Slashdot Mirror
DARPA Announces 2005 Grand Challenge Semifinalists
Mockingbird writes "DARPA announced 40 semifinalists for the 2005 Grand Challenge autonomous robot race today. Notable remaining teams include the Carnegie Mellon University Red Team, Stanford Racing and a high school team, the Palos Verde Road Warriors. 78 teams missed the cut. The race, which will take place on Oct. 8, 2005 features a $2 million prize for the first team whose robot crosses 175 miles of the Mojave in under ten hours. The robots must be fully autonomous, with no team intervention allowed once the vehicle is launched. The first race was held in 2003, when the most successful team managed to log only 7.4 miles."
Yeah...I've seen their machine. It's pretty freakin' sweet. I do love the Linksys equipment glued (or something) to the top. At least they have good taste in technology :P
WASTE - The Secure P2P
And it's street legal! As far as I know, the first such vehicle to make that claim.
Support the First Amendment. Read at -1
The Roomba actually only made it 4 miles, but it cleaned up the competition...
(groan)
I have entered my Terminator this year as I think he needs to interact with other machines more, I still expect him to destroy all the other competitors but a day out and some challenge-response kaboom action will probably do him no harm. Also if your name happens to be John Conner I would recommend staying away from the competition site.
Do not try to read the dupe, thats impossible. Instead, only try to realize the truth
What truth?
There is no dupe
7.4 miles, eh? I bet they'd all be broken down, wrecked or stolen in a few blocks* if they had to deal with Chicago rush hour.
* It takes an hour to go a few blocks in Chicago rush hour. Nothing like starting your commute, only to find out on the radio it will take you 2.5 hrs to escape the city. Yuck!
Do they all run Linux?
*duck*
My little site.
my lego mindstorms robot will rock these guys anyday. its solar powered and has a gauss rifle that can be used to disable the enemy.
"The vehicles must travel approximately 150 miles
over rugged desert"
Where did the 175 number come from?
"Ain't I a stinka..." - Bugs
Today's Indianapolis Star. The mention of Scott Jones - the guy who invented voicemail - has a good project background.
People have been coming from all over the state (literally) to work on the project (just down the road a piece) on a very regular basis, just for the fun of it.
I've talked to several people who have been tinkering with it and are having a good time. Sometimes, bordering on obsession.
I'm jumping for joy--This puts my team one step away from winning, and on top of that, it's all over Slashdot!!!
while(true) { follow_road(); }
"The race...features a $2 million prize for the first team whose robot crosses 175 miles of the Mojave in under ten hours."
;)
The winning design: Hummer + cinderblock.
I'm a coder, not an AI or image-processing geek, so these might be dumb questions... but...
Why the need for so many sensors? I can understand a use for them in low-visibility, eg dust or darkness, but the current models seem excessive to a layman. I mean, can one not use steroscopic cameras (scanning the field, as our eyes do), run edge and shade detection over the frames, and generate 3D terrain models in real time?
How does a vehicle determin terrain density and route selection? Can terrain texture be estimated based on reflection or image matching, so the vehicle can decide not to drive over some water or a bog, for example?
Even a good human driver is going to get stuck in the deset without learning how to handle a truck offroad. Is it feasible to train a neural-net system to select a likely course, possibly with a set of hardwired rules as a base? Eg, make your own way, but don't sink the car.
I've no doubt this stuff is Hard, but much of this appears to be done via brute force...
Forget thrust, drag, lift and weight. Airplanes fly because of money.
What does preventing the government from making a law abridging freedom of speech have to do with reading at -1?
IMHO it was the coolest; but I'm too lazy to look it up.
Team DAD and The Golem Group to name only two.
Seems like we've had GPS navigating cruise missiles for quite a while.
Of course. The purpose of the military is to kill people and break things. Ask any Marine. If you can't accept that, you shouldn't be in the arms business. Entering the DARPA Grand Challenge is being in the arms business.
The Golem I last year finished fourth, travelling 5.2 miles. It had the lowest budget of only $35,000 dollars (whereas some other teams' have a reputed budget of over a million...). And based on this image here, what I believe makes it uber-awesome is that they are cheating the competition by installing an elf under the hood and letting him drive.
Not 2003...
Kevin Fox
I'm somewhat curious as to who the 78 teams are that were cut from this.
