Slashdot Mirror
DARPA Grand Challenge 2005 Rules Announced
Mr_KnowItAll writes "Our friends at DARPA have released the
proposed
rules for the 2005
Grand Challenge. They learned their lesson from the first one, now they will expect teams to submit a video demonstration of their vehicle's ability to complete the course before being invited to participate. Good, but they're also advancing the timeline to the point that it will be very hard to any team to start now and hope to participate. After all the fuss at DARPA's last-minute rule changes in '04, it's interesting to see that they're offering the proposed rules for community review and feedback."
-Terrain navigation
-Obstacle navigation
-Other short races that highlight various systems and their proficiencies
-An overall skill winner
-And THEN the race
Each of the teams would be able to see and learn from the other various technologies and make changes/improvements in the coming years.
This thing was just a joke this year with many of the vehicles crashing before they could display their skills.
you bozo's /.'ed DARPA...
flinging poop since 1969
I guess you could do better? Oh, I forgot, this is slashdot, everything is easy...and if you fail you are stupid. Look, the best team broke an axel which caused them to stop. I bet you would have a hard time driving that course yourself. Have you ever driven somewhere where breaking an axel is a hazard? This isn't like your daily drive to work. Oh, and the CMU team drove ~3000 miles across the US in '94. I guess we should all just give up since nobody completed the course though...
If it was easy, idiots like yourself could do it, but it is hard so it will take a few tries to get it right.
DARPA has been funding development of this technology for many years. The Grand Challenge was designed to expedite the process by placing a seemingly large sum on the table for the winner. They, however, have chosen to not allow any previously government funded software.
While that levels the playing field for all the teams, it certainly does not represent the state of the art in autonomous robotic software that has been developed under DARPA's own contracts in the past. Many of the complex image understanding algorithms for road operations, obstacle avoicance, terrain classification, etc. have required millions of contract dollars to develop. This is far in excess of the prize being offered for the completion of the GC.
The repeat teams will certainly have a development advantage because the bulk of the necessary work is software development and integration not overall vehicle development. Most of the critical algorithms already exist but cannot qualify because they were developed under DARPA (and other government agencies) contracts.
From the article, someone has to read it...
"The route (see definition, p.28) will be no longer than 175 miles. It may include paved roads, unpaved roads, trails, and off-road desert areas. The route contains manmade and natural obstacles, both above and below the surface of the average terrain. Examples of obstacles include ditches, washboard, sandy ground, standing water, rocks and boulders, narrow underpasses, construction equipment, concrete safety rails, power line towers, barbed wire fences and cattle guards. Every obstacle on the route can be either traversed by a commercial 4X4 pickup truck or avoided entirely. DARPA will place on the route one or more obstacles that are designed to disable tactical vehicles. These obstacles must be detected and circumnavigated for a vehicle to successfully complete the route. The route will be wide enough for vehicles to bypass these obstacles."
Still think this is easy? If someone can't make a robot to navigate this they are stupid? By that standard nobody is smart, except you of course...
We've got Top men working on it right now. Who? Top Men
A spider like robot or a flying robot would have a much greater chance of success with the current level of technology, and a flying robot could get over rough terrain much faster than any wheeled vehicle and automatically avoid enemies.
I'm not sure what use you think this is going to give people. All of Slashdot's subdomains all pass through one load balancer. If you go to warez.slashdot.org, 3dwww.slashdot.org, or plain old slashdot.org, the physical server you get directed to is based on distributed load on all of Slashdot.
No, the only useful thing is to subscribe and go to https://slashdot.org
You like splinters in your crotch? -Jon Caldara
Looks like all teams will require an alleged Site Visit in order to participate in this next Grand Challenge. Last time, their PreChosen Few did not need to have a site visit. As a result, you can see what happened! The newer rules will still allow them to game it any way they really want to. Actually, after reading the newer rules, it looks like it will be easier for them to refuse teams for non technical reasons. Before, they had to explain exactly why the teams technical papers did not pass their muster. Now, all they have to do is not like any teams Video Presentation in order to disqualify them! They will not even have to give a a reason why they were not chosen like they had to before! The implied answer is that they did not like your video for some reason. What once was a technical challenge has now turned into which teams are able to produce the slickest and flashiest video of their alleged race vehicle. Something tells me that technical challenge was never the case. Rest assured, they are not going to like our teams video for some, as yet, unknown or dreamed up reason even though we already fulfill and surpass the current GC requirements. We shall all see.
