About the only way your next car has much chance of driving itself is if you live in a gated community or on a college campus where it won’t have to deal with too many variables like other traffic. Just as voice recognition systems work best with limited vocabularies, autonomous vehicles will probably be limited to such constrained environments for the foreseeable future. That’s the conclusion from the recent ITS World Congress 2014 in Detroit. Increasing levels of vehicle automation were a major topic of discussion during the annual conference on intelligent transportation systems.
Google has been pushing the idea that self-driving vehicles will hit the road within the next 5 years. Google had no official presence at the conference, but a lot of companies that build cars, parts, and infrastructure systems were there, and no one that I spoke with was in agreement with Google’s timing projections. The general consensus is that we won’t see widespread use of full operating range autonomous vehicles until closer to 2030.
That’s not to say that no one believes in automated driving; quite the opposite. It’s just that in engineering circles, there’s a rule of thumb known as the 90/10 rule. That is, 90% of the technical challenge of a project takes about 10% of the time and effort. The last 10% takes the other 90% of the time. In the realm of self-driving cars, we have just begun that last 10% phase, where the basic hardware elements are all worked out but a lot of software decisions have yet to be made in order for autonomous systems to be truly robust.
Much of the on-road development by Google and other companies has been occurring in places like California and Nevada, where environmental factors like snow and even rain are a rarity. In order for autonomous vehicles to be both commercially and legally viable, they’ll have to work reliably under any weather and road conditions.
General Motors (GM), Volkswagen, and other automakers have been working on autonomous technology much longer than Google, and they understand these limitations. When GM rolled out a two-seat self-driving pod car known as the Electric Networked-Vehicle, or EN-V, at the 2010 Shanghai World Expo, program leader Dr. Chris Borroni-Bird acknowledged that, while this type of vehicle would eventually be an ideal way to deal with the congestion problems of megacities like New York, Shanghai, and Mumbai, the first feasible real-world applications were likely to be in restricted environments, such as campuses and gated communities.
As powerful as computers have become, they still don’t deal with the nuances of the real world very well. That’s why voice recognition systems still struggle to understand what should be simple natural language commands on a smartphone. The most successful applications of the technology have been for tasks like medical transcription, with limited and specific word vocabularies and little ambient noise. Similarly, automated vehicles function best in constrained spaces, such as buses over fixed routes or the aforementioned commuter pods.
Google hasn’t actually made any major breakthroughs in the technology that we know of. It just jumped into field relatively recently, hiring many of the engineers and scientists that worked on the autonomous vehicles fielded by automakers in the DARPA Grand Challenge and Urban Challenge competitions of 2006 and 2007, and leveraging the cost declines of the required sensors.
Where Google has outdone the incumbents is getting the technology media to talk about their efforts – but that’s unlikely to put full-function self-driving cars into consumers’ hands any sooner.
Tags: Autonomous Vehicles, Clean Transportation, Electric Vehicles, Smart Transportation Program
| No Comments »