Forgot your password?
typodupeerror
Robotics The Almighty Buck Science

DARPA Grand Challenge 2005 164

Posted by Zonk
from the gobots dept.
fishdan wrote to mention that the Darpa Grand Challenge is getting underway again. The qualifying rounds started yesterday. National media has picked up on the story, with pieces at the Washington Post and Seattle Times. From the Post: "The autonomous robotic vehicles began competing Wednesday in the first of a series of qualifying rounds at the California Speedway. Half will advance to the Oct. 8 starting line of the so-called Grand Challenge. The grueling, weeklong semifinals are designed to test the vehicles' ability to cover a roughly 2-mile stretch of the track without a human driver or remote control. Participants ranging from souped-up SUVs to military behemoths will be graded on how well they can self-drive on rough road, make sharp turns and avoid obstacles _ hay bales, trash cans, wrecked cars _ while relying on GPS navigation and sensors, radar, lasers and cameras that feed information to computers."
This discussion has been archived. No new comments can be posted.

DARPA Grand Challenge 2005

Comments Filter:
  • by HugePedlar (900427) on Thursday September 29, 2005 @11:56AM (#13676736) Homepage
    than a soccer mom driving her only child in an SUV it's an SUV driving no one.
    • than a soccer mom driving her only child in an SUV it's an SUV driving no one.

      *eyeroll* Oh, dear goodness, that is one of the most rediculous +4 insightful posts I've ever read.

      Right, because using an SUV chassis for a project that advances our knowledge and technological capabilities in the Computer Science fields of robotoics and AI is such a major problem in the US. Scientific research... bah! It's a perfect example of conspicuous consumerism! After all, using an SUV for it's original design spec

      • Actually I was aiming for +4 Funny. I'm as surprised as you that it's modded insightful. Personally I think it's refreshingly great that these vehicles are being used for what they're supposed to.
        • Actually, after I spewed my rant, I reread your post, and thought, "wait, uh... Oh, he's joking. Ah. uh... oops. :-}"

          I hearby unflame you. Sorry about that, heh, I've got this flamethrower and an itchy trigger finger sometimes.

          Of course, whoever modded you insightful instead of funny is still on crack. ;)

  • by JakiChan (141719) on Thursday September 29, 2005 @11:56AM (#13676743)
    I had a chance to see the Volkswagen / Stanford entry while getting my VW serviced. That cart is pretty cool. There's a rack and a half worth of gear in the back and the shift knob has been modified to allow a robot arm to be attached. The engine is a 5 cylinder TDI and the VIN says it's a factory prototype. I heard that when the challenge is over the car will have to be destroyed since it certainly isn't US legal. And in a parody of the "Drivers Wanted" slogan it says "No Driver Required" on the side. :-) Seeing it in person certainly made waiting for my oil change fun.

    On a side note...I wish they'd let more diesel cars in the country. The chase car is another Touraeg but this one is a Canadian V10 TDI. It has something like 500 lb-ft of torque but gets about the same highway mileage as my small VW does.
    • by op12 (830015)
      Here's a picture [cnn.com] of the modified VW Touareg.
    • by itistoday (602304) on Thursday September 29, 2005 @12:17PM (#13676966) Homepage
      I just got off the phone with a team that's there. Apparently Stanford did the best in the semifinals so far, making it through the obstacle course without hitting a single cone and cruising at a comfortable 40 mph. Carnage Mellon, a favorite last year, actually did surprisingly bad and ended up hitting a lot of cones. The University of Florida also had a good run, only nicking a cone or two. It seems like it's gonna be a worthwhile race this year. And trust me, it is really difficult to make one of these machines.
      • And trust me, it is really difficult to make one of these machines.

        Bah...I have three of them in my garage.

      • Carnage Mellon

        Was that pun intended? Given the circumstances regarding their test run, that's pretty funny.

        Keep in mind that CMU has two entries this year - Sandstorm (last year's design with upgrades) and H1ghlander (still a hummer, but using different systems and software). I haven't heard anything about Sandstorm's qualification run, but their website [redteamracing.org] says "H1ghlander nudged a gate and tipped one cone."

        I went to CMU, so obviously I'm cheering for them however I have a tremendous amount of respect f

        • No offense to them, but they lied about their run then. They drove for a solid 15ft on top of a pile of haybales when they ran into them and proceded to "stay the course" while on top of them. But maybe they left that off of their webpage (I didn't look, just taking your info). They still had a good run in my opinion, but they certainly did more than nudge obstacles. I'm not a fan of any of the big teams, but Stanfords run really has been the best so far.
      • The University of Florida also had a good run, only nicking a cone or two

        Really? I didn't even know UF had a DARPA team. Well, that fills me with some serious school pride. :)

        After searching UF's website, I found some more info about the UF team, Team CIMAR: http://cimar.mae.ufl.edu/grand_challenge/ [ufl.edu]

        Go Gators. :)

    • the shift knob has been modified to allow a robot arm to be attached. This may be a naive question, but wouldn't an autonomous vehicle be one of the few really good applications for an automatic transmission?

