Slashdot Mirror
DARPA Grand Challenge Teams Submit Videos to DARPA
doughnuthole writes "The deadline for DARPA Grand Challenge teams to submit their videos to DARPA just passed and some have posted them online. Some of the teams with these videos posted are Team Caltech, Axion Racing, Virginia Tech (on the Media page), Insight Racing, and UMass Dartmouth. The Grand Challenge is a 175 mile race run by fully autonomous vehicles. Since no teams completed the 2004 race, DARPA decided to run it again, this time for $2 million."
Let's hope the extra money makes the difference this year.
Get a free iPod Nano 4GB!
they have systems like this, it uses metal rods embedded in the road and sensors on the car to figure out where exactly in the lane it is.
Your hair look like poop, Bob! - Wanker.
but where I live it seems most of the cars on the road are self driven. The person behind the wheel is usually taking care of more pressing things like chatting on the phone, eating, putting on makeup, etc.
The problem is figuring out when it needs to leave cruise control and slow down, speed up, avoid collisions, turn, etc. You know, the 99% of the time when you aren't on a straight, flat, open road with no traffic.
I still have more fans than freaks. WTF is wrong with you people?
I think this is pretty cool and all as an engineering effort, but what specific purpose does this have?
I think that the government is secretly trying to create an army of autonomous pizza delivery drones. So long as you live in the desert, one of these few complex machines crafted by some of the world's best and brightest might get you your pizza within 30 minutes or less.
Flying cars and cruise control situations would, in reality, both be considerably aided by being in an electronic environment with constantly updated information from the road / flight path about the environment and changes in it, such as the fact that a blind corner/cloud is coming up, and there is a vehicle stationary out-of-site around it.
What DARPA wants is a combat-ready system that can drive cross-country with little or no outside help.
Different problems, and they require different solutions, and a system that could pass the DARPA test would be overkill and unsuitable for the daily commute to work.
One of my buddies is on one of these teams (which will remain nameless). According to him, one of the major problems last year was the enormous amount of time spent on sensor and hardware development vs. the incredibly small amount of time spent on software development and testing. When you have an autonomous vehicle with great realtime terrain mapping capabilities, but with no collision detection code (!), you get a hunk of metal with great vision that likes to run into the same bush over and over again. Other teams did have collision detection in, which would back the vehicle up when a potential collision was detected. However, the vehicle would pick up too many potential collisions (lots of bushes and rocks in the desert) and continously back up! Just goes to show you that developing the algorithms for these things is pretty important. Let's hope they get it right this year... although the chances are pretty slim.
This thing is the SportsMobile from Team Caltech: just imagine the advances in science if Snoop Dogg had entered the DARPA Grand Challenge! Pimpin' hard but somebody's gotta do it I guess...
"A door is what a dog is perpetually on the wrong side of" - Ogden Nash
http://www.stackworks.net/TASVCD.mpg
It takes an amazing amount of work to get to even this stage, but we're making very fast progress.
We're a new, entirely student run team with a very limited budget, and always looking for sponsors. If you know anyone who can provide money, equipment, supplies or other assitance, let us know!
Interesting that the DARPA Challenge rule precludes hovercrafts from competing as they do not primarily use traction with the ground for propulsion.
I hope a team enters with a very large metal hollow sphere with gear teethmarks lining the inside and the machinery be a shaft, gyroscope and gears on both ends of unequal weight. The rest of the contents can be fuel/energy cells to power the rotating motion of the machinery inside -- it essentially just has to "throw" itself in a direction and keep rolling and steering. Travelling 175 miles ought to be possible as long as the terrain isn't ridiculous.
Dossy's Blog
A load of drone cars should be driven behind these competitors with models of druglords with machine guns, outlaws, corrupt police officers, and people to whom the teams own money.
That way, they should be able to defy gravity enough to land from a great height at a mere 30 degrees to the horizontal without breaking their suspension or driveshafts, and continue to turn even after front side collisions which would leave bushes and pinions bent (or wheels set to cambers which would normally cause them to no longer turn). They might even get speed boosts beyond the maximum engineering speed expected for the motor, gas and gearbox actually installed.
For an added bonus, they could have critically wounded people in the back, and an accomplice who absolutely has to jump out at 40mph somewhere mid course in order to continue some secret mission.
Conversion Rate Optimisation French / English consultant
What DARPA wants is a combat-ready system that can drive cross-country with little or no outside help.
Different problems, and they require different solutions, and a system that could pass the DARPA test would be overkill and unsuitable for the daily commute to work.
