• International Workplace
  • 6 February 2018

Self-driving cars: the safer option for your drivers?

Futurologists love the open road because transportation acts as the perfect metaphor for progress. So it might seem to some like the self-driving car concept is just the next stage on our journey to making the tech in Star Trek and The Jetsons a reality. But it is actually a product of thinking about improving road safety. Volvo had been working on this project for a great many years before being joined by other manufacturers, and their basic premise was that vehicle accidents have a great deal to do with a loss of attention and variation in the speed of response to stimulus; something computers do not suffer from. Such cars are a reality now, and in future are expected to be commonplace. But many drivers find the idea of giving over control to a computer unsettling, despite the fact that aircraft (which are considerably safer per se) mostly fly themselves these days. Indeed, we entrust computers to make stock trades in microseconds that affect the entire world economy if they go wrong. Our anxiety about self-driving cars stems from the way we understand roads as mindbogglingly complex environments, fraught with existential risk, that require conscious decision-making to navigate. A computer, despite its speed and complexity, has not yet developed the capacity to be thrilled by speed, or to fear its own demise, which are at the root of our own conscious relationship with driving.

DeepMind, which is IBM’s super computer, plays video pinball 2,539% better than a human being, but performs dramatically worse at Ms Pacman (only 13% as well). The difference? The computer is good at predicting the trajectory of an object adhering to rules of momentum with the intention of anticipating where it will be. What it is not so good at is decision making that requires anticipation of the behaviour of an apparently random actor (and one that is only able to travel in two dimensions in one of four cardinal directions). Self-driving cars are programmed with the rules of the road and with some principles of proximity detection and collision avoidance, but their designers are finding out there is so much more to driving than that. For a start, the human brain is able to relegate activity to lesser brain functions to free up computing power. Most of us find driving hard at the start: like rubbing your tummy and patting your head to a four-year-old child. Muscle memory and familiarity allows us to improve dramatically and in doing so we free up time for tasks like putting on make-up and answering the phone, and therein is a journey to becoming a much worse driver. The charity Brake this month expressed their concern that fleet operators are still not doing enough to discourage distracted driving. A survey last year (by asserted 70% of motorists eat and drink behind the wheel, while 11% of motorists admitted to putting on cosmetics and 51% of motorists admitted to changing their satnav while driving. This is perhaps an area where the autopilot function might intervene.

The ten-year period from 2006 to 2016 saw a 44% reduction in UK road deaths; over 1,300 fewer deaths per year on the roads. Strikingly though, this data (from UK constabularies) shows a sharp fall from 2006 to 2010 and since then a levelling off and even a possible trend back upwards in recent years. Whilst overall falls in road deaths have been happening since the 1970s (starting with compulsory seat belt laws), a curiosity of vehicle accident statistics has been the decline in deaths tallying with falling and low GDP growth. Road traffic, it seems, is an indicator of prosperity; safer roads are a partial product of austerity. Furthermore, prosperity brings with it an increase in sales of high-performance vehicles for leisure use, which may be contributing to a degree of carelessness. The human driver, of course, has the capacity to make decisions and to learn from them that is currently unmatched by computers. A study by the RAC of enforcement activity showed that 1.4 million people who committed driving offences found themselves on offender retraining courses instead of losing points or being in the dock. They are particularly effective in the area of speed reduction, which is where a great deal of enforcement activity is focused.

The overall decline in road deaths owes a lot to technology. A recent proposal to increase the length of time before a new car must undergo an MOT reflects the increasing safety performance and reliability of new vehicles. This measure was eventually rejected; the safety argument was not quite strong enough to prevail: yet. Michelin and Kwik Fit technicians have found that 40% of cars they checked had at least one tyre that was ‘dangerously’ or ‘very dangerously’ underinflated. Meanwhile, we share the roads with a terrifying legacy of passenger vehicle tyres that are legal even when ten years old (and some illegal tyres older than that). We must also contend with a worrying increase in uninsured drivers; Churchill estimated the number at over one million in 2016.

The use of speed cameras is itself a technological innovation that can claim a proportion of the recent road death reductions. A study of 2,000 camera sites by the RAC in 2010 showed a substantial improvement in compliance with speed limits, a particular reduction in extreme speeding, a marked reduction in average speed at fixed sites, and an appreciable, though more modest, reduction at mobile sites. Whilst acknowledging that speed is not the only determining factor in motor vehicles accidents, it is both directly related to the energy involved in collisions and the level of damage (the consequences in a risk assessment), and it reduces decision making time, which influences your likelihood of having an accident. A study by the RAC late last year reflected on the wide variation in enforcement of speeding that still exists across different constabularies. There were 184,654 offences recorded in Avon and Somerset last year, 83% of which were captured on camera. The county has over 1,800 miles of main trunk roads and motorways, but by no means the most of any county. Sleepy Dyfed meanwhile booked only 1,355 offences, none of which were caught on fixed roadside cameras. Whilst such variations exist, the driving public will continue to view cameras with suspicion.

Through the looking glass, one can speculate that road deaths will continue to decline, not least because the use of personal vehicles must have a limited future, given the environmental costs we are all paying for the privilege. It’s not too big a stretch to see the self-driving cars of the future being networked into long caravans running nose to tail for miles along a slick solar energy-generating freeway, and if you’re going to go that way, you may as well get on a train.