Regulating the safety of autonomous vehicles using artificial intelligence

In the last five years, there has been increasing enthusiasm for autonomous (also called self-driving or driverless) vehicles on public roads in the UK. There are many benefits claimed for them: lower energy consumption, better use of infrastructure, fewer accidents and mobility for all, including those who, for reasons of age or infirmity, are unable to drive. The UK Government has been encouraging, consulting on and facilitating the development and introduction of driverless cars and “connected autonomous vehicles” (CAVs) that can communicate with each other and the infrastructure. In February 2017 they published a draft Bill to legislate for compulsory insurance of CAVs, with explanatory notes showing that the Government expects driverless cars to be introduced in the UK in “five to ten years” – that is, between 2022 and 2027. However, there appears to have been no realistic analysis of how the safety of such vehicles would be regulated or how their interactions with other road users could be controlled. This paper reviews the safety regulatory regime for other automated transport systems and discusses how similar processes might be implemented for CAVs. It concludes that the government’s timescale targets are unrealistically optimistic