      • The transmission *is* automatic. However, it still has a gear select lever. The arm is so that it can do things like select a gear and go into reverse.
      • Actually it is the opposite - a good application of a manual. Manual transmissions still get better milage and handle torque better than automatics, when all else is equal. A computer can shift the manual exactly when required, (in fact that is what an automatic is, a fluid computer that shifts gears) with no issue that it takes a hand that should be on the wheel or some such. So why not put in a manual transmission and get those benefits?

        Note that I qualified things with when all else is equal. Au

        • by lowrydr310 (830514) on Thursday September 29, 2005 @02:03PM (#13677996)
          BMW has a production version of a fully automatic manual transmission. It can be found in the M3 and it's called the SMG (semi-manual gearbox?). You can let the computer control all shifting or you can do it manually with paddles on the steering wheel. I drove a SMG equipped M3 and it's a strange experience. In the fully auto mode, it feels like you're driving a manual transmission but someone else is shifting for you. The shifts are a bit jerky - nowhere near as smooth as your typical automatic transmission. Ferrari and Lamborghini have this available as well, and I'm sure there are a few others.

          Modern automatic transmission are very good. I have a 2003 Accord and it's the best automatic I've ever driven. Shifting is very smooth, and downshifting occurs when it is supposed to. It uses what Honda calls "Grade Logic Technology" which basically detecs when you're going uphill or downhill to determine whether to downshift or upshift much sooner than older automatic transmissions. It's also a 5 speed automatic which helps a lot. I believe Mercedes has a 7 speed automatic in their newer cars though I haven't driven them.

          • BMW is not the only one.

            Daihatsu has a similar steering wheel paddle-controlled gearbox as an option on most of their pensioner utility vehicles (YRV, Sirion F-Speed, Terios, most Japanese specific models, etc). And guess what - the manual is still better :-)

            As a matter of fact the paddle controlled gearbox idea was first implemented in Formula 1 cars very long ago. It is by no means a BMW idea and it is not limited to extortionately priced M3 erectile disfunction compensators.
      • It still has to get into Park or Reverse somehow. And they aren't allowed to use trained monkeys to move the lever... what would you propose?
    • On a side note...I wish they'd let more diesel cars in the country.
       
      I'd hate to see more diesel cars since many people are too cheap to properly maintain the things.
      • How is a modern diesel engine more difficult to maintain than a comparable gasoline engine?
        • It's not, but some people don't maintain them as well as they should (in the same way some people don't maintain gasoline engines as well as they should) -- the issue with the diesel engines is that when people do this (I'm talking about engines that are 10 years old) they force all their neighbors that are behind them to breath the diesel particulates (I've seen more than a few smoky Mercedes over the years)

          • I've seen my share as well. I regularly see a mid-90s Mercedes 300SD near my house that has a thick layer of black soot all over the back and whenever it accelerates, a big dark cloud comes out the tailpipe!

            My understanding is that modern Diesel engines don't have this problem as much, though they still emit tiny particulates. The low-sulphur diesel fuel mandate in the US (2008?) should help with this.

            I'm a big fan of diesel engines, though I don't need or like the models offered in the US. The Mercedes


    • Comment from a member of the Smart Sensor team:
      Go IRV! [indyrobotics.com]...(countdown in milliseconds to the race)


      Last year, most of the entrants died in the first 100-200 yards. There was even a motorcycle entry which went about ten feet. It's supposed to be back this year. This year's race will definitely be interesting.


      A novel form of hurricane relief [70bang.com]
    • The irony is the outcry over a few diesel cars that have very advanced emissions control systems, and are even cleaner than most cars on the road today. Yet nobody blinks about the unregulated (NAFTA) travel of tens of thousands of Mexican trucks, with absolutely no emissions controls whatsoever, plying American roads delivering goods from across the border.

      It makes NO fucking sense.
  • by lightyear4 (852813) on Thursday September 29, 2005 @11:57AM (#13676765) Homepage


    This will be a MUCH more interesting contest if the teams do better than the last time around. (the best team only got 7 miles [imagiverse.org] out of 175 total.) Granted, even that is impressive given the circumstances.

