DARPA Grand Challenge Kicks Off March 13th

GillBates0 writes "A quick reminder that the DARPA Grand Challenge is due to kick off March 13, the coming Saturday." He points to this "quick recap of the teams participating in the event," as well as details about the available satellite feeds. "The Atlanta-Journal Constitution is running a story about the event today. Quoting Frank Dellaert, co-director of Georgia Tech's robotics lab from the article, 'I would have trouble driving some of these roads myself. I think it's beyond the capabilities of autonomous vehicles today.' (shameless school plug). We'll see if the participants can prove him wrong." Iphtashu Fitz adds a link to the New York Times' coverage of the trans-Mojave race, whose participants include "among other things a seriously tricked out motorcycle. The race is being run by the Pentagon, who is offering a $1 million prize to the builders of the first robot to successfully navigate a 200 mile route across the desert. ... a blog on ScienceBlog about the race has just started as well."

Comments

so you telling me .... $$$

[Quadfreak0]

So anyone planning on hi-jackin... uh I mean borrowing some equipment thats just rolling around the desert? Nobody is physically standing there watching right?

Re:so you telling me .... $$$

[ethx1]

All vehicles will be monitored by DARPA. I saw a story about it on Next@CNN yesterday that talked about the race.

Btw, next sunday's episode of the show will have more coverage of the race and the results. Should be fun to watch.

Re:so you telling me .... $$$

[Anonymous Coward]

Sweet Jesus, you can't stand there! That's bat country!

Researchers are wrong

[tyrani]

They're looking in the wrong place. I see a half dozen idiots drive "autonomously" to work everyday while eating, reading the frickin news paper, shaving, applying make-up, etc.

Researchers should be looking to these people for the artificial intelligence that they need!

the real problem...

[Anonymous Coward]

I would have trouble driving some of these roads myself

The real problem is that his turn signal is on for 150 of the miles and confuses all the autonomous vehicles.

Re:the real problem...

[4of12]

his turn signal is on

I've always regarded this driving gaffe to be the moral equivalent of leaving one's fly open.

And the steering wheel auto shut-off after a turn is completed is not enough.

I swear, Caddies and Town Cars ought to be equipped with ramp function for loudness (up to DEF CON 5 buzzer level) and interior brightness (disco strobe light intensity) for turn signals as a helpful reminder that need to be shut off.

Education

[gid13]

In vaguely related news, this Friday, my Eng Phys class (okay, not mine, I did it last year) is requiring the students to slalom autonomous vehicles around pylons of arbitrary position (though powered devices are allowed on top).

As I said, I did the course last year (it was easier at the time), and let me tell you, it's harder than it looks. Hats off to anyone who even comes close to finishing this.

I was all excited until.....

[tyrani]

...I read the articles and found out that there would be no hack-saw blades nor pneumatically controlled spike hammers.

Now that would be an autonomous vehicle I'd pay to see.

The Real Purpose Of This Contest

[Rhett]

Is to prove to the pentagon that terrorists with a million dollars in funding can't build this.

Unfortunately, no one will will this contest.

Re:The Real Purpose Of This Contest

[Neo-Rio-101]

More likely is that they want other people to research how to build a future battledroid to capable of military-level spy/front-line army work... capable of survivng desert conditions for a given number of miles.

Red vs. Blue

[Anonymous Coward]

Red Team: "Muahaha! Bow before the greatest CS school in the universe and our giant Hummer that is smarter than you!"

Blue Team: "Feh! We can beat you with two wheels tied behind our backs!" *obscene gesture*

Red Team: "Come a little closer and say that."

Re:Some info on my team

[Animats]

Yes, mod parent down. I'm the author of the original article.

Whatever happened to the wilderness areas?

[Whatsmynickname]

How will these robots be routed around wilderness areas generated by the California Wilderness Protection Act?

Wasn't the Barstow to Vegas motorcycle race cancelled due to declaration of these same wilderness areas? How is DARPA ensuring these vehicles aren't going to run over some tortoise?

Dont' get me wrong, as I'm no tree-hugger. However, it seems the Wilderness protection act only applies to people who cannot afford a congressman...

