous driving (III), the app Blitzer.de is not applicable, as autonomous cars will probably drive only with permitted speed. Consequently, speed control warnings seem to be needless in the future. As the app Carmunis is also a speed control app like Blitzer.de, the app is suitable for scenarios I and II, but not for scenario III. PedSafe, the app for pedestrians that only provides the function of a blinking flashlight to protect the pedestrian from being hit by a car while crossing the street at night, is not transferable to any higher level of automatization than in scenario I. As it is still the pedestrians’ choice to use this app, there is no benefit for driver assistance or autonomous driving. Autonomous cars will be able to detect pedestrians and other “obstacles” by their own sensors without forcing them to use a light signal. The Wildwarner App provides warnings about wildlife-vehicle collision hotspots via GPS location and is, therefore, useful for all three scenarios. An algorithm calculates the hotspots based on spatial-temporal analyses of historical wildlife collision data. The warnings can be provided to the drivers, but also to the automated driving system, e.g., by providing the warning data via a cloud platform, which is advantageous for security and verification reasons (information system–human and information system-assistance driving communication). The service may connect the warning information to the cars’ driving assistance systems and initiate that some functions like overtaking or using speed control can be blocked. Enlarging the minimal headway distance can also be an assisting function. The possibility of using collision hotspot data for autonomous driving is given, because the algorithm data can be coupled with the sensors of the car, improving its function or sending warnings in advance (information system-automated driving communication) in case of high risks for wildlife collisions. The driver could still have an active role by reporting wildlife sightings, being an implicit actor within the data delivery process.

Apps with image recognition are applicable as additional driving assistance in scenarios I and II, but they will be replaced by the vehicles’ own sensors such as camera and sensor techniques in the future. CarVi, for example, collects data from real-time recordings and ADAS algorithms. By video processing, the app helps to improve the driving behavior of its user. It warns against lane departure and front collision. Regarding the fact that the price of such an app is relatively low, it can revaluate a car without any driving assistance system and equip it with safety systems other cars already have. Wavyn and iOnRoad nearly provide the same functions as CarVi. They all use dash cams to supervise the current traffic situation and control the headway distance in real-time. Like in the case of CarVi, Wavyn and iOnRoad, the apps can be used for driving assistance but are not suitable for higher levels of automation. No practicability for autonomous driving is given in any of the three. Sensors of the car will take over the video analysis function of the app.

KATWARN and NINA are both useful apps for all levels of automation, as they provide official warnings from the (German) government, such as dangerous weather conditions or floods, but also serious traffic situations. The warnings are officially announced by the government and verified for accuracy. Publishing a warning via the mentioned apps means that a high number of traffic participants will be affected by the situation that is warned about. Regarding the provided warning types, the apps are appropriate for driving assistance, but also for autonomous driving, due to the fact that the weather warnings depict external shocks and the warning only shows up, if a certain level of danger for the population exists. The app provides behavioral advice that can also be used by automated systems. For scenarios I and II, this may be focused on advice for drivers on how they should react when reaching a region where dangers like floods, fire, hurricanes or leakage of toxic substances occur (information system-human communication). The same is true for autonomous driving, where the information is considered by the car’s routing algorithm. Even if the passengers do not have any impact on driving, there should be a way to interact in crisis situations and an interaction with the warning system will remain as a minimum HCI.

The prototype app X2Safe provides a car-to-pedestrian and pedestrian-to-car communication. Linked smartphones issue a warning from which direction a pedestrian or a car is approaching that one can focus attention on the car or the pedestrian crossing the street. For autonomous driving, this system would work perfectly if everyone was linked to each other. As this is not the case yet, X2Safe is not practicable for driving assistance. Even if the mobile system itself may not be applicable for autonomous driving, the information for the pedestrians and the passengers of the upcoming danger is beneficial. Especially, the passengers in autonomous cars may still be interested in the reason why the car acts in a certain way. The note of an upcoming obstacle may be helpful to better understand the action of the car.

In the following table (Table 3), the above-mentioned suitability of the apps for driving assistance and autonomous driving is summarized. These results helped us to identify the business models for the apps to be successfully applied for driving assistance and autonomous driving in the next chapter. As not all business models are applicable as a whole, new models may result through a new combination of existing models.