    I wish the best of luck to all of those competing.
    • I'm sure there will be more than a few teams that do better than the 7 miles last year. I'm on one of the teams and we have done many miles fully autonomous in the Anza Borrego Desert (very similar to the conditions at the DGC). The NQE is going well, on the first day many teams passed on their opportunity for their first run becaues they weren't ready. Of the teams that did do their run, about half made it and half didn't. There were a couple of highlights, one of the favorite team's vehicle flattened
  • Finally... (Score:5, Insightful)

    by evil agent (918566) on Thursday September 29, 2005 @11:59AM (#13676776)
    ...we're putting the "auto" into automobile.

    I for one am very happy to see this technology advancing. It's not gonna take much intelligence to make an autonomous driver better than most human drivers.

    • by GuyMannDude (574364) on Thursday September 29, 2005 @01:04PM (#13677439) Journal

      I for one am very happy to see this technology advancing. It's not gonna take much intelligence to make an autonomous driver better than most human drivers.

      The benefits of having cars that drive themselves will be enormous. First, these cars can be programmed to drive in a manner that conserves gasoline (e.g., no jack-rabbit starts, limit speeds to 55 mph, time their accelerations between stoplights so they don't have to come to a complete stop at every one). Second, cars that drive themselves in a rational manner -- instead of the emotional, irrational manner that people drive them -- can significantly reduce traffic jams. There is an insightful analysis of traffic jams at this page [amasci.com] which explains that jams are larely the result of people not letting other people merge into their lane coupled with the relatively-slow reaction time of humans. Cars that can synchronize their motion in relation to nearby traffic could make traffic jams a thing of the past.

      Not to mention that if the car drives itself, I can read slashdot on the commute home (or watch Natalie Portman movies).

      GMD

      • There is an insightful analysis of traffic jams at this page which explains that jams are larely the result of people not letting other people merge into their lane coupled with the relatively-slow reaction time of humans.

        The "merging traffic" analysis on that page is flawed. The "neatly merging zipper" fails to account for the fact that the newly merged cars must slow down in order to re-establish their previous following distance. Furthermore, those two animations are not actually accurate depictions of

      • You're forgetting one huge thing that self-driving cars could do for us- automated deliveries. Once we have cars that can drive themselves, I would think it could very quickly follow that we might have stores which only exist online and deliver items to us via automated courier. It's convenient for us, protects the retailer's merchandise and saves them the overhead of making a physical store a customer-friendly experience. We might even have automated garages allow delivery vehicles to enter, drop off mater
    • ...we're putting the "auto" into automobile.

      The "auto" in "automobile" refers to the ability to propel itself, not steer or navigate itself.
  • by fantomas (94850) on Thursday September 29, 2005 @12:04PM (#13676823)
    Only in the USA could it say FROM souped-up SUVs :-)

    Here in the UK it would probably be FROM a bunch of lego bricks and a clockwork motor UP TO a Sinclair C5 (or possibly an Austin Mini with an Aibo gaffa-taped in)...
    • by millahtime (710421) on Thursday September 29, 2005 @12:18PM (#13676968) Homepage Journal
      The challenge takes place in off road conditions. Existing vehicles like SUVs can handle the conditions where legos most likely can't. They didn't pic SUVs to pick SUVs. They picked them because they are vehivles that can handle the terrain
      • The challenge takes place in off road conditions. Existing vehicles like SUVs can handle the conditions where legos most likely can't. They didn't pic SUVs to pick SUVs. They picked them because they are vehivles that can handle the terrain .. and carry very large weapons and troops. They want the next robotank. That same tech will then trickle down to the consumer (whisper: for an outrageous amount of money of course ;).

        I can honestly say that I do welcome this technology. Anything beats the daily drive fi
      • They didn't pic SUVs to pick SUVs. They picked them because they are vehivles that can handle the terrain
        They changed the course to manicured suburban streets this year??
    • You are probably wrong there. Yes, it would be down market but the UK has a very strong tradition of home built vehicles. The problem bit would be getting sponsorship for the IT and sensors.

      The other bit would be finding sufficient countryside where unmanned vehicles could be let loose!!!!

    • It's a race, so there's sport.
      It's for research, so there's a utility aspect.
      They transport things, so they're vehicles.

      One could argue that every one of the competitors is an SUV.
  • This is very cool (Score:5, Insightful)

    by zappepcs (820751) on Thursday September 29, 2005 @12:06PM (#13676840) Journal
    The software and use of sensors, as well as the sensors themselves are being driven to places that they probably wouldn't have gone if not for this contest. Sure, the 2 million dollars is a big-ish prize, but bragging rights are bigger.