Re:Whatever happened to the wilderness areas?

[Anonymous Coward]

From the Atlanta Journal-Constitution article:

And some fear that, at speeds that might reach 50 mph, the robots also pose a threat to the desert tortoise --- a federally threatened species and the official state reptile of California. Sluggish after a winter of hibernation, the tortoises usually emerge from their burrows this time of year.

Under orders from the U.S. Bureau of Land Management, teams of biologists will sweep the race corridor before the competition, moving any tortoises out of harm's way and fencing their burrows until the robots pass.

The tortoises, which have resided in the Mojave for 60 million years and, as individuals, often live to be 100, probably won't pay much mind.

Re:Whatever happened to the wilderness areas?

[Wesley Felter]

The robots have to follow a predefined route or they are remotely deactivated; I assume that DARPA has chosen a route that doesn't go through environmentally-sensitive areas.

Re:Whatever happened to the wilderness areas?

[Whatsmynickname]

If they have to route through this, then all I can say is "good luck"...

Re:Whatever happened to the wilderness areas?

[Whatsmynickname]

See, that's what confuses me...

I thought the wilderness act didn't allow any vehicles at all. I can't even ride a mountain bike through a wilderness area. Yet, they allow this race with unmanned vehicles?

Ironically, I wouldn't be allowed to test my own USV, if I were in the process of developing one, in the same area. Even if I were testing a UAV, the UAV would not be allowed to land on any wilderness property!!! I shouldn't even ask about riding a motorcycle through here...

Yet, they're going to let these vehicles rip through a supposedly fragile ecosystem? I ask, where is the fairness in all of this?

The ultimate RC car

[daemonc]

I want one of these for Christmas, damnit! You could run over your neighbor's house with that mo fo...

Current status?

[Animats]

We (Team Overbot) dropped out over a month ago. We couldn't deliver a safe vehicle in time. Two of us are flying down tomorrow to watch.

At least two other teams have formally dropped out, and we expect some no-shows.

CMU is the favorite. Fifty people, $3.4 million spent to date, direct support from aerospace companies, and a team leader who expects people to work all night, day after day. (Read the article in the current Scientific American.) But their technology is rather disappointing. The whole route is preplanned by hand, using a bunch of people at workstations in a big trailer with maps obtained by overflying the route with LIDAR-equipped reconnaissance aircraft. It's not very autonomous. They found a loophole in the rules and exploited it very effectively. There's no breakthrough there.

Anthony Lewandosky, with his self-balancing motorcycle, has the most innovative technology. We've met him, and are impressed.

Palos Verdes High School has a viable entry, using a Honda Acura. We've loaned them some hardware. They've had autonomous driving working for months. They started by having handicapped driving control actuators put into a car, which simplified their mechanical problems. They debugged using a golf cart. Very nice work.

Caltech tried to qualify today, but their vehicle made an unexpected turn and bumped into something. They get a second chance on Wednesday.

Most likely, no one will finish. Nobody has really done enough field testing yet.

John Nagle

Re:Current status?

[Avionics Guy]

Caltech tried to qualify today, but their vehicle made an unexpected turn and bumped into something. They get a second chance on Wednesday

Yeah, our software freaked when it saw the "cow catcher" and tried to backup the vehicle, but drove forward instead because the gear selection software was messed-up. We should be in good shape for our run on Wednesday.

Re:Current status?

[Animats]

We've had trouble with that. The transmission in the Polaris Ranger won't reliably go into a gear selected by position only. There's too much slop in the linkage. We need to open up the transmission and put a sensor inside it.

Re:Current status?