In summary, most of the warning services are transferable to autonomous driving, although the interaction between human, information system and automated driving system has to be reconsidered. In the case of Scenario 1, warnings are only delivered to the driver without any direct connection to the vehicle, as visualized in Figure 2 (see Scenario S1 1). The driver receives the notification and decides independently whether and how he or she will react to the received warning, meaning, the driver executes alone by intervening in the driving (information system-human interaction) (S1 2). The vehicle receives a warning due to the driver’s action, while it does not have any connection to the warning services nor to any data. Nearly all warning applications are operating in scenario 1, excluding the app PedSafe (ID = 11), as the app is characterized as a pedestrian-to-driver (human-human) communication and the warning derives from person to person, not from the warning service to the driver. Also X2Safe (ID = 18) is excluded (prototype stage) because of the implicit car-to-car interaction.

Figure 2. User interactions due to the scenarios I, II and III of driving automatization levels

In Scenario 2, in the case of an automation as a driver’s assistance, the warning information is provided to the driver as well as to the driving system (S2 1). Coupling the warning service to the vehicle is required for a system reaction. Nonetheless, also the driver receives the warning from the application. Since the driver still has the main driving control, he can decide whether a reaction to the warning is carried out or not (S2 2.2). When the vehicle is not able to solve an operation, the control and reaction are transmitted to the driver (S2 2.1). Otherwise, the vehicle responds to the warning. An interface passes the warning information plus previously stored information (own sensor data) to the driving control system. If the system does not know how to handle a warning, the task is transferred to the driver (2.1).

In Scenario 2, the driver can act independently if he or she wants another action as the driving system offers, or if the driving system transfers the task. Direct communication exists from the service application to the driving system, and also to the driver. All classified apps can be assigned in Scenario 2, excluding PedSafe (ID = 11) and X2Safe (ID = 18) for the same reasons as in Scenario 1. X2Safe only works, if a complete network between all vehicles on the road is available.

In Scenario 3, the warning information is provided to the vehicle (S3 1), not to the driver. The vehicle responds to the warning according to the prescribed parameters. In addition, the vehicle can inform the driver about the reason for the changed driving behavior, as she or he is no longer actively involved in the decisions, but could get trust by being informed about the reasons of the vehicles’ reactions (S3 2). The vehicle occupants, no matter if they are active drivers or passive passengers could obtain the warning either by voice recognition from the board computer or in form of a pop-up on the smartphone just for information (trust) or for manual intervention. Besides, the passengers could also provide information, e.g., by reporting obstacles, or even, when a warning (e.g., against congestion) is incorrectly detected (S3 2). The driver still has the opportunity to take control of the vehicle (SAE Level 4). Only in level 5 of autonomous driving, he would be unable to take over control.

We can conclude that HCI is still possible and desirable (for user acceptance) during automated or autonomous driving, even when differences exist between the interaction of humans, information systems, and automated driving systems depending on the level of driving automation. Almost all warning services can be generally operated in the three scenarios, even when they must change users’ interaction. Especially the apps integrated in navigation systems can be used as warning services while standalone apps can also be integrated in navigation or rather automated driving services.

4.3. Case study Wildwarner – effectiveness and potentials of a mobile warning service

To better understand the usage and user frequency of warning services as well as the subsequent driving behavior, a questionnaire was sent to 452 registered Wildwarner App users. Wildwarner is a smartphone application – available in Germany -, which warns car drivers based on the current position, daytime and season through an acoustic signal and vibration. The app applies a statistical analysis of the historical WVCs. The WVC data stems from police recorded accidents as well as areas at risk, reported by hunters. It is mentionable that due to the different WVC data sources, the users can choose which data they trust in and, therefore, what data their warnings should be based on. The response rate was 21.9% (99 responses) within the three week time span of the survey. All of these 99 questionnaires were filled correctly and are suitable for further analyses. The persons were asked the following questions (additional questions are not presented in this article). (Q 1) How often do you use the Wildwarner App? Answer categories: Daily/1-3 per week/1-3 per month/only once/never/only downloaded because of the wildlife-vehicle service; (Q 2) How many warnings do you receive on average? Free text response, with nomination per unit (day/week/month/year); (Q 3) What do you think about the number of warnings you receive? Answer categories: adequate/not enough/too many; (Q 4) How do you react after a warning from the Wildwarner? Answer categories: Slow down speed/drive more attentive/slow down speed and drive more attentive/drive on unaffectedly; (Q 5) Which warning source from the Wildwarner do you have the highest trust in? Depending on the priority, please order the answers according to the importance for you. Answer categories: warnings inscribed by hunters, warnings from the Wildwarner algorithm, warnings inscribed by driver users of the Wildwarner App; (Q) Additional questions concerning the person (age, sex …).