    I've seen some hobby roboticists building smaller robots for a scaled down version of this that are just amazing. Even on smaller scales, this is pushing technology. The good part? Much of the hobby stuff is pretty much shared in an OSS kind of way. That means that the technology behind all this will not belong entireley to the military, and will soon find its way into our vehicles and homes.... THAT is very cool!
  • by Elrac (314784) <carl@sLAPLACEmotricz.com minus math_god> on Thursday September 29, 2005 @12:07PM (#13676848) Homepage Journal
    I read the reports once in a while: The winners, or close-to-finishers, are huge SUVs filled with computers and special-purpose sensory equipment. What this tells me is that today's computer technology still has trouble, in many cubic feet of space, and with practically unlimited electrical power, to find realtime solutions for a problem that even severely IQ handicapped humans handle routinely while balancing a McMeal on their knees and keeping up a cell phone conversation. I would wager that, with a fair amount of training and suitable controls, even a dog could handle the task. So...

    Did AI research implode for lack of funding, or is it really that hard? Will we need Cray-like computing power to handle the sensory input quickly enough to work a steering wheel, brake and gas pedal? Or has this problem simply never been tackled by sufficiently big money? And, given the obvious military implications and a $400 Billion military budget alone, why not?

    All these questions are quite serious, and I'd be interested in hearing answers.
    • by zappepcs (820751) on Thursday September 29, 2005 @12:19PM (#13676979) Journal
      YES, this task is THAT hard. If the military could simply throw money at the problem and get the solution, there would be no DARPA Grand Challenge competition at all.

      The simple fact is that while we use senses in our bodies to do things, the similar versions for robots and autonomous vehicles are crude, expensive, and no-one is quite sure how to make them work the way we think they should. Computer vision is becoming a big thing, and despite the millions of people working with it or on it around the globe, there is still no standard way to immitate what the human does with one eye, let alone two. Humans have that inner-ear thing, and this tells us many things: if we are vertical, falling, rising, moving forward or sideways... Our eyes do way more than a movie camera does. People are only now beginning to understand how many ways that we analyze the visual data presented to us through our eyes.

      The problems of autonomous ground vehicles are greater than that of planes because there is so much to run into, get stuck on, fall off of etc. Just writing some code to keep a toy robot from getting stuck under the kitchen table is a huge task without boatloads of sensory data and processing power.

      The tasks the DARPA GC vehicles are trying to accomplish ARE that difficult.

      There are two groups you can try if you are interested in finding out more about hobbyists that are working on these problems http://www.dprg.org/ [dprg.org] and http://www.seattlerobotics.org/index.php [seattlerobotics.org] . There are many others, of course, but these two are fairly active groups.
      • Humans have that inner-ear thing, and this tells us many things: if we are vertical, falling, rising, moving forward or sideways...

        Great! So in the next Grand Challenge, someone can just hook up a Nintendo Revolution controller to the car and be guaranteed a win! Better pre-order my console now...

      • Inner ear? My $300 digital camera has an orientation sensor that can tell if it is vertical, horizontal, or moving. And it's solid state!
    • AI has been oversold time and time again. /. participates in publicizing the extravagent predictions of AI-ers.

      There has been tremendous progress in building software and hardware systems to do things that previously only humans could do well. Chess is an example; so is spam detection; so are various forms of pattern recognition.

      Where AI efforts have been singularly unsuccessful at is in replacing humans entirely for complex tasks in an unpredictable envirnoment. Also to do anything resembling "understand
    • by MOBE2001 (263700) on Thursday September 29, 2005 @12:23PM (#13677018) Homepage Journal
      Did AI research implode for lack of funding, or is it really that hard?

      None of the competitors are doing true AI. They are not using learning systems as far as I know. This is just good old fashioned programming where the designers/programmers try to think of all possibilities in advance. I don't see how this contest is advancing our understanding of intelligence. I think that the qualifying rules should have been more stringent and should have prohibited non-learning systems. Otherwise it's the same old traditional stuff.
      • If this were an AI contest, I would agree. But the goal is practical driving, instead, by any (software) means necessary.
      • by eclectus (209883)
        I hate getting sucked in by a troll like this, but... Please, can we quit having the argument of what is the one true AI? 30 years ago, making computers understand a man-made language of written words was True AI (TM). Now its called compiler design. Later on, True AI was making expert systems that mimicked the behaviour of experts. Now it's called rules-based systems. Lets face it, many people want to define AI to be 'that which we humans can do that computers can't", which is a ever-moving definitio
      • by acaspis (799831)
        the qualifying rules (...) should have prohibited non-learning systems.

        On Judgement Day, you'll fell sorry you wrote that.