I suppose that depends on what one considers a combat zone; dealing with "drivers" who are busy eating a bagel, reading the paper, drinking coffee, talking on a cell phone, shaving, and putting on makeup might require a combat-ready system.
this is my sig
This is the Defense Advanced Research Projects Agency. In case every forgot , their job is to kill people more efficiently. I seriously doubt any amount of money sunk into this will enable robots to kill more than humans on america's highways. What does the military need a robo-car for anyway?
No matter where you go , there you are.
I think the big problem is one of cost and transisition. The most obvious solution is that the road and car talk to each other. However, this would mean that just upgrading the largest highways would be very expensive and take a long time. This would mean that there would be little benefit to those who own cars that are equipped with the ability to drive itself. A guess this is the whole chicken and egg problem.
The other issue is what to do about oder cars not so equipped. In addition to the road and car talking to each other, I also see cars speaking to each other about where they are going and position. Assuming that there is 90% adoption of cars with autonomous gear, what do we do about the older cars? Not allow them on the roads? Maybe they could be retrofitted, but the value of that would probably outweigh the worth of the car.
I'm a member of the Cornell team. I've decided not to post our site, knowing that it would receive a brutal Slashdotting, but here are a couple of the off-site press releases.
m l2 005 /03/02/42255e9b352d0n ews/display.v/ART/2005 /02/24/421d88d386293e ws/25/7487/
http://biz.yahoo.com/prnews/041216/sfth012_1.ht
http://www.cornellsun.com/vnews/display.v/ART/
http://www.cornellsun.com/v
http://www.automotive.com/n
Feel free to Google us and come visit our site.
Who does DARPA think they're kidding, anyway? Add a zero or two, and you'll have the kind of money the DoD would hand to a defense contractor just to work on a problem like this, never mind with a deliverable that had to work. No sane defense contractor would promise such a thing, anyway.
Not Members only anymore. Heres a page with all of them. http://www.titaniumforums.com/torrent/torrentmain/ guest.php
enjoy!
Just forget it. Automotive control innovation is
not going to be taking any leaps. The auto makers
had enough trouble with anti-lock breaks.
Thank the lawyers in congress and the trial lawyer
association for this.
From Carnegie Mellon, I think The "Highlander" a modifed '99 hummer and "Sandstorm" Torrent of both videos http://www.titaniumforums.com/torrent/torrentmain/ guest.php
It's interesting, 50 undergrads took the course in 2003 at Caltech that did technology surveys for the vehicle, and there are about 1000 undergrads at Caltech total. So, 5% of the undergraduate population took part ...
One of our biggest problems in Silicon Valley has been finding a big open space in which to test. We now have access to a huge parking lot built during the dot-com boom, adjacent to an unfinished building complex. So we have the Overbot winding in and out among the parking islands. We'll be testing there today in a few hours.
In terms of technology, Team DAD probably is most innnovative. Everything runs on digital signal processors. They're building their own laser rangefinder this year. Last year, they got further than anybody else without wrecking. (CMU crashed three times; their HUMMV was able to survive the first two.)
Nobody seems to have a breakthrough in sensing. (By this I mean sensing good enough to evaluate terrain. Detecting big, solid obstacles is trivial.) LIDAR line scanners are too limited, stereo vision doesn't register well on dirt, and strong intelligent vision processing requires a breakthrough that twenty years of research has failed to produce. The breakthrough needed is flash LIDAR, which exists, but wasn't ready soon enough for this year's Grand Challenge. (The rules prohibit the use of Government-funded patented technology not available for general sale by last August, and the good flash LIDAR wasn't available in time.) CMU has a LIDAR line scanner on a giant gimbal, and we have a LIDAR line scanner on an overly large tilt head, but that's an interim solution and a technical dead end.
On the other hand, GPS and inertial gear gets better and cheaper every year. It's surprising how good it is.
This year, everybody who makes it to the starting line should disappear over the horizon at the start. The minimal level of performance to make it through the "site visit" hurdle is above that demonstrated by most of the vehicles entered last year.
And this year, DARPA is putting tank traps on the course.
Apparently Team Robo Monster is looking for programmers in the Los Angeles area. The team is headed by the guy from the noteworthy robots that jump blog. From their page:
We are currently seeking additional team members with the following skills and/or resources:
* Sensor integration
* DSP programming
* Gumstix hardware/programming
* Autonomy/AI - CYC/DAML or other ontology-based programming experience
* Game engine programming (for robotic simulation)
* Desert racing/driving experience
* Hobby robotics competitions (e.g. Robo-Magellan, Robotica, Robot Wars)
They've got quite a nice vehicle, and so far have things at the "drive by wire" stage.
Not sure what's with the blink tags on their web site, though.
Don't forget the Ripsaw used by TeamTMT. Their vehicle looks much cooler, and more fun to drive, than the wimpy one from UMass Dartmouth. All they need to do is figure out that AI problem and they're done...
to select reverse.