The results show that the users accept the Wildwarner App, as the majority of registered people activate their app multiple days per week. More than 28% of the respondents answered that they use the warning app daily, further 22% use the app one to three times per week. Thus, 50% of the users can be categorized as heavy-users (Q 1) (Figure 3). Regarding the warning frequency, 54% of the users valued 3.75 warnings per week on average as adequate. Thirty-nine percent marked that 2.99 warnings per week are “not enough,” while only 7% noted that 7.45 warnings per week are “too much.” The aim of this question is to get a clear view on the habituation effect in relation to the warning intensity. We consider the perception of warnings that the app user sees as too frequent as an indicator of the habituation effect. This can be linked to the awareness of irritation due to the frequency of warnings (Q 2, Q 3). Ninety-seven percent of the users indicated that they slow down or drive more attentively after receiving a warning from the app, while only 3% declared not to react actively by reducing speed (Q 4) (Figure 4). Thus, an effectiveness of the Wildwarner App for a more secure driving could be noted.

Figure 3. Frequency of usage

Figure 4. Driving behavior in case of a warning

As the perception about the warning frequency (too less/too much) could depend on the driving route, on the route length, or on the frequency of using the app, we first combined the indicated warning frequencies with the usage frequencies (Figure 5). We assumed a negative relation between both indicators, meaning that the heavy-users will be warned more often, they will be accustomed to the warnings and thus ignore them. This would also strengthen the assumption that habitual effects are the reasons why static wildlife-crossing signs beside roads do not have a long-term attentive effect (Bond & Jones, 2013; Huijser et al., 2015).

Figure 5. Frequency of usage combined with the perception about the amount of received warnings*

*combination of Q 1 and Q 2, remaining answers not included due to the low app experience (meaning Q 1 answers: usage only once/never/only downloaded because of the wildlife-vehicle service); only counted if in both section answers were filled in

In contrast to preliminary assumptions, we found that independent from the frequency of using the app, the numbers of warnings are mostly perceived as adequate. In the case of daily users, the category “not enough” is even 12 times nominated (48%), assuming that a habituation effect is not a consequence. Forty-four percent (n = 11) of the daily users classified the amount of received warnings as adequate. Sixty-eight percent (n = 13) of drivers using the app 1–3 times per week have the perception that the number of warnings is adequate, while for users applying the Wildwarner only 1–3 times per month, the warning frequency is sufficient for 58%. Nearly half of the daily users, but only about one-third of the other two groups wish to receive more warnings. With no significant difference between the user groups, between 5ive and 8% of the users perceive the warnings as being “too many,” again showing that the perceived number of warnings of the Wildwarner App is not too high and does not lead to habituation.

Second, by comparing the driving distance (kilometers per person on average) and the perception of the frequency of received warnings, a correlation can be identified. The user group rating the received amount of warnings as “too many” has with 150 km (median value) a notably higher daily driving distance than the two other groups (60 km for “not enough” and 74 km for “adequate,” median values). Thus, the app users with frequent app usage perceive the intensity of the warnings as more appropriate than the frequent drivers, which seem not to be the same user group.

The majority of the participants in the survey indicated that they are long-distance commuters (45%) and/or 43% are leisure drivers (multiple answers were possible). Furthermore, short-distance commuters (32%) and also professional drivers (11%) were identified. Ninety-five percent of the users are older than 24 years and 85% are employed or freelancers. Half of the users (50%) are between 50 and 65 years, 24% are between 35 and 49 years old and nearly three quarters (73%) are male users.

