DARPA Grand Challenge



Hi Bob, you might be able to go back about a year and a half on this news group and pick up lots of posts on it.
Basically, DARPA expected a few dozen teams. They got around 400. So they made an arbitrary announcement they were sending out inspectors, and they wanted to see all the contestants plans and prototypes, and they'd decide who was worthy and who was culled. The culled were not actually allowed to compete, even though they were still just part way into the design stage and months before the contest.
I've heard, many of those inspectors now have been hired in better paying jobs in the industrial military complex, by the way. But this is hearsay, and I have no first hand knowledge of such practices. However, this isn't Denmark, but I smell something familiar.
--
Randy M. Dumse

Caution: Objects in mirror are more confused than they appear.
  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Randy M. Dumse wrote:

I didn't see that, and I headed one of the losing teams.
This year, you had to submit a video. Those who looked like they had a clue in the video had a site visit. At the site visit, you had to run a very simple obstacle course. That eliminated those who didn't have anything working. Then came the NQE, and obstacle courses. Everybody who had a clean run at the NQE went to Primm.
What's the problem?
                John Nagle                 Team Overbot
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Sun, Oct 16, 2005 at 11:43:38AM -0500, Randy M. Dumse wrote:

In the post I responded to you complained about what was sacrificed to put on the DGC - that the money and resources should have been devoted to other areas.
Hate to break it to you, but people are spending money right now this very instant on things that you probably don't agree with, and that money could be spent on your pet projects. It's happening all the time, at far greater tax payer expense than this little DGC project.

The government offers prize money all the time - for turning in criminals because it's for the greater good. One can argue that developing robot technology is for the greater good. Would you protest if the government offered a prize for a cure for cancer?

You didn't mention this in the post I responded to. Your arguments were economic based.

I don't know about civic duty, but I certainly would not knowingly participate in any event where anyone was excluded based on race, religion, etc. That would just be the proper thing to do.
But your phrasing the hypothetical that way implies that DARPA did that. Being an actual participant in the event, I will tell you that from my experience and observations, DARPA bent over backwards to accomodate all teams. You've already said that your hypothetical is hearsay. Why don't you have the person(s) affected come forward and file a lawsuit against DARPA for discrimination - if there is a case it should be heard. Wouldn't that be your civic duty to bring this forward so that those who did it can be held responsible and not continue to do this to people?
Regarding the process of who participated, and who didn't, it went like this. First you had to do the paperwork to file your intentions to participate - 195 teams did that. Then you had to submit a video that showed specific features of your vehicle that DARPA asked to see - if you couldn't do that, you didn't get a site visit. If you did, you got a site visit - 118 teams did that. DARPA actually sent out representatives to 118 teams! Ours was very short and to the point - all teams had run a 200 meter course, with at least 2 30+ degree turns and avoid two obstacles (30 gallon trach cans). You got three chances to run it. We did it all three times, we cleared all gates perfectly and cleared 5 out of the 6 obstacles, each in under a minute. Based on site visit performance, DARPA selected 40 teams.
Some teams were on the edge and DARPA even gave them a second chance - a second site visit. Of those, 3 teams were added so there you have 43 teams.
Those 43 teams showed up at the NQE. Before you were ever even allowed on the NQE course, you had to pass a "static" inspection - vehicles must conform to DARPA specs with regard to caution light, audible alarm, brake lights, etc, etc. Then you had to pass a "dynamic" test - a simple 200 ft straight line run using a couple waypoints - the DARPA operator shuts down your vehicle mid-way and verifies your e-stops.
Some teams at the NQE did extremely well - Stanford, CMU, Cornell, VA Tech, etc, etc. Other teams did not even get through the first gate. At least one team was several days into the event before they could even pass their initial "dynamic" test.
DARPA gave the teams that were struggling _every_ opportunity to run the course, even to the point of creating an alternate course for those teams on the next to the last day so that they could at least complete _something_.
We ran the course 5 times. On our first 2 runs we had PC power supply going bad which caused our main sensor computer to go down both runs. It took a while to figure out because the power supply worked intermittantly and we could never seem to catch it while it was acting up. But once we discovered that problem, we sent someone out to a local CompUSA to get a replacement power supply and then we ran the course 3 more times with one of those a near perfect run clearing 49 out of 50 gates and avoiding 5 out of 5 ostacles. On our other runs we cleared in the high 40's out of 50 gates and avoided 4 out of 5 obstacles. Those three runs were plenty good to get us into the top 20.
So from what I saw, as a member of a team who was there from the beginning to the end, I didn't see the things you spoke of above. The process was elimination by competition. The accusations you've made are very serious and you really should be very sure of your facts before accusing someone or an organization of such a thing. Like I mentioned above, if you have the facts, I think it is _your_ civic duty to come forward and expose them. But USENET is probably not the optimal place for that.
-Brian
--
Brian Dean
ATmega128 based MAVRIC controllers
  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

You too.
http://www.theregister.co.uk/2003/11/06/darpas_indecision_threatens_integrity /
So you never heard about any of this when it was happening?
--
Randy M. Dumse

Caution: Objects in mirror are more confused than they appear.
  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Sun, Oct 16, 2005 at 06:15:31PM -0500, Randy M. Dumse wrote:

I don't recall accusing anyone of anything, not in this thread anyway.

