By Professor Oliver Carsten, Professor of Transport Safety, Institute of Transport Studies, University of Leeds
By far the most significant differentiation in the levels of driving automation is between high-level assisted driving, where the driver is assisted in longitudinal and lateral vehicle control, and the lowest level of automated driving, where the driver can hand over full control of the driving task to the vehicle, but has to be ready to comply with a request to resume manual control when instructed to do so by the vehicle. In assisted driving, the human driver is required to maintain attention to the road traffic scene; in automated driving the human user is allowed to divert attention until instructed no longer to do so.
The most obvious and simple way to indicate that distinction to users is to require hands on the wheel in assisted driving and permit hands off the wheel in automated driving, where it is not needed. Until recently in the UK and Europe, no driving assistance system was approved that broke that distinction by permitting a driving assistance system that do not monitor drivers’ hands on the wheel and consequently allowed hands-off driving, while still leaving the driver responsible for safety.
However, two recent events have broken that crucial differentiation. They indicate that relevant authorities both in the UK and in the EC do not always give priority to safety considerations when making deciding which types of driver assistance should be permitted for sale in new vehicles.
The first event was that in April 2023 the UK Vehicle Certification Agency (VCA) gave approval for the first hands-off driving assistance system to be permitted for use on UK roads. That system is Ford’s BlueCruise system which is an option on the Mustang Mach-E. The system allows drivers to take their hands off the steering wheel, but they are supposed to maintain their attention to the road and traffic scene and are also required to intervene in vehicle control (e.g. brake or intervene in steering control) when they determine that the assistance system cannot handle the situation. Ford calls this “hands-off, eyes-on” driving. Continued driver attention is supposedly maintained by means of a camera-based driver monitoring system which detects when the driver makes long glances away from the forward roadway or the mirrors.
This approval was heralded with the headline “First driverless car approved for British motorways” in The Times on 13 April 2023. This headline is an illustration of one of the major problems with the decision. This is not a driverless car: the human driver retains responsibility for safety at all times, whether the system is enabled or not. The Ford BlueCruise system is a driving assistance system; it is not an automated driving system which, when enabled, would take over responsibility for the safety of the driving task. The approval was greeted with enthusiasm by Transport Minister Jesse Norman who declared: “I am delighted that this country is once more at the forefront of innovation. The latest advanced driver assistance systems make driving smoother and easier, but they can also help make roads safer by reducing scope for driver error.”
What was the legal basis for this approval? Normally the UK follows technical rules established by UNECE WP.29, the World Forum for Harmonization of Vehicle Regulations. However, there are currently no international regulations for hands-off driving assistance systems. Instead, the procedure adopted via Article 39 of the GB type approval regulations, inherited from the European Community’s Article 39, which allows a manufacturer to apply directly to VCA for approval of a novel system that is incompatible with existing regulations. The GB and the EC Article 39 procedures require that the manufacturer “describes the safety and environmental implications of the new technology or new concept and the measures taken in order to ensure at least an equivalent level of safety and environmental protection to that provided by the requirements in respect of which an exemption is sought.” So, all that is needed is paperwork to describe the safety measures adopted in the vehicle system. No track or on-road tests of the system’s capabilities and behaviour are needed.
Beyond muddying the distinction from automated driving, there are a number of additional serious concerns:
- Hands-off driving is entirely reliant for safe operation on the detection of driver inattention by the driver monitoring system (DMS). That means that the DMS must be 100% reliable with no misses in detection of inattention, an impossible target.
- How do we know that eyes on the road means that drivers are paying attention and processing the information? In fact, we don’t. Missy Cummings, a well-respected human factors and safety expert at George Mason University has summed this up nicely. She states that “hands-free equals mind-free.” (https://www.youtube.com/watch?v=Hz-kufFcE8E)
- By removing direct interaction with controlling the vehicle, drivers with such assistance are placed in supervisory control, required to watch closely for inappropriate actions of the system or for situations that the system is unable to handle. The driver is supposed to realise when an intervention is needed. How reliable will be that determination? There is a very large body of human factors literature on supervisory control; one major point from that literature is that it can be more onerous than direct control.
- Finally, there is a fundamental ethical problem here: if the system makes an error, the human is still responsible because that human should have noticed and intervened.
The second event was more recent. Ford’s BlueCruise system can only drive within a lane on a mapped motorway, but a yet more flexible system was approved behind closed doors by the European Commission’s Technical Committee on Motor Vehicles (TCMV) under the EU Article 39 procedure. The TCMV is managed by the industry directorate of the European Commission (DG GROW) and its membership consists of representatives of the EU Member States. A BMW system called “Extended Lane Change Assistant” was approved. With this system, the vehicle can change lanes automatically without a driver confirmation. Or rather a driver’s glance to the side mirror is deemed to be confirmation that the vehicle can make the manoeuvre in the next few seconds. Here we are on dangerous territory: the driver has pretty much lost authority over the vehicle’s manoeuvring into an adjacent lane and the vehicle is performing the action automatically. But the driver still remains responsible for the safety of the vehicle’s actions.
So questions need to be asked: What was the need for approvals before these systems had been regulated at international level? Why was a fast-track process employed? And why were legitimate safety concerns ignored?
Professor Oliver Carsten is based at Institute of Transport Studies, University of Leeds and is a special adviser to PACTS. He has raised concerns about the distinction between assisted and fully automated driving, particularly as it relates to uncertainties about driver attention, the burden of supervisory control, and the ethical issue of human responsibility for system errors in his report ‘Driver Attentiveness to the Driving Task During ADAS Use‘