[Freeptop]

I have to take issue with the ideas expressed that, a) the Team Red robot is not autonomous, and b) there isn't any impressive technology in their robot.
First, I'll address a). Autonomous means that it drives itself with no outside control. I'm assuming that you are implying that having detailed maps constitutes outside control. I disagree. When a person drives somewhere they've never been before, they usually use maps themselves. If they've been there and are familiar with the area, they basically already have a mental map that they consult. Pre-mapping the terrain and giving that map to the robot is providing essentially the same kind of information. The robot still must perform extensive obstacle avoidance, and must be able to deal with the rough terrain that it will encounter, which happen to be the harder tasks than simply knowing what the general path to take is beforehand. Addtionally, I might point out that the general aim for the whole challenge is to produce vehicles for the Military - do you really think that the Military would not want to be able to provide detailed maps to any autonomous convoy or fighting vehicle ahead of time?
Now to address b). If there wasn't any impressive technology in this robot, the Grand Challenge wouldn't be very much of a challenge, would it? One of the parts I find pretty impressive is the sensor array on that robot, which stabilizes itself, and "looks" the direction of the path the robot wishes to travel, much like a human does when driving. That's actually pretty cool, in my book.

Honestly, all of the teams have pretty impressive technology. I don't think it is really appropriate to insult another team in the competition.

The favorite?

[Innominate Milquetoa]

So how does the favored team train for a 210 mile race through the Mojave Desert?? Why, by testing it in the SNOW of course!

DARPA and Gregg

[joemontoya]

I love that show! DARPA with her wacky ideals and her lawyer partner with a corncob in the butt.

Entries too complicated?

[evilviper]

Is it just me, or are these entries too complex?

I admit I haven't made anything like this (although I've made some very advanced machines before) but it seems to me that a half-dozen laser range-finders connected to a laptop would do just nicely.

You could tell how far you were from the left/right of the road, and how far in-front of you an obsticle or change in terrain is (and can slow-down appropriately).

I suppose you'd also have to throw-in a $200 GPS reciever, since they have a "course", and you'll need to do more than just follow the road. But that seems to be all you'd need to accomplish this (yes I'm glossing over the basics, because they're just the basics).

So please, find fault in my idea. I'd like to know why this $5000 solution wouldn't work, and why 3+ million is required.

Re:Entries too complicated?

[Anonymous Coward]

Some parts are off road, thus you need to navigate around rocks, hills/gullies, potholes etc etc etc. Here are some previous generation stuff that the Robotics Institute at CMU has done over the las 20 years, many of which involve autonomous highway driving, but driving off road is much harder as you cant follow any lane markers and you need to determine the best path so you don't flip over.

Re:Entries too complicated?

[Paul Komarek]

I don't think you're going to get six good laser rangefinders for $5000. I don't remember the numbers as well as I would like, but I think the current favorite rangefinders (I think the brand is SIC?) are well over $1000 each. And you will quickly exhaust your laptop's computational power just denoising the output from crappy sensors. Heck, maybe even for the best sensors.

Autonomous vehicles have already driven across the country on highways, 98.2% of the time without human intervention. The roads it drove on are (I'm guessing) likely to be much nicer than those in the desert. Furthermore there was a human available to handle the surprises. For humor value: I believe one of the self-driving vechicles from CMU has a learner's permit from the state of Pennsylvania. See No Hands Across America for more info on this project.

The hard part of any project like this is uncertainty in the environment. The road may "disappear" completly from your sensors, or you may spot multiple roads. Maybe some mica on a rock screws up your rangefinder. Maybe your vehcicle's transmission gets a little "funny" and you can't shift properly anymore (I saw such a comment attached to this article). And we aren't even talking about genuine malfunctions like a failing rangefinder or sticky throttle.

I think autonomous systems might be the best example of the best laid plans of mice and men not succeeding when the slightest thing goes wrong. In fact, Steinbeck's story seems directly analgous to the problems of self-driving vehicles.

-Paul Komarek

Re:Entries too complicated?

[evilviper]

I don't think you're going to get six good laser rangefinders for $5000.

Okay, I can live with that. Let's just say I lowballed it, and move on.

And you will quickly exhaust your laptop's computational power just denoising the output from crappy sensors.

I have a hard time beliving that. With multi-multi-GHz processors available, I think it can be done. You'd only need to sample each one about 10x each second.

The road may "disappear" completly from your sensors

At which point you quickly slow-down to a crawl until your sensors are able to re-locate the road. At worst, you'd have a 5MPH accident that your vehicle would have to recover from on it's own.