Lastly, it was also of interest, which warning source of the app users trust most (Q 5). As possible sources, hunters, algorithms (based on historical accident registrations) and drivers were ranked (from less to most important). The results show that the app users rather trust hunters as experts in the field. Hunters were named first by 36.9% of the users. The algorithms and computer technique take second place (33.5%), crowdsourced information by users rank third (29.6%) (Table 4).

Table 4. Results of Q5: In which warning source the app users have the highest trust in

Rank	Type of warning	Total points (3 points for first, 2 points for second, 1 for third nomination)	Percentage (%)
1.	Warnings inscribed by hunters	177	36.9
2.	Warnings of the Wildwarner algorithm	161	33.5
3.	Warnings inscribed by drivers/users	142	29.6

The survey shows that humans use and trust the warning app and that the Wildwarner App has positive effects on the drivers’ behavior as almost all users drive more attentively and more slowly in case of warnings. The positive effect of apps, also for other warning types, is also reported by studies from Albert et al. (2016), Botzer et al. (2017), and Cardamone et al. (2016).

4.4. Business models for warning services

We analyzed the 55+ St. Gallen’s Business Model Pattern Cards and identified which of the models were applied by the above classified warning app. The Business Model Navigator was developed at the University of St. Gallen in Switzerland (Gassmann et al., 2017) and is used to develop business models for a new business idea. Fifty-five concepts plus additional five concepts that were added later, are available and can be combined to develop a business model. The pattern cards were generated on already existing business models from newly founded businesses. With this Business Model Navigator, a business model for warning apps does not have to be reinvented, but elements of established business models can be rearranged and newly patched together. We do not consider the listed models as complete business model solutions but give an overview of which characteristics and attributes can be applied on our case. In this chapter, we present the 21 models that are currently used by warning apps or can be applicable for the integration of warning apps in higher levels of autonomy (Table 5) to show how the change of human-machine interaction might impact the underlying business model.

Table 5. Applicability and success evaluation for the selected Business Models for warning services for autonomous driving

The model “Add on” is a possibility for enterprises to choose some app features as basic features, which can then be upgraded by additional Add-on features. TomTom GO Mobile is designed after the “Add on” concept with, for instance, the specific navigation to contacts or photos as Add ons. Economically, the “Add on” model retains an extra fee and is transferable to automated and autonomous driving. Car manufacturers could provide warning solutions of third parties as Add ons through the onboard computer system or as part of the autonomous driving system. Through an app that is connected to the autonomous vehicle, the driver remains informed about the driving behavior, even if the person has no influence on the ride (Level 5).

The business approach “Crowdsourcing” is used but only implicitly by several warning services, e.g., by Blitzer.de, Carmunis, Sygic GPS Navigation & Maps, and Waze, using e.g., tracking data based on information of the smartphone app. This is also a possible model in case of changed communication (information to driving system) relations. Tracking data from the automated driving system can be used but the power of the car manufacturer and his access to sensor data might change the situation of app providers. Involving the driver in reporting relevant content, while being driven by an autonomous vehicle might be a solution, if drivers are willing to invest the available spare time in reporting.

The model “Customer loyalty,” used for instance, by Carmunis, offers a premium version. Prime users pay for downloading the warning app as well as for extra features for frequent users. The Wildwarner provides a special product package for hunters which could exist further on also in times of autonomous driving. The model “Digitalization” is related to all of the warning applications as it is the basis for all models providing digital services. Especially with regards to the warning for locations where wildlife crosses the street regularly, the app Wildwarner can be seen as the digital substitute to the classical warning signs along the road, providing the benefit of more realistic warning over time instead of a sign as a 24 × 7 warning solution.

