Testing an Autonomous Vehicle for Safety is Impossible
The claim of every manufacturer is that their cars will be safer when the car is driven autonomously. Google has logged nearly 1.3 million miles in their self-driving cars since 2009 and has the most miles so far with relatively little error. Even so, what will it take for us to trust that the vehicles being developed to actually be autonomous vehicles? Will we need miles on the tires to show the car is safe or are we ready to nearly blindly trust that the technology installed will be adequate enough to ensure our safety while driving on any road?
Let’s take a step back and look at this a different way for a minute. When was the last time your computer at home didn’t work right? Have you ever had to reboot your phone because a download did something to change your settings? When you go to a store have you ever experienced the credit card system being down which meant you had to wait because you didn’t bring cash with you? All of these situations are electronically controlled and use the internet and over the air signals to complete the task they are designed for. What makes you think a car that drives itself by using GPS, sensors, cameras and the internet to control the vehicle will be any different?
Moving forward in the conversation, let’s assume all is well with the technology of an autonomous vehicle. This means the car can literally drive itself and doesn’t need your input at all. How many miles will actually need to be logged before you can trust this vehicle? If you think about what the Google cars have done it’s pretty impressive, but it’s nothing compared to how many miles we as Americans drive each and every year.
Americans log close to three trillion miles on the road each year, apparently we don’t walk anywhere anymore. With this number of miles and 2.3 million injuries and 32,719 deaths in 2013 we see that the rate we enjoy today is only 77 injuries in 100 million miles driven and only one death per 100 million miles driven. That means an autonomous vehicle would need to drive a full 100 million miles with no fatalities to be able to make the claim of being safer to drive than the vehicles we currently have on the road today. If Google never increased the number of miles driven this would take them nearly 700 years to complete, which is certainly not logical at all.
So far during the time the Google cars have been driving there have been seventeen crashes and sixteen of them were the result of human error. This leaves only one where the car caused the crash which sounds pretty good considering the number of miles, but what happens when autonomous vehicles become more common and more cars are linked online and need the same inputs in the same amount of time. Have you ever had a video buffer on you for even a few seconds; this isn’t something we could endure on the road.
While we will likely have autonomous vehicles on the road in the near future it seems a recipe for disaster even if the vehicles are supposed to be safer. Not only do we not know exactly how safe these cars will be, we don’t know who is responsible if the system fails and causes a crash. How often will these cars ask the human drivers to take over and under what conditions? In my opinion we just have too little data to make an informed decision regarding trusting these cars just yet; but soon we may be faced with making this decision and then we’ll see exactly what these cars have to offer us and how safe they really are or aren’t.