Google Cars: Great for bicyclists, but fit for the road?

Tom Paine

Follow @phillytechnews

 Subscribe in a reader
Subscribe to Philadelphia Tech News by Email

I was recently contacted though email by a member of a small, multinational group that publishes a website (specifically by the editor, based in Copenhagen) who had seen a recent post ( this was the article referenced) and perhaps a little discussion on my site about self-driving autos and their practicality and safety. He pointed me to an article on their site, written by Luke Ameen, that highlighted Google's prototype car and the several ways it would improve bicycle safety, which I imagine is what readers of their website would be most interested in.

Their website, IceBike.org, is beautifully done, and the article, "8 reasons cyclists should love Google Cars", is highlighted here. I guess Ice Biking is something I'm not very familiar with (is it in the Winter X games?), but the scope of the website is broader than that, generally focusing on biking in cold weather climates where snowy or icy conditions are more common. The post does a great job of highlighting all the computer-aided features of the Google Car that could reduce car / bike incidents.

The post on Google Cars is so well done, in fact, that I might have almost thought it was astroturfed (placing things on social media that appear to be grassroots generated but are in fact put together by the organization) which Google is often accused of, though I don't know if there is any truth to those claims. I asked the editor, Mads Phikamphon, "did Google or its representatives provide any assistance, or compensation of any kind related to the preparation of this blog post"?. His response was no, and I accepted that.

The competition over self-driving cars is reaching a feverish pitch, with would-be entrants (both auto makers and tech companies) competing against each for talent, better test results, and the goal of gaining approval in cities, states, or even nations. There is a strong desire among participants to establish an early leadership position. Some are also pushing timetables up to levels that frankly seem unrealistic to me, talking about having largely self-driving cars on the road in the hands of the general public by 2020.

Phikamphon, who says he does some IT consulting in his day job, and I had an interesting email exchange about the issue. I stated my perhaps luddite conviction, based on what little knowledge of software engineering I possess, that I doubted whether all the variables involved in an open road environment (opposed, say, to parallel parking) could be reliably solved in the short term. I used the example of driving on I-287 in New Jersey during rush hour; perhaps the neural net of a lucid human brain is best suited to handle the rapid decision-making required. And how comfortable would people be in an environment like that where there was a mix of driver-operated and self-driving cars?

I-287: So many decisions / Creative Commons

Of course, if all cars were fully automated the answer might be different. But the political will to accomplish that, as well as the economic switching costs, would cause massive problems.

I told Phikamphon (something he wss probably well aware of) that in America driving - the right to drive where you want in the way you want - is considered a fundemantal freedom by many, something that would be very hard to pry away.

Perhaps, in parts of California, and metro ares like New York, self-driving cars might be an acceptable alternative to the pain felt by rush hour (often lasting much more than an hour) drivers. One thing operations researchers found was that if you could regulate a system so that all the moving parts going in the same direction were going the same speed, throughput increased dramatically. So a lot more vehicles could be pushed through the Schuylkill during a typical one hour rush hour period. And most bottleneck accidents could be eliminated, at least in theory.

In Europe (and I hesitate to speaak for a continent I've spent so little time in), public works are more organized, and people seem more willing to accept a system for organizing if they believe its in the public's (and their own) best interests. Also, we briefly talked of Asia, which contains at least half the world's population, and is rapidly catching up to the West in automobile ownership. I think China might be the ultimate market where such a system might be needed. India, with its rather chaotic public works history, might be a tougher sell.

But all this discussion is built around the predicate that a universe of fully autonomous vehicles can work, an assumption I still question. And I think too often when we look to a centrally planned solution to a perceived problem (such as electric cars), the results are middling at best, a total failure at worst. And a complete changeover to self-driving cars is not going to happen solely from the bottom up. Many new public regulations, and public works expenditures, are going to be required.



Carnagie-Mellon's Role and Uber's efforts in Pittsburgh


An interesting side issue is the role of Carnegie-Mellon University in championing the development of "autonomous cars",  as they are  sometimes referred to.

Its Robotics Institute has been working on the problem for 31 years, and much of the development being implemented now originated there. Chris Umson, who is featured in the IceBike post, taught and worked in the CMU program until lured away to be lead engineer of Google's program. Umson will remain its technical director while John Krafcik, who previously led Hyundai’s business in the U.S., was hired last month to step into a new CEO role.







Earlier this year, there were reports that Uber had hired as many as 50 researchers out of the Robotics Institute and set them up across the street to kickstart Uber's own program.

I'll leave you with two views of the autonomous car issue, each of which I agree with in part.


"Features of automated self-driving cars will appear incrementally and organically, with vehicles eventually driving themselves. This will make the cars affordable and encourage public adoption," said Raj Rajkumar, a professor of electrical and computer engineering and the co-director of the GM Collaborative Research Lab at Carnegie Mellon. (Rajkumar also holds a courtesy appointment in the Robotics Institute.)

The second is from a recent interview that David Mindell, an MIT professor, did with MIT News. Self-driving cars should not be fully self-driving, he says, based on studies of past experiences with automation in extreme environments. “Google’s utopian autonomy is a more brittle, less functional solution than a rich, human-centered automation." While "it’s reasonable to hope” that technology will help to “reduce the workload” of drivers in incremental ways, he says, total automation is not the logical endpoint of vehicle development.


Updates:
Tesla Releasing Autopilot Software for Model S Cars (Wall Street Journal)

Microsoft looks to stop bike crashes before they happen, testing Minority Report-style predictive intelligence (Geekwire)