Road Train Test

Back in January of last year, I posted a note here about some work being done by Volvo to develop the technology for “road trains”.  Using a variety of technologies, including cameras, radar, and laser tracking systems, along with wireless networking, the idea is that a group of specially equipped vehicles can travel together as an ensemble.  One “lead” vehicle, with a skilled driver, will lead the way, and the other will follow along using automated controls.  The motivation is that a road train system could reduce fuel consumption, increase safety, and possibly even relieve congestion, by allowing cars to travel safely in closer proximity.   The work is part of a European Union project, SARTRE (Safe Road Trains for the Environment).

A recent article at the Register (a UK-based technology news site) reports that the first tests of the system have now taken place on public roads: 200 km [~ 125 miles] of Spanish motorways.

Three cars have successfully driven themselves by automatically following a lorry [truck] for 125 miles on a public motorway in the presence of other, normal road users.

The average speed during the trip was slightly more than 50 mph.  The three cars stayed in line behind the lead truck, with an average separation of 6 m [~19.6 feet].   Considering that the speed (50 mph) is 70+ feet per second, this spacing would be dangerously close for human drivers; if these results hold up, there would seem to be some validity in the claim of closer proximity travel via this “platooning”.

As with Google’s “self-driving” cars, I think this technology has the potential to make auto travel safer and more efficient.  Both these technologies will also require some changes to traffic laws and people’s attitudes.

Update Tuesday, 29 May, 23:05 EDT

The “Autopia” blog at Wired also has a report on this test, which gives a bit more detail.

Living with Driverless Cars

A couple of days ago, I posted a note on Google’s receiving approval to test its driverless cars in Nevada.  If the technology proves successful, and is adopted to any significant extent, it will probably change how we drive, and how we think about driving, possibly in unexpected ways.

The BBC News Magazine has an article on how some of those changes might play out.  Some of the changes would very likely be positive.  An automated driver will not be talking or texting on a cell phone,  sightseeing, or otherwise diverting its attention to something other than driving.  It will not be sleepy or intoxicated, and will have faster reaction times than a human driver.  That should mean fewer crashes.  Automated driving might also allow more traffic to be carried on the same roadway, because cars could travel closer together, particularly if vehicles are able to communicate with each other, as in the “road train” experiments.  The automated driver can also be programmed to avoid human drivers’ dangerous behavior.

When the car is on self-driving mode, it doesn’t speed, it doesn’t cut you off, it doesn’t tailgate.

Some of the changes suggested in the article seem to me a bit more problematic.  It suggests that the technology might make auto use available to people who currently, because of physical infirmities, cannot drive, such as the elderly, or people with epilepsy.  The article also suggests that a self-driving car could, in effect, run errands on its own.

A car could take the children to school early in the morning, then return home on its own to pick up the parents for their commute.

After a late-night carouse, a drinker could find and reserve a hire car on their phone, then have it pick them up and drive them home.

I’m sure that, properly used, the technology could make driving safer, and allow some people to drive who would otherwise be only marginally capable.  But the idea of having the vehicle operate without at least one competent adult along, in case of emergency, strikes me as ill-advised; I hope that we will require a lot of evidence of the technology’s reliability before we begin that experiment.

Google Gets a No-Driver License

I’ve written here a few times about some of the developments in automotive technology that are aimed at turning over at least part of the driver’s job to an autonomous computer system.  Specifically, I’ve mentioned experiments, mostly in Europe, with “Road trains” (in essence, car convoys), and Google’s work on a self-driving vehicle.   There are also efforts underway to gradually incorporate some of the more mature results of these experiments into production autos.

According to a report posted on the “Law and Disorder” blog at Ars Technica, Google has now received approval to test drive its autonomous autos in Nevada.  The conditions for the license, established by legislation and regulations earlier this year, are not very different from the rules in the initial Google tests.STOP

The state previously outlined that companies that want to test such vehicles will need an insurance bond of $1 million and must provide detailed outlines of where they plan to test it and under what conditions. Further, the car must have two people in it at all times, with one behind the wheel who can take control of the vehicle if needed.

As I’ve said before, the idea of automatic control is not new.  Aircraft have had autopilots and instrument landing systems for many years; trains have had signaling and switching systems for a very long time, too.  Some rail systems, like BART in the San Francisco Bay area, have centralized control systems.

Considering major transportation systems together, it’s clear that, in some sense at least, automobile transportation is the anomaly, consisting of a large number of vehicles operated more or less independently by more or less independent agents.   We have traffic laws, of course, such as speed limits and STOP signs, but travel by car provides a lot of flexibility and freedom to the individual driver.   That comes at a cost, however; in the US, somewhere between 35 and 40 thousand people die in auto crashes every year.  (For what it’s worth, I think that the most important Driver’s Ed lesson almost never taught is “It’s not a Race”.)

Part of the move to “smarter” autos, of course, is about testing and refining the technology, and that has to  be done.  The bigger part of the problem, though, will involve changing how we think about driving.

