British artificial intelligence (AI) and autonomous shared transport company FiveAI is testing autonomous cars in south London.

The company, which seeks to develop a European integrated smart transport solution, is testing five autonomous vehicles in Bromley and Croydon, in advance of planned passenger trials in 2020. The cars are running day and night with safety drivers onboard.

The self-driving systems use data that was gathered in manual tests on UK roads last year. FiveAI co-founder and CEO Stan Boland said, “Safety and trusted partnerships are crucial to everything we do. We’ll keep residents informed along the way, working closely with the London Boroughs and Transport for London.”

FiveAI believes that a European solution is essential in a world in which much driverless technology development is being spearheaded by US and Chinese companies. This is because of the more complex environment of “narrow, curving streets, parked cars, cyclists, pedestrians, unpredictable weather, and unique, localised behaviours” that typifies most European cities. To serve these cities safely, we “must develop highly intelligent and robust tech in Europe for Europe”, says the company, which operates from six locations in the UK: London, Bristol, Cambridge, Edinburgh, Oxford, and Millbrook in Bedfordshire.

FiveAI was a member of a consortium of companies that was awarded a £12 million grant to develop the StreetWise autonomous system in 2017, under the government’s new industrial strategy and its Future of Mobility vision. Future of Mobility minister Jesse Norman said, “The long term potential for self-driving vehicles is huge – to improve road safety, tackle loneliness and isolation, and create economic opportunity.”

The programme has been welcomed by at least one AI and autonomy expert – Dr Sally Epstein, Machine Learning expert at Cambridge Consultants. She said, “This deployment by FiveAI is exactly the kind of rollout that should be used to move through the levels of automation. “The company is approaching the problem of full autonomy sequentially, mastering each level of autonomy as they go, while understanding the nuances of the environment that they hope to eventually operate in.” However, Epstein warned that realism and transparency are essential, and that true autonomy is a long way off for the transport sector. She said, “The goal of passenger trials in 2020 is ambitious, but consumers should know that we’re nowhere near to having genuinely driverless cars on public roads.”

According to Epstein, the outstanding challenges are:

  • Mapping data will become out of date as soon as it is collected. Street furniture is constantly changing, with roadworks, accidents, and more, while human movements in urban environments will remain stubbornly difficult to predict.
  • Mapping data alone is not sufficient for the challenge of driverless cars. Maps must be intelligently integrated with sensor data, in order to keep drivers and other road users safe.
  •  Policymakers need to establish how vehicle decision-making will be understood after an incident. There must be adequate transparency in the system for consumers to trust it, she said.
  • And when fully autonomous cars do eventually arrive, explaining how their decisions are made – particularly following accidents – will be much more important than any statistical proof that they are safer than traditional, driver-operated vehicles.

By some estimates, there are 1.2 million deaths every year on the world’s roads – or one fatality for every 1,000 vehicles in use. Human error is responsible for over 95 percent of those accidents. However, public support for – and trust in – driverless cars has fallen significantly, according to the American Automobile Association (AAA), in the wake of a number of fatal accidents last year involving cars that were running under software control. At present, an overwhelming majority of US citizens – including so-called millennials – would not feel safe in completely autonomous vehicles.