Are we ready to 'handover' to driverless technology? - VENTURER Insurance and Legal Report 2017/18 - AXA UK
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Contents About VENTURER 04 Foreword 05 Chapter 1: 06 Introduction and summary – Legal and insurance implications of Trial 1 findings Chapter 2: 08 Background – Policy and legislative developments – Overview 08 – Industrial Strategy and CAV initiatives 08 Insurance framework and 09 legislative developments – Automated and Electric Vehicles Bill 09 – Civil Liability and insurance model 09 Chapter 3: 12 VENTURER Trial 1 results – Trial 1 – key results 12 – Safety 14 – Traffic Management 15 Chapter 4: 16 Handover: outlining potential issues – Liability in the handover phase 16 – Time lag in handover and regaining baseline control 18 – Unpredictable handover 20 – Conclusion 21 – Safety implications of the handover period 21 – Driver distraction 22 – Driver competence 23 – Safety of other road users 24 – Traffic implications 25 – Impact on underwriting 25 Chapter 5: 28 Perspectives on insurance and civil liability – The current model of civil liability in motor claims 28 – Government position on civil liability 29 – International dimensions 32 – Autonomous driving technology, 33 safe systems and liability 2 3
About Foreword VENTURER VENTURER brings together public sector, private sector This report is our second as part of the VENTURER project and draws and academic experts to understand the blockers and on the excellent work done by our academic partners as well as the enablers to wide scale adoption of Connected and emerging legal and insurance environment. Autonomous Vehicles (CAV).1 In the 2017 Autumn Budget Whilst some motor are both very important. The VENTURER trials are intended to develop understanding of the insurance and the Chancellor of the manufacturers have stated Once we get to Level 4 we have legal implications of increased vehicle autonomy. The project is now in its third Exchequer confidently their intention to go straight to true autonomy and insurance and final year and takes place in the Bristol and South Gloucestershire region. this higher level of automation, protection for the ‘driver’ as well stated that automated many cars will be coming to as their passengers, but this work vehicles (AV) will be on market in the coming months and these reports have relevance the UK’s roads by 2021. and years which will be Level and importance not just when This should excite everyone 3 and capable of allowing the we reach that end state, but who believes in the positive driver to hand over control to at every stage of the exciting societal impact that these the vehicle, but only in certain journey as things progress. VENTURER is made up of organisations from various sectors: vehicles could have, saving circumstances. Should the driver lives and offering mobility wish to regain control or should solutions to those who are the vehicle identify a reason for currently unable to drive. handing back control, this will take place as part of the dynamic driving task i.e. whilst the car We need to ensure, however, is still moving. It’s important that we are as clear as we can to realise, therefore, that we be about how we define the are discussing in this context David Williams, terminologies surrounding high-level driver assist features. this technology. People must Technical Director, With these, fundamentally, AXA understand what the vehicles the driver remains responsible are capable of and, very for the vehicle in the eyes of importantly, what the law the law and potentially liable allows us to do (or not do) from an insurance perspective when travelling in them. in the event of an accident. For example, the Automated This is what makes the findings and Electric Vehicles Bill that of VENTURER’s Trial 1 results is currently making its way so fascinating. Those trials have through Parliament states a real-world application now as Chris Jackson, that the Secretary of State will the handover element will be an Head of Transport Sector, create a list of vehicles that will increasingly common feature on Burges Salmon LLP be deemed to be “automated”. many vehicles in the near future. As it stands, we anticipate that How we, as drivers, react in those this definition is likely to fall circumstances and how motor within SAE Level 4 (the most manufacturers incorporate the widely used definition of the technology in the first place to various stages of autonomy).2 take account of that behaviour 1 VENTURER: A New and exciting autonomous vehicle project, http://www.venturer-cars.com/ 2 SAE International, Automated Driving - http://www.sae.org/misc/pdfs/automated_driving.pdf 1 VENTURER: A New and exciting autonomous vehicle project, 2 SAE International, Automated Driving - http://www.venturer-cars.com/ http://www.sae.org/misc/pdfs/automated_driving.pdf 4 5
Chapter 1 Introduction and summary – Legal and insurance implications of Trial 1 findings The VENTURER consortium incidents arising as a result From a legal and insurance The regulatory background of drivers as issues around is expected of them to support has reported separately of system failures whilst a perspective, the Trial 1 results defining requirements for allocation of liability between uptake of the technology and the results of Trial 1, in vehicle is in autonomous confirm that the allocation of handover of control will insurers, drivers and OEMs are underpin the development of which it investigated the mode. This is intended to liability that accompanies the also need to reflect human resolved by the courts on an autonomous driving systems transfer of control between factors. OEMs, insurers and emerging case by case basis. which are safe by design. handover of control from avoid the procedural issues, system and human driver consumers will all benefit from an autonomous driving delay and complexity which needs to be considered clear standards informed by For these reasons handover It is also clear that all the system to a human driver.3 would result from affected carefully. Increased attention rigorous experimental and is a live issue. The results work being done in the UK must This report considers the third party motorists needs to be paid to the empirically-obtained data of VENTURER Trial 1 and be seen in an international legal and insurance aspects otherwise being faced with consequences of automation around how human drivers other similar trials should be context, with the UN Economic of handover in the context the prospect of claiming for fault-based systems of interact with automated driving considered for relevance to Commission for Europe of the VENTURER Trial 1 against original equipment negligence. Proper regard must systems for the purposes of the development of a suitable (UNECE) actively looking results and Government manufacturers (OEMs) for be had to the difficulties for passing control between an handover protocol as part of the at whether the handover proposals for insurance of accidents caused by product human drivers in responding