Overview
This project combined my place of work and university and gained a high merit grade.
Examining how improved UX can help the user transition from conditional automated driving (Level 3) back to non-assisted driving (Level 0) - AUGUST 2022
Autonomous vehicles may in future replace human drivers, hopefully leading to fewer accidents. We currently have conditional automation and while it has its benefits there are a few issues around fear of the technology or alternatively over trusting the technology. Because cars have not reached fully self-driving capability, over trusting them can lead to accidents.
As autonomous vehicle technology is improving rapidly, there is also a need to improve the User Experience of autonomous vehicles. To also improve education around what these vehicles are capable of and improve safety and engagement around transitions between automated states.
Automated vehicle technology has received a lot of global attention and it is a rapidly evolving technology. Due to the rapid evolution, consumer understanding of the technology is low.
The Society of Automotive Engineers (SAE) have a set of driving automation levels, level 0 is no automation at all and requires the driver to always have full human interaction and attention, Level 5 is fully autonomous which will require no attention from the driver. Level 3 is when the automation features of the vehicle drive for the user under limited conditions, automation will not operate when all the conditions for automation are no longer met, and a human driver will need to resume control of the vehicle. (SAE, 2021).
Fully automated vehicles will not be widely available for some time, researchers predict with the best data available that full automation will take more than a decade. (Leonard, Mindell, and Stayton, 2020). The control of driving can currently be shifted between the human driver and the automated driving system, this is called conditional automation. Transitioning between level 3 automation and manual control can be a safety issue but the transition also affects the user experience. The performance of this transition has been a subject of rigorous research and has highlighted the need for new design solutions to improve drivers take over performance and enhance the user experience during this transition from automated back to manual driving. (Johansson et al., 2021).
I will discuss issues around consumer understanding and how this public knowledge of automation influences the user experience. It will cover some of the issues faced with the transition between automated states and how improved UX can help the user to safely and comfortably transition from conditional automated driving (Level 3) back to non-assisted driving (Level 0) As autonomous vehicles are in relatively early stages of development, fully autonomous vehicles are not currently available in the UK this study should provide some clarity around paths for UX going forward and improvements for conditional automation and what further research can be done.
methods
primary Research
- User Experience Survey (not using demographic data in this study as this can be analysed in further research)
- Expert Interviews
- User Interviews
Secondary Research
- Consumer Survey Data Thatcham Research (not using demographic data for this study as this can be analysed in further research)
- Literature Review
References
- See the bottom of the page for a full list
Data Sets
Have you driven a car with automated technology?
Do you understand the different levels of automation? (as defined by the sae levels of automation)
How safe would you feel in a vehicle with self driving capability?
Do you think there should be different levels of alerts based on how urgently human control needs to happen? ( for example emergency vs recommendation)
Would a reward system make you more likely to resume control of the vehicle within a set time frame? (a gamified system)
Do you think there should be a penalty for not taking back control of the vehicle within a specific time frame?
Do you think it is possible to purchase a car today that can drive itself? (By ‘drive itself’ we mean a car with technology that can drive the car completely autonomously, as safely as a competent human driver would, and allows you to remove your hands from the steering wheel)
When using a current Assisted Driving System such as Tesla Autopilot, Volvo Pilot Assist, Nissan ProPilot or BMW Active Driving Assistant, the driver is responsible for all driving? (Select one)
How comfortable, if at all, would you be driving a vehicle which has a Self-Driving capability, where the vehicle is performing continuous steering and speed control, if you had to be prepared to take over control instantly if something went wrong?
How would you spend your time in a vehicle that has a Self-Driving capability, where you as the driver can take your eyes off the road and the vehicle will handle everything that might occur whilst driving in situations that the Self-Driving capability can handle? (Tick all that apply)
What do you see as the key benefits, if any, of vehicles with Self-Driving or Autonomous technology?
What impact, if any, would an independent rating scheme for vehicles described as having Self-Driving Capability or full Autonomy have on your confidence to use the function?