Maybe I should have been clearer. I am a CMU alumni. I don't worry about the others who mess up a hundred feet from the starting line. I guess they look foolish. But the CMU Red team went ~7 miles and broke an axel on a nasty road hazard that many human drivers couldn't navigate. But people like you make sweeping statements about nobody being close. BS, the CMU team is close, everyone else, well, I don't know. Some are better than others. Most are very smart. Shadows are a bitch for recognition algorithms and plants play havok with other sensors. Many times it is more about which sensor is most accurate, which is a difficult thing that humans are uniquely good at.
I've been following the GC for a couple of months now, with the aim of setting up a team at school.
While there are some very tough technical barriers to be overcome, getting a vehicle in 'race' shape shouldn't be THAT hard. All of the pieces needed are available, they just need to be integrated. A dedicated team of college students (engineers) should be able to start in september and still be competitive.
That being said, let the naysaying commence.
Oh man, I guess my Lego Mindstorm kit won't be quite enough....
....for God's sake DON'T enter a motorcycle, have it go for like 1 foot and have it fall over like last time.
"When the going gets weird, the weird turn pro" -- HST
And what would that be? A hovercraft?
If I point out that you are incorrect, making me a foe does not make you any more correct.
It would be interesting if they tried more unconventional vehicles. Berkeley's motorcycle was a good idea (very very efficient), but it just couldn't balance right. Maybe a tricycle design is in order? Like those 666 mile per gallon freakshows? Or a car that can flip over and still run (like some of those fancy toy RC cars)?
I hope there will be teams who will think it's worth it to do a really radical design: we might see some strange and exciting stuff on the battlefield later on.
Please put your video on the web. Thank you.
John Nagle
Team Overbot
We're recruiting. Programmers, this time; we have most of the hardware working. Silicon Valley only; we're in Redwood City. Send us 1000 lines of C++ code that you're proud of. We'll be having an open house in late August. Watch the Overbot web site for details.
Last I heard, minivans still had axles...
If I point out that you are incorrect, making me a foe does not make you any more correct.
There are going to be landmines out there this time!!!
And as you tread the halls of sanity, You feel so glad to be, Unable to go beyond. I have a message, From another time..
Sorry, but how to make the video. Run it over your course three hundred thousand times, and save the one time it works. Send to DARPA. I'm still upset that non-American entries are not allowed :-(
Quoth the competition rules;
"The manual emergency stop must be easy to identify and activate safely, even if the vehicle is moving at a walking pace. The operation instructions for manual emergency stop actuators must be clearly labeled in English and Spanish."
Who'd have thunk it, government organizations requiring instructions in English and Spanish, bilingually, for vital instructions on the vehicles. Is this a sign of the U.S. going bilingual (adopting Spanish) at snail's pace?
...no mine doesn't. it's got independant suspension on each corner: we leave "axels" to farm vehicles along with ancient tech like leaf springs. Oh, apart from the ones you USians fit to your new sports cars, like the new Corvette.
You said: You could try driving an axel-less(sp) vehicle like I do!
Minivans still have axles (look it up in the dictionary, dumbass), even small ones called stub axles. If you minivan runs on wheels, it will have axles.
If I point out that you are incorrect, making me a foe does not make you any more correct.
What's with this requirement that teams must be US citizens or at least lead by a US citizen?
:-)
Are DARPA afraid that some foreigners might win or something?
Surely this would be like Junkyard wars where the best competitions are those involving teams from all around the globe?