I think this is probably just about impossible. The plain fact is that you have to get 150 miles in 10 hours. That's an average of 15 miles per hour minimum. You obviously need to go faster than that.
So I'm going into geek mode now and say that most of the time you need to be going 30 mph. So in one second (if i did my math right) you need to be able to see something ~14 meters in front of you. You most likely need at least a 5 second headstart since you have to break or turn. So about 60 meters sound right? The more the better. You need sensors/image processing that are powerful enough to be accurate out that far and then you need to process it. Don't forget AI so you know what to do.
Don't get me wrong, I think this is very cool and would love to try to do this, but I'm realistic. If someone suceeds this time I will be very surprised.
"Is life so dear, or peace so sweet, as to be purchased at the price of chains and slavery?" - Patrick Henry
I think they made a spelling mistake. It is "Palos Verdes High School". That's cool that they made the news. This is just their second year as a new high school and they seem to being doing great. I wonder why my school hasn't gotten into this...oh well we kicked their ass in football.
$sig$
I'm studying a course in 3D Computer Vision right now, at TUHH. It's part of the Erasmus exchange program I'm having here - the eigth and last semester (excluding the thesiswork) of my master of engineering in automation and mechatronics at Chalmers in Gothenburg. I can easily say this course is the most difficult one of all I've been taking for all of my study time, hopefully the three weeks I have between that exam and the last of my others, will be enough to learn what doesn't stay in my head during the lectures...
In fact, I have the course book right beside me. To begin, the description of it would be more or less along the lines "an orgy in linear algebra, mathematical statistics, with some flavouring of image processing, geometry, optimization and algorithms". Basically, it's 30-40% mathematical formulas, 650 pages, some containing things not even all MSc even learn like tensor notations etc. Not something I'm even sure is a good thing to recommend to very many slashdotters, even. You'll get its name though - "Multiple View Geometry in Computer Vision", by Hartley & Zisserman. ISBN 0-521-54051-8.
What I see as problems in the book, is that almost everything is working on corner detection. This is great, if you want to make 3D-models of houses or other man-made objects (at least half of the examples in the book are architectural, I would say). It's not so great if you want to image bushes, rocks and other things with not so obvious corners on them. Also, the process involves quite heavy processing - both image processing, finding all those corners, statistical processing (to sort out outliers, which there will be), and optimization to find the best fitting backprojection of the image planes). I don't have a sure grip on the needed processing power but I doubt, when considering realtime demands in a car, that it'll hardly be easy to get it working.
Also, it's still to a big deal itself an area under research. The situation with using 5+ images (from different cameras och just consecutive images from the same, moved camera), isn't very well known. Using more images, of course would mean a bigger chance to get a decent 3D model of the scene...
And still, you would at least need two cameras to do anything useful. You can't reconstruct 3D space without having at least two images of the object to reconstruct. And probably you will need more - you would probably want to reconstruct all the way around (ie more cameras on the sides and backwards), and add extra sensors like radar etc for extra checks.
And then you really haven't solved the problem of driving the car. You have only built a decent mapping of the 3D surroundings of it. You have to add AI/some kind of steering logic, which only in itself is a demanding task. Just look at all FPS games out there - if it would be easy to construct good AI, with a known 3D-world, tailormade for the figures, would we really be seeing that many games with crap-AI? I'm happy I ain't taking an AI course too, for sure!
I have a really elegant proof for Fermat's last theorem. If this sig was only a bit longer...
Isn't that kinda hard for our current excuse for technology? This has never been done before, right? I welcome our new killbot overlords.
Zhrodague.net - I do projects and stuff too.
Wired magazine has an pretty funny article on the results of the 2003 race with a description with what went wrong for each team.m l?pg=15
http://www.wired.com/wired/archive/12.05/start.ht
Don't be an asshat. The end goal is to build automated supply convoys.
Automated supply convoys? Dude, you're dumber than a bag of bricks.
A friend of mine does robotics work at CMU. He was building a "autonomous search and rescue vehicle" for the DoD, went through all the vision and mapping and what-have-you work and was appalled when he was supplied with the final control system and discovered that it sported a big red "Weapon" button.