Which of the following, if any, best predicts your actions when vehicles with Self-Driving capability are available? (Select one)
Which of the following, if any, describe your expected attitude when vehicles with Self-Driving capability become available and you could buy a car with one? (Tick all that apply)
Which, if any, of the following statements do you believe to be true? (Tick all that apply)
If you were to purchase a car with Self-Driving capability, what, if anything, would you miss about driving a manually operated car? (Tick all that apply)
The results from both data sets above showed a clear pattern in certain areas that will provide a clear path for discussion and ideas for further research. Coupled with interviewing industry experts, users of the technology and an in depth literature review I came to the following conclusions
Research Summary
AUTONOMOUS VEHICLES AND HUMAN FACTORS
Autonomous vehicles are becoming increasingly more capable, shifting the user’s role from active driving to passive monitoring of systems. As these technologies advance, consumer acceptance and adapting the way the driver interacts, depends on designing user experiences around trust, comfort, control and safety.
Avoiding unrealistic expectations and confusion around automation is important creating a user experience that clearly indicates the strengths and weaknesses of the current available technology. From traditional Human Factors perspective, advantages and downsides of the autonomous vehicle have been identified (K. Weir, 2015).
Changing from active participant to a passive observer and back is not something humans are efficient at. In fully automated vehicles the machine does all the work, so that switch between passive and active is not an issue as everything is automated, but with conditional automation it poses the risk of passive fatigue and creates a safety issue around allowing the car to automate easier tasks and needing human engagement during more difficult or emergency situations.
Data has confirmed that, over time, passive fatigue is expressed as decreasing task engagement. Furthermore, drivers in the passive condition had slower response times to unexpected events and were more likely to crash than those in the active and control conditions, we can see signs of it after only 10 minutes. (Saxby et al., 2007).
PASSIVE TO ACTIVE CONTROL TIMINGS.
A study conducted by University of Southampton calculated the length of time needed for drivers to switch from automated control to manual control for maximum safety.
This is an important consideration in system design to consider the time needed to take control of a vehicle. It suggests focusing on the range rather than average time needed for a person to switch. The reason for looking at range rather than a specific time is that there are so many variables in driving conditions and driver behaviours, narrowing it down to a specific lead time is very difficult to accurately predict.
Under non-critical conditions, drivers needed between 1.9 and 25.7 seconds to take control from automation. Such a large range reflects a variety of driver behaviour and environmental conditions.
(Eriksson and Stanton, 2017) Eriksson explained:
“Too short a lead time, for example seven seconds prior to taking control, as found in some studies of critical response time, could prevent drivers from responding optimally.
This results in a stressed transition process, whereby drivers may accidentally swerve, make sudden lane changes, or brake harshly. Such actions are acceptable in safety-critical scenarios when drivers may have to avoid a crash but could pose a safety hazard for other road users in non-critical situations."
Having such stressed responses in could potentially result in more accidents
CONSUMER UNDERSTANDING OF AUTONOMOUS VEHICLES
One of the main issues around automation is consumer understanding, creating confusion as well as over confidence in available technology shown, in the data above it found that 62% of drivers polled mistakenly believe that you can currently purchase a fully automated vehicle.
This sort of thinking could lead to safety issues as people could believe that they don’t need to pay as much attention when driving a car with automated technology. While many people polled believe that fully automated vehicles could be a benefit, safety concerns are an issue and 47% of drivers would be uncomfortable driving an automated vehicle and 21% would avoid purchasing an automated car completely.
Interestingly 74% of drivers would rather keep their eyes on the road when the first cars with self-driving capabilities like Automated Lane Keeping Systems (ALKS) are made available, despite the technology allowing them to take their eyes off the road.
This poses an interesting paradox of over confidence in current technology by consumers mistakenly thinking that fully automated cars are currently available, misunderstanding the current automation levels, but also having a mistrust of the current technology available and not feeling safe enough to allow the car to control certain aspects of driving due to not understanding the limitations and capabilities.
In more research conducted by IAM Road safety discovered 59% of drivers agree that the growing ability of vehicles to drive themselves is a serious risk to their personal safety. Opinions don’t vary much by age, but women were particularly concerned with 67% rating it as a threat. (Traffic Technology Today, 2022)
Education is the key to ensuring safety in future and the UK has given the green light to allow using systems such as ALKS (Automatic Lane Keeping Assist) on motorways.
Clear user interfaces, instructions and experiences are vital in supporting this education to ensure that when people take control of a vehicle, they fully understand its limitations and capabilities.
USER EXPERIENCE IN AUTONOMOUS VEHICLES
While automated vehicle technology moves forward psychologists and UX designers can play a huge role in making them as safe as possible. There are many challenges to designing UX for autonomous vehicles.