Evolving Autonomous Autos

A little over a year ago, I wrote about Google’s research project to develop and test a self-driving automobile.  This was not a totally novel idea; the DARPA Grand Challenge, a prize competition for driverless vehicles, had been running for several years.  Though Google has tested its technology on normal California roads (always with a human back-up driver on board), there are still technical, legal, and cultural obstacles to be overcome before we will be able to sit back and enjoy our coffee while the car drives us to work.

Technology Review reports that a more gradual approach to automated driving technology is being taken by auto manufacturers, especially in Europe.   I’ve written here about a European project to allow communicating vehicles to form up as “road trains” on highways, to improve energy efficiency and safety.  And the automakers have already begun to introduce incremental driving assistance capabilities.

[BMW's Werner] Huber and executives at other European automakers say the automated driving revolution is already here: new safety and convenience technologies are beginning to act as “copilots,” automating tedious or difficult driving tasks such as parallel parking.

The expectation is that these features will be introduced first on high-end models, then gradually make their appearance on a broader range of cars, depending of course on their reception by customers.  There are models offered now that offer parallel parking assistance, and other capabilities are being offered as well.

For example, for $1,350, people who purchase BMW’s 535i xDrive sedan in the United States can opt for a “driver assistance package” that includes radar to detect vehicles in the car’s blind spot. For another $2,600, BMW will install “night vision with pedestrian detection,” which uses a forward-facing infrared camera to spot people in the road.

Probably one of the marketing goals for these features is to get people more accustomed to the idea of the car “thinking for itself”.  One doesn’t have to look at very many car advertisements to realize that the product is often sold as an extension of the driver, probably not ideal for selling a fully autonomous car.  There are also legal obstacles to be dealt with.  Traffic codes assume, at least implicitly, that a person is in control of the vehicle while it is moving.  There will also be interesting issues of software liability, if it appears that a failure of the automatic system caused a collision.

Still, this is potentially valuable experimentation.   Travel by automobile is a big consumer of fossil fuels, as well as being fairly dangerous (compared to other forms of transport).   Anything that might make it safer and more energy efficient is worth a look.

 

Elementary Watson

Early last year, I wrote several times about IBM’s Watson system, and its victory against two human champions in the popular TV game show, Jeopardy!.  IBM has also announced some initiatives to use Watson’s technology in medical diagnosis.

In December 2011, at the USENIX Large Installation System Administration Conference [LISA] in Boston, MA, Michael Perrone, Manager of Multi-Core Computing at IBM’s Thomas J. Watson Research Center, gave an entertaining talk on Watson’s methods, and how they were developed in the context of the Jeopardy! game.  He presents some statistics and examples to show how the game is difficult, compared to something like chess, and how what Watson does is different from a search engine, like Google.  Inference from natural language is tricky.  For example, what do these three things have in common: shirts, TV remotes, and elevators?  Well, they all have buttons.

Mr. Perrone also gives examples of some of the questions Watson got wrong in its early incarnations (it also made a couple of bloopers in the match against human champions).  For example, in the category “New York Times Headlines”, the clue was:

An exclamation point was warranted for the end of this! in 1918.

Most people would probably realize the correct answer is “What is World War I?” — Watson came up with a silly but eminently logical answer: “What is a sentence?” The talk also includes some information about the hardware used to implement Watson for Jeopardy!.

The talk has now been made available on YouTube; it runs about one hour and twenty minutes.  The slides from the talk [PDF] are also available.

Building the Analytical Engine: Update

Last December, I wrote about the project, launched by John Graham-Cumming, to build a working implementation of the Analytical Engine, designed by the British mathematician Charles Babbage.  Babbage’s design was arguably the first for a stored-program computer, although it was a completely mechanical — and steam powered! — device.  Nonetheless, it incorporated features common in today’s computers, including branching, looping, and expandable memory.

The BBC site reports that the Science Museum in London is undertaking a first step in organizing the extant design information for the project.

A project to construct one of the earliest mechanical computers based on sketches by its designer, Charles Babbage, has received a major boost.

The Science Museum in London has agreed to help by digitising the mathematician’s original plans.

The aim is to make the designs, which in their original form are in the Science Museum’s archives, more accessible to a wider range of people.  Once the designs are available for study, a follow-up step will be to develop a computer simulation of the Analytical Engine.  This will allow those studying the plan a chance to test their interpretations of Babbage’s notes and drawings, without building the full-scale machine, which Mr. Graham-Cumming estimates will be about the size of a small steam train.

Mr. Graham-Cumming has also developed some rough estimates of the Analytical Engine’s potential computing power.

Its memory would be equivalent to around 675bytes, or just over half that of Sinclair’s ZX81, released in 1981. A later proposal by Babbage called for 20KB of storage.

The machine’s clock speed would work out at around 7Hz, compared to the ZX81′s 3.2MHz.

Although it will take several years before a physical implementation of the machine is attempted, a tentative goal is to complete a working Analytical Engine by 2021, the 150th anniversary of Babbage’s death.