Suzy Charman acknowledges nervousness about her first trip in a "robotaxi".
Being driven around by “the invisible man” is a little disconcerting, especially when your head is full of all the statistics on road collisions and casualties you can imagine.
As a road safety expert, Ms Charman was given a test ride in a Waymo autonomous taxi in Atlanta, Georgia, before their anticipated roll-out in London sometime this year. Several companies are running tests in London while they wait for the government to give its approval.
After a few minutes of nervousness, a little like the first time on an plane, she felt comfortable.
“The displays in the vehicle are very reassuring,” she said. “You can see what the car can sense.
“There’s still a steering wheel turning, which is quite funny. But it’s really quick to feel ‘It’s got everything under control, I can relax’.
“My car was very courteous and confident, not hesitant. If there was another vehicle, pedestrian or cyclist it slowed down slightly or gave space, as you’d hope it might.
“Where it knew there was a gap it didn’t hold back, or dither.”
By the end of her 24-minute journey, the executive director of the Road Safety Foundation was convinced the technology could not only work in London, but could make the roads safer. The lack of a human in control, as disconcerting as that may be to some, means no speeding, no tiredness, no distractions. In short, no rule breaking.
So, can driverless cars really make the roads safer?
Medieval streets
Waymo test vehicles, white Jaguar i-Paces with large rooftop sensors, are currently mapping the city’s neighbourhoods – effectively their own version of ‘The Knowledge’, the test that Black Cab drivers take to prove they can find their way around – learning the roads, the routes, and their quirks.
The US company, which has received investment from Abu Dhabi’s Mubadala Capital, gave an update to industry executives last month and told The National it has a “limited fleet of vehicles in London and will steadily expand the fleet as we grow our operations in the city. We've been driving in London since late last year, with trained specialists operating the vehicles. Our intent is to open to the public later this year, contingent on government approvals”.
They are not the only ones. Chinese firm Baidu has signed deals with ride-sharing firms Lyft and Uber. UK firm Wayve is also in the vanguard, with UK government funding and a tie-in with Uber.
They are each getting used to the UK capital.
Compared with the grid system of America or the planned roads of Abu Dhabi – cities that were early adopters of autonomous vehicles – London is, to put it bluntly, a bit of a mess.
As anyone who has attempted to navigate its streets can attest, it’s not a straightforward city to drive in. It’s slow, narrow, winding, full of roadworks, and has as many dead ends as it does roundabouts. It also has more pedestrians and cyclists, neither of whom are known for sticking to the rules of the road.
It’s understandable that people may be sceptical that a robotaxi could cope.
But David Wong, head of technology and Innovation at the Society of Motor Manufacturers and Traders, says the cars’ abilities should not be underestimated. They have passed every test so far.
He travelled in a Waymo in downtown San Francisco at rush hour. “It’s was 5pm, quite chaotic, and hilly,” he said. “It was near the financial district with plenty of people coming out of offices, vulnerable road users everywhere. It handled everything with aplomb. Those were chaotic streets like London.”
The cars just need time to become attuned to local traffic, he said.
He has witnessed huge improvements as the technology has developed for more than a decade, from jerky, slow vehicles that could be outpaced by walking, to commercial operations on private land such as airports or opencast mines, to now being as good as humans.
“But there is a huge difference to humans in that they won’t do the things humans do,” he said.
“The American statistics show 94 per cent of all accidents were down to human error and the British [version] 88 per cent.
“Two thirds of all collisions in Britain in 2023 were down to human faults, such as speeding or being distracted. So, in theory, two thirds of all collisions could have been avoided.”
He also points out a broader safety issue, that many people prefer to be alone in a vehicle without the need to talk to a stranger, particularly female workers finishing their shifts at night. “A lot of nurses in America prefer being driven by Waymo,” he said.
Digital background
But there are inevitably pitfalls. The American experience has thrown up examples of cars blocking emergency vehicles after getting stuck, coming to a standstill after bollards were placed on the bonnet, or even being paralysed after customers left doors open.
The commissioner of the Metropolitan Police, Mark Rowley, was in San Francisco last week for meetings with the FBI, but took time to visit Waymo to discuss how it would abide by London’s traffic laws and to be reassured that the taxis would get out of the way in a blue light emergency.
Other cynics have questioned how the cars would cope if there were temporary traffic lights or unplanned diversions signalled only by a sign on the pavement.
The process should be used as an opportunity for the entire road network to be improved, according to Ms Charman.
“London’s roads have evolved, rather than being designed. We’re trying to cram a lot into our space, such as cycle lanes on the road, for example,” she said. The AVs’ data could be used to improve the road experience for all users by synching with the network that already exists.
“We need consistency about how we lay out signage and make sure they’re clear for humans, as well as autonomous drivers.
“The Met Police, the borough councils and Transport for London are highly likely to have a protocol for exchanging information with the AV operators.
“But we do need better digital background information on our roads. That would be a good by-product.”
By taking a human out of the loop, she is hopeful that the number of accidents will fall. The road Safety Foundation campaigns for a safe system that reduces the likelihood of serious harm if there is a crash, providing evidence for reduced speed limits where appropriate and improved road layouts.
“Now, we have to think about what a safe system means with AVs in it,” said Ms Charman. “We should use this as a reset for the rules of the road, and think whether our roads are safe enough generally.”
For her, obeying the rules can still result in serious injury. “This is an opportunity to think about speed for AVs that may calm the rest of the traffic. Ideally, they would not necessarily reach the speed limit, because the limits are a touch higher than is satisfactory in terms of survivability,” she said.
