Autonomous now: the shift to self-driving
This article, a version of which first appeared in the July 2018 issue of IMI Magazine, was the spark for Cars of the Future. IMI editor, Tim Kiek, said: “I’ve featured articles on autonomous vehicles throughout my tenure but never one which explores the topic with such forensic rigour.”
Autonomous Now… Neil Kennett explores blockbuster themes in the shift to self-driving
In our May issue, Traka’s Paul Smith outlined the six levels of autonomy, detailed how Audi’s A8 and Tesla’s Model S are already around Level 3, and noted that driverless motoring could be a reality on UK motorways by 2021.
Let that sink in for a moment. 30 years ago, this was the stuff of science fiction. Gen X children across the world dreamt of owning KITT from Knight Rider – a talking, self-driving, bulletproof Trans-Am. For many, that dream will nearly come true.
While the leaders in this new wave of cars are ready to roll, a few thorny hurdles stand in their way: the elimination of digital ‘not-spots’; devising a new liability framework; meeting the challenges of a mixed (autonomous and non-autonomous) car parc; and addressing legitimate cybersecurity and ethical concerns.
Then there’s the skills shortage in intelligent mobility; data governance issues; what all this means for those who actively enjoy driving; and the impact on other road users.
Before investigating out how we might get from A to B in just three years, let’s get the acronym sorted. For the purposes of this article, we’ll call them connected and autonomous vehicles, CAVs.
Now for a definition: According to The Institution of Mechanical Engineers, when the term autonomy is applied to a car, it refers to decisions taken by embedded intelligence in the vehicle systems.
This is rules-based software. For example, if the forward-facing camera image contains a pixel pattern associated with a car, and the radar confirms this, and a collision is predicted, then a solution will be deployed, such as emergency braking.
Relentless progress
“If I had asked people what they wanted, they would have said faster horses,” Henry Ford is often quoted as saying. Vehicle manufacturers (VMs) have been pushing the envelope ever since.
A cautionary tale comes from Dan McComas, former senior vice-president for product at Reddit, who witnessed “…a complete breakdown in the kind of thought process behind how your technology is going to affect the users that use it and the world at large.”
VMs and tech giants cite many attractive benefits in shifting to self-driving: increased mobility, improved road safety and reduced congestion.
However, the excellent 2017 House of Lords Science and Technology Select Committee report, Connected and Autonomous Vehicles: The Future?, pointed out: “While we cannot say with any certainty what the impact on congestion will be, it is possible to imagine a situation of total gridlock… while some of our evidence has suggested that CAV could have huge economic benefits, we are not convinced that the statistics provided have been properly substantiated.”
That’s a lot of uncertainty, but the committee also heard compelling evidence in support of CAVs, not least the Association of British Insurers submitting that human error is a causal factor in 90-95% of road traffic accidents.
That stat originates from the US and some claim the UK figure is lower. Here, there were 27 road deaths per million inhabitants in 2017, a 5% year-on-year improvement. Within the EU, only Sweden had a better record. The point stands: humans are fallible.
The BBC’s Tomorrow’s World boldly stated: “Over a million people are killed worldwide each year by cars, with 90% of accidents caused by human error. Several million miles of test drives have shown driverless cars to be safer.”
One firm advocate is Antonio Avenoso, executive director of The European Transport Safety Council. “We are calling for safer vehicle standards such as mandatory fitment of automated emergency braking and intelligent speed assistance; better infrastructure safety rules and a solid framework for the safe rollout of automated driving,” he said.
Bob Lutz, former vice chair of General Motors, is more caustic. “Human drivers are distracted. They drink. They text. They take drugs. Autonomous vehicles do none of that,” he said.
Further to the safety argument, there’s cold hard cash. Morgan Stanley has estimated that autonomous cars could save the US $1.3 trillion annually through lower fuel consumption ($169bn), reduced crash costs ($488bn) and productivity increases ($645bn).
In the UK, a 2015 joint report by KPMG and the SMMT, Connected and Autonomous Vehicles – The UK Economic Opportunity, estimated socio-economic benefits “in the region of £51 billion per year by 2030”, if we consolidate a leadership position. Most of this weighty sum is predicted to come from increased ease of travel, fewer accidents and improved productivity.
Other key findings were that CAVs could create an additional 320,000 jobs in the UK by 2030, 25,000 of which would be in automotive manufacturing, and that CAVs could save over 2,500 lives and prevent more than 25,000 serious accidents in the UK by 2030.
“Already more than half of new cars sold are available with at least one semi-autonomous driving feature,” said SMMT chief executive, Mike Hawes.
