Just how good are Google’s driverless cars? – Quartz
The sun has barely risen as we set out on the Hana Highway, headed to Maui’s “wet” side. It’s a long, magical drive punctuated by sea vistas, waterfalls, Lindbergh’s grave, and, finally, the Seven Sacred Pools.
Mid-afternoon, we pack up for the drive back to Kaanapali. We can turn around and risk an unpleasant ride on the now congested Hana Highway, or we can forge ahead along the “dry” side of the island via the Piilani Highway:
“[A] must drive road … a twisty, up and down roller coaster of a ride … in some places only wide enough for one vehicle, and in many places bordered by a drop of hundreds of meters over the sea, unprotected by guardrails.”
Piillani it is.
Midway, we turn a slow corner and face a gravel truck coming towards us. As advertised, there’s no room to pass. Remembering the advice from a local waiter, I dismount, walk up to the truck with an embarrassed smile and ask the driver for help: “Sorry about being in your way. A friend said you’d know what to do…”
The gent chuckles. “There’s more room than you think. Back up twenty feet, fold in your side mirror, I’ll guide you so you hug the hillside real tight.”
A few minutes of slow motion terror as the massive vehicle inches past, and then a honk and a wave. The truck disappears around the corner.
Let’s revisit the scene aboard a fully autonomous Google Car, one that doesn’t have a steering wheel. (We’ll see if the Google Car moniker sticks, or if the Alphabet holding company gives it another name.)
What happens when we meet the gravel truck? How do we tell the car to squeeze within a breath of the hillside rock? Will its sensors even allow the maneuver?
This may be a fanciful example, but let’s consider a quotidian analog that I recently experienced when I came face to face with a Lexus on a narrow Mountain View street turned single lane road because of the curb parking. We were both holding up a line of cars, but we negotiated the bottleneck through a combination of patience and courteous hand signals —this is California, not Paris.
A Google Car would have been checkmated. It can’t move forward, it can’t go back:
The autonomous car’s passenger has to get out with an embarrassed smile and ask the conventional car drivers to help him out, to clear the road so as not to confuse his vehicle. That’s not what I think of as self-driving.
In an ideal future world where all cars are autonomous, this checkmate situation wouldn’t arise. Our cars will “talk” to each other and get a “god’s view” of traffic all around. When and how we get to this world, how long we tolerate the transition, isn’t clear.
(The difficult intermediate period reminds me of an old joke: A European country considers changing their traffic from right- to left-side driving. After much back and forth and no resolution, a politician comes up with a genius idea: “Let’s take it a step at a time. We’ll start with just the trucks…”)
In a Slate article subtitled “The Autonomous Google car may never actually happen,” Lee Gomes describes some of the obstacles that stand between us and our dreamed-of autonomous cars:
“… before the company’s vision for ubiquitous self-driving cars can be realized, all 4 million miles of US public roads will be need to be mapped, plus driveways, off-road trails, and everywhere else you’d ever want to take the car … The Google car can’t consistently handle coned-off road construction sites, and its video cameras can sometimes be blinded by the sun when trying to detect the color of a traffic signal…”
Gomes ends up surmising that the Google Car will someday be an exhibit in the Museum of the Future That Never Was.
Even if Sergey Brin’s prediction that self-driving cars will be a reality by 2017 isn’t about to come true, it’s a splendid PR exercise. The putative Google Car is consonant with the company’s achievements in Machine Learning and its preeminent position in computerized mapping. Google: Technically ahead for your well-being.
The exercise also validates the value of partial solutions, of computerized driver assistance solutions. One feature at a time, our cars are becoming semi-autonomous, performing safety and comfort tasks under the control of a driver sitting behind the steering wheel. This isn’t some hazy future, it’s a present we can buy today. No magic required, no science, no infrastructure changes, just applied technology that the marketplace will sort out.
For example, a Prius I once drove in France featured a smarter cruise control and collision avoidance system. As I came upon a slow moving truck in front of me, the Prius automatically decelerated from programmed speed to keep me at a safe distance. When I turned the wheel left to get into a passing lane, the car re-accelerated back to cruising speed. Nice, especially when looking at chain collisions on foggy freeways.
Newer high-end cars offer traffic sign recognition that notifies drivers of the traffic restrictions that are in effect. Other cars offer lane-departure warnings that, in the example of my spouse’s car, send a rumble through the steering wheel if you stray too far. (I’m a bit skeptical about this one because it cries wolf each time you switch lanes, voluntarily or not, thus diminishing the alert’s value.)
My next car offers autonomous driving in start-stop, bumper-to-bumper traffic and is loaded with a bevy of other driver assistance goodies such as infrared night vision enhancement and a 360-degree view and warnings. This will be a big jump from my five-year-old car from the same maker, and while I’m a bit concerned about the bugs that are sure to lurk in the massive increase in on-board electronics, it seems there’s no going back.
As luck would have it, a Guardian story titled “Documents confirm Apple is building self-driving car,” just came out. According to the article, the Cupertino company appears to be in negotiations for access to a high-security testing facility near San Francisco, a fact that would give additional substance to past rumors of an Apple Car development project.
A few thoughts come to mind.
First: Bravo Google! Minds are now bent and any car development shrouded in some sort of secrecy must be a self-driving vehicle.
Second: The GoMentum station mentioned in the Guardian article is affiliated with a public agency, the Contra Costa Transportation Authority. Besides an Autonomous Vehicle activity, the agency promotes a number of connected vehicles programs, some of which might be of interest to Apple’s Car Play developments:
While many, yours truly included, would like to see an Apple Car, these rumors don’t mean that the company will actually ship a car. The experiment could be just that: An opportunity to learn what not to do, a chance to fail to great advantage.
If Apple is developing a production electric vehicle, the project will certainly include driver assistance functions — but it will definitely not be an autonomous, self-driving car. The company likes to ship products, not concepts.