User Experience of On-demand Autonomous Vehicles
Part 3: Would you feel secure and comfortable in a robotaxi?
Human Factors Specialist, Ictech, Göteborg.
When the first article in this series was published, On-Demand Autonomous Vehicles (OAV) were still something to be considered quite far from a commercial product or service . By the time of the publishing of the second article , Waymo had taken a leap towards commercialisation of On-Demand Autonomous Vehicle Services (OAVS) by launching its first commercial robotaxi service, albeit with human safety drivers behind the wheel.
Another seemingly small but very important step has been taken since the latest article: Waymo is now offering completely driverless rides for the public – with the passenger/user being the only person inside the car during the ride . The service is presently geofenced to a limited area in Phoenix, AZ – but still, this must be considered another significant advancement towards the widespread introduction of OAVS.
What is particularly interesting with this new development from a user experience (UX) point of view are the problems, challenges, and opportunities that are uncovered from actual real-world usage that are hard to observe or even predict in a lab, or even in field trials with human safety drivers . Observing people using OAVS will yield different results compared to if you just ask people what they want or how they envision using certain products. For example, Waymo apparently tackles the issue of how to handle things such as drop-off locations, sudden changes in plans and other types of interactions that typically can be smoothly solved through a simple conversation between a passenger and a human driver:
“User experience turns out to hold some surprisingly thorny challenges once humans are removed from the equation […] Computers […] lack the human flexibility and adaptability needed to be a good mobility provider.” 
Although the deployment of a service such as Waymo One, completely without any human behind the steering wheel, marks an important milestone in terms of the technological advancement towards a large scale introduction of OAVS, Waymo now focuses on how to “weave” their computer driver into the complex fabric of society and solving all these problems that may arise when real people interact with their product .
Waymo is however not the only company realizing that the user-robotaxi interaction may hold difficult challenges which may compromise a wide-scale introduction of OAVS. We see other companies such as Renault  and Daimler  spend significant efforts in trying to uncover the user-related issues that may arise in the OAVS business and more actors will likely follow as they transition from field testing to actual service deployment.
In our previous article, we focused on the central topic of understanding and building users’ trust in driving automation . In the current article we continue unfolding user experience aspects of OAVS, such as cybersecurity trust and intra-passenger trust, as well as other topics that we believe are important to consider in the UX domain: physical ergonomics, comfort, motion sickness and keeping the vehicle interiors clean.
Trust and feeling of security
Trust may be defined in many different ways but specifically within a human-automation interaction context, it can be defined as “the attitude that an agent will help achieve an individual’s goals in a situation characterized by uncertainty and vulnerability” . Our discussion so far has concerned whether or not users are being able to trust that the automation performs the driving task in the safest possible way . While automation trust mostly refers to the car being able to maneuver properly and handle all traffic situations, the driving task also includes decisions taken on a higher, strategic level, such as deciding which route to take in order to get to the destination as soon as possible (see e.g. Michon’s classical model of driver behavior ). In this vein, studies have shown that users may be skeptical about the car’s ability to find the best route and whether or not it will get stuck in traffic  which may be a sign of distrust in the automations’ capabilities on the strategic level.
There may however be aspects of human-automation trust that are not directly related to the driving task. One such concern could be whether the autonomous vehicle will be resistant to cyber threats such as remote hijacking of the vehicles . Naturally, the vehicle automation system needs to have a solid digital infrastructure that prevents such intrusions from happening. At the same time, users also perhaps need to be made aware of the vehicle’s capability of resisting cyberthreats to fully trust the vehicle? Could it be that companies even need to market these features to potential users to avoid cybersecurity distrust?
Moreover, given that the mobility service may strive for delivering a highly user-individualized experience, there may also be concerns amongst users whether their personal data, travel patterns, in-car app usage etc will be recorded and/or used in an unauthorized way . The service provider must ensure personal data protection and prevent the user’s personal integrity from being infringed by either the mobility provider or some other entity in the mobility service chain . Mobility providers must also ensure that users’ travel data (e.g. start and destination point) can be protected while at the same time offering the possibility for sharing the ride with other users. This problem has been referred to as privacy-preserving route matching and could potentially be solved using blockchain techniques .