In the “Experience-Selling” model, the customer’s experience with the purchased product contributes to the added value. A warning app, for instance, only provides a benefit if the users activate the app while driving. The experience is the higher safety given by the warnings while driving. “Flat rate” stands for a fixed fee-based app with a monthly or annual fee. A simple cost structure for the user leads to a constant revenue stream for the app company. Unlimited access to offline maps is one flat rate example from e.g., the Tom Tom GO Mobile app for one or three years and is also suitable for future mobility. The “Freemium” model is given when apps provide basic features for free and additional extra features for a certain price. Tom Tom Go Mobile and Carmunis use such a model. The users can download the free version but can also buy the premium version.

With the model “Guaranteed Availability,” the general functionality of an app has to be guaranteed, for example, even without mobile data, which is partly essential for autonomous driving. PedSafe is an app with warning functionalities without data provided online and, thus guaranteeing availability at any time. Apps such as Here WeGo or Google Maps provide the possibility to download offline maps for an area-wide service availability, which is important for autonomous driving when navigation by car is completely data-based.

With the business concept “Ingredient branding,” companies use well-known brands – in case of warning apps, e.g., TomTom Go Mobile – to increase the attractiveness of their services through the high recognition factor. In an integrated system with communications between information systems and automated driving systems, the brand effect does not longer provide benefits for the client. The concept “Layer player” will receive positive responses by the warning service providers, as they will only provide one value-adding step. The integration and action preparation for drivers or rather automated driving systems will not be developed by the warning providers. Especially apps classified in the first app group, namely CarVi, Google Maps, TomTom Go Mobile, Waze, Sygic GPS Navigation & Maps, follow the business concept “Leverage customer data.” They collect data about driving behavior for further developing and improving their products in order to increase the clients’ satisfaction, which is even possible and useful in the automated driving stadium.

Also the “Long tail” concept will be profitable for warning services as every provider for apps, such as Wildwarner or Wavyn, offers niche products which are able to be integrated in larger navigation and security products without a large number of competitive products. Applications of the identified group one (see Table 1) could be the comprehensive service providers including niche services from other groups, such as group two of Table 1 which warns against one specific obstacle. Additional services (“Make more of it”), e.g., ADAC apps and TomTom GO Mobile offering educational training and consulting, use existing know-how. This will not have a large potential if automated driving systems will take over the leading role. Instead, “Mass customization”, thus the individualization of mass production, will be prosperous. Single services could be activated or deactivated, depending on the passengers’ wishes. “Open business models,” including the collaboration between partners in an ecosystem are also a suitable way of product positioning on the way to a larger integrated service system for autonomous driving. Data (Point of Interests) of the Wildwarner are already available and integrated into navigation services (e.g., POIbase) of other providers, while Sygic GPS Navigation & Maps is based on data from TomTom Traffic. As it is a free decision to install and use a warning app, downloading and setting up may be seen as a self-service. “Self-service” is only partially a prospective concept in our digital period. It is already provided by all classified apps, having the intention to be downloaded by the users. This will no longer be necessary if the services will be integrated in the driving system and are shipped with the car. “Solution provider,” in contrast, could partially be useful in case of an accident, such as the Wildwarner’s additional accident services. Additional services for improving driving skills, such as in CarVi, will not be needed in a self-driving system.

The “Subscription” model, where a customer must pay a recurring annual or monthly price, will remain, but presumably within an integrated pricing system for total navigation and security system for autonomous driving. As mentioned in the introduction, security packages will constitute one of the largest revenue streams according to European Commission (2017). Offering more than one service, maybe in collaboration with other services, will be a possibility, named “Supermarket.” For autonomous driving, a car seller will work with a supermarket model, meaning all kinds of warnings are available at one store. Either an app already provides more than one service, e.g., Via Michelin (navigation, traffic jam, speed and police control) or Wavyn (collision, lane departure, speed and police controls, weather and natural disaster), or a niche product. Both will work as a supplier in collaboration with one of the larger providers. According to the Supermarket model, the provider also has to offer a large variety of products, in our case warning types, for low prices. For a customer, it will not be useful to have just a small and incomplete selection of warnings.