It's amazing that people still fall for DARPA's B.S., after last time when they changed the rules over 25 times and stil haf not admitted that they had changed the rules, after they ruled out my S-10 that was 25% race ready.
/. is forgetful enough to cover their B.S. "race" that rules out 99% of the teams after promising the a spot.
It's too bad they don't have ANY integrity, and it's worse that
Andy Out!
Hey,
I am Steve Jones of UBC Team Thunderbird. We are a Canadian team based out of the University of British Columbia. We have a very clear plan for how we are going to use Fuzzy Logic to deal with the unique challenges of this vehicle and are very actively implementing those plans at this time. We have been making an incredible amount of progress over the past few weeks as we have moved into the full implementation stage after a lot of planning. We are already well beyond what is visible in the Discovery Channel segment. You can find out more about our team at http://www.ubcthunderbird.com/. (We should have our entrance video posted later today but right now you can see a segment that was broadcast on Discovery Channel and CTV.)
Thanks,
Steve Jones
Instrumentation and Sensors Group Leader
steventy@gmail.com
Axion Racing is looking for talented software engineers. If you live in the San Diego area, and are interested in partipating in this historic race, please contact mdumas AT axionracing.com.
Hope to see you at the race!
Well apparently team Simple Solution Inclusive had a much harder time of it.
They have an amusing write up of theirs trials and tribulations (and electrical fires) in getting their video ready. http://ssinc.us/TooSimple.htm
Maybe they should be called Team Murphy's Law
KD
KD Founder Simple Solutions Inclusive Not to Over Complicate the Matter
why is that a problem? use a simple graphics system to determine needed things, like other vehicles, edge of road, lanes, etc. use laser or radar to determin relative speed to other vehvicles. Simple modifications to the car for steering and speed control. I imagine I could do this for under $5k with off the shelf materials, and I'm not even a real engineer.
like I said, considerable thought.
Good judgment comes from experience, and a lot of that comes from bad judgment.
If you were a real engineer, you'd know it isn't that easy. How do you determine another car in front of you? Shape, color? Computers are NOT good at doing graphical pattern matching. Its non-trivial to recognize slow moving identical objects, much less fast moving objects of widely differing shapes and sizes (like a sedan and a big rig).
Detect lane edges? How? You can't assume special equipment in the road to tell you, it doesn't exist yet (and could fail if it did). Look for patches of white lines? Sure- if they aren't faded out, if they exist at all.
How do you detect a pot hole? No two are identical in shape or size. What about road debris and litter- how do you program a computer to know whats safe to hit and what isn't? Computers are bad at such judgement calls, they aren't wired to make assumptions like humans do.
Remember- some of the top universities in the country, most of them with fairly large budgets, competed last year. None of them made it a mile. This is not a simple project, it goes right to the differences of how computers think vs how humans do. Short of an Einsteinian leap in AI, this is a problem that won't even come close to a general solution soon. At best we can hope for a partial solution in a highly controlled test environment in the next few years.
I still have more fans than freaks. WTF is wrong with you people?
Okay, so I didn't say it was easy, it isn't trivial, but maybe that's the problem with "real engineers" is that the solution has become far more complex than it needs to be. Us engineering tech types (kind of like real engineers but get paid less) like to keep things simple.
Let's look at how us human types drive. We basically determine what is road and non-road. More specifically, what is our lane and what is not-our lane. We keep our speed at the speed limit (in theory). We don't hit anything directly ahead of us. We have to stay in our lane and on road.
Ok, so we know about the size, location and most important shade of color of the road. Anything that is drastically different than this is NOT road and we can't drive on it. Factor in some exceptions for known things like shadows and other vehicles on the road. Detecting the edge of a lane or the shape of a car in front of you is trivial for a visual system. You don't need to know if it's a sedan or a big rig, just don't hit it.
How do you as a human detect the edge of a road if the lanes are faded out? I can guarentee that a computer with a camera and a half assed algo for finding the edge of something can pick them out 1000% times fainter than you or I can.
To be fair, I do a lot of programming of visual inspection systems, so I have a bias on how easy these things are to work with. I've seen a computer inspect the solder joints on a part in light that was so dim that the part wasn't recognizable to the naked eye.
A few years back a group (don't remember specifics) got a car to pass the German driving test. Having taken the German driving test myself, I can attest to the fact that this is quite a feat. So I would have to believe that the DARPA contest last year would of been somewhat more challenging than your run of the mill Sunday drive.
Damn, this makes me want to fire up that spare P3 750 I've got at home and dig some CCDs out of the scrap bin here at work and get started on something like this for this weekends drive to KS to pick up my kids.
Good judgment comes from experience, and a lot of that comes from bad judgment.