Nobody in the military is going to ask for funds to build an autonomous bloody kill-bot, for the obvious political reasons. However, once you *have* an autonomous platform and a human-tracking system (so, you know, you don't "accidentally drive into your soldiers", or so you can "find them on search and rescue missions"), you stick a couple of servos and a gun on top and you have a basic version of your kill-bot. Once you have *that* done, it's a much smaller political step to request funding for improvements to "help ensure that none of our boys get accidentally shot", or something along those lines.
Remember the unmanned Predator? Yeah, that "unarmed", "reconaissance" aircraft? The one we stuck a Hellfire air-to-ground-missile to and used to assassinate some al Quaeda suspect (and a couple of people that happened to be in his car) in Yemen, waaay back in 2002? Once you have a vehicle platform, sticking weapons on it is not hard.
You really think that the armed forces would allow their budgets to be threatened by unmanned combat vehicles?
Naturally, you mean *other* than the abovementioned Predator (and the other unmanned aircraft that can serve as weapons platforms). And other than cruise missiles. Come on, man. Face reality. They've been doing this for a long time.
Commanders only think about how many men they command.
Wow. You have one heck of a simplistic view of the people in the military. If that was true, we wouldn't have most manpower-reducing military hardware we do today -- just lots and lots of infantry.
If I'm a general, and someone offers me something that means that I don't have to send people out into a meat grinder, suffer political pressure, pay death benefits, pay to replace someone, deal with photos of dead soldiers making their way into the press...no, I'd have to say that I'd be pretty quick to look into those kill-bots.
That doesn't mean that it's necessarily *bad*, having robots do killing for the United States. We just have to be aware of what we're getting into, and be sure that this is a path that we want to take. Occupations of a country may someday consist of just dumping robotic sentry vehicles all over it, things that shoot on sight anyone that isn't carrying the appropriate radio identification device (or is moving outside of their prescribed limits). No American soldiers need be killed, but we have to consider what happens when wars are cheap and easy to fight, and the media doesn't stir up the public over them.
Any program relying on (nontrivial) preemptive multithreading will be buggy.
Of course, if you *don't* build arms, you don't get the funding to do competitive bleeding-edge robotics work. Bit of a Catch-22, eh? If the NSF got the funding that the DoD is provided with to make tomorrow's killing mechanisms, we wouldn't have these kind of dilemmas, would we?
Any program relying on (nontrivial) preemptive multithreading will be buggy.
DARPA, and the DOD would *love* to have semi to full autonomous "kill bots" - in a way, today, they already have them for some tasks - they are called "cruise missles", which can be launched, told to stay on "hold" above possible targets, then commanded to strike on located targets. I would assume "located" likely means some form of lat/lon coordinates or painted with a laser (either by troops or from the air).
The exact same thing could be done with a kill bot: send it to a predetermined position, and tell it to "hold fire" unless acted upon agressively, or if non-friendly comes into position (at which point it could bark a series of commands in different languages to the offender - think of it as an active landmine with intelligence that can move on command), which if not heeded, shoots a warning, then if continued, shoots to kill. Friendlies are identified by RFID or similar tags. Equip them with the ability to identify each other, as well as to flock or coordinate efforts with one another. Other commands could be something like "fire on ident", where they could be set up, then when a target is painted with a laser (perhaps from a troop's rifle), it fires on that target.
You better bet that the DOD and DARPA would be all over such a system if it was proven field safe (to our troops) and easy/quick to use, and rugged. They are half way there with the TALON robots already, they just lack the rest of the package, which the Grand Challenge is dealing with...
Of course, one can also easily see the potential of scaled up versions - robotic Humvees and M1A tanks, as well as robotic quads, and perhaps legged versions...
BTW - this last was actually funded by DARPA back in the 1980's, which culminated in the Odetics, Inc. (now known as Iteris, Inc. - based in Anaheim, California - interesting the strange things going on at this company, whatwith name changes, etc - plus, they are developers of an "electronic highway" concept - I am sure there is no relation to the Grand Challenge - wink, wink) ODEX-1 legged walker - a very unique leg design that proved to be fairly robust and strong, while keeping outboard weight (on the legs) to an absolute minimum by moving all the electric motors inward toward the torso of the machine.
