EXPERT INVESTIGATION OF AUTOMATED VEHICLE COLLISIONS AND AEV ACT 2018 CLAIMS
The UK is involved in an international race to develop, demonstrate, validate and safely deploy automated driving technologies. The benefits of connected and automated mobility (CAM) deployment are expected to deliver a step change in road safety, significant job and knowledge creation, improved access to mobility, increased productivity, higher road network efficiency and provide a means to tackle transport’s environmental performance. These benefits are tantalising, with the UK market alone to be worth £42 billion by 2035. In response to this opportunity and the disruptive potential of CAM, the UK Government has, since 2015, enabled and supported a sustained programme of activity to develop, test and ultimately permit the deployment of CAM technologies. This wide ranging support, coordinated through the Centre for Connected and Autonomous Vehicles (CCAV), has included: innovation funding for consortia to develop and test technologies for connected and automated mobility; the creation of guidelines for on-road testing; supporting the development of a network of on- and off-highway test facilities (such as London’s Smart Mobility Living Lab); funding standards development; understanding the legal framework for automated vehicles and enacting legislation to permit the use of approved automated technologies on our roads.
To further accelerate CAM testing, trials and adoption, the Government has encouraged manufacturers to apply for exemptions to Construction and Use regulations, to undertake limited scale fleet trials and enacted the Automated and Electric Vehicles Act 2018, which provides manufacturers with a route for CAM technology approval and thus permission to operate on the UK’s roads. This legislation will enable the approval of Automated Lane Keeping System (ALKS) technologies as well as other automated driving technologies as their readiness for commercialisation matures.
Against this background it is widely accepted that independent CAM safety monitoring and collision investigation capabilities are required to measure the safety operation of vehicles operating in automated modes and to investigate incidents involving these technologies when they occur. This approach allows timely feedback to industry and Government and forms an essential component of a regulatory system for new technologies, where their real-world performance can only be fully understood with large-scale deployment. We are now preparing to implement this approach by creating frameworks for the in-service safety monitoring of CAM technologies and by developing processes for in-depth investigation of incidents involving these systems. These processes will guide and support Police investigations, independent safety analysis and future claims where expert advice is required on the real-world performance of CAM systems.
The importance of preparing for the expansion of automated driving on the UK’s roads has been highlighted by Zenzic (an organization established by the UK Government to coordinate the UK’s Connected and Autonomous Mobility Testbeds (CAM Testbed UK)) which has set out a roadmap for the UK to achieve connected and automated mobility at scale by 2030. This roadmap points towards a significant acceleration of the automated technologies entering public road environments in the UK over the next 5 years, from advanced on‑road trials of automated driving (from now until 2024) to national approval and licensing schemes for automated driving systems in 2025. Through this period, it is predicted that the environments, or domains, in which automated driving systems will operate will move from the current low complexity to high complexity in 2026. In parallel with these developments automated driving systems will be deployed into increasingly sophisticated service settings, whereby they move from small-scale solutions in 2021-2023 to preferred public service solutions in 2026 and beyond, whilst demonstrating commercially viable applications in freight within the same timescales.
We can expect, therefore, a wider geographical spread of automated vehicle trials and the prevalence of consumer vehicles with advanced driver support functions such as Tesla’s Autopilot to increase in the short to medium term. As has been observed in other geographies, the risk of incidents involving vehicles with automated driving technologies will increase as the testing of developmental systems and the deployment of consumer systems scales up. In recent years we have seen a number of high‑profile collisions reported internationally, involving vehicles with developmental automated driving systems and vehicles with advanced driver support systems driven by consumers. The most notable of each being the fatal incidents involving Elaine Herzberg (18 March 2018, Tempe, Arizona) who was struck by an Uber automated vehicle during routine development testing, and Joshua Brown (7 May 2016, US Highway 27A, Williston, Florida) whose Tesla, driving in Autopilot mode, struck an articulated vehicle which was turning across his carriageway.
In the US, the investigation of these incidents has been undertaken by the Police and other agencies, including the National Highway Traffic Safety Administration (NHTSA) and the National Transportation Safety Board (NTSB). Whilst a civil claim was avoided by a settlement reached between Uber and Mrs Herzberg’s family, news reports suggest that civil claims against Tesla in relation to the alleged use of the Autopilot system are being investigated and pursued by lawyers in the US.
In the UK it is recognised that the capability to independently investigate automated vehicle collisions, including vehicles which provide advanced driver assistance functions (hands off, feet off driving) is now required. These in-depth investigations will strengthen our understanding of causal and contributory factors in CAM incidents and will ultimately allow us to refine the lead indicators that can be monitored to intercept safety issues before collisions occur.
The first step to be taken in this direction is to provide support framework for the Police which provides rapid access to specialist collision investigation and automated vehicle expertise. To this end the Government has recently awarded TRL, the UK’s Transport Research Laboratory, a task within the Road Accident In Depth Studies (RAIDS) programme to provide support and assistance to UK Police forces when investigating an incident involving an automated vehicle.
TRL’s role will enable the Department for Transport (DfT) to both facilitate the forensic investigation of automated vehicle incidents for evidence of offences, whilst also monitoring occurrences of such incidents. This approach will provide DfT with data from which an understanding of emerging patterns of automated vehicle incidents can be established. As the data resource develops this information will inform future policy around priorities for AV testing, approvals, in-service monitoring legislation and regulation, provide a factual reference for AV vehicle incident numbers, incident types, casualties and trends, and provide a route to establishing deeper case studies which can provide high quality evidence of causal and contributory factors in AV collisions.