Problems with trust in OAVS is not only technology-related but may also occur on an interpersonal level. In general, sharing- and person-to-person services (such as Airbnb and ebay) rely on that users can build mutual trust in their interaction [12, 13]. OAVS should be no different and would be even more critical in this sense since the users share a quite small space, possibly leaving you quite exposed and with few escape ways, should you end up sharing a ride with someone with the intent of harming you.
A study conducted by Salonen investigated the attitudes among travelers in a low speed automated shuttle bus . It was found that while the perception of traffic safety was better in the automated bus compared to a conventional bus with a driver, the respondents were lacking personal in-vehicle security. A majority (64 %) of the survey respondents answered that their sense of in-vehicle security in the driverless shuttle bus was worse or much worse than in the conventional bus and this held especially true for women who felt significantly less secure than men. This finding is remarkable especially since the study was conducted in Finland which has been ranked as the safest destination globally according to World Economic Forum, and a land where people generally trust each other .
Indeed, we have to consider that all sorts of people are going to want to travel with the OAV. The well-behaved user must be able to trust that the service is being able to reject those who have malicious intent or are ill-behaved from using the service. The user needs to feel confident that the service can be trusted to perform the necessary checks on passengers before it lets them in on the ride. Could you for example be sure that you don’t have to face sharing your ride with a bunch of drunken, violent hooligans on their way home from a lost soccer game? Or a maniac on a murdering streak?
Design firm Teague envisions that automated vehicles can be equipped with technologies that can detect and respond to any potentially dangerous situation, e.g. by analyzing the sound or visuals recorded from inside the vehicle and automatically alert emergency assistance in case needed . Teague further suggests that users should be able to discreetly select new “safe” drop off locations should they feel threatened. Thus, vehicle/service interaction devices inside the vehicle that are easily accessed by all the passengers such as shared touch screens may not be good in this sense. The interaction should rather be performed only through a personal device, smartphone or similar. This type of interaction is more or less standard in all ride sharing services today (Uber, Lyft, etc.), but it could be noted that the only OAVS on the market today, Waymo, uses a combination of personal device interaction and interaction with onboard screens. It could be questioned whether or not these onboard screens will be advisable from a ride sharing/user privacy perspective.
The reasoning above also implies that users may request that the service allows the user to control with whom s/he wants to share the ride with. This naturally requires that the users have some way of assessing whether or not a potential co-rider is trustworthy or not. Creating interpersonal trust is something that has already been struggled with by the services relying on person-person relationships such as ebay, Lyft and Airbnb, who have discovered that it is necessary to invest in a “trust infrastructure” which to some extent relieves users from the responsibility of assessing each other’s trustworthiness [12, 13, 16]. This trust infrastructure can involve employing analytics of user behaviors, to require that users rate other users, to require that users have non-anonymous user profiles, a back-office that quickly can handle complaints etc.
While such measures can be effective in catching the most obvious bad actors, it can also lead to a system where users are “hyper-accountable” and risk being shut out from the service as a result of getting some (perhaps unfair) poor ratings (cf eg. the TV-series Black Mirror – Episode “Nosedive” where a woman seeking to raise her ratings to be able to purchase an apartment, by mistake ends up death spiraling towards poor ratings and eventually becomes arrested ).
On the other hand, it may be the case that people also prefer opting out on sharing the ride with others. After all, people who today primarily commute by car do often travel alone and the reason for this situation may simply be that people like to spend some time on their own. The concept 360c by Volvo Cars  is for example presenting a vision with one person pods that can travel over a longer distance and allow the passenger to sleep on their journey. Such solutions may not be very beneficial if the general goal of automated vehicles is to reduce congestion and increase energy efficiency of traveling, but may be what people really want to have for them to consider autonomous vehicles over normal cars.
UX designers in different companies are nonetheless making efforts in trying to make the sharing experience more pleasant. The vehicles used within Volkswagen’s ridesharing branch MOIA is an example of where special attention has been paid to creating a personal space within the vehicle that will make riders feel more comfortable, and similar features are also implemented in NEVS’ vehicle Sango that will be used in NEVS’ OAVS called “Pons” [18, 19] (see also Figures 1 and 4).