Drivers are not all one demographic they are different ages and experience levels with different levels of cognition, different temperaments, personalities and abilities. How to keep such a variety of users safely engaged is a difficult task. As vehicle manufacturers add more alerts, alarms and dashboards to communicate information to the driver, if this information is presented poorly, it could cause confusion or distraction.
Having a multitude of systems trying to communicate lots of information can become counterproductive very quickly. Self-driving cars roaming the streets will affect all of us and everything around us. Making this an opportunity to design for interactions that exist both inside and outside of AVs: riders, pedestrians, cyclists, animals, other vehicles, traffic conditions, and surrounding transportation infrastructure. (Punchcut, 2020)7
Researchers are looking at the best way to present drivers with information about driving conditions without distracting them.
“The last thing you want a driver doing in an event where they need to take over control is to not look at the road,”
says Gregory Fitch, PhD (Fitch, Bowman and Llaneras, 2014)
There are many challenges and opportunities that are apparent when designing UX for autonomous vehicles. Some of which are driving style measures. The five factors include Angry driving, Anxious driving, Dissociative driving, Distress-reduction driving, and Careful driving. Depending on different driving styles.
UX designers can use different approaches and strategies for example:
Angry Driving:
- Calm and Soothing Interfaces: Design interfaces that emit calming visuals, colors, and sounds to help reduce anger or frustration in aggressive drivers.
- Emergency Assistance: Provide clear and easily accessible emergency features to assist drivers in stressful situations, such as road rage incidents.
Anxious Driving:
- Clear Information Display: Present information and alerts in a clear, concise, and reassuring manner to reduce anxiety in nervous drivers.
- Adaptive Automation: Offer adjustable levels of automation, allowing anxious drivers to maintain some control and gradually adapt to autonomous features.
Dissociative Driving:
- Engagement Prompts: Implement features that periodically engage the driver, ensuring they remain aware of their surroundings and the driving task.
- Transition Alerts: Use clear and unmistakable alerts when transitioning from automated to manual driving to re-engage drivers effectively.
Distress-Reduction Driving:
- Comfort-Enhancing Features: Provide options for drivers to personalise the vehicle’s environment, such as temperature, lighting, and seat settings, to reduce stress.
- Adaptive Music/Entertainment: Offer stress-reducing audio content or music playlists and adapt these based on the driver’s stress levels.
Careful Driving:
- Advanced Safety Information: Display detailed information about the vehicle’s surroundings and safety systems to reassure cautious drivers.
- Predictive Alerts: Use predictive analytics to anticipate potential hazzards and proactively warn careful drivers about upcoming challenges.
Incorporating these strategies would allow UX designers to create interfaces and systems that accommodate a wide range of driving styles and user behaviours. Personalisation and adaptability are key principles when designing UX for autonomous vehicles, ensuring that the technology meets the unique needs and preferences of each driver, but as ever further research would be required in this area.
conclusion
One of the aims was discover how people would react to a reward and consequence system (gamification) to encourage safety as part of the UX in autonomous vehicles to see if that is something that could be used to influence the transition between automated levels. While people seemed to be more open to the consequences of not responding to request for reengaging, they did not seem to respond well to the idea of a fun reward system, stating it could make them feel like a child or patronised. One interview subject did comment “VW does a blue eco points thing. I like that and think that if you shared the car with someone, I.e., a partner, you could compete to be a safer or more eco driver based on the points” which could be an interesting concept to explore, but the consensus was that a lot of respondents would feel patronised or like a child despite the “fun” element so it would need to be investigated further to ensure if a system like that was in place it was “grown up enough” and not patronising.
The combination of Thatcham research’s set of data and my own has shown that there are clear areas of mistrust around the technology, education is needed to ensure consumers understanding of currently available technology. From a safety perspective in conditional automation, avoiding passive fatigue until AV’s are developed enough to be fully automated in all environments and to potentially consider automation from a different angle, instead of designing a system which is primarily automated during easier tasks, where the user only assists during an emergency, instead design systems to compensate for driver error and to potentially have automation as a safety net. (Apa.org, 2020).16
The research also implies, users understand the need for safety and consequences for not acting safely by taking back control of the vehicle in a timely manner, 39% even to go so far as to say a consequence should be to report it to insurance companies with one respondent during interview suggesting loss of licence if safety measures were not met.