To create a Solution provider with a Supermarket system, it is necessary that providers cooperate and White labels might result. Under a “White label,” services could be integrated, where the brand of the service does not appear but gets sold under the label or brand of the platform provider. KATWARN, an official governmental app is such an app, where warning services of different private and public service providers are presented in one governmental app. As the White labels are still autonomous providers, just offering their product under another company’s name, another alternative could be an Open business model, not unconditionally debarring a White label model. As there are various warning types and also a lot of warning possibilities not discovered yet, it would be beneficial to work with an Open business model, meaning that the actual business model can be adjusted more flexibly to environmental conditions, especially in the transition period from human driving to autonomous driving. The “Virtualization,” as last selected business model, could still exist due to 2- or 3D Navigation independently from the location or device, but will be useless someday when the navigation is undertaken by the automated driving system (level 5).

4.5. Discussion

The overview of currently applicable business models gave insights which warning services and business models will persist and which ones will get lost due to the conversion of HCI from a user to smartphone to a user to information system concept. Autonomous driving will change the interaction even more. We wanted to discuss and evaluate the business modeling for road safety applications based on this shift of interaction. Guaranteed availability has to be named as one important and obvious model in times of driving automation. Without continuous warnings, while driving, an app does not seem to be reliable but guaranteed availability might provide a different level of safety although it is not required. The apps can be made available either through a Flat rate or through the Freemium model as long as basic safety aspects for autonomous driving are covered by the basic version. The warning services are not available everywhere, not because of the functionality, but because of the limited regional spread, even though basic functionalities which are necessary for autonomous driving must be made available comprehensively. Additional functions shall be placed at the disposal as Add ons. Experience selling and Customer loyalty as two supplementary models emphasize the fact that the warning apps’ benefit is disproportionately higher than the price of these apps; some apps are even free-of-charge. In return for installing a warning service, the user will receive driving assistance and, thereby a safer ride. The accident risks with the associated health aspects are related to much higher costs than the installation costs. By utilizing Crowdsourcing or rather crowd-based data, the algorithms underneath the warning detection can be improved, e.g., with a machine learning approach. Using warnings comprehensively from apps, which partially provide only one functionality, a Solution provider model is essential. Even if there are various sources of warnings, a central source should exist which can centrally be tapped by automated driving systems. The various sources of warnings can be titled as White labels. Small providers should cooperate with each other and try to scale.

The apps show different spatial coverage that ranges from national via international to global spread and also depends on the age of the app. Spatial coverage indicates the global market penetration and global success but reflects also the company’s strategy based on its resources, data availability to run the app, legal constraints or legal gaps, and, last but not least, the ability of the company to bring the development from one step to the next. While data for an app usage are not available, the download number together with the user feedback can be taken as a success factor for an app.

The presented business model patterns can be used as a starting point to further develop the strategy of traffic warning services. Especially for startups, it is essential to optimize the alignment of technical development and business development to be cost efficient and not risk the existence of the company by a larger incorrect decision. Nevertheless, it should be considered that the presented business patterns are a combination of models derived from existing business cases. With the changes in the business cases, it may be that an innovative business needs to reconsider the models and invent a new one because the existing patterns are insufficient.

4.5.1. Observations about the business model of the Wildwarner App in times of autonomous driving

One exemplary case can be the Wildwarner App. The app has 50,000+ downloads (available for Android and Apple iOS) and in the survey, mentioned above, we mainly obtained positive users’ critiques. As the way of collecting data (hunters, drivers, and algorithms) is complex, it takes a lot of time to scale. Cooperating with other projects dealing with wildlife-vehicle collisions in other countries may therefore be helpful. Even if every project has its own way for collecting data, there might be a possibility of how to combine the results. Equipped with sensors, an autonomous car will be able to recognize wildlife close to the road. Maybe, however, the detection takes place too late and the wildlife-vehicle collision is inescapable because an animal is hidden by trees or buildings. An autonomous car will solve the situation itself, not dodging but breaking sharply, as is the best way to avoid an even worse situation. But still, there might be damage to the car. Involving the data from the Wildwarner might improve the cars’ reaction. Knowing that the region is known for a high risk of wildlife-vehicle collisions at a certain season or time of the day, the car may slow down (slower than the current speed limit) to be prepared for wildlife crossings. If the passenger does not want the car to slow down, but to go on at normal speed, he can just reject the preventive measure. Implementing the warnings as described, the Wildwarner will work as an Add on, which the user can purchase in addition, but does not have to. Data still may be collected via crowdsourcing (users and hunters), by the police, and the already used algorithm. As the people in the car will change their role from drivers to passengers, they will have more possibilities to use their smartphones while driving. Passengers’ real-time warnings can be added to the system and other cars may receive the warnings as soon as it is added. Possibly, the Wildwarner will not be distributed from an independent company (as the wuidi GmbH is nowadays) but by using Ingredient branding, automotive companies may own all warning apps and sell them as Add ons to Solution providers.