But looking at it from a more positive perspective, the shared space created by OAVS can also be an opportunity for increasing interpersonal trust in society. Maybe an autonomous car service could offer new ways of meeting people, just like in Rinspeed’s concept, “Oasis”  where a matchmaking app is used to pair users together. And maybe the OAVS could not just find you a date for the evening; it could also let you meet the person you would never meet otherwise, spark a conversation that would make you look at things differently and teach you something that can’t be googled.
Physical ergonomics and comfort
To a great extent, traditional passenger cars are currently used as the vehicle platform for investigations in the OAVS domain. Waymo for example uses Chrysler Pacifica minivans, GM/Cruise uses Chevrolet Bolts and Uber uses Volvo XC90s. It is likely that these types of vehicles are not optimal for OAVS operations since they are primarily designed with the driver’s ergonomics and experience in mind. Newcomers in the automotive business that focus solely on autonomous vehicles, such as Navya , take a different approach to the physical design of their vehicles, and we can also see from the big manufacturers’ concept vehicles, such as Renault’s EZ-GO  and Volvo Cars’ 360c , that they envision a need for new types of vehicles in the fully autonomous era. From a physical ergonomics and general comfort perspective, there are some specific aspects of these and other purely autonomous concepts that are interesting to analyze.
Ingress, egress, door design. First of all, an OAV would likely benefit from having a different way of entering and exiting the vehicle compared to today’s cars. Just like a regular bus, riders will enter and exit the vehicle frequently and this process has to be optimized not to create an inefficient ride. Analysis of users’ ingress and egress behavior in today’s passenger cars show that inappropriate doorway height, sill height as well as sill width can cause problems for users . OAVs would likely face similar problems, but due to the fact that a driver’s seat and all the related controls and interfaces are no longer needed, the vehicle can be designed more freely in terms of ingress/egress and thus avoid the problems encountered in today’s cars.
Doors would likely need to open and close automatically, or at least be powered (by hydraulics or electric motors) to make ingress and egress as quick and easy as possible and allow all types of riders to use the vehicle – both young and old users as well as those with disabilities and impairments. Ingress and egress should furthermore be facilitated by using access ramps and by having low instep or suspension that can regulate the height of the vehicle when the vehicle is stopped. Needless to say, doorway height and width should be enough to allow easy access for all users. Furthermore, the way the door opens will affect the ease of ingress and egress. The optimal door design moreover depends on how the seats inside are arranged.
A variety of door designs have been proposed by the different OEMs’ products and concepts, see Figure 1 below. Common to most of them is that entry is only possible through the sidewalk-side of the vehicle, just like in most buses. The reason for having this arrangement is obviously that entry through the roadside of the vehicle would be both cumbersome, potentially dangerous and unnecessary.
Side-hinged doors such as those found on traditional passenger cars can easily be in the way during entry and potentially also obstruct the sidewalk (see e.g. the Mercedes F015 concept, Example 1). A common solution otherwise seems to be sliding- or “pantograph” doors such as used in the Cruise Origin or the NEVS Sango (Examples 2 and 3). Upwards-swinging doors such as NEVS InMotions’ side-hinged door or Volvo 360c’s gullwing door (Examples 4 and 5) are more uncommon but could potentially allow for a better, more open entry. Renault’s EZ-Go (Example 6) presents perhaps the most unique solution with its combined roof/door opening that allows for almost unlimited ingress/egress height. The concept also shows a ramp solution that makes entry with prams, wheelchairs and etc easy. The decision to place the entry at the front of the vehicle is unconventional and could require a special pick up/drop off spot designs; thus, such a vehicle would be more difficult to integrate into existing infrastructure. It could be noted that it, in contrast to the other vehicles, allows for a different seating configuration with seats along both sides of the vehicle.
Figure 1: Some different door solutions. Note that Examples 2 and 3 are vehicles intended for production while the others are concept vehicles.
Seat design and seating arrangements. The way the passengers are seated as well as the seats themself inside the vehicle will affect how the ride is perceived and thus also the overall experience of the ride. The work by Bengtsson  suggests different parameters of importance in seat arrangement and seat design for automated vehicles, as well as some interesting concept designs. Bengtsson identified the following most important factors relating to physical ergonomics and comfort when it comes to seating design:
- Provide stability: Support in curves when driving at higher speeds by e.g. using side cushions and tilting seats.