Geez, I'd have a go but I'd want to do it flying my own country's flag.
I was modifying my 2001 Chevy S-10 to be controlled by a computer. They changed the minimum time to copmlete the course form 1 day to 10 hours to 4 hours to 8 hours to 6 hours, and then back to 4 hours in the span of 2 weeks. They refused to admit (even after certified mail letters) that they had changed this rule at all. My vehicle would operate at a "speed" of 30 MPH, but including all the loopbacks, I could not make the course in 4 hours. Therefore, we scrapped all our (months of) hard owrk and dropped out.
Andy Out!
Ahhh, I need more time to enter my Bolo Mk I.
:^)
I have some details to finish before starting the fusion reactor and driving to the race site.
I predict I will crush the competition.
Then I have an idea to add a WOPR, and with the new AI the Mk III will be really impressive.
They Live, We Sleep
From: Paul Robinson <Postmaster@paul.washington.dc.us>
Subject: Rules Clarification - 3.6.4 Manual Emergency Stop Unit
With respect to the following section:
I ask the following:
Respectfully Submitted
Paul Robinson <postmaster@paul.washington.dc.us>
The lessons of history teach us - if they teach us anything - that nobody learns the lessons that history teaches us.
"I am a CMU alumni."
:-)
I found the problem. The prints showing how to make the axles strong enough were annotated in Latin, and so nobody at CMU could read them.
A Purdue Alumnus.
To be more precise the problem can be restated in that the teams are trying to reengineer the biologic process of vision while what they should be doing is attempting to reverse engineer the process. The teams are coming into the problem backwards beginning with the concept of GPS navigation and then attempting to pick a route based upon a camera input data field without properly compensating the camera data for the differentials of the human eye. So they rely too much on GPS and not enough concentration is given to the camera data.
The first problem is that human eyes see about 14 powers of 10 from the darkest to the brightest in differential. The best of our cameras see about 5.5 powers of 10. There is if anyone wants to write me as paulnoel(at) (my server is) => knology.net. Sorry about the spam paranoia...! I will discuss it with them in more detail. The essential problem with shadows is solved by developing this range of differentials or even wider.
The other problem is one computer programmers have a real bitch of a time with. It is understanding that natural systems operate because of inaccuracy and not because of accuracy.
When we get these two things together and fix them the problem is essentially gone.
Never Politically Correct ~ I prefer the facts If you don't like what I say, get a life, or comment yourself.
Actually, the CMU team didn't stop because they broke an axle - they broke an axle because they stopped. What happened was that they took a hairpin turn too sharply, got high-centered on a berm on the side of the road, and started spinning their front wheels extremely fast. Apparently, this caused their front wheels to catch on fire, throwing flaming rubber everywhere. The chase vehicle saw this and hit the e-stop, but their e-stop was a little too good - it clamped down on their front axle to stop the wheels from spinning, and the sudden force snapped it.
You have to complete the course within a certain time frame. Besides the mars rover platform wouldn't exactly handle quick obstacle avoidance at 50 mph. Technically the software in CMU's sandstorm probably could have won the competition. But the poor vechicle had a terrible roll-over 2 days before the competition. This fatuiged many of the mechanical components, and sadly the front axle broke 7 miles into the course. What these teams are doing is on the right track, give them another year or two, and we'll have winners.
Instead, check out SRS/SERVO Magazine Robo-Magellan contest. All you have to do is traverse 300 (straight line) feet - 300 grueling, twisted, Seattle Center obstacle covered feet. Of course, actual covered distance may be longer...
Think you can do it?
Reason is the Path to God - Anon
The CMU front axle broke because they high-centered on a berm while taking a switch-back too tightly, their front wheels were off the ground and started spinning enough to set the tires on fire, and so DARPA hit their e-stop, which was a system which clamped down on their front axle - and the sudden force snapped it. CMU failed not because of their rollover (and the fatigue on their mechanical parts) but because their software relied more on pre-planned maps than on obstacle avoidance, and they hit crap and high-centered on stuff.