Linux Hits the Road

An anonymous reader writes "Vicroads does regular surveys of the roads in Victoria, Australia, to determine where they need to be patched or otherwise repaired. It used to be done in a vehicle travelling at 20 kph: slow, tedious, and hazardous to the traffic around it. Now, thanks to Linux, it's being done at speeds of 80 to 100 kph. The Melbourne Age has the details. Short version: the cost has fallen from $1.2 million Australian to $850,000. Not bad..."

Comments

car video guidance

[bryanthompson]

I've been wanting to build a system that'd use cameras to find the lines on the road and keep my car between them. Now GPS would probably be an easier way to guide a car down the road, but i'd still like to see if it's possible, safe, and reliable.

Re:car video guidance

[HBI]

I have often dreamed of a system whereby I could have four missiles stored in the quarter panels of my car. They would be vertical launch tubes with caps painted the same color as the car, to be less conspicuous.

At the appropriate moment, I could press a steering-wheel mounted firing button to launch the missile. The missile would launch, locate the double yellow line in the middle of the road, and track it until it found the person blocking traffic. It would then break left till it circled around, slightly above traffic levels, do a pop-down maneuver, and lay into a slight up-angle as it strikes the driver's side rocker panel of the offending vehicle, knocking it off the road.

Unfortunately, I don't believe the kinetic energy imparted by a small missile of the 50 lbs variety can actually do that to a 2 ton SUV (it's a given that someone in an SUV is blocking traffic. Ok, maybe a minivan). But I can dream, can't I?

In the process of thinking about this, I realized a couple things that may be of interest to you. First, not all roads have the same kinds of painted lines. Some have shoulder marks. Some have buzz strips. Some have single yellow lines, and some have double. Some have single yellow lines with a dashed line on the other side (signifiying a passing lane). Assuring an optical sensor would be able to digest all these differing inputs would be challenging, to say the least.

Also, what happens when a road has a middle passing lane with double yellow lines, dashed on the inside? Those confuse human drivers, I can guarantee an optical sensor would not be happy with that.

What about turn lanes? NJ is famous for those stupid jug handles. Obviously a 'turn/go straight' decision would have to be made. But what happens when the primary road turns slightly at the point of the turn lane? Some interesting behavior of your automated system could result.

I also doubt that GPS has the resolution to actually handle driving down a road. The promised CPE is big enough that you could ram into a telephone pole at just about any time.

This is a really tough problem. I ultimately think that a passive response device along the lines of an RFID would be necessary to keep vehicles travelling in the correct direction. These would need to be installed along all road surfaces. For those which aren't equipped, we'd be stuck with the current method.

Re:car video guidance

[Eric Ass Raymond]

Ever considered that a bike has several advantages over a car:

1) You'll get your daily exercise. It does not stress your joints so even if you're overweight you can start cycling straight away.
2) No more getting stuck in a traffic jam, breathing the fumes and sweating your ass off.
3) It's environmentally friendly.
4) It's inexpensive.

But where will this technology go from here?

[cvk]

This high-speed video capture is definitely the way to go for a first step, but of course the situation will be hugely improved when all that data can be taken back to the lab an scanned for drivability by software instead of by human brainpower.

Perhaps when the sun is low shadows would be cast over potholes that would lead to lower temperatures inside the crater than on the surface of the road. That would make infrared cameras an obvious choice for picking out the cold-bottomed potholes.

Or perhaps a rear vehicle could shine a light at an acute angle to the ground that would turn potholes into shadowy pits for easy detection by a forward vehicle on the other side of the pothole?

So many possibilities. (So many challenges!)

Re:But where will this technology go from here?

[EvanED]

That misses the potholes you don't run over directly with your wheels.

Linux not the answer

[DerekLyons]

Actually reading the article shows that Linux is incidental to the 'breakthrough'. The improvement comes from video processing software, not from the operating system of the computers that perform the processing.

This is a newspaper article?!?

[Anonymous Coward]

A newspaper article about Linux that backs up its claims with details, has not one hint of FUD.

Soeriously now, an nwspaper article that mentions
limitations in the firewire drivers.

I mean the readers are expected to know what drivers, RAM, firewire, is.

They call Linux robust and hint that windows isn't.

There is no catch!!!

Now this is unbelievable!!
This must be a hoax article.

You don't really get stories like this in the newspaper

MILLENIUM train

[Anonymous Coward]

Full day drive away in Sydney, we're suffering through the MILLENIUM train fiasco, technology supplied by, no surprise, a Microsoft Operating system.

aarrh!

Won't Make Much Difference...

[Victa]

VicRoads recently resurfaced the main road near my house... After 4 weeks working on a 400m (1/4 mile) stretch of road they went away. Leaving a worse surface than they had started with...

It's fairly typical of VicRoads to resurface perfectly good roads regularly (every 6-12 months) and the roads that are actually in bad shape get ignored, or made worse... I guess it must have something to do with where the money lives...

Re:RTOS?

[Snoopy77]

Why aren't they using a RTOS?

Cause they don't need one. The shutters are triggered every metre (or 20 metres) and it works just fine.

Forgive me for asking such a silly question but did you actually read the article?

Re:I use linux...

[edwdig]

> I mean, is it *really* that much harder to grab some video in Windows vs Linux?

Although I've never tried video programming in Windows, I did try it in Linux. I ended up giving up fairly quickly, because although the individual V4L API calls are documented, there is no documentation stating which calls are necessary to get something to happen, or in what order you have to call the different functions. Getting something working involves a lot of trial and error. So I'm sure for video purposes, Windows would be easier to code.

Anyway, that completely misses the point of the line you quoted. The author chose Linux because he wanted an OS that wouldn't crash, not one that was easy to program for. If OS #1 provides an easy to use but crash prone API, and OS #2 provides a harder to use but stable API, #2 is the better choice.

Oh, and from my personal experience on lots of systems, Windows NT/2000/XP are terribly unstable when doing video capture. Both with consumer and professional grade capture devices.

pity that they dont actually fix the roads

[narkotix]

have a look at this interesting article

A pity, though

[NaveWeiss]

that searching their site for "linux" results in nothing.

I don't usually respond to anonymous cowards but..

[donscarletti]

right after he admits having no knowledge of Windows systems. WTF?

He said he had limited, experiance, not no experience. And with bill gates confession that five percent of windows systems crash more than twice a day the chances of him knowing all two well the habit that MS systems have of failing

and don't tell me that KDE never crashes

Why would a survey device with the job of simply recording data use KDE, or for that matter GNOME, or even X? Somehow I seriously doubt that KDE's crashing effects this thing in the slightest.

Back in the days when I used KDE I saw it crash a number of times (not as frequently as most MS user interfaces I have used), but even if they were running KDE they cirtainly woudn't be stupid enough to have the data monitoring program running on top of it, and so therefore the mission critical application would continue to run perfectly. This is not as much of an option for windows setups when a UI failure or a falure of a non-essential subsystem can take down the entire system.

Re:Go figure

[drinkypoo]

Also... it was chosen because of familiarity, not because of any features or cost comparison (You cannot call saying "Linux is free!" a cost comparison.) The author-to-be of the software said "I don't know shit about windows programming, but I've done a lot of work with Linux. Let's use that!"