Think about it - if you could, in addition to GPS coordinates, vision systems, etc - also bury in the ground or nearby some form of active or passive "locator" beacons, such as what Odetics - oops, I mean Iteris - is developing - wouldn't the problem become just a little bit simpler...?
Nah - DARPA hasn't been thinking about this, not at all, not at all...
Reason is the Path to God - Anon
Last year's DARPA challenge was very good for breaking things. Killing people might happen if they stand anywhere around the course, and the robotic drivers decide you are just a sagebrush.
My rights don't need management.
Please stop calling it a kill-bot. Hello...Simpsons?
Shutting down free speech with violence isn't fighting fascism. It IS fascism!
"The race...features a $2 million prize for the first team whose robot crosses 175 miles of the Mojave in under ten hours.".
Too bad most teams have already spent more than that. Oh well, it's the experience I guess. It makes my FIRST robot seem like legos! Personally, I will be rooting for PV high (1) because they are a high school team and (2) they are 5 miles away from me. However, it won't be done. I predict a maximum of 20 miles.
Idiot, "kill-bot" is from Futurama, not the Simpsons. :)
keep dreaming...
http://www.navy.com/navyofthefuture/ddx
"Crew: Many of the functions performed by crews on conventional destroyers will be automated on the DD(X). That means a reduction in crew size - 330 fewer sailors than the Spruance class destroyers and 200 fewer sailors than the Oliver Hazard Perry class frigates. The crew will also focus on fighting versus ship maintenance."
http://www.navyleague.org/sea_power/apr_05_46.php
"Part of the 1,200-man reduction in the size of a carrier crew planned for the CVN 21 next-generation carrier will be accomplished by installation of the new launching system and arresting gear."
Everyone knows that a clone army will respond more creatively to situations that would stump a droid army.
No, the purpose of the competition is to find a replacement for UPS.
(see the newer
Hey all. I'm in charge of the actual pathplanner software for the OSU WAVE team. All of you should probably know that this whole DARPA thing is a joke. They don't want people to succeed. MIT, Princton, and even Ford didn't make it, but somehow my school did. Don't get me wrong, our software is freakin' amazing, but it was never used. The dumbest technology known to man that got lucky and passed 1/3 of the site visit tests got us through. I'm happy to be a part of this whole thing, but this whole thing is weird.
There's a zillion ways to destroy a heavy vehicle, or parts thereof.
SNAAAAAAAAAAAAAAAAAAAAAKE!
you can have my violent video games when you pry them from my cold, dead hands.
Prime UID Club
The U.S. Military is going to a model where the "heavy lifting" of combat will be done by autonomous or semi-autonomous units. It is all about reducing training costs and the "political cost" of human lives.
A V. You will F-22 fighters for Air-to-Air Superiority to control the skies as F-35 Joint Strike Fighter squadrons augmented with UCAVs take out ground targets. The UCAVs get the dangerous missions and the F-35s do clean-up work. Targets are found and designated by UAVs like the Predator and the Global Hawk. 80% of the Air Force's costs are training and personnel. Even when you get to hard equipment costs, the UCAV is 1/3 the price of a JSF and 8 - 9% the cost of a F-22.
The Air Force is going to this model with the http://www.fas.org/man/dod-101/sys/ac/ucav.htm/UC
The Army and the Marines will have the "packbot" and Talon robots to engage "hardened targets". Ground robots will still require human guidance and will most likely be an extension of the eyes, ears, and rifle instead of autonomous soldiers.
As a veteran of the Gulf War and a former paratrooper of the 82nd Airborne, I can not even begin to describe the advantages of a highly mobile force that can air-deployed or sea-deployed in 1 - 3 days anywhere in the world. The 82nd Airborne can have 1/3 of the Division anywhere in the world in 18 hours or less of the President saying "Go." Imagine them having that same deployability but now having that same size force able to employ more firepower than the entire Division thanks to robotic augementation.
The problems here are still those of logistics and command and control. You still need boats and planes to put your forces on site. Commanders still need to be able to get an assessment of the tactical situation from the field and give orders to their forces. Robots put that much more of a strain on command and control and logistically are yet another thing that must be transported to the front, kept running, and repaired as needed.
I wonder what the MOS of a robot repairman is in the Army?
"The purpose of the military is to kill people and break things."
It's obvious you know nothing about war history or the military in general. Just stop posting, makes you look much less stupid.
