Skip to main navigation Skip to main content
The University of Southampton
Engineering

We're making driverless cars safer

Published: 12 June 2017
Driving simulator
Testing AV's in the driving simulator

Once the stuff of science fiction, driverless cars are set to revolutionise road transport in the coming decades. Partially automated vehicles (AVs) are already on the roads, and manufacturers are racing to bring out more advanced models.

Driverless cars have the potential to make road transport safer, more efficient and more accessible. However, in the shorter term, the development of AV technology brings numerous challenges, the most critical being safety. This was brought into sharp relief last year by a fatal accident involving a Tesla car in ‘autopilot’ mode.

The human-technology interface

Our researchers are helping to address this challenge by exploring the human factors in automated driving, with the aim of creating synergy between driver, vehicle and environment. Neville Stanton, Professor of Human Factors in Transport, explains: “The big unknown is how the driver will behave in an AV – for example, how they’ll cope with monitoring what’s happening when the car is driving itself, and how they’ll handle giving control to the car or taking it back. Our research focuses on this interface and interactions between the driver and the AV system.”

The findings – which include significant new information about how quickly people can resume manual driving – are informing manufacturers’ designs and helping to shape the regulatory and legal frameworks that underpin safety on the roads.

Neville has been at the forefront of AV-related research for more than 20 years, working closely with motor vehicle manufacturers and research institutions around the world. “One of our early projects was the development of adaptive cruise control for Jaguar, resulting in the launch in 1999 of the first car to use this technology,” he says.

More recently, he has led studies that use a sophisticated driving simulator, as well as AV cars on test tracks and on the road, to gain new insights into human responses to highly automated driving systems.

Partial automation – the risks

While highly automated cars are on the horizon, they will still require some level of oversight by a human driver – it will take longer for truly ‘driverless’ AVs to become commonplace. The Society of Automotive Engineers has defined different levels of vehicle automation, from manual driving (level zero) to fully automated cars (level five). Currently available systems, such as Tesla Autopilot and similar cars developed by Mercedes-Benz and BMW, are at levels two and three.

“The problem with this level of automation is that the driver has to continually monitor the automated system and take back control of the car when needed. It doesn’t allow you to do something else, which rather defeats the object of automation,” says Neville. “Our studies show that monitoring a car is actually harder work than driving it yourself. It’s very difficult to sustain the necessary attention levels for long periods of time, and we know that when drivers’ attentional focus moves to non-driving tasks, their awareness of what’s happening on the road deteriorates.”

Key findings

  • On average, AV drivers take five times as long to respond to emergency braking compared to manual drivers.
  • It takes AV drivers up to 25 seconds to take over driving from the automated system – much longer than previously thought.
  • A system that provides continuous relevant feedback to the driver (analogous to a chatty co-driver) can improve drivers’ awareness of what the AV is detecting in the wider environment, helping to avoid errors.
  • A comparative study between on-road and simulated AV driving showed similar driver response times, validating the use of simulators for this sort of research.

In order to inform the design of safer AVs, the researchers used the simulator to test how long people take to resume manual control following a period of automated driving. “The commonly held view was that people take between seven and 10 seconds to take over the controls,” Neville explains. “In fact, our study showed that they take anything from 1.9 to 25.7 seconds. The drivers in our study weren’t distracted by any non-driving tasks – when people have been immersed in reading, watching videos or replying to emails for a couple of hours, it will take even longer for them to take the controls.”

This study was part of the EU-funded Human Factors of Automated Driving programme, which was supported by car companies including Jaguar, Volvo, BMW and Toyota. The findings, crucial for the development of safer systems, will influence future AV designs. “The challenge for manufacturers is accommodating the full range of response times rather than limiting parameters to mean or median transition times,” says Neville. “Our findings can also guide policymakers in setting guidelines for how much lead time a driver will need when changing in and out of automation.”

Optimising the handover of control

A new project, part of an £11m programme funded jointly by Jaguar Land Rover and the Engineering and Physical Sciences Research Council, is finding out the optimum level of information that needs to be exchanged between the AV system and the driver when control is handed over, in order to come up with a design specification that is safe and workable. Neville says: “The nature of the handover will very much depend on how long the driver has been away from the controls – if it has only been a few minutes, not much will have changed but if it’s an hour, more of an update will be needed.” The project, led by Southampton and conducted in partnership with the University of Cambridge, draws from protocols used in other domains, such as aviation and air traffic control, where effective handovers are critical for safety.

Neville’s wealth of expertise is informing national policy as well as manufacturers’ AV system design. Last year, for example, he presented written evidence on safety issues to the House of Lords Science and Technology Committee on driverless cars.

Our expertise in this field is contributing to the regulatory and legal frameworks that will make highly automated vehicles a reality on our roads.

Neville Stanton - Professor of Human Factors in Traffic
Privacy Settings