All you’ve been told about driverless cars is wrong

Autonomous vehicles have long been proposed as a solution to the biggest problem in transport: human beings. The figures are stark and hard to argue with. According to the World Health Organization, 1.35 million people die every year on the planet's roads and 93% of those fatalities are caused by human error rather than systems failure.

I've said it before, but it bears repeating. Roughly 1.4 billion vehicles on the road give us a simple calculation: for every 1,000 of them in existence someone dies, thanks to a careless driver.

And that's not all. Country by country, road fatalities have actually been increasing, despite cars getting smarter and more connected. In the US, for example, 2019 deaths in traffic accidents totalled 41,694. Five years earlier, the figure was just over 36,000. In the UK, 2,169 people lost their lives on the road in 2019, while in 2015 it was 1,927, a broad trend that is reflected nearly everywhere, with occasional blips.

But we are living in a blip, of course - 2020 is likely to show a significant drop in road traffic deaths, as we were forced out of our towns and cities by the Coronavirus. Numerous studies have shown that remote and home working will have a long tail: lockdowns have permanently changed our behaviour as some organisations realise that they can function without expensive premises. That means fewer commutes.

Nevertheless, the conclusion remains stark: if traffic was a virus it would be regarded as a global health emergency, one that kills between two and seven times the number of people lost to seasonal flu.

But as a Westminster Energy, Environment, and Transport (WEET) Policy Forum revealed this week, there are problems with any simplistic vision of autonomous, on-demand transport sweeping into our cities and saving lives. For one thing, some of the statistical models unveiled by speakers seemed to omit humans entirely from their calculations.

Andrew Pearce, Practice Director for Intelligent & Smart Technology, UK & Europe, at engineering and project management company Atkins, said:

Why am I still building gantries and traffic lights if self-driving cars are just around the corner? And does this matter? Well, it matters because of how long our infrastructure lasts. The vast majority of what we're building at the moment is still going to be here when we get to the Net Zero threshold in 2050.

A fair point: we need to consider the design of our future infrastructure needs, rather than let them slowly accrete, as the streets of London did over centuries. But no local authority will spend money on infrastructure for a hypothetical future. Electric vehicles are with us already, but good luck finding a charging point.

Pearce then shared some projections of the savings that the UK could make when we hit Level 4 and 5 autonomy at scale, the point when humans stop being active participants in transport and are no longer expected to grab the wheel. At that point, private, human-operated cars may be pushed out of our cities completely. The US has already redefined the legal concept of a driver to include an autonomous system.

If we move on to AVs [autonomous vehicles] only, you'll see the traffic lights have disappeared. So, in a world where there are no manual cars, where people cannot drive the car themselves, there is now an opportunity for savings. We are starting to get a benefit because the traffic lights aren't required.

But if we really trust the cars, then we don't need the loops, the radar, the detection part of the system either, because the car tells us ‘I'm here, I'm coming' and we save more money on infrastructure.

Pearce went on to claim that streetlights and other urban features would no longer be necessary either. In total, he claimed this could create UK savings of between £320 million a year and 12 times that amount - a vague scale if ever there was one.

Yet the problem with this big-picture view is that it is bunkum: driverless cars may not need that infrastructure, but people do. Humans still need to see at night and be able to cross the road safely, so it is more likely that we will need entirely new infrastructures to help manage the relationships between people and traffic.

That would mean more, not less, expense. After all, part of the promise of driverless cars is that they supposedly enable us to focus on people's needs, on designing cities around pedestrians rather than traffic. Putting the needs of autonomous cars first - or rather their lack of needs - is an inversion of the technology's promise.

I put this to Pearce, who clarified his point, but then appeared to reinforce it.

I'm not advocating a future of AV only, but it was felt that someone needed to model and cost this one extreme scenario. There may be some difficult decisions to come by policymakers if we want the benefits of new technology. We may not be able to maintain all the current rights [for human beings?] as well - not a popular thing to say, perhaps.

Why autonomy equals more cars on our roads

Interesting. But there is another, much bigger problem with the futurist view that autonomy is a cure for the global infestation of lethal tin boxes - sometimes driven by drunks, road-ragers, and people staring at their phones. It too is nonsense. The reality is that AVs will massively increase the infestation of tin boxes, it's just that humans won't be driving them.

Why is this? I'll explain. It has often been claimed that on-demand AVs would mean less traffic and therefore less crowded, friendlier streets - an occasional pod full of rooibos-sipping hipsters drifting by, an Amazon robot or two, but little more. In fact, the reverse is true. There would indeed be fewer cars - a lot fewer, in fact - but a huge increase in traffic: by nearly 50 percent (more on that figure below).

This counter-intuitive outcome is because of the inconvenient truth about private car ownership: the average driver-owned tin box is stationary in driveways and car parks nearly all of the time. (This is why some organizations rely on parking revenue - councils and hospitals among them.)

But on-demand, autonomous vehicles would be in constant use, with many filling the revenue void left by the closure of car parks. So, while there would be fewer vehicles overall (once private ownership becomes an expensive leisure pursuit), AVs would be on the road most of the time.

Even motoring organisation the RAC accepts this view. Steve Gooding, Director of the RAC Foundation, explained:

The cars that many of us have got used to owning, having in our driveway, having to drive to work, that prized piece of ownership spends most of its life being prized statuary, rather than a useful device. The average British car spends well over 90 percent of its life still.

Atkins' Pearce confirmed this in his presentation too, quoting a recent Lisbon study by the PTV Group on what would happen to that city if private cars were no longer allowed in. The conclusions are fascinating.