autonomous driving system and Government’s drive to enable the process should be entrenched where an automated system a human driver. The standards introduction of AV technology in the system at the autonomous (or ‘automated’) failure. We are supportive of passes them control of a vehicle introduced need to reflect the to UK roads, understanding that manufacturing stage. vehicles.4 Government has Government’s proposals for the immediate application will and the circumstances in capabilities it is reasonable for signalled its willingness to legislative reform to facilitate which they will be considered us to expect of human drivers, apply to SAE Level 3 vehicles extend the current regime an insurance regime for liable for events around and respect the limitations and form part of Advanced of driver insurance to autonomous vehicles. the handover period. of human performance by Driving Assist Systems (ADAS). designing in safety. This in Standards need to afford drivers turn will inform expectations sufficient certainty as to what 3 VENTURER – Trial 1 Findings are available at http://www.venturer-cars.com/. This report is one of a series produced by the VENTURER partners and is the follow-up to the first AXA Insurance Report produced in July 2016 (also available at http://www.venturer-cars.com/). 4 “Automated” vehicles being the term used in the Vehicle Transport and Aviation Bill, previously introduced as the Modern Transport Bill and now intended to be introduced to Parliament as the Automated and Electric Vehicles Bill. 6 7
Chapter 2 Background – Policy and legislative developments Insurance framework and legislative developments Government has recognised the This was followed on 7 September relies on an automated system The purpose of this is to obviate Overview potential benefits and taken a 2017 when the Government launched Automated and and is not therefore directly the need for an affected third The House of Lords, Science proactive approach to CAVs and has MERIDIAN, aimed at creating a Electric Vehicles Bill in control of the vehicle. party to seek redress from an made it a policy priority to position cluster of excellence along the OEM, maintainer or other third and Technology Select the UK a global leader in this area. M40 corridor between Coventry The Automated and Electric The Government chose to party who might have some Committee published a and London and accelerating the Vehicles Bill, originally address this in Part 1 of the degree of responsibility for the report on ‘Connected and This activity has included: development of CAV technology, announced in the Queen’s Automated and Electric defective performance of the Autonomous Vehicles: The growing intellectual capital and Speech following the election, Vehicles Bill by defining a automated driving system, future?’ on 15 March 2017 • Providing funding for in excess of attracting overseas investment. was published in October framework for how CAVs where liability is ultimately with outlining the potential uses twenty driverless car projects; 2017. The Bill replaces the involved in accidents will be the manufacturer of the vehicle, and benefits of connected Cementing all the above, the previous Vehicle Technology treated for insurance purposes. the company that programmed • Publication of the DfT code of Government published its Industrial This proposes to extend the the algorithm operating the and Aviation Bill, which failed and autonomous vehicles: practice for testing driverless Strategy White Paper7 in November to complete its legislative requirement on the insurer to vehicle in autonomous mode cars and ‘The Pathway to 2017 which committed the UK to pay out to affected third parties or a person who in some passage through Parliament Driverless Cars: Summary becoming a world leader in the way where the system, rather than other way is responsible for due to the General Election on report and action plan’; people, goods and services move. the driver, is at fault. the system’s failure. 8 June 2017. The substantive The possible applications In the paper the Government stated • Establishing the Centre for clauses are nearly identical of connected and Connected and Automated its intention to see fully self-driving to those put forward in the autonomous vehicles Vehicles (CCAV); and cars, without a human operator, on previous iteration of the Bill. (CAVs) are far-reaching, UK roads by 2021. It will therefore • Introducing legislative proposals We expect automated cars to appear from the 2020s. straddling a variety of make ‘world-leading changes’ to The Bill aims to support to modernise relevant insurance They present an enormous opportunity for the UK: different sectors. The the regulatory framework, including innovation in self-driving legislation, through the Automated securing high quality jobs and investment; creating new examples provided in updating the code of practice for technology in the UK and to and Electric Vehicles Bill (which mobility solutions that can transform lives; and, as I said our evidence included— testing automated vehicles to ensure that the UK remains a replaces the Vehicle Technology earlier, improving road safety. In 2016, human error but were not limited allow developers to apply to test world leader in new industries. and Aviation Bill which was was responsible for a very significant proportion of all to—aerial, marine, public their vehicles nationwide without It contains two parts: dropped as a result of the reported accidents. Automated cars will radically change roads, private and public a human safety operator and Part 1, which sets out the General Election 2017).6 carrying out a project with the Law that. To support consumers and businesses involved in transport (including broad parameters of how Commission to set out proposals for automated vehicles involved automated vehicle accidents, they will need an insurance metro and rail), space, Industrial Strategy and a long-term regulatory framework in accidents will be treated framework that is fit for purpose. Currently, they may not military, warehousing, for self-driving vehicles. for insurance purposes; and be covered for collisions that result from vehicle failure, ambulance services, CAV initiatives Part 2, which covers electric because in the UK only the driver is insured. Victims might precision agriculture, vehicles and charging. have to take vehicle makers to court, which would be inspection and As part of the Government’s time-consuming and expensive, undermining the quick monitoring of resources, long-term strategy in this area, and easy access to compensation that is a cornerstone working in dangerous and it announced a new ‘Connected Civil Liability and of our insurance system. Not tackling this problem risks hazardous environments and Autonomous Vehicle Hub’ (such as nuclear facilities) or ‘CAV Hub’ at Loughborough’s insurance model jeopardising consumer protection and undermining the automotive industry’s competitiveness.8 and the delivery of Olympic Park Campus, which will Under current legislation, humanitarian supplies.5 act as a co-ordination centre for the insurers of the driver who – Rt Hon John Hayes MP, Minister of State for Transport Legislation the development of autonomous is at fault pay out to third vehicles in the UK, as well as looking parties who have suffered at the wider framework to support damage. The question then CAVs, such as data management, arises as to where liability cyber-security and insurance. should sit where a driver 5 House of Lords Science and Technology Select Committee, Connected and Autonomous Vehicles: The future? (15 March 2017), p.11 https://www.publications.parliament.uk/pa/ld201617/ldselect/ldsctech/115/115.pdf 6 HMT, The Queen’s Speech 2017, The Automated and Electric Vehicles Bill, p. 27. 