Britons’ bad driving habits are to be taught to driverless cars to train autonomous technology to think like a human.
Oxford University scientists are analysing footage from thousands of CCTV cameras and drones.
The Telegraph reports that scientists are analysing the footage with technology that uses computer vision to track road users’ movements.
Machine learning technology then extracts the predictability within natural human behaviour to create real life simulation scenarios and teach driverless cars how to cope with them.
According to The Telegraph Oxford University professor Simon Whiteson said that a large amount of the footage the team uses to teach driverless vehicles is of bad drivers.
‘It’s raw video, so the quality of the driving is exactly what you see when you are on the road yourself.’
‘We don’t want to learn perfect behaviour, we want to learn naturalistic behaviour, and if people drive badly we want that to be what they (driverless cars) capture and learn,’ said Whiteson.
The technology also analyses the movements of pedestrians, cyclists and motorcycle users to predict their behaviour in an infinite number of scenarios. This technology is being used in two government-funded trials worth £7m, which aim to test driverless vehicles on real roads from the start of next year following virtual trials, reports The Telegraph.