A new study by the University of Nottingham and the RAC Foundation suggests that behavioural training for drivers is paramount for the transition into the next stage of automated vehicles, known as level 3 automation1.
Researchers from the University's Human Factors Research Group studied two groups of experienced drivers in a high-fidelity driving simulator to observe their behaviour while 'driving' a car with level 3 automation.
The study found that drivers who received behavioural training were more measured in their behaviour and better understood the car's capabilities and limitations.
The behavioural training included the provision of a checklist known as 'CHAT', pioneered by Emily Shaw in the University's Faculty of Engineering. The other group trained by reading an operating manual.
Key findings found in the behavioural training group:
- Significantly more likely to notice a potential hazard during the transition from automated to manual driving (in this case, a tailgating car), with 90% of drivers noticing the car in this group, compared to 24% of the operating manual group
- Made more measured decisions in lane change manoeuvre shortly after taking back manual control
- Spent more time preparing (e.g. acquiring knowledge of the road environment through mirror checks) before physically making the lane change
- Made more mirror checks in the run up to and during the lane change manoeuvre
- Checked their mirrors more frequently, even while the car was driving autonomously.
During a period of automated driving, participants in both groups could decide whether to engage in a non-driving task, such as looking at their phone, tablet, laptop or reading materials, or nothing at all. When notified by the vehicle to take back control and transition from automated driving, the group who were trained with only an operating manual took almost 10 times longer to pay full attention to driving, continuing glances at their non-driving task for an average of 11.2 seconds, compared with 1.8 seconds in the behavioural group.
The behavioural group was also markedly faster at making their first glance at the road when notified to take over — on average 7.3 seconds, in comparison to 21 seconds in the other group.
Transforming the role of the driver
This research demonstrates that the complex and changing nature of the role of the driver in an increasingly automated car should not be underestimated.
Emily continued: "It's clear that behavioural training has a positive effect on the driver's behaviour and their understanding of the car's capabilities. Our proven training procedure, 'CHAT', can be used to help drivers understand their shared role and responsibilities with automated vehicle technology, so that they develop the right state of mind for interacting with these vehicles.
"The 'CHAT' procedure's unique messaging is designed to support and motivate drivers to rapidly recall the checks and assessments they need to undertake before making a controlled transition from automated to manual control."
While future vehicles will deliver more automation than ever before, these vehicles are likely to remain visually the same as the current cars we are used to, and therefore may not present an obvious step-change in development.
David added: "Drivers, as well as those responsible for delivering driver training — and indeed, those who manufacture and sell these vehicles — may therefore be forgiven for assuming that no new skills are required to use them.
"However, this research clearly shows there is a lot of work to be done around preparing drivers for this next stage of automation. Without this, drivers are more likely to form inaccurate or inappropriate mental models of automation capability and competence, which can lead to misuse or incorrect use of driving automation."
Steve Gooding, Director of the RAC Foundation, said: "Many of us have been uncomfortable with the very concept of a partially automated car suddenly seeking to thrust control on a human driver whose mind, understandably, will have been on other things."
The good news in this report is that there are ways drivers can be trained to handle the take-over safely. The challenge is going to be in making that training happen. The risk is that we leave it up to drivers to read their new car user manuals and when was the last time you met someone who'd done that?
The research was funded by the RAC Foundation and led by researcher Emily Shaw, working with Dr David R. Large and Professor Gary Burnett, in the Human Factors Research Group, Faculty of Engineering.
Emily Shaw said: "To date, driver training for automated vehicles is no different to that provided for manual vehicles. The assumption being that prior manual vehicle training and experience, alongside instruction from the dealer and/or the vehicle's operating manual, is more than sufficient. However, the introduction of intermediate levels of automation into vehicles means that the driving task is shared between the driver and system, fundamentally changing the role of the driver."
Emily Shaw developed the driver behaviour training ('CHAT') as a proof of concept to test during the study, designed as a standardised operating procedure, to motivate and support drivers in remembering and applying explanatory knowledge acquired about level 3 automated vehicles.
The letters represent the actions required and help to guide the driver in making the appropriate checks and assessment of their internal and external environments prior to taking over operational control of the vehicle:
- CH = CHeck
- A = Assess
- T =Takeover
The use of 'CHAT' provides a way of standardising this procedure, similar to 'mirror, signal, manoeuvre', which drivers are already accustomed to as a result of driving lessons.
Emily Shaw said: "We chose the acronym carefully to ensure the procedure was easy to remember, as well as the semantic association with the word chat - giving a purposeful nod to the necessary communication that needs to take place whenever control is being passed between members of a team, as a further reminder of the shared control drivers have with the automated systems."
Understanding the capabilities and limitations of automated vehicles
Cars on the road at present are equipped with levels 1 and 2 of vehicle automation, meaning several simultaneous activities which assist steering or acceleration are partially automated — this includes, lane centring, adaptive cruise control, blind spot warning and automatic emergency braking.2
At level 3, automation capability extends to the monitoring task, allowing drivers to switch their attention towards non-driving related tasks (NDRTs). However, the human driver remains responsible for the vehicle's actions and must be ready to intervene in the event of a system failure or where the boundary of the automated system's operational design has been reached.
Planned takeovers — when a vehicle issues a takeover request to the driver because the boundaries of automated driving capability have been reached — will become a normal part of the driving task with level 3 vehicles.
In a takeover scenario, drivers need to be aware that having their eyes on the road and their hands on the wheel is not enough, their minds need to be back on the driving task.
Emily Shaw continued: "If and when a driver is issued with a planned takeover request they need to be making observations in order to orient themselves, making those checks and assessments, so they can start engaging their mind on the driving task, making decisions on what actions they may need to take, over and above controlling the steering and speed of the car, before they take back manual control."
Dr David R Large said: "The upshift in automation capability at level 3 introduces a new task to driving where the driver will be required to transition in and out of the driving control loop. This brings additional challenges where driver attention is divided between the non-driving related tasks, such as using a mobile phone, and the driving task, which can lead to missed information and error when it comes to switching from automated driving mode back to the driver being in manual control. Drivers, therefore, need to learn how to interact with these vehicles safely and effectively to be able to smoothly transition in and out of the loop of the driving task."