8 Second reading of the Automated and Electric Vehicles Bill (23 October 2017) 7 https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/662508/industrial-strategy-white-paper.pdf https://hansard.parliament.uk/commons/2017-10-23/debates/BDAB60DC-D67C-44CF-B0CB-9FBE8DAE3F30/AutomatedAndElectricVehiclesBill 8 9
Chapter 2 This represents the minimum Additionally, in a situation where The question of liability step possible to accommodate there is a product failure when becomes more complex automated vehicles within the vehicle is in an autonomous when we consider the the existing insurance function, the driver may be handover point between framework, which speaks to the the victim of a personal injury autonomous and manual consequences of fault through that he or she did not cause. driving phases. Determining reliance on the insurance policy Under existing legislation, a at what point fault lies either of the driver. For insurance driver’s insurance only needs with the driver (driver error) purposes, the system will to cover third parties and not or with the system (product/ effectively ‘become’ the driver the driver themselves. After maintainer error), will be critical under these proposals (to the extensive consultation, the in determining where liability extent the behaviour of the Government has decided to and financial responsibility, system triggers a claim), with the widen insurance cover so that and potentially criminal liability, insurer then separately being this includes damage to the ultimately rests. This issue entitled to pursue the OEM or driver where the automated has also been raised during other third party responsible vehicle is driving itself. The the ongoing passage of the for the system’s failure to intention behind the legislation Automated and Electric recover any compensation is to emphasise that if there is Vehicles Bill. We look at this which the insurer has paid to an insurance ‘event’ (accident) in chapter 5 of this report. affected third parties. the compensation route for the individual remains within the motor insurance settlement framework, rather than through a product liability framework against a manufacturer.9 Summary Part 1 of the Bill (clauses 1–7) for third parties for personal addresses the insurance issues injury or property damage due to that will arise when responsibility a driving related incident. Such for a vehicle is shared between principles need to be extended the driver and the system itself. to cover automated vehicles. The application of ‘intelligence’ The Government believes to cars is gathering pace and there that answering the insurance is a strong push by manufacturers questions sooner rather to develop automated vehicles, than later will encourage which will drive themselves. manufacturers to develop Currently it is a requirement that transport technology in the all (human) drivers have to have United Kingdom with the insurance when they drive in confidence that they can order to provide compensation exploit market opportunities.10 9 The House of Commons Library, Automated and Electric vehicles Bill 2017-19 (20 October 2017) p.3 10 Ibid, p.9 10 11
Chapter 3 VENTURER Trial 1 results As we outlined in the first VENTURER insurance report,11 switching control between a human VENTURER Trial 1 – key results driver and an autonomous driving system raises difficult questions around the allocation The following conclusions drawn by the investigators are of particular relevance of liability. The proper allocation of responsibility and liability between the system and from the perspective of allocation of liability around the handover event: the driver requires a clear understanding of the handover process, when and how control switches to the driver and vice versa. The findings from the driving simulator study, supported by VENTURER Trial 1 involved performance, the initial literature The results of Trial 1 indicated similar experimental conditions used in the road study, suggest simulator and road experiments at review undertaken by the better performance in frequent that designers of AV technology with handover functionality Bristol Robotics Laboratory on the VENTURER team having identified handover scenarios than might need to proceed with caution. The experiments highlight the University of the West of England that studies to date had focused have been expected by reference need to consider human performance under multiple driving campus. The experiments tested on driver performance following to earlier studies. However, delays conditions and scenarios in order to plot accurate takeover drivers’ ability to retake control long periods of disengagement in regaining control of various and handover time safety curves.14 following a period of automated from the driving task. The trials lengths were observed at different driving and measured takeover also sought to cover a wider and speeds, with other effects on time (the time it took participants slower range of speeds and to baseline driving being observed. to put their hands back on the involve less experienced drivers These included slower driving by The results of Trial 1 confirmed the difficulties for human drivers executing the controls) and the handover than in previous studies. Using participants following handover handover task. This highlights the need for a rigorous evidence base informing the period (the total time to achieve a road vehicle alongside a simulator and a marked delay in achieving development of appropriate standards around handover. a level of control consistent with the scenario parameters baseline performance when withbaseline indicators). set to replicate the road vehicle, retaking control at speeds ranging combined with a simulator from 20-50mph. Previous studies have shown environment which represented that drivers experience a delay the real-world test circuit, and Full details of the Trial 1 results Handover process in regaining effective control a within-subjects design for the and a separate summary following handover of control road and simulator conditions, document can be found Autonomous from the system.12 Trial 1 was amounted to a novel experimental viewed online.13 Transition Manual driving conceived to test the effect of design for CAV research. driving more frequent handover on driver 1. Handover request to transfer control from autonomous driving to manual driving 2. Takeover time from 3. Handover period driver request to contact with performance measured as controls significantly different to ‘baseline’ 4. Total handover time = takeover time plus handover period 5. ‘Baseline’ manual driving achieved 11 VENTURER, AXA Annual Report 2016.(https://www.axa.co.uk/uploadedFiles/Content/Newsroom_v2/Media_Resources/Reports_and_Publications/Downloads/ Driverless_Cars/VENTURER%20-%20AXA%20Annual%20Report%202016%20FINAL.pdf) 12 Morgan, P., Alford, C. and Parkhurst, G. (2016) Handover issues in autonomous driving: A literature review. Project Report. University of the West of England, Bristol, UK. Available from: http://eprints.uwe.ac.uk/29167 13 http://www.venturer-cars.com/. 14 VENTURER - Trial 1 Summary, page 3 http://www.venturer-cars.com/ 12 13
