21 June 2014

Self-driving cars: the long road ahead

Known for a couple of years, Google’s vision for self-driving cars has materialized recently into a couple prototypes on the streets of California. As one who doesn’t have a driver’s license and who enjoys reading on my way to work, I would certainly welcome a vehicle that could take me there on the shortest route possible while allowing me to relax and disconnect. Nevertheless, I see three major problems with self-driving cars, which will considerably delay their adoption and success.

Google driverless car prototype
When reporters pressed Dr. Urmson about when there might be commercially available cars and pointed out that the Google co-founder Sergey Brin has been quoted as saying the technology would be available by 2017, he demurred. While noting that Mr. Brin is his boss, he said his overriding concern was safety. He also suggested that his personal goal was to have self-driving car technology by the time his son turns 16, five and a half years from today. John Markoff

Despite Google’s recent test drives, the technology is far from being ready. An article in The Atlantic describes the way it works and, as far as I understood it, the prototype needs a preloaded, extremely detailed map of the area in order to navigate it. Take the car out of this predefined area and you’re left with a giant, silly-looking toy car. That could still be fine if you only planned to use this for the daily commute to work; you would load the directions into the onboard computer and let it drive you there. But as soon as you wanted a new destination you would need to preload another map – or to take the wheel again… I’m not sure Google’s strategy of digitizing every minute detail of every possible road will scale; streets change all the time and the maps will have to somehow keep up with that. The article mentions the special maps for self-driving cars cover about 2000 miles, less than 0.05% of the total U.S. road network. This sounds like a good solution for vehicles who only travel fixed routes, like busses and trucks, not for personal cars. A ‘real’ self-driving car should be able to find its way anywhere by recognizing the road and other vehicles around it and make constant decisions: when to stop for traffic lights, to avoid pedestrians and other cars that might not be controlled by logical computers, but by impulsive humans. It should be able to react in real-time to unforeseen situations without the need for human intervention; if the passenger has to constantly supervise the car and be ready to take control, he might as well drive it himself!

We tell it how high the traffic signals are off the ground, the exact position of the curbs, so the car knows where not to drive, he said. We’d also include information that you can't even see like implied speed limits.

Google has created a virtual world out of the streets their engineers have driven. They pre-load the data for the route into the car’s memory before it sets off, so that as it drives, the software knows what to expect.

Alexis C. Madrigal

But let’s assume these technical constraints will be solved at some point. How will a truly self-driving car change our cities and mode of transportation? Many people assume that it would mean the end of car ownership: when you need a ride, you just tell your smartphone to order one (or wink at Google Glass) and within minutes a car will arrive to take you where you need to go. But this is harder as it sounds and highlights another problem with cars in general, the logistics. Much of the daily traffic is caused by people going to and returning from work, causing peak traffic at specific times during the day. During those hours, demand for transportation is very high, making it difficult for people to find available cars, causing traffic jams. Or think about the rush to get home after a big concert or soccer game; last year after a Depeche Mode concert I couldn’t get a taxi for at least an hour. Having self-driving cars available doesn’t change any of this dynamic; they will still get stuck in traffic, people will still be late for work and get annoyed about the delays. The only difference will be passengers can do other things while in the car – hopefully productive – than stare at the road and wait for the line to advance.

if you think about it, something very similar already exists, it’s called taxi!

Given the current rate of 33,000 or so fatalities per year on American roads, if we assumed every single person suddenly bought a Self-Driving Car and those cars miraculously experienced no faults, no pedestrians walked out in front of them and they collided with no cyclists, we'd still be looking at 3,300 road fatalities a year. Know how many people die in rail transit related accidents every year? 106. And 18 of those are where rails cross roads. How many bus occupants die a year? 54. Even if Google's car eliminates all user error from roadways, thousands more people will still die on roads than will be killed while riding a bus, train or other forms of public transportation. Statistically speaking, riding in a car, even one driven by Google, will always be considerably more dangerous than riding on public transportation. Wes Siler

Like any new technology with potential to massively change our way of life, social acceptance will certainly prove difficult for multiple reasons. First of all there is a certain kind of status associated with driving a car, a sense of control and freedom that disappears if a piece of software is driving you around. Despite extensive security testing, I imagine many people will still be frightened to use a driverless car, especially if they will populate the streets at the same time as regular ‘dumb’ cars. As someone remarked on Twitter, in the case of an accident, who will be reliable for the driverless car? I somehow doubt that Google will be willing to pay damages for software glitches or mapping errors. If self-driving cars start to replace transportation trucks, buses and taxis, a whole range of people will find themselves unemployed and society should ideally be ready to relocate them and provide support until they find new jobs. There are also possible monopoly issues; given the massive investment in mapping and servers, very few companies will be able to replicate Google’s efforts, giving them effective control of this emerging market; something that Google is probably aiming for, but can’t have positive effects for the customers and innovation. On the other hand, with few large competitors on the market, you could run into other issues; if you urgently need to be picked up by a car, can you order one from any supplier or will you be bound by contract – a monthly rental fee for example – to a specific company and cannot get one from the competitors if available? I am excited about the prospects of driverless cars and hope they can take shape in a way to improve our whole society, not just the large corporations that built them.

Google says it began its self-driving car project with a “blank sheet of paper”, and yet, all it created was a slightly different variation of our current transportation problem. This isn’t “disruption”; it’s tinkering. Rather than find a truly transformative solution, the company has acquiesced to the basic failings of suburbia with an innovation so incremental that it epitomizes our national short-sightedness, and failure of imagination, when it comes to improving mobility in America. What’s more, Google fundamentally misunderstands that, miserable as Americans are behind the wheel, they still love cars because they love being in complete control of a powerful machine. Take away the wheel and the pedals, and you’ve taken away whatever joy there is to driving. Ben Walsh
Self-Driving Car Test: Steve Mahan

Update: Interesting details about the safety requirements for car hardware and how different they are from PC and smartphone chips:

Even as the systems proliferate and software developers such as Google Inc. and others roll out plans for connected entertainment and mapping systems, carmakers have been slow to switch to unproven chip suppliers because their products are governed by rigorous safety requirements. When a computer crashes, a user might lose some data. When a car crashes, people can get hurt. Ian King

Post a Comment