“We” are on the cusp of a profound change in how we move people, goods and services around our towns, cities and countryside. This is driven by extraordinary innovation in engineering, technology and business models,” reads the government’s industrial strategy, published last year with the mandate of building a Britain that is “fit for the future”. Connected Autonomous Vehicles (CAVs) – vehicles capable of navigating their environment without human input – are name-checked 12 times in the white paper, which claims that the UK has an opportunity to “assert global leadership” in the field.
Self-driving cars will soon become a reality on Britain’s roads. In 2015, the government launched the Centre for Connected and Autonomous Vehicles (CCAV), a joint policy team between the Department for Business, Energy & Industrial Strategy and Department for Transport. CCAV is providing over £250m in funding, matched by industry, with the belief that CAVs “could change the way we travel”. A £5m test trial, conducted as part of the government’s 5G strategy, found numerous benefits associated with self-driving technology, including improved productivity. The government has also established the “MERIDIAN” initiative, a “co-ordination hub” that seeks to convene UK industry around a set of “long-term strategic priorities” aimed at keeping the UK on the cutting edge of autonomous vehicle development. Chris Grayling told the Association of British Insurers last year that the UK is “well positioned not just to follow changes in motoring technology over the next couple of decades, but to lead them”.
One of the British companies leading the CAV revolution is Cambridge-based FiveAI, founded in 2015. Launched by a group of tech industry veterans, the firm is planning a “fully automated shared mobility service” – a taxi fleet, effectively, of self-driving cars – as an alternative to driving to work. The firm, which is already testing its CAVs in Bedfordshire, made headlines in August when it announced that it will begin testing on the streets of London, collecting data from roads in Croydon and Bromley.
The company’s co-founder, Ben Peters, says his team has always wanted to do something “technically difficult and technically interesting”. After reading “about 2,000” research papers on AI between them in a year, they drew up a list of six problems with commuter cars that their technology will aim to solve. FiveAI’s research has found that cars, taxis and private hire vehicles account for almost half of the street space used in London, but only 13 per cent of distance travelled. A third of commuters in the capital have no “better option”than their own vehicle, and up to 5,900 Londoners die a year from air pollution. TfL recorded 3,881 people as having been killed or seriously injured in London traffic collisions in 2017, and 17 days in a London drivers’ year are spent in traffic. According to the ONS, UK households spend more on transport (mostly buying and fuelling cars) than they do on food, clothing, recreation, health or education.
“This model of car ownership we’ve had for the last 100 years is being exposed as a net negative thing for society,” says Peters, who contends that adoption of CAVs will increase seat use, reduce congestion, and free up the “economic burden” of owning a car.
Peters says the roads of the future will be driven by “heterogeneous” transport – bus-like vehicles for rush hours, sleepers for long-distance travel and smaller occupancy units for night-shift workers. FiveAI plans to start with shared, zero-emission shuttle services that travel a “relatively fixed route” in areas on the outer rims of London where public transport is limited. It’s for this reason that its cars are currently learning the roads of Croydon and Bromley, where it hopes to begin offering services in 2020. As more routes are learned and services added, Peters says, “eventually those routes cross over, and become more and more dense, until you can go from anywhere to anywhere else.”
FiveAI doesn’t manufacture self-driving cars, but builds systems to be integrated into existing models. The current models use 14 cameras, three lasers, six radars and GPS, all of which combine to gather data on pedestrian movement, road conditions and a huge multitude of other factors. With this data, it constructs an image of the surroundings, “deep learning” – a type of machine learning based on repeated processing of different scenarios – to build an understanding of the environment and the responses needed to navigate it.
This requires serious computational power: the on-board computers are able to process 300 trillion calculations a second. The hardware is spread across the vehicle’s roof-box, boot and bumper. Large-scale data collection is needed to allow the car to deal with an effectively limitless number of scenarios. Cyclists, for example, are one category of road user, a courier might cycle in a different way to a commuter, who in turn will use the road differently to a tourist on a hired bike. The job of FiveAI’s CAVs is to learn how to use the roads in co-operation with all possible types of cyclist, and anything else it might meet on the road.
Alongside real-world testing, the firm has built simulations in which countless variables are created to test and train the AI. Weather conditions, road surfaces and even shadows can be added.
But while FiveAI’s technology is impressive, Peters explains that “a whole series of different techniques, sciences and disciplines have to come together” to solve the engineering problem of autonomous vehicles.
And while around 40 out of the top 50 experts in fields associated with CAVs are European, Peters says other countries are beginning to pull ahead.“If we don’t solve [autonomous vehicles] in Europe,” he warns, “our transport in cities and systems will be delivered by US or Chinese companies, which I don’t think is necessarily a good thing”.
For Peters, this problem is frustrated by two factors, the first of which is regulation. In March, the government announced a three-year review of driving laws, which will examine the legal issues of widespread CAV use, such as responsibility and associated risk. Peters says this “regulatory clarity” is needed fast, or companies will be more hesitant to commit to the technology.
“From a regulatory perception, China and the US could be ahead of us if we don’t get to grips with setting up certification, licensing and other regulatory instruments to make these services real,” he warns.
The second factor is a lack of investment. In 2018, the two largest funding rounds for automotive technology have been the $2bn and $600m raised by Faraday Futures and Lyft respectively: both firms, based in the US, are testing CAVs. Other companies, such as Waymo, have the backing of technology giants such as Alphabet (Google’s $748bn* parent company). The cash offered by the UK government pales in comparison, and private investment is harder to come by.
Peters suggests that European firms follow the leads of firms such as Google and Amazon, which target their technology directly at consumer demand, winning large markets quickly. There is a tendency, he says, for European companies to lack focus; because they work across many fields, they “spread their science thinly”. Success, for FiveAI, involves a “laser sharp” focus on application.
That said, FiveAI is far from the only UK company in the game. The government’s £120m in CAVs supports over 70 projects and 200 partners. Britain’s considerable base of automotive manufacturers, universities, and insurance companies give the country a head start.
On the roads between London and Oxford, FiveAI rival Oxbotica is testing its own technology after a £14m investment – the same amount raised by FiveAI last year – in September. Oxbotica is part of the Driven consortium, which is looking not only at how CAVs navigate, but also how they communicate with each other. The government has predicted that the CAV market in the UK will be worth £28bn by 2035, and automakers that sell and manufacture in the UK are looking for ways to get in early.
Peters is, as you’d expect, bullish about the transformative potential of CAVs. Implementation is for him a question of when, not if, and it will be felt most by those in the (non-driving) seat: “This is honestly the most transformative change that is coming over the next decade. It’s going to completely change the way that transport systems in our cities operate.”