Regulatory red tape a roadblock for driverless cars
Driving and insurance rules will likely delay the rollout of autonomous vehicles — and that means more needless road fatalities
By 2025 self-driving cars will be fully autonomous, allowing drivers to kick back and get on with more pressing tasks like catching some sleep or watching a film instead of focusing on the road.
It is a sci-fi dream that is turning into reality, and — in the Arabian Gulf — Dubai has been in the driving seat of advancing such technology.
Dubai launched its Autonomous Transportation Strategy (ATS) in April 2016, which aims for a quarter of all journeys to be made on driverless forms of transport by 2030, including the Dubai Metro. It has begun testing driverless minibuses, while Dubai-based ride-hailing firm Careem last July announced a partnership with a US tech firm to launch mass-transit driverless pods in Dubai by the end of the next decade.
But while the necessary in-car technology will be complete for what will be the greatest transportation revolution in more than a century, driving and insurance regulations and essential supporting software are struggling to keep pace and may slow the roll-out of these vehicles.
Delays would mean unnecessary road deaths, given that McKinsey forecasts crashes will decline 90 percent from current levels once autonomous vehicles are the main means of road transport. This is especially relevant in the Gulf, which has the highest traffic fatality rate outside Africa.
These driverless vehicles will also have a positive impact on the environment and economy, as they will be vastly more fuel-efficient. Morgan Stanley predicts widespread use of such autos would provide $5.6 trillion in annual savings worldwide. Less than a generation ago, autonomous cars were the pre- serve of science fiction, but development has accelerated with 11 manufacturers testing prototypes in California last year including Google, Nissan and General Motors (GM).
Google has opted for attempting the “moonshot” of creating a fully autonomous car without a steering wheel, pedals or other humanactivated controls from the outset.
Auto manufacturers, in contrast, have taken an incremental approach, steadily adding features such as lane guidance and adaptive cruise control to progress through the five levels of automation.
“We believe consumers will more easily adopt autonomous technologies if they are phased in over time versus a strategy that goes straight to autonomous,” said Josh Clifton, a Nissan spokesman.
“Advances in artificial intelligence (AI) are making vehicles smarter, more responsive and better at making decisions. But we are still not at a point where autonomous vehicles can know exactly how to handle unpredictable situations. This is one of the roadblocks to realizing a fully autonomous future for driving.”
Level-one automation includes functions such as cruise control, for example, while level five is entirely driverless. Level two, which would allow for autonomous driving on highways, should be operational next year, with the same function available in cities from 2020, Nissan forecasts.
But just because the car is at the helm does not mean it is time to crack out the Sudoku.
“Level three is spooking people because at this stage the car will do the automation, but if it needs the driver to take control, the driver must be ready — the problem is whether the driver will understand the buck still stops with them,” said Matthew Avery, director of research at Britain’s Thatcham Research, an insurance industryfunded vehicle testing firm heavily involved in autonomous cars. “Manufacturers are about 95 percent there … (but) we are not sure the world is ready.”
In terms of driving rules, the main sticking point is Regulation 79, drawn up by a UN body and in practice enforced in most countries, which requires drivers to always have their hands on the wheel.
Germany’s Judicial Ministry in late 2015 baulked at making legal changes that would allow drivers of autonomous vehicles to do other activities and California, the main testing location, has taken a similar approach, insisting the driver is responsible for monitoring their vehicle and able to seize control at all times.
Yet being unable to mentally switch off removes the main allure of self-driving cars.
“The purpose of automation is to improve safety and allow humans to make better use of the time moving from point A to B,” said Bryan Reimer, an MIT researcher who studies driving behavior. “We as humans have been shown time and time again to be terrible overseers of automation. We tend to over-trust systems and as such fail to be ready when asked to resume control.”
Insurance is another regulatory quagmire. The main quandary is around assigning fault: Who is responsible if a car was in autonomous mode at the time of a crash — the driver, the manufacturer, the software provider, or none, some or all of them?
“Some insurers are talking about the need for a strict liability insurance system — at the moment, car insurance is with the driver, but we might need to move to the principle of the car being insured not the person,” said Thatcham’s Avery.
Insurers want access to an autonomous car’s data history in the event of a claim.
“The way of establishing who is to blame is finding out what mode the vehicle was operating in and whether it was operating correctly — we won’t know that without information from auto manufacturers,” said David Williams, technical director of AXA Insurance.
“Six months ago, German motor manufacturers told me in no uncertain terms that we would get nothing at all, it was their data. Now they seem to favor data being made available through an independent third- party server. That could be fine, but we need to make sure there is no hidden agenda like we cannot have the data for six months or we have to pay millions for some translation software.”
Vehicle-to-vehicle (V2V) systems for autos to communicate with each other and road infrastructure are also essential to making driverless cars a reality. Such technology will guide vehicles through unpredictable situations such as bad weather, dangerous human drivers and roadworks.
The US, Europe and Japan are all developing V2V systems, but these remain a long way from being road-ready.
“In theory, it would be possible for somebody to implement a highly or fully automated vehicle without V2V communications, but it would have disastrous consequences in terms of safety, traffic flow and energy and environmental impacts,” said Steven Shladover, a transportation engineer at the University of California, Berkeley.
“For these systems to function well, V2V communications are a prerequisite.”
Nissan is among the first auto manufacturers to acknowledge that external human help will be required for autonomous cars in addition to V2V. The Japanese manufacturer in January unveiled its Seamless Autonomous Mobility (SAM) system, which will connect its cars to staff at a high-tech call center that will guide vehicles through tricky situations such as an accident scene.
Berkeley’s Shladover added: “Before it will be feasible for vehicles to operate without continuous driver supervision, a great deal of fundamental work needs to be done on software safety engineering.”