On the plus side, the survey found that up to 90% of cars would no longer be necessary, the equivalent of 210 football fields of land could be freed up by reduced parking needs, and CO2 emissions would fall by 27% : all great things in themselves that satisfy Net Zero objectives.

So, fewer cars on the road, right? Wrong. That's because the remaining vehicles on the road - autonomous taxis, for example - would be on the move constantly and often running empty. In total, the study found there would be a near-50% increase in traffic, Pearce told the forum. And remember, in his view this uptick would be vehicles that don't need traffic lights or street lighting.

Full-beam nightmare

That's hardly a vision of a pedestrian-friendly city that resets the balance between people and machine. Yes, it would remove human-driver error as a cause of most accidents, but certainly not systems error or pedestrian error when faced with a 50% increase in moving vehicles.

Factor in the colossal boom in urban air traffic envisaged by many technologists - as my recent report revealed, more autonomous delivery drones in the sky in one day than there are planes in a year - and our future cities are beginning to look like a nightmare.

Consider the real-world output of this generational thought experiment about our transport needs: thousands of noisy rotorcraft in the sky, many over parks and public spaces to keep providers' insurance costs down. Plus, a constant stream of autonomous cars - twice as many on the move as there are private cars moving in cities today. All of this, apparently, is intended to make those cities more liveable for ageing, isolated humans. Clearly that's nonsense.

That outcome would be nothing like the stated reasons for developing these technologies in the first place: to make our cities safer and greener. And don't forget Elon Musk's announcement this week that you can now buy a Tesla with Bitcoin, thus wiping out the carbon benefits of owning an electric car by paying for it with a coin that's largely mined via Chinese coal.

Crossing a busy road to the future

The fact is the future is messy and rarely as simple as technologists claim, which brings us to the next challenge. Most of the claims made for autonomous transport's benefits concern an era that has not arrived yet at scale. But it needs to arrive big for any of this to work.

For the claimed benefits of AVs to become reality, we need trustworthy, reliable Level 4 and 5 autonomy that's as safe in narrow, bendy streets full of people, in rural areas with poor network connectivity, and in urban canyons of skyscrapers, as it is in open, grid-like cities of wide, straight roads, low buildings, and few pedestrians (ie much of California).

It also needs to be safe in all weather conditions: a 2021 study showed that LiDAR sensors are confused by heavy rain, and so see drops of water as objects to avoid. That's fine in LA, but less so in London or Lahore.

And it needs to be standardised: there's little point jumping in an AV from the UK to France, or from Texas to Mexico, if the vehicle doesn't work when it gets there, because of different standards in the supporting infrastructure.

And it demands careful regulation, given that consumer data will be the real currency of this new transport economy. And it needs a consistent, robust, reliable, fast, and always available data/communications infrastructure.

To get to that far-off point will be messy, complex, wasteful, and almost certainly dangerous, because it will mean sharing our roads and cities with a mix of entirely different technologies and communications systems. Not to mention Level 3 autonomy, where the driver may not understand that he or she still has responsibility for vehicle control and safety. The inherent dangers of that world may damage consumer confidence in the whole project.

Pearce said:

The nightmare scenario in commerce is having mixed systems. Take the motor industry when it comes to propulsion. They don't care if it's petrol or diesel, they don't care if it's electric, they don't care if it's hydrogen, but they absolutely do not want to produce all of those things and have the spares and the support. It's the same with autonomy. The mixed scenario is the most expensive. Don't forget that.

For the RAC Foundation's Gooding, we may be thinking about AVs completely wrong, and so need to change our direction of travel. We could design them to take us between places on motorways, for example, then offer a different type of vehicle or service within cities.

He said:

We need to be thinking more about the commercial applications. Think about the vehicles that actually keep our cities functioning, where the value of autonomy can come in reducing operating costs […] with vehicles on predetermined routes that can be intensively digitally mapped.

We could be thinking of getting away from deliveries happening during the times when the rest of us are moving around, and perhaps have them made on an automated basis at night, quietly in vehicles that don't have to be driven and don't need headlights on. And let's think about other vehicles, like refuse collection.

The way our public transport services operate, we are still struggling with. The vast majority of our railways are gated, geofenced, discrete networks, but they don't run driverless. Or the bus, maybe we can have that vehicle operating autonomously.

We don't need a driver, but let's bring the conductor back. We're all thinking about how we can make our streets, our cities, safer and more women-friendly at night. Wouldn't we feel safer if a night bus didn't have a driver, but it did have a conductor - not sealed in a bulletproof glass box, but sat amongst the rest of us?

And we can rethink what that might mean for the design of a vehicle in places where mass transit is still an issue. It needs to be easily accessible, easy to keep clean, quiet, comfortable, and non-threatening. In a rural context, a bus could look more like a Volkswagen camper van than a Routemaster: lightweight, efficient, environmentally sensitive, and showing a smiling face to the world.

What we need here is joined-up thinking.

My take

No one denies that autonomous vehicles could democratise transport by making it available to everyone, regardless of age, disability, wealth, or sobriety. And we must acknowledge that electric transport and connected vehicles are part of a move to make our roads greener, smarter, and more sustainable.

The model of private ownership of expensive, depreciating, oil-guzzling cars that are parked nearly all of the time has long been absurd, wasteful and - the death figures prove - dangerous when drivers finally get behind the wheel.

Our car addiction is killing us and the planet. But that doesn't mean we should abandon common sense and critical thinking about what sort of world we are designing to take its place.