4.5.2. Trust in autonomous driving through human-computer interaction

Still, many doubts about autonomous driving exist for drivers. Using warning detection, especially via apps that are able to analyze the driving behavior, are a good approach to lead people to autonomous driving. Warning apps might be even helpful in the transition toward higher levels of automation and autonomous driving. When the interaction of humans is no longer needed, the drivers’ trust, the possibility to intervene and have the feeling of a supervisory actor (as it has been until now) is necessary for the acceptance of this technical development. Also, Biondi et al. (2019) argue that there is “[.] the need to move beyond human supervisory roles, and embrace the possibilities of peer-to-peer relationships in models that guarantee system flexibility in coordination and disambiguation of roles and control tasks.” Trust in automation is an important threat in terms of technology acceptance. Hence, interaction between humans, information systems and automated driving systems should still be able not less than the whole transition process is passed through.

The analysis of the survey provided a first insight in the usage of warning apps and shows that the warnings have an impact on drivers’ behavior and consequently on the driving safety. The perception of warnings and especially the adequate amount of warnings per time unit represent the users’ feelings and lacks a quantitative basis of accidents in contrast to avoided accidents. Even when the survey shows that the trust of the users on other users’ information (see Q 5 results of the survey) is not as high as on experts (hunters) or algorithms, users’ information (tracking, crowd-based data) are nevertheless inevitable for warning algorithms beyond the warning services.

5. Guidelines

The transition from human driver to automated driving will force the providers of warning services to adapt to follow this trend. The providers ought to change in the following ways:

• Startups and app provider should change from an app-based service to a technically integratable service.

• Car manufacturers need to provide the car as a platform and provide interfaces to integrate third party warning services and provide warning detection to the user. Solutions for a secure communication between external service providers and the car as a platform must be considered during implementation.

• Startups and app provider will have to change the business model to better integrate as a service provider in the car’s information system.

• The transition phase from manual to autonomous driving requires existing services providers to adapt to the new situation, but also open the door to completely new services and see the end of selected services which become of no avail.

This way toward fully automated driving will significantly change the user interaction with the car itself and especially with the onboard information systems. Regarding the warning services providing several solutions available as smartphone apps, the transition process of interaction between human, smartphone-based information system, onboard information and automated driving system is complex. Smartphones or their underlying services will be stronger integrated in the onboard information systems, while direct user interaction with the smartphone will be reduced for drivers to reduce the risk of distraction. On the other hand, the integration of services in the onboard information systems as well as systems for driving automation will lower the direct interaction of drivers with the service itself. Results of the warning algorithms could be directly transported into the automated driving system which may act immediately – without explicit warning elements -, leading to a faster decision process and involving sensor data of the car, while HCI might remain only for informational purposes – possibly with auditiv and visual warnings depending on passengers’ purposes, independent from the warning type. Some of the analyzed warning services will disappear, as the service will be invaluable for automated driving or it will be substituted by the car and its integrated sensor systems, while some other warning services will provide added value to passengers in the car. For other service types (navigation, speed control), we were able to show that a high level of service integration and, consequently, changes in the user interaction seem to provide a significant benefit to users and will also increase the usage of the service. Besides the external services, sensors within the automated driving system can improve security for the car itself, but also by giving their data to the external services for improving the service’s algorithms.

