Will software verification issues keep the driverless car off the road?

Recalls are, almost always, hardware related. A quick scan of a recall website shows just one example of a software based recall; a particular model is said to have faulty tyre pressure monitoring software. Yet today's cars feature huge amounts of code - and testing that code exhaustively is time consuming. There are apocryphal stories about how a particular car's ABS can be disabled if someone opens the rear offside window whilst Radio 4 is on. Similarly, in the early days of electronic control, the braking system of a British luxury car was alleged to 'fail dangerous' in some instances instead of 'fail safe'. Stories such as these highlight the challenge in finding the unlikely combinations which might cause problems. The challenge increases when the software is required to drive the car. Last week, I had the opportunity to drive a Porsche Panamera equipped with Bosch's Electronic Horizon package. This uses digital mapping technology to tell the car when to brake for a bend, what speed to take the bend at, when to accelerate out of the bend and where speed limits apply. It was an interesting experience to be required only to steer, but you choose whether to use the system or not and could disable it if you suspected it was not performing properly. But what about the software in Google's prototype self driving car? Here, things are taken a step further by doing away with the steering wheel and pedals. Now, you're entirely dependent upon software operating correctly under any combination of circumstances. It may well be some time until we see the truly driverless car.