Understanding Google’s Self-Driving Cars
As Alexis Madrigal of The Atlantic recently reported, Google’s ambitious plan for self-driving cars requires first that Google collect and upload a massive amount of data. Why data? Because Google is actually creating a “virtual track” out of our highways. Well, 2000 miles out of 4 million so far, so Google has a lot of work to do. But it is undaunted.
What is this “virtual track”? Remember trolleys that move along rail tracks while connected to electric power? Or those rides at Disney or other amusement parks that travel along pre-set groves in the constructed track? They had routes. They could not deviate from them, other than to shake a bit left and right. Imagine the airport tram, but with a much much larger “track” on which to operate.
Now, imagine those “route” kind of tracks, but add them via computers, the internet, GPS, and a myriad of sensors, and not with asphalt or bulldozers. The “virtual track” is what the self-driving car relates to first, not the real-world environment ahead per se, but with the map of the environment that is already loaded into its data system or beamed to it . That frees up the cars to travel along known (as in the data is fully mapped and recorded in the car and the system database) tracks on known streets with ease, so the car can its sensors to focus more on variable hazards, such as other cars, bikes, dogs, and so on.
In other words, at least for the time being, the self-driving cars have to drive on a pre-mapped virtual track of the road. And Google is mapping it down to the centimeter, even down to the curb level. What does that mean for consumers?
Pretty much nothing right now as the technology is not finished, nor is the virtual track remotely complete. Currently, there’s only about 2000 miles of roads mapped and uploaded for the track. Thus, even if you have a self-driving car, it cannot go off track and remain self-driving The car would not know where to go. It would be like you trying to take the Disney ride off its track or having your trolley cut loose from the power lines and tracks and head off in an entirely new direction. It can’t happen unless you discontinued the self-driving mode.
So, until your self-driving car has a mapped section of the road, you won’t be taking too many detours or using it in towns or even parts of the country that are not mapped. At least not yet.
Also, if the virtual track, or the data-carrying device in the car or in cyberspace, goes down or suffers a glitch, then your car won’t drive on its own. You will have to assume control and finish the job. The good news is that Google is building the cars to “learn” from the users when they take over. So, maybe next time you won’t have to take over. The car will “remember” what you did when last confronted with a particular obstacle or event.
The most obvious first use of this technology would be taxis or trams that go from Point A to Point B on a fairly regular basis. Mass self-driving transit would benefit people whose routes do not vary much, including city dwellers and commuters, bus riders, cab takers, and students. The blind and disabled should benefit greatly and see their personal autonomy greatly increased.
It is promising technology, but there is still much to do. But Google appears to have the excitement and resources to do it.
What’s the end game, according to Ms. Madrigal? It eventually might be robots, namely, smart robots that can operate within or “on” the same “virtual track” in the future. Recall too that Google is also starting to work on mapping the inside of various structures as well.
Is this the droid you are looking for? Yes, Officer, that is my droid. I sent him to the grocery store for some cold beer and pretzels.
– Pat Dawson