Robocar projects, laws and other news


An update on the backlog of robocar related news caused by my recent travel and projects:

Nevada law

Many people have noticed the new law recently passed in Nevada which directs the Dept. of Transportation to create guidelines for the introduction of self-driving cars on Nevada roads. Here is the text of the law. Because Google, whom I consult for on robocars, helped instigate this law, I will refrain from comment, other than to repeat what I've said before: I predict that most transportation innovation will take place in robocars because they will be built from the ground up and bought by early adopters. The government need merely get out of the way and do very basic facilitation. This is very different from things like PRT and new transit lines, which require the government's active participation and funding.

You'll find lots of commentary on the story in major news media.

Volkswagen announces simple self-driving car

While VW has a long history supporting the Stanford team in DARPA challenges and funding the VAIL lab at Stanford, it was in Germany that they recently announced a "Temporary Auto Pilot" as a result of the EU Highly Automated Vehicles for Intelligent Transport project.

The VW autopilot is a combination of adaptive cruise control and lane-following. They plan operation in stop-and-go traffic as well as highway driving. These are all technologies that have been around for a while, though adding automatic steering to the lane following crosses the threshold into automatic driving. As designed, the system requires constant supervision -- take your hands off the wheel but don't take your eyes off the road, at least not for long and definitely not for long at speed.

At the same time, back in silicon valley, VW is sponsoring an effort for automatic valet parking in specialized parking lots. We first saw this in Junior 3 at VAIL.

Both of these are interesting approaches to the problem of how to deploy prototype robocars that are not yet ready for fully autonomous safe operation mixed with pedestrians and drivers. In the first case using human supervision, and in the second case, going very slow in a controlled environment.


BMW has also built a course-following car on top of a 330i. This is mainly an accurate-GPS vehicle used to show people how to drive a course at a racetrack.

Obama pushing robocars?

In his recent speech at CMU on advanced manufacturing, President Obama said:

"Imagine if America was first to develop and mass-produce a new treatment that kills cancer cells but leaves healthy ones untouched, or solar cells you can brush onto a house for the same cost as paint, or flexible displays that soldiers can wear on their arms, or a car that drives itself" Interesting that the White House is now identifying robocars as a key technology for U.S. success. We'll see how far they go with that.

Formal verification

For decades, there has been research into what is known as formal proof of the correctness of software. The rigour of a mathematical proof is applied to code so you can be sure it won't fail, at least if you trust the proof. This is valuable but very difficult, and judged too difficult with the full complexity of real world software systems. It has been done to C compilers and microkernels, however. At CMU, however there has been an effort to formally verify an automatic cruise control in a limited section of the problem space.

While I doubt that formal methods can be applied to all the systems in a robocar -- and further feel an attempt to do so would slow down development greatly -- it might well be applied to certain subsets and to safety check systems which monitor less verified systems. If there is a breakthrough in AI based program verification this also may bear fruit.

Greenhouse gas report

The most recent report on sources of greenhouse gas emissions confirms again what a large role cars play in that. While electric generation is also huge, electric cars on most power grids do reduce GHG substantially, even when they are as heavy as regular cars. (They are not quite as much of a win when it comes to total energy consumed.)

Because I believe that robocar technology can make the electric car a viable, marketable product this is good news for those looking to reduce GHGs. Indeed, if the issue of gasoline cars is not solved, there is very little chance of stopping the climb in GHGs.


>> The government need merely get out of the way and do very basic facilitation. This
>> is very different from things like PRT and new transit lines, which require the
>> government’s active participation and funding.

The point is that Robo-Cars can simply integrate into the existing grid without any additional infrastructure. First replacing a small amount of traffic and eventually subsuming all of it. With little or no intervention required (outside of legal structures to allow it and set a framework for insurance.)

Long term as they take over the roads infrastructure requirements actually decline. For example reduced requirements for signalization at intersections once human driven cars are eliminated. That coupled with increased throughput (with no additional investment) will give governments an incentive to push this agenda. Much easier to adopt than to raise bonds to fund more road building.

"the text of the law" link is downloading a zip file of a PHP website for me, not the text of the law. But I'm very interested in reading that law. Got a better link? Thanks.

Whoops: Never mind, found it at

I was thinking that Google could develop something more legally and wide acceptable before introducing a complete automated car.

For example in order to use and gain a massive amount of driving data G could use the hardware and software created to monitor and make decision to drive the car in a different way.
In my opinion this way could be "big external airbags for car-car collisions".

The great problem of external airbags was the impossibility to foresee a crash and its entity, but Google automated driving system does it quite well.

- External airbags would be a great way to reduce injuries and deaths.
- There are no legal issues. You keep driving the car.
- Could be sold in many different versions: -from side only protection to complete 360 protection (number and position of airbags) - little add-on for older car - carmaker collaboration to sell built-in system in new cars...
- Nobody would say: "I don't want it cause I love to drive".
- And last but not least: Google would have millions and more cars testing (internal simulation) G's self-driving system and possibly sending a huge amount of driving data from all over the world.

I think it could be a great deal and a step forward within the self-driving car roadmap.
What do you think?

Thanks for the very interesting blog
Best Regards,

Tommaso Gecchelin

Add new comment