It should certainly be a good time to be in the sensor business. ABI Research forecasts that as many as 36 million LiDAR units will ship in 2025, with a market value of $7.2bn.
A recent study by The University of Greenwich found that 43% of 925 respondents “felt positive” towards the concept of CAVs. 46% were undecided, with road safety (51%) and cybersecurity (44%) their primary concerns.
Cyber threats
These concerns are not without foundation. In a series of experiments from 2013 onwards, Charlie Miller and Chris Valasek showed that a hacker with wired or over-the-internet access to certain vehicles could disable or apply the brakes, turn the steering wheel and cause acceleration.
“No matter what we did in the past, the human had a chance to control the car,” said Miller. “But if you’re sitting in the back seat, that’s a whole different story. You’re totally at the mercy of the vehicle.”
While a malicious attack could have horrific consequences, Craig Smith, a security researcher who runs Car Hacking Village at Defcon, the world’s largest hacking convention, believes CAVs are generally more secure.
“They have to use lots of different sensors,” he explained. “The interesting thing is that each sensor doesn’t trust the other. It’s closer to the way humans figure out whether something is an illusion or not. And that’s harder for a hacker to deal with.”
What if a CAV owner wanted a trusted third party to gain access, an independent workshop, for example? The aftermarket received a boost from The European Parliament in March, when 633 MEPs requested that the European Commission publish a legislative initiative to guarantee fair, unrestricted and in real-time access to in-vehicle data before the end of the year.
Level 3 tipping point
The halfway stage of automation, dubbed Level 3 or Conditional Assistance, is pivotal. These vehicles can monitor their surroundings, change lanes, and control steering and braking, but the driver must be ready to take back control if required.
To illustrate the current state of play – how these are no longer theoretical conundrums but real world problems – news broke in April of a very modern dangerous driving incident.
A driver who put his Tesla into autopilot and moved into the passenger seat while at 40mph on the M1 was disqualified for 18 months after footage was posted online. Bhavesh Patel admitted that what he had done was “silly”, but insisted his car was “amazing”.
PC Kirk Caldicutt, of Hertfordshire Police, said: “What Patel did was grossly irresponsible and could have easily ended in tragedy. He not only endangered his own life but the lives of other innocent people using the motorway. This case should serve as an example to all drivers who have access to autopilot controls. I want to stress that they are in no way a substitute for a competent motorist in the driving seat who can react appropriately to the road ahead.”
Professor Neville Stanton, of the University of Southampton, has highlighted that, in simulated emergencies, up to a third of drivers of automated vehicles did not recover the situation, whereas almost all drivers of manual vehicles were able to do so.
Further still, CAV drivers took, on average, six times longer than manual drivers to respond to the emergency braking of other vehicles. “This is particularly true if they are engaging in other activities, such as reading, answering emails, watching movies or surfing the internet,” he said.
Steve Gooding, Director of the RAC Foundation, expressed concern that a Level 3 vehicle could hand back control to a driver who “might well be asleep at the time”. He suggested the risk could be managed by skipping this level and requiring CAVs to be capable of coping with any eventuality.
Who to save?
Unfortunately, in many crash situations, there is no win-win; it’s a case of the least worst option. Which brings us to ethics, and a thought experiment called The Trolley Problem.
The scenario is this: There is a runaway trolley and, ahead, five people are tied to the track. You are standing some distance off, next to a lever. If you pull it, the trolley will switch to a track only one person is tied to. What do you do?
Referring to our definition, CAVs work on rules, which must be coded. In surveys published in the journal, Science, researchers in the US and France set out to canvas opinion on how driverless cars should behave in no-win situations.
76% agreed that a driverless car should sacrifice its passenger rather than kill 10 pedestrians. The pinch came when they were asked if they would rather purchase a car programmed to protect them instead of pedestrians.
The waters get murkier still. If regulations forced manufacturers to install moral algorithms that minimised deaths, the majority of respondents said they’d buy unregulated cars instead, potentially undermining the much-vaunted safety benefits.
One of the report’s authors, Azim Shariff, of the University of Oregon, commented: “Would you really want to be among the minority shouldering the duties of safety, when everyone else is free-riding, so to speak, on your equitability?”
Alan Winfield, of the Bristol Robotics Laboratory, added: “Without transparency you cannot regulate, and without regulation, driverless cars are unlikely to be trusted. There’s a strong case for a driverless car equivalent to the Civil Aviation Authority.”
Public perceptions
What’s needed, of course, is more testing, but a fatal crash involving an Uber test vehicle in March did little to advance that cause.