When thinking about consumers feelings around safety, training people how to safely use their automated vehicle is one of the things that I pondered after conducting my research. The interviews gathered further data around which warnings and notifications people would prefer to grab their attention. One interviewee said “commercial aircraft use multiple warnings together – an alarm, a light, a voice saying “terrain! pull up!”. I think multiple warnings would be better especially if there were seconds before you hit a wall”. This comment led me to consider pilots who are trained for hours in simulators before they can fly alone so perhaps when purchasing a vehicle there could be a training mode for the initial stages of driving to get users used to their new system. I am not sure what consumer response would be, so that would require further research.
A seamless and intuitive UX becomes pivotal when the system requires the user to reassume control. A well-designed UX should provide clear and timely alerts, offering information about the state of the vehicle, road conditions, and potential hazzards, ensuring that the user remains alert and ready to intervene. By focusing on enhancing the user experience during these critical transitions, we can foster greater trust and acceptance of autonomous driving technologies, advancing us closer to a future of safe and efficient transportation.
The user’s cognitive disengagement from the driving process is a critical concern. So it should not only incorporate user-friendly interfaces but also training mechanisms that assist in the reacquisition of driving skills that may have become dormant during automated driving, ultimately ensuring a safe and smooth transition between automation levels. To address this, advanced UX design must prioritise clear and unambiguous communication. Visual, auditory, and haptic cues should be carefully crafted to inform the user when their intervention is needed. These cues should not only convey urgency but also provide context regarding the reason for the handover, such as adverse weather conditions, road construction, or technical limitations of the automation system.
The transition from automation to manual control demands a reactivation of the user’s driving skills, which may have become rusty due to prolonged periods of inactivity behind the wheel. Enhanced UX should incorporate these training elements, such as simulated scenarios, interactive guidance, and real-time feedback, to help users regain their confidence and competence in driving. This could include adaptive driver assistance systems that gradually relinquish control, allowing users to become accustomed to the dynamics of the road before full manual control is reinstated.
Additionally, fostering trust and user acceptance is pivotal. Transparency in system behaviour and decision-making processes is crucial for users to comprehend and trust automated systems. A well-designed UX should provide insights into the vehicle’s perception of the environment, the status of sensors, and the prediction of potential dangers. This not only keeps users informed but also encourages them to stay engaged and attentive.
Improving the UX for transitioning from Level 3 to Level 0 automation is a complex task that goes beyond mere interface design. It entails a comprehensive approach encompassing effective communication, skill reactivation, and trust-building measures. By addressing these aspects, we can ensure that users feel confident and secure as they return to non-assisted driving, ultimately promoting the safe and successful integration of autonomous vehicles into our transportation landscape.
Further work
Future research I would like to delve further into potential for initial “tutorial mode”. Secondly a system of consequences/rewards for conditional driving systems which could include having to reinstate “tutorial mode” if safety measures and reengagement rules are not met and thirdly to look at systems of automation that could support emergency situations to assist the driver in an emergency rather than handing control over and potentially increasing risk factors due to human error.
Future research should take into account demographic information, which I do have, but was not relevant for these initial stages of discovery. I would like to delve further into how to educate people on the systems available to them and how to provide consequences and rewards without patronising the user. By improving questions with more visual interface prototypes and with hypothetical situational scenarios along with observing users driving live vehicles with current automated technology, it will get a better picture and this research will provide the basis for future projects looking into the critical area of safety and gain a deeper understanding of what users would like to gain from and be comfortable with in conditional autonomous vehicles.
I would also like to ensure things like reaction times and the human factors are accurate as intuitive design is about timing and predicting when a user will need something. This will involve more in depth testing and task analysis.
From the next set of research data, low fidelity prototypes and user flows can begin to be built and tested in further research phases.
Future Research plan
Future Research Plan: Understanding and Enhancing User Experience (UX) in the Transition from Level 3 to Level 0 Autonomous Driving
Introduction:
The aim of this research plan is to address critical issues surrounding the user experience and safety aspects associated with transitioning from conditional automated driving (Level 3) back to non-assisted driving (Level 0) in autonomous vehicles. The rapid evolution of autonomous vehicle technology demands comprehensive research to improve user understanding, trust, and safety during this transition. This plan outlines the objectives, methods, and areas of investigation for a second phase of research, building upon existing knowledge and findings.
