Will we ever prove autonomous vehicles are safe?

1 min read

One of the problems with safety critical software is being able to test it rigorously. The larger the code base, the greater the problem as the ability to exercise every IF and OR statement becomes increasingly problematic.

It’s something of which those writing software for the aerospace industry are acutely aware and even the auto industry is affected; a story which is something of an urban myth suggests that if someone opens the rear offside window whilst Radio 4 is on, the ABS system fails. Similarly, in the early days of electronic control, the braking system of a British luxury car was alleged to ‘fail dangerous’ when the sunroof was open, instead of ‘fail safe’.

Confirmation of the task facing those developing autonomous vehicles comes in a report from research organisation RAND, which says autonomous vehicles would have to be driven ‘hundreds of millions of miles and, under some scenarios, hundreds of billions of miles’ to create enough data to clearly demonstrate their safety.

Susan Paddock, senior statistician, noted: “The most autonomous miles any developer has logged is about 1.3million – and that took several years. This is important data, but it does not come close to the level of driving that is needed to calculate safety rates. Even if autonomous vehicle fleets are driven 10m miles, one still would not be able to draw statistical conclusions about safety and reliability.”

You could argue, of course, that manual cars have been driven the requisite ‘hundreds of billions of miles’ and we’ve proved conclusively they, and their drivers in particular, are dangerous.

Time for a red flag in front of an autonomous vehicle? Probably not.