In a recent survey of consumer attitudes, the American Automobile Association found that 73% would be too afraid to ride in a self-driving car, up from 63% in the previous survey. The biggest slip in confidence was among millennials.
In Arizona, where the crash happened, the previously supportive Governor, Doug Ducey, suspended Uber’s testing. In Minnesota, Senator Jim Abeler set about drafting legislation to ban automated driving systems until the companies behind them can prove they’re safe.
Harald Proff, of Deloitte, commented: “The US is in no way reckless when it comes to autonomous driving tests, but in Europe, and especially in Germany, rules are a notch stricter when it comes to putting cars on public streets.”
Indeed. A flagship project in Berlin involves four autonomous buses ferrying doctors and staff across the private grounds of the city’s Charite hospital, along pre-defined routes, away from public streets, at a maximum 12mph.
Here, roads minister Jesse Norman announced a three-year review of driving laws – an “extensive regulatory reform programme” intended to ensure “the right laws are in place before the widespread use of these vehicles on UK roads”. In the meantime, public tests continue.
On UK roads
Early UK road trials include: The GATEway Project in London; Venturer in Bristol; and UK Autodrive in Milton Keynes and Coventry.
GATEway is entering its final phase, which will see a fleet of driverless pods providing a shuttle service around a 3.4km route on the Greenwich Peninsula. In a world first, members of the public are invited to take part in the research, by riding in or engaging with the pods and sharing their opinions.
Developed by British companies Westfield Sportscars and Heathrow Enterprises, and powered by Fusion Processing technology, the pods have no steering wheels or typical driver controls.
Venturer has so far published the results of two trials involving a Wildcat road vehicle. The first – to understand handover of control between vehicle and driver – concluded that the functionality should “proceed with caution”.
The second – to consider how CAVs interact with other road users and junctions – suggested that making them drive more cautiously than the average human driver could create a traffic calming effect, resulting in safety and congestion benefits.
UK Autodrive, with partner Jaguar Land Rover, has sought to address the fact that, in times of heavy congestion, up to 30% of traffic consists of vehicles looking for parking spaces.
“In the future, connected features will alert drivers to empty spaces and autonomous vehicles will be able to drive straight to them,” said Tim Armitage, Arup’s UK Autodrive project director.
“Valet parking systems will enable autonomous vehicles to drop passengers at convenient points, after which the vehicle will leave by itself to undertake a further journey, or park out-of-town. As well as making parking less of a hassle, these new ways of parking and drop-off will allow cities to radically redefine their use of space, with far less land potentially needed for parking spaces.”
In February, Highways England announced the HumanDrive Project to undertake “the most complex journey across the UK, without driver input”, taking in country roads, high speed roundabouts, A-roads and motorways.
Business and Energy Secretary, Greg Clark, commented: “Low carbon and self-driving vehicles are the future and they are going to drive forward a global revolution in mobility. This revolution has the potential to be worth £52bn to our economy by 2035, and the opportunity to be at the forefront of this change is one we cannot afford to miss.”
In April, the Driven consortium, which is in receipt of an £8.6m government grant, unveiled its plan to run a fleet of Level 4 vehicles in urban areas and on motorways.
Level 4 CAVs might occasionally ask for a manual input, but will continue self-driving if they don’t get one. The project will culminate in journeys between London and Oxford in 2019.
Give us a date
So, if Level 4 testing goes to plan next year, when can we expect Level 5, full automation? Bloomberg New Energy Finance asked 300 automotive, energy and technology executives to name the year when US consumers will be able to buy a Level 5. Nearly 75% predicted the milestone won’t be reached before 2030. Suddenly, that sounds quite conservative.
Carrie Morton, deputy director of Mcity, the University of Michigan’s purpose-built test facility for CAVs, was more precise. “You’re going to see in the next couple of years isolated pilots of shared, automated Level 4 vehicles,” she said.
“And the distinction between Level 4 and Level 5 is that Level 4 have to operate in a very specific operating domain – it can only go certain places. You’re going to see that really soon. All the while the personally-owned vehicles you and I drive are going to have increasing levels of automation.”
As IMI CEO, Steve Nash, noted in the April issue of this magazine: “It requires around 2.5 terabytes of data to enable an A380 Airbus to fly autonomously across the Atlantic, whereas 45 terabytes of data are required to equip a Level 4 autonomous car to handle routine driving tasks.”
Is this an achievement comparable with putting a man on the moon? No. Is it as ground-breaking as the invention of the motor car itself? Probably not. It is simply progress. In the next decade, humans could set foot on Mars. Here on earth, cars will no longer need drivers.
