A UK-based artificial intelligence company has a fleet of data-collecting cars that will gather information from roads in Croydon and Bromley as it looks for an alternative to driving to work.
FiveAI hopes its technology can be used to run a driverless car service to replace the commuter car and help reduce congestion, emissions and the cost and time of journeys. Ben Peters, co-founder of Five AI, told Spotlight: “We talked to municipalities, and their number one problem was individual vehicle use… personal car use is still really high.”
He added: “We’re trying to figure out a shared mobility service, initially for six to eight people, that is so attractive to consumers that they give up using their cars to commute.”
A study by INRIX found that London commuters spent an average of 74 hours stuck in congestion during peak commuting hours in 2017. The study estimated that the total costs of congestion to all UK motorists at £37.7bn.
Peters said: “Commuters are paying a lot of money for their cars, and end up spending most of their time frustrated and stuck in traffic. As you get to the outer boroughs the distribution of people’s origins and destination is so high that it’s hard to offer a public transport option that meets their needs.”
This is an issue Peters hopes FiveAI’s driverless commuter car service can solve. The government has committed £120m into the research and development of driverless vehicles through the Centre for Connected Autonomous Vehicles, founded in 2015. Through the scheme FiveAI has led a consortium of companies including McLaren and TRL, which were granted £12.8m to assess the safety and validity of using driverless vehicles to replace the urban commuter car.
In order to make the driverless service a reality, there are several unsolved pieces of science that need to be resolved. Peters explained: “Firstly you need to have an accurate view of the world around you. You then need to understand it, so you can predict how the situation will evolve in the next three to five seconds and decide how you will interact with it. It is novel science, which is really only now finding its way from academia to industry.”
In order to improve the capabilities of its autonomous vehicle AI, FiveAI must test its model of road use against real-world situations. The company will spend ten months gathering information in Croydon and Bromley by driving cars fitted with radar, LIDAR (a 3D digital imaging system) and video cameras that will constantly collect images of the roads around them. This will then be analysed against FiveAI’s “naive model” of what they expect the roads to look like, to see how they can improve the AI’s ability to interpret and understand how to navigate the roads.
Peters said: “For our service to eventually operate in London, we have to learn about the road layout, topology and traffic flow. Because the system is under development, and will be for the next few years, there are areas where our model’s understanding of the world needs improvement and driving our vehicles around London lets us do that in a large-scale fashion.”
Through this process, the team at FiveAI will teach the system to drive the car. “The way we design the system allows us to introduce a corrective action, to make sure that we’ve fixed any issue and fixed it for good,” said Peters. “This method gives us the confidence that eventually we’ll have performance that is safe to use in a city.”
The next step will be to test the AI across “billions of miles” in a simulator. In 2019, Five AI will begin a supervised trial on London roads with a safety driver behind the steering wheel with the aim of testing the autonomous vehicles without a driver in the early 2020s. “By the 2030s,” Peters predicted, “we should expect to see these vehicles operating in most of our cities.”