Chapter 3 Safety Traffic Management For a safety-critical system, In the development of standards underlines the need for a process the average response time is and assessment of liability, around handover, and does not It is important that driver assistance features such as ‘lane limited as a valid measure: the due regard will be needed as to support a liability model which keeping’ are retained and functioning within vehicles, system also needs to be able to how a wide range of individuals expects a human driver to take particularly during handover. Rather than handover being account for the slowest expected perform. As a minimum, these responsibility immediately when from autonomous system to human driver, in practice it might responder. Indeed, this would be findings emphasise the need for a previously disengaged from the be from autonomous system to human driver supported a failsafe in the case of an upper managed process which respects driving task. by driver assistance. acceptable limit to takeover. slower responders: it may be the It is reasonably likely that the case that a lower bound criterion Some commentators see high-speed, limited access roads as the extremes of driver behaviour and based on a statistical analysis most natural first niche for AVs. Where handover is concerned, performance during handover needs to be set to capture the however, the findings suggest that lower speeds, similar to those have not been measured in Trial 1 majority of responses without in urban areas, are moderately safer. It may be that AV systems given the modest sample size. trying to mitigate for all. It also should follow procedures to slow the vehicle to a lower, safer speed, such as 40.16 An additional finding from the road trial that was not evident Some commentators see high- These findings underline the need a system in which safety is not from the simulator trial was the time lag between accepting speed, limited access roads as the to design safety into the handover dependent on an individual’s responsibility for control and beginning to take control. most natural first niche for AVs. procedure. Acknowledging split second decision-making Where handover is concerned, and accommodating driver capability is likely to result in This represents a risk of a different nature, during which however, the findings suggest that shortcomings will be necessary to safer journeys. the autonomous system has ceded control, the human has lower speeds, similar to those in reduce the scope for human error. signalled acceptance, but in practice he or she has not exerted urban areas, are moderately safer. This type of systems approach More broadly, the VENTURER control and could not be regarded as being in control. It would It may be that AV systems should draws on the lessons learned Trial 1 results and the team’s therefore be important that future handover design relies not follow procedures to slow the in the development of modern analysis indicate that further on the human signal for the passage of control, but evidence vehicle to a lower, safer speed, safety culture and acknowledges technical investigation will be such as 40mph.16 that individuals may not perform needed before it is possible to of active input into the human-machine interface by well in safety critical situations. define an appropriate use case the human driver.15 To the extent possible, designing and supporting liability model. Noting that the UK has not when it indicated it considered may wish to consider how best to ratified the Vienna Convention, the human was not performing minimise unnecessary handovers. Mean time to takeover STISIM controls following international standardisation is well enough. Designers and regulators will also a handover request strongly desirable where possible wish to have regard to the limited This may be an area for regulators and ultimate driver responsibility human capability to manage to assess in the creation of for control reflects the traditional the transition between regulations or guidance around liability model. A requirement controlled and uncontrolled for evidence of active input (and autonomous driving – e.g. where environments. 3 Takeover time (seconds) a vehicle is in autonomous mode, any assessment of the quality of input) raises interesting liability the current law would mean 2.5 that drivers should be obliged considerations. Query for instance to remain engaged to retain 2 how a requirement for the system a degree of ‘feel’ for the vehicle to validate that the input of the and awareness of its surroundings, human driver is of sufficient even if not actively controlling the 1.5 quality sits with the requirement vehicle (i.e. when driving using for human override assumed by ADAS in a Level 3 vehicle). At level 1 the Vienna Convention, or the 4, it may be very difficult to require potential difficulty of determining a human to retain this degree 0.5 liability consequences if a human of ‘feel’, and as such regulators driver were to override the system 0 20-mph 30-mph 40-mph 50-mph 15 VENTURER – Trial 1 Summary, page 3 http://www.venturer-cars.com/ 16 Ibid. 14 15