Research Objectives:
The primary objectives of this second phase of research are as follows:
- To further explore user perceptions and understanding of autonomous vehicle technology and its capabilities, addressing issues of overconfidence and misconceptions.
- To investigate the user experience during the transition from Level 3 to Level 0 automation, focusing on UX design, alerts, training mechanisms, and trust-building strategies.
- To examine the potential for implementing a “tutorial mode” for initial engagement and training.
- To explore the feasibility and acceptability of a consequences/rewards system for encouraging safe user behaviour within conditional automation.
- To investigate alternative automation strategies that support drivers during emergency situations rather than transferring control abruptly.
Research methods:
The research will employ a mixed-methods approach, combining quantitative and qualitative data collection techniques. The methods include:
- Surveys and Questionnaires: Conduct surveys to assess user understanding, perceptions, and preferences regarding autonomous vehicles. Gather demographic information to analyse variations in responses.
- Interviews, Focus Groups and Usability tests/task analysis in conditionally automated vehicles: Conduct in-depth interviews and focus group discussions to gain deeper insights into user experiences, concerns, and suggestions. Observe users in vehicles and how they interact with infotainment systems and react to changing automated states. Collect qualitative data on attitudes toward consequences/rewards systems and training modes.
- Prototype Testing: Develop and test visual interface prototypes to evaluate their effectiveness in conveying critical information during the transition from Level 3 to Level 0 automation. Use hypothetical situational scenarios to gauge user comfort and comprehension.
- Literature Review: Continue to review existing literature on autonomous vehicle technology, human factors, UX design, and safety aspects to inform the research and identify best practices.
Areas of Investigation:
The research will be divided into the following key areas:
- User Understanding and Perceptions: Investigate user knowledge and misconceptions regarding autonomous vehicle capabilities and limitations. Analyse how this understanding influences their attitudes and behaviours.
- UX Design and Transition Experience: Examine the impact of UX design on the transition experience. Evaluate the effectiveness of alerts, information presentation, and training mechanisms in enhancing user safety and comfort.
- Tutorial Mode: Explore the feasibility and user acceptance of a tutorial mode for initial engagement and training in autonomous vehicles. Assess its potential in improving user confidence and competence.
- Consequences/Rewards System: Investigate the feasibility of implementing a system that incentivises safe driving behaviours within conditional automation. Assess user perceptions and preferences regarding such a system.
- Alternative Automation Strategies: Research and propose alternative automation strategies that provide continuous support to drivers during emergencies, reducing the need for abrupt control transfers.
Expected outcomes:
This second phase of research aims to provide further insights into improving user experience, safety, and trust during the transition from Level 3 to Level 0 automation in autonomous vehicles.
The expected outcomes include:
- A deeper understanding of user perceptions, misconceptions, and attitudes towards autonomous vehicle technology.
- Enhanced UX design recommendations for a smoother and safer transition experience.
- Insights into the feasibility and acceptability of tutorial modes and consequences/rewards systems.
- Innovative ideas for alternative automation strategies to mitigate safety risks.
- Recommendations for educational initiatives to bridge the gap in user understanding.
Conclusion:
This research plan outlines the objectives, methods, and areas of investigation for a second phase of research focused on improving the user experience during the transition from conditional automated driving to non-assisted driving in autonomous vehicles. By addressing issues of user understanding, trust, and safety, this research aims to contribute to the safe and successful integration of autonomous vehicles into the transportation landscape.
Timeline:
- Month 1-2: Conduct literature review and refine research objectives.
- Month 3-4: Develop and distribute surveys/questionnaires.
- Month 5-6: Conduct interviews and focus groups and usability tests in current conditionally automated vehicles.
- Month 7-8: Develop visual interface prototypes and conduct prototype testing.
- Month 9-10: Analyse data and identify key findings.
- Month 11-12: Compile research results and recommendations into a comprehensive report.
Budget:
- Research personnel salaries and stipends
- Survey and interview software/tools
- Prototype development and testing resources
- Travel and logistics for focus group sessions
- Data analysis software
- Report publication and dissemination
Ethics and compliance:
Ensure that all research activities adhere to ethical guidelines and obtain necessary approvals from relevant institutional review boards or ethics committees.