        Joke aside, what's the difference between a learning system and a non-learning system ? Aren't the DARPA entries already immensely more "intelligent" than factory-floor robots operating in a predictable environment ?
        Is a Bayesian algorithm a learning system ? Is it AI ?
        Does AI have to be some kind of automagic algorithm that we can't analyze with the concepts of computer science ?

      • by Spy Hunter (317220) * on Thursday September 29, 2005 @01:28PM (#13677696) Journal
        The goal of the Grand Challenge is to produce useful robots, not "true AI". The designers of the contest realize that's a badly-defined goal that is unlikely to be reached in the near future (after all, people have been failing for decades). Instead they require results and don't specify the methods. If "true AI" is the best way to achieve results, then the people who use it will win. If it is not, then requiring it would be counterproductive.
        • If "true AI" is the best way to achieve results, then the people who use it will win. If it is not, then requiring it would be counterproductive.

          I see what you mean but I have to disagree. By not requiring learning systems, DARPA is not encouraging progress in AI. In fact, it is discouraging it because robot people love to tinker with their robots by progamming the hehaviors themselves instead of giving the machines the ability to acquire their own behavior through trial and error. The US defence department
          • by Zathrus (232140) on Thursday September 29, 2005 @02:25PM (#13678240) Homepage
            By not requiring learning systems, DARPA is not encouraging progress in AI

            Since visual perception and interpretation is often considered an AI related field of research, I'd say you're wrong.

            But, more importantly, you still don't get it. The GC's goal isn't to encourage progress in AI -- it's to develop an autonomous supply vehicle. Do you have any idea how much of the military is involved purely in transport/resupply?

            The US defence department would sell its soul for a truly intelligent system and that's what we should be after.

            Funny. That contradicts a rather large number of public statements from the DoD. And privately I suspect the more sane individuals don't want it either -- we've seen more than enough SF flicks that go into the potential issues with such a thing.

            include big-city driving in the challenge

            Yes, and we should make all toddlers learn to run before walking or crawling.

            It's called incremental progress -- right now the DoD could benefit immensely from a fully autonomous transport vehicle that simply goes between depots in low traffic but highly rugged environments. After that you could look at highway driving (which is already being worked on by all the major automobile companies) and then maybe high-traffic conditions. But that last one is of relatively little use to the DoD, and DARPA is only mandated for Defense related projects.

            As it stands, all we're gonna get is clever engineering which we already know we're good at, but not good enough.

            When it comes down to it, it's all just "clever engineering" -- especially in retrospect. Most progress is made in small steps, not giant leaps.
      • You are correct. What they're designing is the ability to create a three dimensional understanding of the world in our minds based on perceptual input. They're also designing an ability to cross-reference the vehicles capabilities with this internal map in order to identify a navigable path from one place to another. These are things that we take for granted, but without which our intelligence wouldn't be able to operate.

        Before you can understand something you have to be able to perceive it, or at least
        • And, no, a soccer mom with a big mac on her knee talking on a cell phone couldn't traverse the course that they'll be on.

          IMO, a soccer mom could do MUCH better than that, after proper training and sufficient practice in desert terrains.
      • Mod Parent Down (Score:4, Insightful)

        by Illserve (56215) on Thursday September 29, 2005 @03:48PM (#13679077)
        He doesn't know the mean of the phrase AI.

        AI doesn't mean "Learning", it means Artificial Intelligence. Said poster is probably a stage in his life where his visual system is relatively stable from day to day. Whether it got there by being hard wired by his designer or through learning is irrelevant. His intelligent behavior (barring perhaps said post) on a moment to moment basis is the result of his pre-wired system, not some kind of fabulously amazing learning algorithm.

        Some of the engineers attacking this problem are using machine learning, others are using pre-fab algorithm, most are using a combination of both. They're all true AI by any stretch of the definition.

    • by Jeremi (14640) on Thursday September 29, 2005 @12:23PM (#13677020) Homepage
      a problem that even severely IQ handicapped humans handle routinely while balancing a McMeal on their knees and keeping up a cell phone conversation


      Driving across 150 miles of roadless, obstacle-ridden desert is not something most humans do, or even attempt. Don't be so sure that "even severely IQ handicapped humans" could handle it routinely.


      Will we need Cray-like computing power to handle the sensory input quickly enough to work a steering wheel, brake and gas pedal?


      Yes, because being able to take two dimensional sensory input and use it to construct an acccurate three-dimensional representation of the local surroundings, and then plan a viable route through those surroundings, is not a trivial task. People do it pretty well (at least when on foot), but then they've had billions of years of development time put into their massively parallel computational hardware. Computers can do it too, and eventually that "Cray-like computing power" will be squeezed down into smaller boxes, but it isn't an easy problem.