Chapter 4 Handover: outlining potential issues The ‘handover problem’ arises more as a feature of the development of autonomous vehicles, which are ‘high-level driver assist’ but not ‘highly or fully’ autonomous: where drivers do not need to engage with a range of tasks in the driving process when the autonomous mode is selected, but do need to be able to take control when necessary. Manufacturers have identified this of central importance in the development of autonomous vehicles, with Mercedes-Benz commenting that this is the most difficult stage of the transition to autonomous vehicles17 and Ford intending to bypass this stage completely and move directly to level four automation.18 The VENTURER Trial 1 demonstrates that handover continues to present difficulties in the context of more frequent transfers of control between vehicle and driver. Both industry and the Government will need to work to find effective solutions to some of the issues raised below, in order to encourage sufficient public confidence in the technology and therefore support wider uptake. Liability in the handover phase As discussed in Chapter 2, driver involved in an accident with when compensation is payable. Government’s approach to that of a reasonable driver. The Civil liability is a compensatory the development of insurance basis of the doctrine therefore concept. Clearly however the law regulation (as set out in the requires a baseline of what a will in due course need to look Automated and Electric Vehicles reasonable driver is (objectively) at what behaviours and actions Bill) assumes that insurers required and not required to do. are to be defined as constituting would pursue rights of recovery That is well known and long- fault for the purposes of criminal if negligence could be proven established in conventional law. The criminal law is not about in respect of some other party driving. No equivalent baseline compensation but about sanction or the vehicle manufacturer. currently exists for a new situation by the state against individuals However, the handover period – that of handover. These trials and organisations for conduct currently represents an area and the further work which will falling below the level specified as of uncertainty as to where follow on from them will need required. That issue will be looked to apportion liability. to inform that baseline. at in our third and final report. The issue is that negligence as This report focuses on the civil a concept is defined objectively liability and insurance position by comparing the actions of the – which in turn determines 17 Wheels Mag, ‘Mercedes-Benz autonomous tech hits handover speed hump’ (July 2016) https://www.wheelsmag.com.au/news/1607/mercedes-benz-autonomous tech-hits-handover-speed-hump. 18 BBC news, “Ford’s self-driving car ‘coming in 2021’’ (August 2016) http://www.bbc.co.uk/news/technology-37103159. 16 17
Chapter 4 Time lag in handover and regaining baseline control Assuming that an autonomous in completion of an effective either the development or the vehicle was involved in an handover process presents a take-up of the technology. This incident causing damage to complication in respect of the then raises the question of how another road user while in fully basic liability model: where in to judge a reasonable length of autonomous mode due to a the handover period should the time for handover, and the extent malfunction of the vehicle’s human driver’s responsibility end to which the handover process technology, insurers would and the insurers’ right of recovery should be standardised. (if having to pay out to an against the OEM start? affected third party under the Government has signalled its Government’s proposed new It would appear unreasonable intention to designate certain legislation) then be able to to expect people to be able to vehicles, or categories of pursue rights of recovery from disengage from the driving task vehicles, as ‘automated’ vehicles. the vehicle manufacturer (or only to have to resume control Assessing whether a vehicle other person responsible for of the vehicle immediately. On is “automated” (i.e. capable the vehicle or system failure). the other hand, it would also of driving in autonomous appear impracticable, from mode and benefiting from Contrast this with the position the perspective of developing the insurance arrangements where a driver is engaged in the good driving habits as well described in the Automated and driving task and responsible for as encouraging effective Electric Vehicles Bill) will require the vehicle where, if an incident manufacturer and supply chain definition by reference to some arises due to a mistake of the engagement in designing form of minimum standards. driver, he or she (and his/her systems which seek an optimum insurers) will be liable. balance between safety and performance, to expect OEMs The VENTURER Trial 1 (or other potentially liable experiments found time lags in third parties, such as software drivers retaking control following providers or system maintainers) handover from the automated to create a handover system that system (e.g., of just under three took an undue length of time We recommend that as part acceptable handover protocol. liability will rely on case by seconds at a speed of 20mph), once the vehicle had indicated of the process of defining SAE Level 4 should arguably only case development of the law together with other delays in the intention of handing back standards for automated request dynamic driving task – it will be for the courts to returning to baseline driving control to the driver. A balance vehicles consideration is given handover when not safety critical determine where the limits of performance, particularly at between the two will need to be to how an effective handover and a suitable and acceptable human control end and where higher speeds. This time lag struck so as not to discourage protocol may look, the process handover protocol is in place the liability of OEMs begins through which control is handed to ensure that handover, if (in much the same way as the over to the driver (including accepted, is not only safe and courts establish whether a driver probably an element of active effective but potentially also is at fault now). In assessing monitoring and feedback) and optimal (i.e. it could be the same where these boundaries lie, we the time periods for a driver to as the Level 3 normal operation would expect the courts to pay be able to retake control. handover protocol or even close regard to the difficulties more generous to give drivers of handover, including as This may include establishing even more notice and time to demonstrated by the results different conditions and assume control in full command of VENTURER Trial 1. procedures for handover at and a state of readiness to SAE Level 3 and SAE Level 4. drive progressively). SAE Level 3 hands off operation should only have a realistic Notwithstanding the expectation of handing over to a development of standards, driver in normal operation mode defining the boundaries under a suitably established and between driver and system 18 19