Through the integration of services or even the integration of services in the onboard information systems or automated driving systems, not only user interaction but also underlying business models will have to change. Especially through automation, some of the services will provide additional input data for decision-making of automated driving systems. They will not be delivering not any more a decision basis for drivers anymore but provide valuable and optional information for the car users. These services will have a chance to remain on the market, if they consider the paradigmatic changes of user interaction and an adaption of their business model. For the case of the wildlife warning service Wildwarner, we were able to indicate directions how the integration with systems for driving automation could provide a benefit, but might change the business model from a standalone provider to an integrated service using business models such as White labeling or Ingredient branding.

Especially for the transition from actively and individually driven vehicles, where humans interact totally with the vehicle and its onboard-computer toward an autonomous driving stage, people may still feel as having the role of an active driver. They might still want to interact or know the reasons for driving changes such as speed reductions. Vehicles’ drivers and passengers have safety requirements. Until the technique works for 100%, people may be unsure and want to keep control over the vehicle for safety reasons and optimal decisions, especially in the case of emergencies (Biondi et al., 2019). This is also confirmed by the executed survey, in which the app users predominantly trust more in experts such as hunters than in algorithms. In this transition stage, drivers could still be informed by the warning apps for gaining trust into reasonable technical interventions. Situations without a clear visible reason for speed reduction or alternative routing could be explained. Unlike conventional public transportation by fixed bus or train routes, individual but already autonomous driving does not necessarily require a complete inactiveness of passengers. Depending on their wishes or conditions, people could still decide and interact with the vehicle and control it, e.g., if they do not accept a speed reduction because the model-based warning of a traffic jam does not correspond with the real situation on the road.

The study is currently limited to smartphone apps available on the German market, but can easily be extended to other regions. Especially in the transition phase from user driven cars to fully automated cars, different forms of user interaction with the warning services will be necessary and, hence, business models of warning services need to be flexible and adaptable to upcoming changes. The transition from smartphone apps to integrated information systems and the integration of external services into automation systems will depend on the willingness of car manufacturers to provide interfaces and access to external data and services. The development of standardized technical interfaces for a better integration of services into automated and autonomous driving are as relevant as the standardization and integration of user interfaces and warning services into larger platforms. This is not only recommended to provide a better interaction and user experience but also to reduce distraction of the driver. Especially governmental standardization approaches, such as mCLOUD and mdm of the German Ministry of Transportation and Digital Infrastructure, to make core data available and bundle information and warning services, might be promising approaches.

Disclosure of potential conflict of interest

The authors declare that this submission contains original work which has not been published previously and is not submitted for publication elsewhere. No potential conflict of interest was reported by the authors. The funders had no role in the design of the study, writing, or in the decision to publish the results.

Additional information

Funding

This research was funded by the GERMAN FEDERAL MINISTRY OF TRANSPORT AND DIGITAL INFRASTRUCTURE (BMVI) as part of the mFund project “WilDa – Dynamic Wildlife-vehicle collision warning, using heterogeneous traffic, accident and environmental data as well as big data concepts” grant number 19F2014A. We thank wuidi GmbH for their support and the distribution of the questionnaires.

Notes on contributors

Johanna Trager

Johanna Trager is a research scientist at the Technische Hochschule Deggendorf. In the WilDa project, she worked on business modeling and on conceptual development of IT solutions and marketing areas. She is currently studying Economics and Public Policy at the University of Economics in Prague.

Lenka Kalová

Lenka Kalová is a research scientist at the Technische Hochschule Deggendorf. In the WilDa project, she also worked on business modeling and on conceptual development of IT solutions. She is currently doing her Master studies in Business Administration at the University of Passau with a focus on business information systems.

Raphaela Pagany

Raphaela Pagany is a senior research scientist at the Technische Hochschule Deggendorf. Her research is focused on Geoinformatics, spatio-temporal analysis and risk prediction. In the project WilDa, she develops an algorithm for the Wildwarner service. In addition, she is currently doing her PhD at the Paris Lodron University in Salzburg.

Wolfgang Dorner

Wolfgang Dorner is professor for computer science at Technische Hochschule Deggendorf, director of the university’s Institute for Applied Informatics, and co-director of the Institute for Entrepreneurship. His research is focused on spatio-temporal modeling, and as a mentor of the Wildwarner Startup Team, he supported the development of this warning service.

Notes

1. Information about data origin are based on the apps’ websites and Play Store descriptions