Maybe your vehcicle's transmission gets a little "funny" and you can't shift properly anymore

With an automatic transmission, I can't see that as being a real problem...

a failing rangefinder

I accounted for 2 sensors on front, left, right. Perhaps we can say I lowballed that too, and can go for 3 on front, left, right, and rear. The computer can simply then go with the consensus (2 out of 3).

or sticky throttle.

Not much of an issue. If it's going too fast, the brake can always be applied. I would hope the software would have enough smarts for that. (I'm accounting for digital feedback from the car's speedometer).

I think autonomous systems might be the best example of the best laid plans of mice and men not succeeding when the slightest thing goes wrong.

I wouldn't think of deploying this on crowded city streets, but there's less destruction when you make a mistake in the middle of the desert.

Re:Entries too complicated?

[JustinXB]

Let's see. The range finders can't tell you what angle the vehicle is at, can't tell you the size of objects, or if an object is moving. Blah blah blah blah.

There are a lot of faults with your idea. But I do agree, $3 million is a lot to spend on this and the team who spent that barely has an "autonomous" vechicle since they have to pre-plan the whole course the vehicle will take.

Re:Entries too complicated?

[Paul Komarek]

You want to minimize the chance of destruction whether you're on a city street or in the desert. I doubt there's a prize for coolest disaster.

On the other stuff, I think I didn't communicate my point effectively. You can't just read sensors and rely on their data. Sensors return wrong data, even laser rangefinders. Picking out which data is good and which is bad requires more computation than just 10 serial port reads per second.

Until a person has been involved with these sorts of projects it is very hard to even imagine the scope of the problems. I don't have much robotics experience, my field is data mining and artificial intelligence. But even there the data is always screwed up. Nobody ever gives you a dataset that is complete, or training data that is correct. Much of the time it isn't clear what "correct" even means. The real world sucks big time.

And software sucks big time. I did do some robotics programming for a local company, to control elevators used by their robots. The problems involved were insane. We had troubles getting the elevator companies to give us access to the elevator call buttons (the buttons in the hallways) and the elevator direction and destination -- and I can tell you that getting all that info (which we didn't) was going to cost a lot more than $5000 per installation. We had to estimate the elevator's direction according to floor data, which isn't that hard -- until you run into the problem of elevators which don't have any floor info (some cargo elevators). Then there's the issue of communication. We couldn't run a wired network, and wireless did a poor job penetrating into some elevator cabins. That meant that the elevator doors might open on the wrong floor (due to humans pressing buttons), and we couldn't notify the robot. And we didn't have access to the "door open" button so we had to rely on holding down the floor button. There's also the matter of telling the robot that the cabin has arrived to pick it up -- we don't know when the doors open, and different elevators can take a vastly different amount of time to open the doors.
Thankfully the robot could help a bit for that particular problem.

And I haven't even begun to tell you about the robot navigation problems getting into and out of the elvator, especially for elevators that didn't line up right with the floor. And if there's a wide gap between the elevator and the floor, the rangefinders think there's a 10-storey cliff in front of it. And once you get the robot going, it's wheels might not work well crossing a two-inch gap. But you can't change the wheels because their design is critical to the navigation software.

This was what I encountered in 3 months of consulting for a robotics company whose robots drove very slowly along the walls of uncluttered hallways. All of the problems were dealt with eventually, and many of them had autonomous solutions. But now the robots have a webcam that allows humans to drive it out of bad situations that the software still can't deal with.

Did I mention that the robot tracked the wall to help its navigation? Now think about driving across a sand-covered highway through a sand-covered plain. I'm not sure what these guys are really facing, but I'm sure it's a lot harder than we can possibly imagine.

-Paul Komarek

Re:Entries too complicated?

[Textbook Error]

You say I'm wrong without any facts to back it up, and someone is supposed to believe your claim?

You say you're right without anything to back it up, and someone is supposed to believe your claim?

I don't trust this "Pentagon"

[flopsy mopsalon]

It is a well known fact that military vehicle driving is one of the few high-paying positions that less-educated individuals can qualify for.