      • Driving across 150 miles of roadless, obstacle-ridden desert is not something most humans do, or even attempt. Don't be so sure that "even severely IQ handicapped humans" could handle it routinely.

        I think that driving around a big city (New York, London, Paris, etc...) is much harder than driving around the desert, orders of magnitude harder, IMO. Especially during rush hour.

        Yes, because being able to take two dimensional sensory input and use it to construct an acccurate three-dimensional representation of
        • [...] this is precisely how not to do it. The coupling between sensors and effectors should be as short as possible, especially when your processors (neurons) are very slow.

          Good advice for designing a brain made of neurons, but not good advice for a system based on today's computers. Neurons are massively parallel and not very fast; computers are lightning fast and not very parallel. Attempting to implement brain-like processing on today's computer architectures is an exercise in futility (as decades of

          • If mimicking the brain is your goal, then a completely different computer architecture is needed. You need a large memory with embedded massively parallel processing units. You need a non-von Neumann architecture to eliminate the von Neumann bottleneck. Only after this brain-like computer architecture is developed will we be able to implement true AI.

            I think it will require a lot less processing power than most people would think. It is known that the brain can focus on a very narrow subject/concept at a ti
            • We could do amazing things with what we have, if only we knew how.

              That often-repeated claim fails to explain why progress in AI has been so abysmally slow. Are people really that stupid? Have we failed to see the answer lying in plain sight for so long? I believe there is a different explanation.

              We don't need more computing power exactly, we need a different type of computing power. The processing needs to be closer to the memory and massively parallel, but not necessarily very fast. Simulating thi

              • That often-repeated claim fails to explain why progress in AI has been so abysmally slow. Are people really that stupid?

                I think we are that stupid. We've wasted more than fifty years and billions of dollars on the GOFAI symbolic approach.

                Have we failed to see the answer lying in plain sight for so long?

                I think that, when we finally find the answer, we will kick ourselves in the ass for having been so stupid.

                We don't need more computing power exactly, we need a different type of computing power. The process
    • Controlling a car is not particularly hard. Knowing where to drive it is also not all that hard when you know what the terrain looks like.

      In my opinion, by far the hardest part of this challenge is in the sensing. None of the sensors we currently have for robots come close to giving us the same level of useful information we get from our eyes and visual cortex. These robots need to determine what the terrain looks like from noisy sensor data, and they need to do it fast enough and for a far enough dist

      • Off-road driving is more difficult in that there is no regular terrain pattern to follow. If you were on a paved road, or even a cart path, the task would be much easier. Off-road you have to gauge the depth of ruts and holes, the softness of sand, the grade, random obstacles, etc. I agree that the sensing of these things is probably the number one problem, but it is no easy task to deal with them once they have been sensed.
    • Yeah buddy, it's *that* hard. If you don't think so, please tell me how (using a laser or camera) to have a computer tell you the difference between a canyon (no laser return), a water puddle (no laser return, reflection on the camera), and the sky (ie: going through whoops and the vehicle is looking straight up). Yes, a gyro can help you with the latter. Our mind can figure out a lot of things that just take too much time to do with computers. I can see a puddle and look to either side of it to see how
      • So it needs clusters of CPUs working together. There would be stereo color cameras to work with the laser to determine depth at 60fps, along with the gyro sensors for orientation. That would handle the canyon situation. It would of course be combined with GPS and topo maps to have a general idea of the height change in terrain. Additionally, it should know where the sun is and based on the intensity of colors and shadows if a cloud was blocking the sun. Then it would adjust what its been programmed to
      • And the image processing software also knows that the reflection of sky on water is moving too slowly compared to the surrounding terrain, so that's another clue it is water.
    • The winners, or close-to-finishers, are huge SUVs filled with computers and special-purpose sensory equipment.

