Autonomous vehicles, the pros and cons
First publishedin ITS International
Driver interface and human factors could provide the biggest obstacles to autonomous vehicles as Jon Masters discovers.
Autonomous vehicles were momentarily made to look like yesterday’s technology this summer, when entrepreneur Elon Musk revealed his idea for the Hyperloop: a transportation system of passenger cars propelled through sealed tubes by differential air pressure.
Commentators and columnists have since been dismissing the proposal as outlandish and impractical, while the idea of autonomous vehicles on conventional roads has been looking less fanciful. But is the picture of vehicles driving themselves itself realistic and how far is the technology likely to go?
Google’s autonomous vehicles have attracted much media attention worldwide since they were launched in 2010, but others have already demonstrated fully autonomous vehicles steering among traffic without input from a driver. Among them is BMW, which earlier this year announced a new project in partnership with Continental’s Automotive Group.
The main aim of this two-year initiative is to develop the necessary sensory and control systems to prepare for deployment of ‘highly automated driving functions beyond 2020’, according to a Continental press statement. A new fleet of fully autonomous vehicles will be built between now and the end of 2014 for testing on highways.
“Automated driving is a key element in future mobility. It will significantly enhance safety, comfort and efficiency on the roads,” said Continental executive board chairman Elmar Degenhart at the project’s launch.
“The joint research project with BMW Group addresses the need for the enormous amounts of R&D required to realise the vision of automated driving. Driving cannot be automated overnight. It is a gradual process, stretching out over a period of ten years or more.”
Continental’s input will be in providing the sensory systems needed for production of an electronic co-pilot that can sense and respond to everything around it. The company will be employing ‘close to production’ technology in systems architecture sufficiently robust to maintain safety even if malfunctions occur, according to Continental’s statement.
BMW also makes a point of explaining that the principal technologies required have already been developed and fully tested, among traffic on the A9 between Munich and Nuremberg in Germany, for example. According to the company’s latest project announcement, what’s needed next is clear definition of operating strategies. Exacting lane positioning and recognition of everything around the vehicle is needed for this.
‘To help the system progress from vision to reality, a number of teams from both companies will address a range of technical challenges, since only with a complete command of the necessary technology will it be possible to develop a legal framework for automated driving,’ BMW’s statement says.
Dr Siraj Shaikh, head of Coventry University's Digital Security & Forensics research group
This last point is critical. As long as the Vienna Convention on Road Traffic remains in force then the driver must be ultimately responsible and liable for what the vehicle is doing. While there are viable arguments for adjusting or rewording parts of the Vienna Convention in recognition of modern technology, it is likely that the Convention’s fundamental point will remain for the time being (ITS International Sept/Oct 2012).
With the technology pretty much proven, the prospects of sufficient rewriting of the Vienna Convention to allow more vehicle autonomy is likely to hinge on the study of human factors – how drivers behave when the auto-pilot is in control, or as it comes in and out of operation.
The technical people at BMW have previously told this magazine that a principal benefit of the company’s autonomous vehicle programme comes from the knowledge and expertise that it feeds into development of technology that can be introduced to new cars (ITS International Sept/Oct 2011).
These advanced driver assistance systems (ADAS) now available on BMW models include Lane Departure Warning, Active Cruise Control and Traffic Jam Assist – the last two of which control distance to the car in front by braking and accelerating without the driver doing anything but holding the steering wheel.
The driver remains ‘in the loop’ as a central principle of current ADAS strategies, although BMW and others have foreseen the next stage as the vehicle’s computer systems taking lateral and longitudinal control. This could provide a comfort aid when driving in heavy traffic. Or, an emergency stop function will take the car to a safe halt at the side of the road if the driver loses consciousness – a safety feature BMW has stated as a long-term aim for some time.
All of this highlights how important considerations of the human-machine interface have already become. It can be assumed (and both BMW and Continental have officially confirmed) that the technology developers are focusing considerable attention on building a picture of how drivers are likely to react to partially or fully autonomous vehicles.
The same subject is also of great interest for military operations. Researchers at Coventry University have studied drivers’ reactions to ADAS systems for the UK’s Ministry of Defence (MoD).
“Unmanned logistics vehicles are one goal, but the project we have been involved in concerns driver aids for easing fatigue on long distance journeys,” says Dr Siraj Shaikh, head of Coventry University’s Digital Security & Forensics research group.
“The MoD wants to ensure its drivers are supported by every available aid. When driving in convoy a sense of intelligence is necessary for maintaining lateral and inter-vehicle control. Loss of communication between vehicles is common, however, so the proposal was to look at what happens when driver aids fail, comparing driver behaviour when ADAS is on and off,” Dr Shaikh says.
“We found, like many studies before ours, that built-in automation brings additional cognitive load, which generally goes against the design goal of these systems. We found that trust is an issue. All drivers report some sense of mistrust of the technology, but we also re-established some long held principles, that over time familiarity and skills improve. People get used to using the technology and their trust in it builds.”
Dr Shaikh has studied autonomous vehicles further as a member of Coventry University’s Integrated Transport & Logistics Grand Challenge Initiative – a network of researchers looking at major issues facing transport networks. Working with Professor Paddy Krishnan of the Centre for Software Assurance at Bond University in Australia, Shaikh has developed a system of modelling interaction between drivers and ADAS.
As reported in a paper presented by Krishnan and Shaikh, active cruise control (ACC) was used to demonstrate the computer model, with interesting results. “Our analysis suggests that on longer journeys, safe distance to the vehicle in front has to be extended due to fatigue, leading to reduced speed, so lengthening the journey and more tiredness,” Dr Shaikh says.
“ACC was chosen for this study because previous work has shown ACC increases risk of delayed driver reaction, awkward handover and mode confusion. Current ACC systems do not take cognitive state into account. It’s an area that should be looked at more. We’re moving towards a near autonomous state, but are all driver aids helping?
“The feeling is that as we’re making cars safer, technology is gradually defining the driver experience and people are becoming more aware of these systems. It’s not a step change. That would come from something like platooning.”
Affects on others
Interest in platooning – effectively ACC among a line of vehicles – includes a UK government initiative to research the effects of weather and human factors on HGVs in platoon formation.
Human factors researcher at the Transport Research Laboratory (TRL), Magali Goury has taken a different approach, studying how platooning affects the behaviour of other drivers nearby. Goury has carried out the study as part of the wider European Adaptation project, which is accumulating research on human responses to ADAS.
“Platooning has been shown to present advantages of increased safety, fuel efficiency and traffic flow, for vehicles in the platoon, but comparatively little is known about how other drivers react to it,” Goury says.
A simulator at TRL’s headquarters was used to test drivers’ reactions as they encountered platooning vehicles. Most significant of the findings, Goury says, came from study of headways maintained.
“Drivers tended to close up to the car in front, imitating the platoon as they passed alongside,” Goury says. “This shows that platooning and ACC present a potential hazard and that their affects should be investigated more, especially as more vehicles fitted with ADAS systems are mixing with vehicles not fitted with the same technology.”