What Is The Accident Rate Of Self-Driving Cars?

Self-driving cars promise to make our roads much safer. They’ve also been in a number of accidents lately.

In September, officials in California began requiring Google to report any accidents involving the self-driving cars it is testing; since then, there have been three such accidents.

The company has also volunteered that eight other accidents have occurred since the testing program started six years ago—a total of 11, all of them minor and not the cars’ fault, at least according to Google, which has not released actual accident reports.

Google is not the only company testing self-driving cars, nor is it the only one whose cars have had minor crashes. The accident per mile rate works out to 0.6 per 100,000—that’s twice the official rate for normal cars, but since human drivers seldom report minor accidents, it’s quite possible the actual figures are much closer together.

So far, autonomous vehicles have a pretty safe record.

In theory, self-driving cars should have fewer accidents than humans, since computers do not become distracted, fall asleep behind the wheel, drink, or text their friends. Their electronic senses are designed specifically for driving so they have much better road awareness than we do and their reaction times are much faster. In practice these cars still have shortcomings, and their existence opens up major philosophical and legal questions—how does one sue a robot, for example?

For now, self-driving cars always do have a human driver capable of taking over. For the foreseeable future, at least, the best solution may be to keep a human in the driver’s seat while giving technology a greater and greater supportive role.

Already vehicle fleet managers can use telematics to record and analyze employee driving behavior. Although the human driver is completely in control, the human’s supervisor is aware if something goes wrong and can use that information as a training aid and as a way to identify which employees need either special recognition for good driving or more help (or disciplinary actions).  

We’ve also started to see vehicles with a range of features that blur the line between traditional and self-driving cars. These models can slam on the brakes to avoid a collision, warn the driver if there is something in the blind spot, parallel-park themselves, and even stay in a lane unassisted. In the future, we may see cars that communicate with each other before changing lanes, or even maintain safe following distances and speed limits automatically based on road conditions.

Will we ever have cars with no steering wheels or brake pedals? It seems likely the technology will get there, but do we really want to give up the option to control our cars ourselves? Why? Why not?

As with the invention of cars over a century ago, new technology will spur the development of new infrastructure, new legal concepts, and new social conventions. Now is the time to be thinking about what kind of future we want to create.