      You have to realize these are general computing components programmed to do specific task. Most design stage hardware is large for that very fact its being designed. Once the exact software and hardware needs are finalized a production version could shrink this down to much small size, and given moorse law it will be small enough to fit into a smallest car in no time.
    • by blackcoot (124938) on Thursday September 29, 2005 @12:50PM (#13677300)
      I wouldn't say that this is an amazing failure of AI so much as an amazing failure to realistically estimate the real difficulty of AI and the mess of systems integration problems that accompany robotics (I happen to work for a company that's part of a GC team and specializes in autonomous robotics). Firstly, sensors suck. We are just now barely approaching video sensors that have the same resolution as the human eye, but at 9+megapixels a piece, you have an insane amount of numbercrunching to do before you've reduced a frame into useful information. Now repeat that at 60Hz and you now have an appreciation for where a large portion of the computing power is used. Now take three such cameras for multi-baseline stereo and terrain classification and you're talking 1.6 gigapixels per second that you have to process. You also have to find machines which can sustain 3.2GB/s or 4.8GB/s transfer rates (depending on whether you use YUV 4:2:2 or RGB 8 bit per channel imagery). Now toss in a couple LADARs scanning at 100Hz, 360 x 16 bit samples per scan line, a bunch of RADARs operating at 30Hz, an IMU, two GPS units (one for the IMU, one for you to use)... you begin to see some of the problems. You need all those different sensing modalities because the fundamental truth of sensors is that they lie. You can do things to get reasonable estimates up to some confidence, but realistically what you're seeing are random values near the real values. Sensors fail, so you need back-up systems, and some way of determining which sensors failed (or rather, a way to change your beliefs about which sensors are reliable).

      In short, the classical AI part (most folks seem to use D* + reactive controls) is not where 90+% of the processing bandwidth is used, you need that power for sensing and for guaranteeing that your control loops cycle at at least some minimum frequency to guarantee safe operations.

      That said, there's a lot the gov't can do to make this problem a lot easier to solve. Standard bus designs (like FireWire) which can power most of the sensors on the bus are a really great start. Open protocols from the wire up are also important. A push towards integrating more intelligence in the sensors (embedded FPGAs which allow you to do optional processing on the raw signals coming in) can help quite a bit. Research into high-speed busses that allow you to pretend you have a shared memory multiproc will also help a lot. Finding a way to reliably and efficiently move processing algorithms into FPGAs or microcontrollers will also help to distribute the workload and reduce overall bandwidth and processing requirements. Unfortunately, there's still a lot of fundamental algorithm work to be done before you get to that point, but as certain algorithms becomes standardized this will become a lot more feasible.
    • What this tells me is that today's computer technology still has trouble, in many cubic feet of space, and with practically unlimited electrical power, to find realtime solutions for a problem that even severely IQ handicapped humans handle routinely while balancing a McMeal on their knees and keeping up a cell phone conversation. I would wager that, with a fair amount of training and suitable controls, even a dog could handle the task. So...

      The AI systems are competing against 500 million years of evoluti
    • by jd (1658) <imipak&yahoo,com> on Thursday September 29, 2005 @01:27PM (#13677690) Homepage Journal
      There are some interesting estimates out on the web of how fast the human brain can process data [merkle.com]. Current estimates are between 10^13 and 10^16 operations per second, which would put the upper limit at about 10 giga mips (remember, 'mips' is a million instructions per second). If we assume the brain handles 'reals' rather than integer values for data, then this translates to about 10 peta flops.


      In comparison, the world's fastest supercomputer (BlueGene/L) is rated at a maximum of 183,500 gigaflots, which is about 0.2 peta flops, or one fiftieth of the maximum speed of the human brain.


      Now, you don't NEED the full processing power of the human brain in order to drive. That's not my point. My point is that a car-load of computer parts, at the current level of technology, is probably going to drive about as well as a Horseshoe Crab. I'm actually very impressed that developers have actually got as far as they have, as they're very unlikely to be using state-of-the-art technology for this, most are probably using pile-of-PC architectures, not much more than some webcams for vision and basic motors for the robot linkage, most likely continuous for power - steppers have vastly superior accuracy but have no force behind them.


      You also have to look at the power cleaning systems they need - car batteries are NOT smooth and car electrical systems are typically pretty rough. On the other hand, computers need power that is spike-free and ADCs (analog-to-digital converters) rely on a steady reference voltage to be able to do anything useful. A noisy power system would be Bad News for a self-operating vehicle. Oh, and computers don't do well when hot, but air conditioning units - particularly if they switch on and off - are going to add some serious noise to the power.


      Whoever builds a car that can go a decent distance is worthy of vast respect and awe, because there are some massive technical problems that require ingenious hacking of mechanical, electrical and microelectronic systems to operate in some pretty harsh environments.


      I do think DARPA would be foolish to end the contest if there is a winner this year - rather, they should extend the challenge. Have the vehicles go through a wider range of terrains, as a multi-stage rally, perhaps, with cars who succeed in the desert then having to navigate through a forest, swamps, along the tops of snow-covered mountains - pretty much any terrain that a vehicle could realistically encounter if used for military missions.


      If DARPA did that, and the contestents succeeded, then (and pretty much only then) would DARPA have a general-purpose robotic vehicle they could throw into any arena that would be hazardous for humans under combat conditions. Why stop when you have something that could have made things easier three years ago had it existed, but which may be useless in a scenario three years from now, when the dangers may be completely different?