Chapter 4 Unpredictable handover Conclusion VENTURER Trial 1 tested the result of issues with the Further trials on an unpredicted The Government, manufacturers, around when a driver (or an OEM) systems and processes, handover period when the design of the system or how handover may provide data insurers and drivers will want may be criminally liable for including an appreciation driver knew he or she might be the system performs, would on the time lag for regaining to ensure that the allocation mis-managing handover.) of what it means to be a skilled alerted to take control in certain insurers seek rights of recovery effective control in an unpredicted of responsibility between driver driver in an increasingly automated situations, described as ‘predicted against manufacturers? Secondly, handover to assist with this and system during the handover More broadly, as has been world, in order for the allocation handover’. The trial did not test in the event that the driver is final question. However, given phase is fair and proportionate to observed in similar contexts, of liability to remain fair and ‘unpredictable’, unplanned not able to take back control or the difficulty of takeover on a reflect the capabilities of users. where a system is designed not manufacturers incentivised handover situations where the handover is not executed planned basis, it must follow Any other approach risks stifling to require the input of a human to deliver safe systems. a vehicle might suddenly properly, does responsibility that unplanned takeover is yet the appeal of the technology being (or to require limited input), have a technical fault or lie with the driver for this failure more problematical from the and unfairly penalising drivers. if that input is required it is not In the following sections of other issues. In the case of – or with the manufacturer as human driver’s perspective and Separately, a review of the likely to be of high quality.19 this chapter we look at other an unpredictable handover, a result of the failure of the less suitable as a basis for fixing relevant road traffic law and Accommodating the increased outcomes of Trial 1 which the process can be expected to system causing the unpredicted the driver with responsibility guidance for drivers will be difficulties of handover may reinforce the need to view occur in the context of a crisis. handover? Thirdly, assuming for an incident. required (e.g. as to the extent to require development of a more handover in the context the driver is able to regain which clarification is required nuanced understanding of safety of system design. Given that there is a time lag control, what is a ‘reasonable’ This links again to the need in regaining effective control length of time for handover going forward (assuming that in a predicted handover, it can in such a scenario and again, negligence will remain a part be expected that there will also be a time lag in an unpredicted should this be standardised? Finally, consideration will need of the legal attribution of civil liability) to establish a new Safety implications of the handover period handover, and possibly one to be given to liability issues baseline for negligence. of a greater length. This raises in the event that handover is Distance travelled 50mph do not compensate for time to ensure it is safe. Due to the questions for both manufacturers completed successfully but during handover the greater distance travelled at limited sample size in VENTURER and insurers. Firstly, where the the driver is then unable to higher speeds. At 50mph, a vehicle Trial 1, it is reasonably likely that handover is unpredicted as a manage the stricken vehicle. The findings from the STISIM will be travelling at 22.5 metres the extremes of human takeover simulator, supported for some per second, so given the average and handover performance have experimental conditions by the time of two seconds for takeover, not been captured in the data. on-road Wildcat, suggest that the it will have travelled a distance designers of highly autonomous equivalent to 45 metres, or half To solve this problem and ensure vehicle technology will need to a full football field, two lengths that highly autonomous vehicles consider human performance of a typical swimming pool are safe, the system may need under multiple driving conditions or a row of nine parked cars to incorporate speed dependent and scenarios in order to plot before the driver actually begins phased handover periods, accurate takeover and handover to manipulate the vehicle or require the vehicle speed time safety curves. controls.20 Moreover, basing these to be automatically reduced to estimations on average takeover a manageable safe speed before In particular, the differences in times is limited, as any system handover is attempted. 21 faster response to take control would also need to account for shown between 30, 40 and the slowest expected takeover The American Society of Safety Engineers, “Automation vs Human Intervention, What is the Best Fit for the Best Performance” 19 http://www.asse.org/practicespecialties/management/automation_human_intervention/. 20 VENTURER – Trial 1 Findings http://www.venturer-cars.com/. 21 Ibid. 20 21
Chapter 4 While most behaviours measured Given that, following rollout of the To mitigate this risk, it may in the STISIM simulator technology, a scenario could arise be possible to employ driver experiment showed greater where large numbers of vehicles assistance features, such as caution following takeover, the on multilane roads could be active lane keeping, while in findings relating to steering input transitioning from autonomous to human driven mode. This would at all speeds suggested a reduced human control at approximately mean that rather than viewing level of control. In particular, the the same time within the same the handover period as a move positioning within the lane at the space, these behaviours could be from autonomous system to highest speed condition (50mph) problematic. Looking at the most human driver, in practice it might is cause for concern. extreme scenario, there is the be from autonomous system to potential for two human drivers supported human driver, or in to drift lanes during this period, other words a move between endangering each other and levels of automation.22 potentially causing a collision. Driver distraction One of the cited socio-economic even if they are aware that car Chancellor for Research and benefits of autonomous vehicles may hand over control to them. Knowledge Exchange and is that people will regain time Evidence provided to the House Professor of Human Factors at otherwise lost to driving and will of Lords Science and Technology the University of Nottingham, be able to engage in other more Committee inquiry into notes, “it is therefore important productive activities such as autonomous vehicles is relevant to understand the implications sleeping, reading, watching films/ in this context. As Professor of increased autonomy on the TV or replying to emails while Neville Stanton explained, capability of humans to maintain the car is driving autonomously. “even the most observant human vigilance and attention in order It is clear that this benefit is only driver’s attention will begin to to be able to respond to an realistic when fully autonomous wane. Their mind will wander.”23 emergency situation. It may also vehicles that are capable of A driver may well fall asleep be necessary for the rollout of coping with any circumstance by accident or get distracted by highly autonomous vehicles they encounter are achieved, as engaging in conversations with to be accompanied with the Driver competence any handover period in a highly other passengers – these are advice – or even law – that in autonomous vehicle would not problems that can occur with some or all circumstances the As well as risks with driver This could be addressed by there is a need to maintain be safe if the driver was otherwise a manual vehicle so are arguably driver must maintain attention concentration, there are also strategies to ensure sufficient the understanding that people engaged and therefore the time more likely with a Level 3 vehicle to the driver situation and potential negative implications driver exposure to maintain skills. have “an appropriate level of lapse for takeover is slower. where less concentration that other activities should be of highly autonomous vehicles When giving evidence to the competence through a driving is required when the car is minimised or avoided.” 