It is equally obvious that by using this so-called contest, the Pentagon is trying to obtain for themselves a cheap automated replacement for human vehicle operators. No hazard pay, no training no insurance needed for robots. And a bargain at $1 million.

And where will that leave formerly well-paid and regarded vehicle operators? Walkng across minefields with poking sticks, that's where. I for one am shocked and appalled.

Why no news?

[SiliconEntity]

I'm disappointed that there is not more information available about this event as it happens. I've been following it vicariously for months, and now I'd like to hear about what is happening at the speedway in Fontana. How many teams showed up? How many tried out today? Which passed, which failed? I haven't been able to find out any of that information.

The so-called Science Blog article was from February 10! That's not exactly timely, is it?

Nagle's later posting here does present some information about Caltech. The Caltech team web page provides the same basic info, with a little different spin. But I guess we're lucky they posted today; the previous entry on the team's news page was dated November 16, 2003.

CMU has been updating almost every day, but their last entry was Saturday, saying "The curtain goes up Monday morning". Again, what happened?

You'd think in this age of bloggers, when every windbag on the net sees fit to tell us what he had for lunch that day, someone would be watching this event and posting some updates in the evening. If this isn't happening, I beg anyone who is attending to step up and start writing! Maybe I'm spoiled by the usual instant access to information, but I'm passionately interested in this event and starving for news.

Mark Burnett where are you?

[John Harrison]

Somebody really missed the boat here. They should have turned this into a reality TV show. Film each of the teams getting ready, introduce us to the robots, and then film the race, editing it for maximal drama.

I can't believe that somebody didn't buy this thing up. If not a broadcast network then at least the Discovery Channel (science oriented angle) or Spike TV(monster truck robots race across the desert angle).

Somehow I don't think that the military feed is going to reach a wide audience. I won't be able to see it.

Day 1 Field Report

[EvilXenu]

Although the kick-off isn't until March 13th, the days leading up to it include Qualifying and Demoing time. Below is a friend's field report after Day 1, posted with his permission:

Total attendance was probably arround 350-450 people. I think there were about 50 people that attended that were not associated in any way to the race.

Vehicle inspections were performed on several vehicles the morning of the first day. The inspectors were verifying functionality of safety devices including the e-stop buttons, remote e-stop, beacon, and the alarm.

The qualifying and demonstration (Q&D) of several vehicles was done in the near 90 degree heat of the afternoon. The test course was composed of an approximately 1/4 mile serpentine track outlined by short penguines (orange cones). It consisted of about 50 GPS waypoints Obstacles included a sand pit, two gate openings, and two abandoned cars.

The biggest event of the day was the events that didn't happen.

I don't believe the first team, Team Phantasm appeared at the test track for Q&D. If they did appear at the start line of the Q&D course, they did not move.

The second team for Q&D was the Lousiana State CajunBot. The vehicle moved about 1 foot and stopped short of hitting the elevated start line ribbon. According to a CajunBot team member, a DARPA representitive was confused by the vehicle dwelling for 30 seconds before starting and erroneously pressed the remote e-stop while the vehicle was still behind the start line.

The ASI/Florida state vehicle did not move at all.

Team Caltech had the first vehicle to pass the start line. It completed about 3/4 of the course. It was out of view when it stopped moving and I don't know if it was e-stopped or if it malfunctioned and stopped on it's own.

I don't believe the A.I. Motorvators team appeared on the track.

Team D.A.D did not move at all

I don't recall seeing the Golem Group appear at the track; however, I was loosing interest by this time.

The Palos Verdes High School team had a brand new, well polished $40,000 Acura SUV. It would have won 1st prize in a beauty contest. The vehicle passed the start line and immediately went off course and killed two penguins. It drove parallel to a concrete barrier, dwelled for about a minute, turned it's wheels towards the barrier and did a face plant. That was an impressive demonstration from a high school team.

ASI/Florida State was given a second chance to qualify. The second verse was the same as the first... It just sat there. Bummer!

Tomorrow I'll don some SPF 50 to prevent any more sun burning, attempt to acquire either a pit or press pass so I can take note of the technology visible on each vehicle, and hope that at least one team is able to qualify.