Reporting and dissemination:
Publish research findings, present at conferences, and share results with relevant stakeholders in the autonomous vehicle industry to promote safe and user-friendly technology adoption.
references
Eriksson, A. and Stanton, N.A. (2017). Takeover Time in Highly Automated Vehicles: Noncritical Transitions to and From Manual Control. Human Factors: The Journal of the Human Factors and Ergonomics Society, 59(4), pp.689-705. do: 10.1177/0018720816685832.
Fitch, G.M., Bowman, D.S. and Llaneras, R.E. (2014). Distracted Driver Performance to Multiple Alerts in a Multiple-Conflict Scenario. Human Factors: The Journal of the Human Factors and Ergonomics Society, 56(8), pp.1497-1505. do: 10.1177/0018720814531785.
Flisher, T. (2022). Confused UK drivers believe they can buy a FULLY Autonomous car TODAY. [online] Thatcham. Available at: https://www.thatcham.org/trust-in-automation-study-findings-revealed [Accessed 20 Nov. 20221.
Matthews, G. (2002). Towards a transactional ergonomics for driver stress and fatigue. Theoretical Issues in Ergonomics Science, 3(2), pp. 195-211. do: 10.1080/14639220210124120.
Molar, L.J., Ryan, L.H., Pradhan, A.K., Eby, D.W., St. Louis, R.M. and Zakrajsek, J.S. (2018).
Understanding trust and acceptance of automated vehicles: An exploratory simulator study of transfer of control between automated and manual driving. Transportation Research Part F: Traffic Psychology and Behaviour, 58, pp.319-328. doi: 10.1016/j.trf.2018.06.004.
Punchcut (2020). UX Design for Autonomous Vehicles. [online] Medium. Available at:https://medium.com/punchcut/ux-design-for-autonomous-vehicles-9624c5a0a28f. [Accessed 26 Nov.2022].
Jeon, Myounghoon. (2015). UX challenges and opportunities of autonomous vehicles regarding driving styles and automation levels.
Johansson, M., Mullaart Söderholm, M., Novakazi, F. and Rydström, A. (2021). The Decline of User Experience in Transition from Automated Driving to Manual Driving. Information, 12(3), p. 126. do:10.3390/info12030126.
K. Weir, “The psychology behind self-driving cars: Along for the ride,” Monitor on Psychology, vol.46, no. 1, pp. 60-65, 2015.
Leonard, J.J., Mindell, D.A. and Stayton, E.L., 2020. Autonomous vehicles, mobility, and employment policy: the roads ahead. Massachusetts Inst. Technol., Cambridge, MA, Rep. RB02-2020.
SAE (2021). SAE Levels of Driving Automation TM Refined for Clarity and International Audience –
SAE Levels of Driving Automation TM Refined for Clarity and International Audience. [online] www.sae.org. Available at: https://www.sae.org/blog/sae-j3016-update.
Saxby, D.J., Matthews, G., Warm, J.S., Hitchcock, E.M. and Neubauer, C. (2013). Active and passive fatigue in simulated driving: Discriminating styles of workload regulation and their safety impacts. Journal of Experimentai Psychology: Applied, 19(4), pp.287-300. do: 10.1037/a0034386.
Schneider, T. et al. (2021) ‘Increasing the User Experience in Autonomous Driving through different feedback Modalities’, 26th Interational Conference on Intelligent User Interfaces [Preprint]. Available at: https://doi.org/10.1145/3397481.3450687.
Strayer, D.L. and Drews, F.A. (2007). Cell-Phone- Induced Driver Distraction. Current Directions in Psychological Science, 16(3), pp. 128-131. doi:10.1111/j.1467-8721.2007.00489.x.
Thatcham. (n.d.). Automated Driving. [online] Available at: https://www.thatcham.org/what-we-do automated-driving/ [Accessed 20 Nov. 2022].
Traffic Technology Today. (2022). New survey shows UK public not ready for autonomous vehicles. [online] Available at: https://www.traffictechnologytoday.com/news/autonomous-vehicles/new-survey-shows-uk-public-not-ready-for-autonomous-vehicles.html[Accessed 26 Nov. 2022].
www.southampton.ac.uk. (n.d.). ‘Takeover time’ in driverless cars crucial to safety | University of Southampton. [online] Available at: https://www.southampton.ac.uk/news/2017/01/driverless-cars.page.