Although rarely used with the utility for which it was conceived, the horn is a safety feature essential in a car. Yes, it is likely that only listen to along with a window that low, and a lot of expletives, or in the celebration of the title won by a football team, but warn of a danger to other drivers via sound signals can be the difference between an anecdote and an accident. Maybe that’s why Google is concerned about because of the logic of operation of the horn in your autonomous car. And a lot of eye because, even surprise you, is not being anything but simple.
Google reminds us of an essay by Eugene Garfield in which he spoke in 1983 of “The Tyranny of the Horn”. Probably a world without speakers it would be much better. As I write these lines I have had to get out of my job to close the window, because down there is a car parked in a double row, and the driver who has gone to retrieve his car blocked, not for beeping. But the question we have to ask ourselves is the following, what would be safer a world without loudspeakers? when was the last time that evitaste an accident by pressing the horn?
Google think that the speaker, in effect, is required, even in a car that is designed to perceive everything that happens on the road, with a level of attention, and without blind spots, impossible to achieve by us imperfect humans. But Google wants its autonomous car to be courteous, considerate, and just beeps to improve security.
here is the major problem for the artificial intelligence of the autonomous car of Google. The only premise that has been met is to trigger the horn only when necessary to avoid a hazard. In this way, the great challenge of the car of Google is interpret an imminent danger, avoidable, and distinguish it from false positives. The number of situations in which there is no danger, and could be interpreted as such, is almost infinite. And Google wants its autonomous car, be patient, and kind, so they can’t afford it beeps when there is no danger.
In its evidence, Google told us that they began by a loudspeaker of very low intensity, which emitted a warning sound that is only heard in the cabin, so as not to disturb the rest of the drivers. In this way, the technician in charge of the test wrote down in a document if the autonomous car had triggered the horn to an imminent risk, or in effect it was a false positive.
According to Google, their autonomous car is already in a position to operate the horn on the outside, and hear the rest of the drivers. Indeed, we have that have designed your horn with different intensities, according to the urgency of the situation. So if a car comes back to us it will warn you with two puffs soft and quick, whereas if the risk is imminent, and almost unavoidable, beep thoroughly and with great intensity.
In the end, the amount of parameters to control an autonomous car is amazing, and thanks to Google we’re learning a little bit more the work that goes into a development like this. By the way, Google already has 34 prototypes and 24 Lexus RX450h to autonomous shooting in public, that they have already toured in stand-alone mode a total of 2.646.000 kilometers. It’s almost nothing!