Today’s Police collision investigation activities consider a wide range of factors relating to the causes and consequences of vehicle collisions: they consider the physical evidence from collision scenes; witness evidence; data downloaded from vehicles or connected devices and CCTV; they seek evidence of causes and contribution from vehicle damage and condition; the road environment; driver actions and behaviour; as well as the actions and behaviours of other road users. In virtually all cases investigated today a driver will be in charge (if not in control) of a vehicle during a collision and investigations will consider whether the performance of drivers fell below the standards expected, which may lead to criminal charges.
Our Police forces have robust processes for the investigation of these incidents and employ dedicated collision investigation teams to deliver them. However, for the investigation of serious and fatal incidents involving automated vehicles the knowledge and experience of Police collision investigators will be limited, creating a knowledge gap. The approach taken by DfT will support Police investigations by providing access to TRL’s investigations’ capability and specific knowledge of CAM operation and safety management, together with specific guidance around the priorities for automated vehicle collision investigation.
The framework that is emerging for AV investigation extends beyond the analysis of physical evidence at a collision scene which forms the basis of ‘what’ happened (although this remains crucial), towards creating an understanding of ‘why’ an automated system failed to avoid it. New questions are, therefore, introduced concerning a range of issues, which AV collision investigations must address in order that AV operational, performance or component or failures are understood, examples of such questions are outlined below.
The status of an automated system at the time of a collision: was the automated driving function active or not? Clearly this is a fundamental question, the recording and preserving of this information is a function of an automated vehicle’s data storage system (DSSA), a requirement of approved systems and test vehicles under the UK’s code of practice for automated driving trials. A compliant vehicle, will therefore, be able to provide a record as to whether it was in manual or automated control at the time of an incident.
Is the vehicle in developmental testing, part of a supervised trial or an approved system for commercial or consumer operation, and what responsibilities, therefore, fall on the occupant(s) or ‘AV test driver’ (invariably an employee of the testing organisation)?
The management of safety risks for the testing of developmental systems in real-world conditions, was a robust safety case in place and was this being complied with?
The operational design domain (ODD) of the vehicle, was the vehicle being operated on a road that it was intended to be tested on, or designed/approved to be operated on in AV ‘mode’ if in commercial or consumer usage? This may reveal testing being conducted outside of safety case parameters, or unrestricted use of a commercial/consumer AV technology outside of its design environment intentionally or unintentionally.
The map data and localisation, how precisely did the vehicle know its position, what information was provided by its map data, was the vehicle behaving in accordance with this information?
The performance of the vehicle operator and the management of ‘automation complacency’ and/or inattention, was the automated system being supervised appropriately? What measures were in place to manage driver attention and engagement?
The components of the automated driving system, their specification, function and data processing flow; consideration of data from sources such as lidar, radar, cameras, proximity sensors, telemetry, vehicle positioning, systems monitoring, telematics and telecoms which may be required to address what an AV sensed or detected. Including electronic or mechanical failures.
Sensor calibration, condition examination to confirm whether sensors are operational and in‑calibration, together with evidence of testing and calibration frequency, self-calibration, etc. of safety critical systems.
Data access from in-vehicle and off-vehicle systems, including data, visualisation, and interpretation which may demonstrate what a vehicle sensed, the objects and motion that it was aware of before an incident (the performance of object detection, identification and tracking systems).
The classification of objects within the road scene before an incident, what could have been detected, how would the automated system ‘see’ and classify objects within the road environment, how would objects be classified?
Object tracking and path prediction, how would the automated system predict the motion of objects within the road environment?
The motion planning behaviour of the automated driving system, based on the object classification and tracking/path prediction what response was directed by the vehicles automated control system. How effective was this motion planning behaviour?
Cyber security and attack surfaces, evidence of forms of cyber-attack which may affect a vehicle’s real-time or the behaviour of its back office systems, and sensor impairment by localised DOS or similar attacks.
Ultimately it may be beneficial and efficient to have a single UK safety monitoring and investigation body for collision investigation which takes the lead in investigating critical safety issues (rather than criminal offences) with a remit across road transport, including automated vehicle technologies and the in-use safety monitoring and evaluation. Such a body would provide a central hub to monitor safety, analyse data, investigate incidents, and provide timely feedback and recommended actions.
The capability to perform detailed investigations into automated vehicle collisions is a key element in building and sustaining public confidence in automated vehicle technologies. This confidence may be tested as wider trialling and deployment of the technology enters the public consciousness. Demonstrating that learnings from safety monitoring and incident investigation are feeding into the AV regulatory approval process is an important step, along with: the refining of safety goals (so that these can be measured); defining desired automated driving behaviours (so that these can be evaluated), and; feeding real-world scenarios into simulation libraries for ‘virtual’ validation. This proactive approach will drive continual safety and performance improvements, accelerate innovation and ultimately make Vision Zero a more realistic and achievable target.
The development of this capability counters the risk that public (and institutional) confidence may be undermined and deployments curtailed if industry and regulators are not seen to understand and respond to real-world safety issues as they develop. An undermining of confidence in AV safety may delay the realisation of benefits from AV technologies, the impact of which may be significant when considering the promise of improved safety, efficiency, productivity and the economic opportunity of developing and implementing such potentially disruptive technologies.
In TRL’s view, the Government’s approach to supporting the development, testing and ultimate use of automated driving systems on the UK’s roads, is strengthened significantly by the steps that they have taken to provide access to AV collision investigation support. The service to be provided by TRL will support the investigation of automated vehicle collisions by UK Police forces when they occur on the public road network. The framework that will develop from this support will be applicable to future independent safety investigation of AV incidents and the investigation of civil claims arising from the introduction of market ready AV systems under the AEV Act 2018.
For more information, please don’t hesitate to contact me.