24 for drivers’ competence. Drivers Lords Science and Technology test”, and that there is a need to However, it is within the realm in autonomous mode. could become complacent Committee, Professor Natasha consider whether any such driving of possibility that a driver’s and over-reliant on technology Merat from the Institute for test includes an understanding of attention could wane when As Professor Sarah Sharples, as they get used to driving in Transport Studies at the University how an autonomous vehicle will a car is in autonomous mode Associate Faculty Pro-Vice- autonomous mode, creating of Leeds suggested that there behave rather than just focusing the problem of ‘de-skilling’, should be a system of driver on control of the vehicle.27 particularly in terms of a reduction licencing for autonomous in ‘situational awareness’.25 Given vehicles, as well as a need for that a driver may need to take driver training, including for back control of the vehicle, even those drivers who already have with a fully autonomous vehicle, licences for conventional vehicles. 26 this could be problematic. Professor Sharples added that 22 Ibid. Morgan, P., Alford, C. and Parkhurst, G. (2016) Handover issues in autonomous driving: A literature review. Project Report. University of the West of 25 England, Bristol, 23 House of Lords Science and Technology Committee inquiry into autonomous vehicles: written evidence (AUV0029) from Professor Neville Stanton, UK http://eprints.uwe.ac.uk/2916. Chair in Human Factors Engineering, University of Southampton (October 2016) http://data.parliament.uk/writtenevidence/committeeevidence.svc/ evidencedocument/science-and-technology-committee-lords/autonomous-vehicles/written/41762.html. 26 House of Lords Science and Technology Committee inquiry into autonomous vehicles: oral evidence session, question 60 (November 2016) http:// data.parliament.uk/writtenevidence/committeeevidence.svc/evidencedocument/science-and-technology-committee-lords/autonomous-vehicles/ 24 House of Lords Science and Technology Committee inquiry into autonomous vehicles: oral evidence session, question 56 (November 2016) oral/43733.html. http://data.parliament.uk/writtenevidence/committeeevidence.svc/evidencedocument/science-and-technology-committee-lords/autonomous- vehicles/oral/43733.html. 27 Ibid. 22 23
Chapter 4 Safety of other road users Traffic implications The risk of complacency and regional differences in to the presence of automated Another often-cited benefit of Current free flow traffic conditions Further investigation into also extends to other road- pedestrian and cyclists’ behaviour vehicles, particularly if human autonomous vehicles is the typically show average speeds whether the observed cautious users who will interact with are complex for human drivers to drivers know autonomous reduction in congestion and that are at or moderately behaviour would, in practice, autonomous vehicles, such as understand, let alone the sensors vehicles are designed to be risk improvement in traffic flows, above the speed limit. If the be eroded by the competitive pedestrians, cyclists and other and cameras of an autonomous averse and compliant with traffic through the transmission of traffic cautious driving behaviour pressure of other drivers in the drivers. An understanding of vehicle.29 Professor Sharples has rules.31 The study provides some information such as the location and reduced speed found in context of widespread roll-out how autonomous vehicles also pointed out that risks to evidence that, as autonomous of road closures or incidents. VENTURER Trial 1 is replicated of the technology in a real-world will affect the behaviour of autonomous vehicle users’ safety vehicles become more prevalent, However, from the perspective across drivers in general in the environment, or whether caution these other road-users will be may arise as a consequence some human drivers may of traffic management, the real world, and if they persisted would depress traffic speeds, important in developing both of other road users adapting adapt their driving behaviour. findings around delayed with greater experience of is needed. Investigation by the technology and the policy for their behaviour in response to At junctions, human drivers response and cautious driving autonomous vehicles, then this highway engineers on the effect the rollout of these vehicles. autonomous vehicles being on may pull out into smaller gaps behaviour following takeover could cause a build-up of traffic of cautious driving following the road.30 In the reverse, other between vehicles when there are could be important. and therefore have a reverse handover on traffic flow would In particular, it is important road users may need to adapt more autonomous vehicles in effect on road networks.33 also be valuable to gain a fuller to understand how human their behaviour and expectations the traffic, but when overtaking, understanding of the potential behaviour may change as to accommodate the conduct participants typically chose to extent of any problem. a result of the interaction of various types of automated wait until the approaching vehicle with autonomous vehicles. vehicle: for example it could had passed in all instances, The Government previously be the case that pedestrians regardless of whether the vehicle Impact on underwriting commissioned a scoping study would become complacent and was an autonomous vehicle to understand the main social assume that autonomous vehicles or a human driven vehicle. Data on driver performance during performance of drivers with a range who are highly experienced and behavioural questions would avoid them, thereby Comments from drivers who did handover will enable underwriters of age and driving experience. drivers, and in some cases relating to autonomous vehicles, crossing roads at any point. not adapt their behaviour towards to assess the risk class and This is particularly the case given experienced and trained for which identified nearly 400 autonomous vehicles suggest category of a driver based on the current policy debate around simulator studies. However, open questions and concluded The Greenwich Automated they may be motivated to do so their behaviour. This judgment high car insurance premiums VENTURER Trial 1 determined that behavioural aspects have Transport Environment (GATEway) in certain circumstances, such as enables a decision to be made on for younger drivers, 34 and the the age range of participants been