- Provide customization: Support different activities (active, entertainment, working position)
- Provide spaciousness and a clear outside view: Reduce the risk of motion sickness
When it comes to seating arrangements, there are a few different variants that seem to have emerged among car designer’s visions.
Figure 2: Some possible seating arrangements (adapted from )
A traditional seating arrangement (arrangement A, in Figure 2 above) is seldom shown in futuristic concepts, but it is more common to see the face-to-face variant C. Is this really better from a user perspective? For persons seeking privacy, face-to-face arrangements are obviously not advantageous. It is also interesting to note that a commercial vehicle that has been designed specifically with sharing in mind, the VW Moia, has chosen a traditional seating arrangement and also worked with providing extra privacy for its users by utilizing spaced wing chair-type seats. An outlier in this respect is the Volvo 360c concept which envisions one-person pods as one possible solution for longer travels. Moreover, traveling backwards or sideways is not advantageous from a motion sickness perspective since it limits the passengers’ ability to look outside and therefore increases the conflict between visual and vestibular senses and also decreases the passenger’s ability to anticipate upcoming maneuvers (see next section on how to limit motion sickness).
In a study by Koppel et al , 552 persons of ages 18-78 in four different countries were asked in an online survey which of the seating configurations presented in Figure 2 they would prefer as a passenger of a fully autonomous vehicle when traveling alone or with other people. Overall, the main preference considering all age groups was the traditional “A” seating where all passengers are facing in the direction of travel. Younger people seemed to preferred seating arrangement D when traveling with partner or spouse but apart from this distinction, all thought they would prefer the A-seating no matter if they would travel alone or with people they know or not know.
Supporting various activities. One of the main arguments for autonomous vehicles is that they will allow users to spend their time in a more productive way. For example, the 360c concept by Volvo Cars  suggests that the driver should be able to either work, relax, party or even sleep (See Figure 3). Similar ideas are envisioned by other companies’ designers, e.g. the Mercedes F015 concept  which has similar features for socializing and working. Recently released information from NEVS show that their selfdriving vehicle intended for production, the Sango, will have a configurable interior that can be rearranged for either privacy (see Figure 4), social interaction or for family trips .
Investigations into how people think what they would do in an autonomous vehicle show that common activities include read, talk to other passengers, listen to music/podcast/radio, watch TV or a film, and relax/rest/sleep , play games or talk/text with friends and family . People also seem to think that they simply would be watching the road or traffic ahead  – this was even the most anticipated activity in a study by .
Provided that a lot of time is spent on commuting to and from work/school nowadays (an average of 52 min each day in the US ), it seems that it would be very beneficial if autonomous vehicles would allow the commuting time could be integrated in working day and in this way make everyday life more efficient. Autonomous vehicles could offer this possibility and make this way of daily transportation much more attractive than using a regular car for commuting .
Figure 3: Volvo’s 360c concept, arranged for sleeping (Source Volvo Cars).
Figure 4: NEVS’ Sango, in privacy mode (Source NEVS)
The seating arrangement and flexibility thereof dictates how the space within the vehicle can be used for different activities, as will other features of the interior space. For example, efficient working performance would likely require some degree of privacy, ergonomic seating and lighting conditions, as well as a stable and secure network connection and perhaps special teleconference possibilities. Creating a good space for resting or sleeping would require the seat to be comfortable and possible to tilt, and would also benefit from the possibility of blocking out light and reducing disturbing noise. Supporting the interaction between passengers would require face-to-face oriented seats and/or good speech intelligibility. Maybe the vehicle should even have a good audio system and a bar for those trips to the party? Supporting all these and other activities would require that the interior and its features are easily configurable/customizable, maybe even personalized (i.e. automatically adjusted) for each passenger and for each trip.
Avoiding motion sickness
There is a growing concern that the possibilities for performing non-driving-related in-car activities will be limited due to the risk of motion sickness (kinetosis) . Motion sickness is a condition characterized by symptoms of nausea, dizziness, fatigue and other types of physical discomfort . According to  there are three main factors leading to the condition of motion sickness: a conflict between visual and vestibular inputs, loss of control over one’s movements, and the reduced ability to anticipate the direction of movement.