“You can’t have them grinding the network to a halt and frustrating everyone. You have to get the balance right. But AVs have been promised to be a safety gain, so they’ve got to do better than us.”
She is reassured that London is the first part of the UK that will adopt AVs as it has led the reduction to 20mph speed limits.
“It would be a greater concern if they were to begin in other areas where the speed limits are higher, and there are more single lane A-roads with T-junction entries which are the highest risk points. Roundabouts are much more forgiving.”
The mapping by companies such as Waymo means it should expose black spots. “I’m really interested in using data from AVs and their sensors to be able to automatically code the things that are really important for road users, whether they’re driving, walking, cycling or even on horseback.
“It’s all about making sure that we learn during the pilot period. You’re never going to have that 100 per cent guarantee, we just need to keep our eyes open to the data.”
Who’s on the road?
The robotaxi companies are convinced their cars are up to the task.
The government will certainly hope so. It predicts the UK self-driving vehicle industry will be worth £42 billion and create 38,000 more skilled jobs by 2035.
Each company uses AI to control the cars and read the road, but have developed their tech in different ways as they race to be the first to conquer the driverless market.
Wayve’s company’s technology uses artificial intelligence models which learn how to drive using video content, driving data and recognising patterns.
It intends to deploy its “supervised autonomy software” in consumer vehicles from 2027.
It recently raised $1.5bn (£1.1bn) from investors, including Microsoft and chip giant Nvidia. Major car manufacturers Mercedes-Benz, Nissan and Stellantis have also injected cash as part of the deal.
The Series D funding round, which is one of the largest ever for a British start-up, values the London-based business at around $8.6 billion (£6.4 billion).
A £500 million Sovereign AI Fund is set to be launched in April at Wayve “to give British companies access to funding, compute and other support to compete and succeed globally”, the government has said.
Technology Secretary Liz Kendall said Wayve was a “powerful example of the strength, ambition and potential of Britain’s innovative firms”.
The tech
Waymo, owned by Google's parent firm, Alphabet, operates by absorbing all the information it can during test drives, with a driver at the wheel, before its AI systems are unleashed for a live journey. It essentially remembers everything about a city’s streets.
“We’re not at the point where you can throw an AV into an unknown location, and then it figures out that it needs to go left and right,” said Frank McCleary, partner at management consultants Arthur D Little. “They have to understand the streets around them.”
Due to Waymo’s restriction to operating in a known area, this can limit the take-up. “I live in Atlanta and Waymos are available Downtown. But I have to use an Uber to get to the airport.” Like the take-up of Ubers, he recognises they need to start small then expand as customers try it out.
He said as Waymo rolls out city by city, it is getting faster from mapping to full driverless operations, but would still take several months to “learn” London.
After understanding the layout and road system, it is then “understanding the nuances to reduce the number of edge cases that an AV would potentially see so that you can have a larger area of operations or footprint at the starting point,” said Mr McCleary.
Constraints
The real challenge for AV manufacturers is the ability to anticipate, rather than memorise, so they can operate in unfamiliar environments.
“If you think of yourself as a driver, you haven’t seen every single situation that you could ever encounter,” Mr McCleary said. “But you as a human use reasoning based on your learning on how to react to that new situation. That’s what they’re trying to get to.
“The current models are able to resolve 95 per cent of the problems fairly quickly, but you can’t unleash it on to the world until you’ve got 99.9 per cent.
“Whether it’s robotaxis or personal use, if you want to get to the next level it’s going to have to deal with edge cases otherwise we might as well just ride the tube or bus on a known track. We need to get to models that are able to solve a problem that they’ve never seen before.”
Opportunity knocks
The rest of the UK will be watching London’s development with interest. Driverless electric shuttles are planned for Milton Keynes, 50 miles north-west of London, and have begun ferrying passengers near the central railway station under human supervision for now.
Smart City Consultancy, the company behind the scheme, is poised for Waymo et al to get the go-ahead so it can ramp up its operations. It intends to operate on a more limited scale, offering a “first mile, last mile” service on a pre-planned route which is likely to feed public transport.
Like Waymo it uses a Lidar (light detection and ranging) system, combined with satellite positioning and human oversight from a control room, to ensure its eight-seat vehicles can operate safely. For now, it is showing an abundance of caution to help reassure both the local council, and potential passengers, that it is safe.
During a test-ride by The National, it kept below a capped speed of 25 kmph (despite the vehicle having a top speed of 40 kmph) and gave other road users a wide berth.
Like other AVs, the company is finding it is the humans who do not stick to rules. Getting across a yellow box junction took a few attempts as drivers coming the other direction kept encroaching, while a car dropping off a passenger blocked the middle of the road, leaving it nowhere to go. A human driver would likely have hit the horn at least.
All this is flexible, and Ian Pulford, Smart City’s director, said that with time the vehicles would get bolder. He sees opportunities in city centres such as Milton Keynes – already known for adopting innovations such as drone deliveries, pavement gritters and chewing gum removers – to improve public transport as car space becomes limited by planned buildings. Removing the driver reduces costs radically in comparison to a bus, making routes more commercially viable.
It would also work in areas such as industrial parks, large university campuses, and he has had one vehicle operating at the National Exhibition Centre in Birmingham. Their modest size means they also fit in a normal parking space and have easy access for wheelchair users via a step lift.
All aboard
The next step, for both Smart City and the robotaxi giants, is to get customers on seats and prove any doubters wrong.
“It’s dependent on the city, but as Waymo has expanded and had consistent traction, positive press, it has driven more willingness to try,” said Mr McCleary.
“The whole concept really is about eradicating bad drivers. If the technology works, we should all be in vehicles with good drivers.”