under-researched.28 project has carried out a trial when they are in a hurry.32 a driver’s insurance cover and the potential benefits autonomous with a view to including younger, with a TRL driving simulator, premium he or she should pay. vehicles can bring to groups less experienced drivers Professor Merat believes one comprising two driving tasks: Further research is needed whose driving might be restricted as well as older and more of the reasons for the lack crossing a ‘give way’ junction and in this area to gain a fuller It is therefore important through lack of experience and experienced drivers. of research on pedestrians’ overtaking a slow-moving vehicle understanding of the impact of to examine behaviour and limited or declining skills. understanding of autonomous on an urban dual carriageway, in the technology on the behaviour performance when handover vehicles is that it is a very order to understand more about of other road users. is required multiple times Previous studies into the complicated topic: cultural how human drivers might respond throughout a driving scenario, handover period have generally as well as considering the used middle-aged participants Participant sample Trial 1 experiments Experiment UWE STISIM Simulator Wildcat Road Vehicle House of Lords Science and Technology Committee inquiry into autonomous vehicles: written supplementary evidence (AUV0095) from the 28 Government – Department for Transport (DfT) and the Department for Business, Energy and Industrial Strategy (BEIS) (December 2016) http://data. parliament.uk/writtenevidence/committeeevidence.svc/evidencedocument/science-and-technology-committee-lords/autonomous-vehicles/ written/44865.html. Number of participants 31 27 29 House of Lords Science and Technology Committee inquiry into autonomous vehicles: supplementary written evidence (AUV0092) from Professor Natasha Merat, Institute for Transport Studies, University of Leeds (November 2016) http://data.parliament.uk/writtenevidence/committeeevidence. svc/evidencedocument/science-and-technology-committee-lords/autonomous-vehicles/written/43683.html. Powered to detect a Sample size Medium to large effect size. medium-large effect size. 35 30 House of Lords Science and Technology Committee inquiry into autonomous vehicles: written evidence (AUV0049) from Professor Sarah Sharples and colleagues, University of Nottingham (October 2016) http://data.parliament.uk/writtenevidence/committeeevidence.svc/evidencedocument/ science-and-technology-committee-lords/autonomous-vehicles/written/41871.html. 18-69 years of age. GATEway (Greenwich Automated Transport Environment) research project led by TRL, launched in February 2015 https://gateway-project.org.uk/ 31 20-60 years of age Mean = 41.0, Standard ppr807//. Mean = 39.6, Standard Age range Deviation = 13.9, 32 Published GATEway Project Report PPR807, ‘Driver responses to encountering automated vehicles in an urban environment’ (February 2017) Deviation = 12.5, 3 > 60-years of age hence a https://gateway-project.org.uk/ 6 participants ≥ 50-years wp-content/uploads/2017/02/Driver-responses-to-encountering-automated-vehicles-in-an-urban-environment-1.pdf. mean age > 40 years of age 33 VENTURER – Trial 1 Findings http://www.venturer-cars.com/. 34 BBC News, MPs debate £1,200 cap on insurance costs for young drivers (March 2017) http://www.bbc.co.uk/news/business-39327089. 16 male 17 male Gender 15 female 10 female 35 (Cohen’s f = .25 – .4) with power of .8 (determined using G*Power 3.1.7 software: Faul et al., 2007). Faul, F., Erdfelder, E., Lang, A.-G., & Buchner, A. (2007). G*Power 3: A flexible statistical power analysis program for the social, behavioural, and biomedical sciences. Behaviour Research Methods, 39, 175-191. 24 25
Chapter 4 While age has not been pool due to faster handover is more widely taken up, if as considered as an independent times, this could bring a benefit expected vehicle crashes decline variable in the data analysis, if it to them as a group in the form when vehicles are in automated is the case that younger people of reduced premiums. driving mode, it is possible that are able to react faster to a those still driving manual cars will handover request and therefore The effect of autonomous vehicles be priced out of the market for have a quicker takeover time, on insurance premiums across insurance. Insurers, manufacturers then arguably they could see the board will also need to be and the Government will have a reduction in their insurance considered. Given that one of to work together to provide premiums. Typically, younger the key benefits of autonomous a solution to this problem. One drivers face higher insurance vehicles is the increase in safety option could be to help those who premiums because they are (90% of all motor accidents are can’t access driverless technology, typically regarded as carrying caused by human error), those for example by introducing a a higher risk, due to their age driving autonomous vehicles can scheme to encourage people as well as factors such as their expect to see a reduction in their to ‘cash in’ their old, non- driving experience and claims premiums: Telegraph Money has autonomous cars to help them history.36 However, if in the reported that annual premiums afford an autonomous vehicle, context of autonomous vehicles, could be reduced by £265 on known as a scrappage scheme. younger people were found to average by 2020.37 Looking further bring less risk to the insurance to the future as the technology Lots of work has been done on this by insurance companies and by market consultants, and they predict substantial reductions in the total premium pot. There will probably be a slight increase initially because you will have more expensive gadgets strapped around the periphery of vehicles, but once we see a higher proportion of these vehicles on the road, consultants predict a 50%-plus reduction in the total motor premium market. From our perspective, we are planning in that regard. The good thing is that it will not happen overnight, and therefore as we see motor premiums reduce we can move our staff and our capital on to other lines of business.38 – David Williams, Technical Director, AXA 36 House of Commons Transport Select Committee and Petitions Committee one-off oral evidence session on cost of car insurance for young people: question 38 (February 2017) http://data.parliament.uk/writtenevidence/committeeevidence.svc/evidencedocument/petitions-committee/the-cost- of-car-insurance-for-young-people/oral/48201.pdf. The Telegraph, ‘Driverless cars will shave £265 off insurance premiums in five years’ (May 2015) http://www.telegraph.co.uk/finance/ 37 personalfinance/insurance/motorinsurance/11623218/Driverless-cars-will-shave-265-off-insurance-premiums-in-five-years.html. Automated and Electric Vehicles Bill Committee oral evidence session (October 2017) https://hansard.parliament.uk/commons/2017-10-31/ 38 debates/b1d00f88-a22b-4291-937d-8f6a47d335d7/AutomatedAndElectricVehiclesBill(FirstSitting) 26 27
You can also read