I was rooting for a somewhat local team but they didn't qualify, probably will enter next year though.
Their vehicle is pretty awesome, hope they can get some good navigation software worked out by next year. The vehicle can just drive over lots of obstacles that seem to get the various trucks and jeeps hung up, which gives it a good advantage, but they will probably need to "armor" it a bit from branches, posts, and loose rocks.
Videos of the vehicle here: http://www.howeandhowe.com/videos.htm
yup, i saw Real Genius too
you're pretty caught up in your viewpoint here, but i thought i'd point out that the vehicle doesn't have to be autonomous to be unmanned. Wired had a recent article describing how the Air Force is requiring that vehicles like the predator be flown by trained pilots, so it'd seem to be moving in the other direction if they were suddenly this much more miserly about the resources operating their vehicles.
otoh, throw some level of autonomy onto a vehicle that is being piloted remotely for the most part and you've got the ability to continue operation past the point where you're communications are jammed. the knowhow that goes into operating autonomously could also be used as a system to doublecheck the actions of the operator.
anyway, these killbots already exist, it's just that there's someone with a little joystick operating them now. there will probably always be someone targetting their weapons.
I actually used to live where they are doing the testing (Near Fort Irwin, CA) and have drove in that terrain with absolutely ZERO visibility (I'm not kidding, no stars, no moon, no headlights, all you can do is "use the force" and drive slow) and its quite the task.
Of course, 'bots in computer games can easily traverse terrain such as this, wonder why Bungie didn't have a submission...
Seems to me that a bunch of collaborative systems wouldn't be the worse way to go. Use a terragen/landscape generator and some scripts to create thousands of images of vaguely representative terrain. Run the images through a software representation of the visual sensors (including any feature detection/compression routines) to generate visual inputs.
Get some neural networks working on that input and matching it to a number of factors - visible surfaces, object recognition, material recognition, anything you can think of. Hook all of these into a neural guesser that tries to determine the overall 3D-nature of the surroundings (including objects (static and moving) and likely materials).
Connect the output from that to a path plotter. See if the result can naviagate the virtual landscapes in a virtual vehicle without stranding it. NN for two abilities - shortest reasonable path and road-following. Train it to avoid obstacles and dodgy terrain.
Use sound and vibration and pressure and heat and light level and GPS and shock and other sensors to determine the state of the vehicle and immediate surrounds. Limit the vehicle to 0.1mph and load your software. Have it prowl semi-randomly around the desert for a couple of weeks, calibrating its distance-guesses on the nature of the terrain against close-up sensor readings.
Throttle it up to 1mph and implement a time-dependent reward/rating sytem, setting a pair of goalpoints at semi-random locations and running the vehicle back and forth between them a couple of dozen times, then resetting the goal points and so on, until the fastest method is determined. Have the vehicle sensors constantly monitoring for potential damage and marking certain sensor/state patterns as potentially dangerous.
Ramp the speed limit up to 5mph and run a third set of trials. Measure fuel consumption vs different driving/environment patterns.
Ramp the speed limit up to 20mph. See if you can get the vehicle to tip over. Run over rough terrain, make notes of any damage or temporary inaccuracy of the sensors (via vibration, shock etc) vs speed and surrounds. Implement a goal of sufficient sensor/evaluation accuracy to allow undamaged travel over easy-to-rough terrain, plus a generous buffer.
Ramp up the limit to 50mph (and have at least three different remote-stopping systems at this point). Have the vehicle find a speed point over various terrain types that minimizes damage, fuel use and time taken while maximizing sensor/evalutaion accuracy.
Tweak the algorithms and hardware for speed and accuracy of evaluation under a wide variety of conditions. Abridge the NN. Make sure it's mostly only evaluating the terrain along the most likely courses. Increase the raw speed of the computers and shockproof them. Beowulf them, if you have to.
With luck, the safe max speed will be above 20mph on flat terrain. Without luck, you may have to settle for either getting the furthest in 10 hours without finishing, or flooring it.
--Kent Brockman
Child, go away, don't bother me. This was a quote five years before Futurama ever got on the air.
Shutting down free speech with violence isn't fighting fascism. It IS fascism!
Something like five URLS in the story and not a single link directly to the semifinalists?