Thus, passengers are more likely to develop motion sickness than drivers since they do not have control over the vehicle’s movement and therefore also have reduced abilities to predict the vehicle’s direction of movement. Such abilities will also be further limited by passengers who do not have focus on the road ahead. Not looking at the road, but instead looking at a smartphone, book or similar, is also bound to create mismatches between visual and vestibular sensory inputs, which will further amplify motion sickness. A study performed by  showed that many activities that people foresee doing in autonomous vehicles would require focused visual attention – such as reading, social media and watching movies – and thus are prone to inducing motion sickness.
It is known that motion sickness can compromise task performance [34,35], and thus passengers’ ability to effectively make use of the time in their vehicle. Motion sickness can also result in after-effects that negatively affect an individual long after being exposed to the sickness-inducing situation . Thus, apart from being highly unpleasant, motion sickness can limit passengers’ ability to perform tasks both during- and post-trip which would limit the actual usefulness of selfdriving vehicles.
Taken together, motion sickness in AVs is a problem that must be addressed not to compromise the usefulness, acceptance and user experience of selfdriving vehicles.
This is not only important from the perspective of the companies offering the transport service, but also for those aiming at selling products and services to passengers through visual display interfaces within the vehicle.
Diels & Bos  propose design guidelines for minimizing the risk of motion sickness which are based on the principles of 1) allowing occupants to anticipate the future motion trajectory and 2) avoiding incongruent self-motion cues. First of all, the passengers should be able to look out, which can be enabled by having large window areas, minimizing the A- (and other- ) pillars, keeping the shoulder lines low and providing seats of sufficient height. Furthermore, rearward facing seats should be avoided to allow for passengers to anticipate motion in a good way. When it comes to displays, they should be located close to the line of sight out of the window and be limited in size, in order not to remove too much of the peripheral visual motion cues. A solution could also be to use head-up displays, which likely will solve the problem of reduced visual motion cues in an even better way. Finally, Diels & Bos suggest that one should avoid dynamic display content in situations with static velocity situations and vice versa, avoid static display content in varying velocity situations. It is unclear whether there is evidence that back up this last guideline, but there is at least evidence that adding visual cues that are congruent with vestibular motion cues can reduce motion sickness ( & references therein). Minimizing vestibular motion cues by e.g. avoiding extensive lateral and longitudinal accelerations (e.g. high-speed cornering, rapid braking and acceleration) which could be sources of sensory conflicts could be another way of limiting the motion sickness problem. Yet another, relatively simple, solution proposed by Larsson et al.  would be to provide the passenger with information about upcoming maneuvers so that he or she can prepare and look up to avoid visual/vestibular conflict. The information could be provided by subtle auditory cues as proposed by  or by visual or other signals (see the patent by Uber ). It has also been shown that voice messages that allow the passenger to anticipate upcoming motions could reduce motion sickness .
Dealing with filthy cars and pandemics
A topic which has not been widely discussed in the realm of highly automated, shared vehicles, is the question of how to keep them clean on the inside . A regular taxi or car, or bus for that sake, obviously has a driver who can monitor the interior, make sure it is tidy and return the car to the garage for a thorough cleaning when needed. But when the driver is removed, passengers might face the garbage and grime of previous passengers which may result in an unpleasant ride and indeed a very poor user experience.
Manually inspecting and cleaning automated cars at the fleet operations center on a regular basis could of course prevent this from happening, but may be a costly and time-consuming procedure. Relating back to the previous section on motion sickness, the interior can also become quickly soiled and unusable if e.g. someone becomes acutely car sick . Therefore, some kind of constant monitoring of the cleanliness of the interior is likely required. Uber suggests in their patent a robotic vacuum cleaner arm positioned somewhere in the middle of the vehicle which could automatically detect filthiness using cameras or other sensors and reach out and clean up the mess if needed . While these kinds of outlandish inventions probably won’t become standard, it shows that the problem is at least starting to be acknowledged.
Given that the cleaning needs to be done quickly and possibly also in an automated fashion, the need for an easily cleaned interior seems to be of essence. As opposed to normal passenger cars of today which have complex interior designs with many types of different materials and surfaces, simple layouts of durable and easily cleaned stainless steel and hard plastic surfaces will probably be a necessity in interior car design . The challenge for the car designer is then to create a compelling and comfortable interior using such easily cleaned surfaces. Easily accessible waste bins (constantly monitored and emptied at regular intervals, of course) seem also to be a key feature of an OAVS.