    • I'm replying to this, as it came up in my meta-moderation.

      Yes. AI is really that hard.

      If you could get the intelligence of a _cockroach_ to drive these vehicles, you'd win and get all sorts of funding by the DOD. Insects have no problems navigating their environments, and indeed, they even have a built in survival sense - something that is *way* down the road after navigation.

      Yes, you've asked some _very_ valid questions that illustrate that "AI is Bogus".

      --
      BMO
  • by Tom Courtenay (638139) on Thursday September 29, 2005 @12:09PM (#13676869)
    My money is on the team that spent all of their money on identical twin spokesmodels [axionracing.com]

    Yes I know, shamelessy stolen from Cruel.
  • I don't care if these vehicles can drive by themselves 2 miles or 20, just make sure they come with a red LED on the front that alternates back and forth and make it say "right away Michael" everytime I get in the car.
  • Video of MITRE entry (Score:4, Informative)

    by eludom (83727) on Thursday September 29, 2005 @12:10PM (#13676887) Homepage
    FYI there is a 5min introductory video clip of the the MITRE entry here:

          http://www.mitre.org/tech/meteor/ [mitre.org]

    I saw it a few months ago doing it's thing around the
    parking lot. It will be interesting to see how they
    do on a live course.

    • Apparently the MITRE team had a somewhat disappointing first day, like many contestants. It's not necessarily over though, as the judges choose the vehicles based on what they think will work the best, not necessiarly the vehicles that did the best in the NQE.
  • by robyn217 (575679) on Thursday September 29, 2005 @12:17PM (#13676960) Homepage
    Yeah, I just hope someone can finish the race. It looks like the best site out there to track the race is GrandChallenge.org [grandchallenge.org]. They have team write-ups and blogs.

    I know my money is on Austin Robot Technology. Vehicle "(Not Available)" sounds like it'll be a real winner. lol!

    -robyn [gearlog.com]

  • The driverless busses [timesonline.co.uk] are coming!
  • by Anonymous Coward
    http://www.popsci.com/popsci/science/09a7dd9a0cc36 010vgnvcm1000004eecbccdrcrd/3.html [popsci.com]

    "So Thrun pioneered what's known as probabilistic robotics. He programs his machines to adjust their responses to incoming data based on the probability that the data are correct. In last year's DARPA race, many derailments occurred when a 'bot's sensors provided faulty information, causing it to, for example, mistake a tumbleweed for a rock and stop in its tracks. Thrun's car didn't go off the cliff mentioned above, becaus
  • by Druox (911165) on Thursday September 29, 2005 @12:37PM (#13677160)
    Has any of the contestents overcome the obstacle of negative space (i.e. a cliff, a sudden drop, a crater)?
    Its easier to detect something that is there like a bale of hay by radar, but what about something that isn't there (isn't an object sticking out of the ground, in y+ axis)? If not, I can see alot of Wile E. Coyote incidents with these cars flying off cliffs.
    (**poof**)
    • Ok, I wasn't going to post, but this got modded up as "Insightful" and I couldn't resist - this has to be one of the dumbest posts in this thread...

      What do you think that bale of hay is sitting on? Radars recieve ground bounces all the time - even in airborne applications. Usually radar people call that "clutter", since if you're looking for airborne targets it's information you don't want - here it's information you *do* want. Depending on how the radar is mounted, it could create a ground map and tr
  • by spicydragonz (837027) on Thursday September 29, 2005 @01:06PM (#13677459)
    Remember the famouse automaton Mephisto from the 19th century that claimed to be a chess playing robot.
    http://www.angelfire.com/games/SBChess/automaton.h tml [angelfire.com]
    I think I could hide a midget inside an SUV with enough computer looking doohickies to make a cool $2mill.
  • my company is sponsoring a team and happened to book some extra hotel rooms, so i'm going off to watch the race (instead of gambling away my soul in vegas). bunch of other guys are going from work as well -- should be a blast.
  • Tomshardware is posting daily updates live from the Fontana Speedway http://www.tomshardware.com/hardnews/20050929_1259 19.html [tomshardware.com]
  • Once the military has truly autonomous vehicles it won't be that much longer before we cede our control of cars to a computer.

    Just make sure it isn't using an MS operating system. Otherwise a BSOD will take on a whole new meaning.
  • Cornell's Team (Score:2, Informative)

    by spenceM7 (683840)
    My home team, Cornell, is currently in second place at the qualifiers, knocking only over one cone on the obstacle course.

    You can read their blog here [blogspot.com], or find their website (with technology writeups) here [cornell.edu].

A CONS is an object which cares. -- Bernie Greenberg.

Working...