The question of keeping OAVS clean on the inside should of course be brought up high on the agenda as a consequence of the ongoing COVID-19 pandemic. Waymo has increased their cleaning routines  as has Baidu who reportedly provides vehicles that are disinfected before each ride . Baidu also screens passenger temperatures before letting them into the vehicle to prevent the disease from spreading according to . Interior surfaces of a robotaxi could be given antimicrobial coatings, although this only helps cutting the spread of bacterial illness, but has no effect on viruses . Automatic disinfection that combats both bacteria and viruses could possibly be solved using ultraviolet radiation (UV) emitters [45, 46] in the vehicle; obviously these should not be used not while there are passengers in the vehicle.
Developing a vehicle that can navigate by itself in dense city traffic is undoubtedly a huge technical challenge that just represents a first step in bringing an OAVS to the market. Achieving customer acceptance and adoption is a task that holds equally important challenges related to the potential users’ experience – some of which we have tried to identify in our series of articles [1, 2].
Making people trust the service, both in terms of driving safely and to the correct destination may be a problem in the initial phases of large scale introduction of OAVS. Once a general public acceptance for driverless vehicles has been established, achieving user trust in driving automation and its cybersecurity is probably less of a problem, provided that adequate actual safety and security standards can be achieved and maintained.
Creating interpersonal trust necessary to make people want to share rides is perhaps a more intricate problem. Lessons from other services from the sharing economy, such as Airbnb, could perhaps be transferred to the OAVS business – but as initial research indicates, this might be one of the more tricky problems to solve. Nonetheless, solving this problem will be crucial in order to reach the envisioned transportation system efficiency improvements and perhaps also necessary to make the services economically viable. Hopefully, the evolving services will find a common way to tackle the problem of creating interpersonal trust.
When trust in various forms has been established between users and between users and the service providers, how can you, as a service provider, create that unique selling point that attracts customers? Why would potential customers choose OAVS of Brand X over OAVS of Brand Y? Here we think that creating substantial usefulness of the service is key. One needs to keep in mind that these kinds of services initially compete with the regular car and therefore first of all has to offer a similar kind of usefulness as a regular car. In the longer run, the differentiation lies likely in the perceived value created by the OAVS, both in terms of efficiency of getting from A to B but also in terms of that the time spent in the vehicle is as valuable as possible. By giving the user freedom and ample support to perform whatever tasks they prefer – whether it is sleeping, working, socializing, or something completely different – and doing so without inducing motion sickness, true usefulness and of the service will be achieved.
Coming to this point will obviously take a lot of effort and careful considerations along the way. On top of this, recent evidence indicates that people’s general behaviors and preferences may change due to the ongoing COVID -19 pandemic which could limit the widespread adoption of OAVS. For example, a study involving more than 25,000 U.S. adults polled in the month of April 2020 showed that the respondents were planning to drastically reduce or forgo their use of ridesharing services as a result of the COVID-19 pandemic . As a result of the measures taken worldwide to limit the spread of the disease, people have also grown more accustomed to working at home and performing daily tasks such as grocery shopping online [47, 48], thus reducing the need for personal transportation. In the study by , a majority of the respondents indicated they would like to continue to work remotely at least occasionally, or even that they would like this to be their primary way of working.
Therefore, if one can assume that more and more people will continue working and performing other tasks from home even after the pandemic, the general need for personal transportation might decrease significantly in the future which would have a large impact on both potential OAVS businesses as well as on public transportation and traditional car use. This could of course also mean that people who no longer feel that they need a personal car to do their daily commute, could rely solely on on-demand transport services when taking those occasional trips – as long as they feel the services are trustworthy, safe and comfortable.
In sum, it is difficult to beforehand predict the user behaviors and needs in these types of radically new types of services, especially in a world characterized by uncertainty. It is nonetheless clear that the OAVS provider that is best in observing and analyzing their potential users’ preferences and behaviors in their interaction with the service and continuously adapting to the user needs and desires is the one that will succeed and attract the most users in the end.
Would you like to subscribe to the next article in the article series, User Experience of Shared Autonomous Vehicles Part 4? Click here and we will send it to you when it’s published.