Is that the worst you can find? Seriously, an article in the esteemed "The Register" from 2 years ago? Boy, that really does rip into DARPA doesn't it?
Instead of some deep and intricate covered-up plot by DARPA to confound and confuse potential participants with their evil plan to "exclude them from the event", I find it much more likely that DARPA under estimated the number of teams wanting to participate and had to trim the field to what they thought they could logistically handle and they chose what they thought were the most likely candidates to actually have vehicles ready for the event. Never attribute to malice what can be adequately explained by incompetance ... or something like that.
I do remember some of the complaints about the (not unexpected) rule change that limited the number of entrants for DGC version 1. But in trimming down the field of entrants, DARPA did offer a site visit even back then to see what teams had before doing so - that was what, 5 months before the event? If you didn't have working technology to demonstrate by then, a mere few months before the event, you had next to no chance of being ready by the time of the event anyway. So if you were cut because you had nothing to show at the offered site visit, what's the complaint? And you're saying it was discrimination? I think your argument is on life support.
Lots of teams that participated in DGC 1 or wanted to participate but were not selected were back for DGC 2. So everyone got their chance this time around, with plenty of forewarning, and folks don't have much to complain about if they were even remotely serious about competing in the first place. DARPA planned better and the selection process was better for DGC version 2.
So I don't think there is an evil conspiracy here, in DGC version 1 nor in version 2.
-Brian
--
Brian Dean
ATmega128 based MAVRIC controllers
  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Yes, Brian. That's the worst. Scoffing at it doesn't make it untrue. Vilifying the source reporting it there doesn't mean there weren't others. Violation of fair play and "the American way" means more to some than others. Cutting off people told they had (what was it?) 18 months when they were less than 18 months into a preparation, eliminating them by capricious, arbitrary standards not related to competition, even requiring the submission of white papers and submit to a "search" of their facilities as a condition to entry, based on having listened to many Supreme Court rulings to the best of my understanding of Constitutional law were illegal. Who knows what breakthroughs might have happened in the final days before performance? Was anything modified on your entry in the last five months? Eliminating anyone before the contest, from the competition, which offered public funds for demonstrating performance of a task at the contest, was an act of discrimination. I'm sorry that is a hard concept. But ridiculing it does not change its veracity.
Now, since I can't stand your condescending attitude, about something I take seriously, without feeling the strong need to reply in kind, I think instead I'll stop discussing this with you, with my closing comment: People often get the government they deserve, and that should scare the devil out of most of us.
--
Randy M. Dumse

Caution: Objects in mirror are more confused than they appear.
  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Sun, Oct 16, 2005 at 11:57:44PM -0500, Randy M. Dumse wrote:

My response was not intended to be condescending. I believe that you believe what you believe and I respect your right to believe that. That doesn't mean I agree with it. I'm just offering a different explanation of the data available.

Villify me if you must, but I just don't see any foul play here. That doesn't mean there aren't other instances of aggregious misconduct. But I don't see it here. One should choose one's battles carefully, otherwise you risk becoming the boy who cried wolf.
-Brian
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Brian Dean wrote:
.........

Hi Brian. Can you expand a little on what kind of obstacles there were? Maybe mention what sensors were used to detect them.
thanks, - dan michaels www.oricomtech.com ===================
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Sun, Oct 16, 2005 at 07:32:50PM -0700, dan michaels wrote:

DARPA presents the course to teams in RDDF form - an ASCII text file containing lat/lon of each waypoint, segment width defining the boundary between two waypoints, and a suggested speed for the segment (not to be exceeded).
The first RDDF was 2.2 miles and consisted of passing between several gates to a "high speed" section which whose speed was 40 MPH. That fed into a hay bail funnel. While the hay bails were not considered obstacles, they gave most of the teams their problems, oddly enough.
The hay bail funnel led into a long tunnel lined with metal to block GPS signals. After the tunnel, the course wound near a stacked pile of tires and to the first official obstacle which was a parked "rental" car as the announcer described it. Vehicles had to avoid the card and move on to the "rough terrain" section which was composed of 4x4 and 2x4 posts and car tires lying flat on the ground with hay bails lining the sides.
After that, vehicles had to navigate a simulated "mountain pass" where the course hugged up tight agains a long curved row of k-rail barriers. Following that was a few twisty turns and then into the home stretch which had two more "rental" cars on it that needed to be avoided. The last section of the course had some rumble strips (speed bumps) immediately followed by a tank trap that also needed to be avoided.
The official obstacles were the rental cars and the tank trap - 4 for RDDF 1, though the hay bails, tunnel, k-rails were hit a lot but did not count as obstacles.
On RDDF 2, which I beleive was 2.7 miles and the first one we successfully navigated, I think that one included a new "rental" car and re-arranged the order of some of the sections. And instead of the high speed section which was on pavement, it included a parallel section but on grass instead of pavement and the speed was lower and much rougher. That RDDF had 5 obstacles instead of 4.
Similarly with RDDF 3 - I can't remember the specifics - all the courses were similar, just hooked up the sections in different orders and the inclusion of the van as a 5th obstacle. We actually hit the van, twice I think. First time we barely scraped by it - we actually avoided it OK, but turned back into the centerline too early and scraped down the left side of our truck. The second time we hit, we were steering around it, but didn't steer quite hard enough. Had we been 4 inches to the right, we would have missed it. As it was, we tagged it not too hard, our bumper actuated, detected the impact, we shifted to reverse, backed up a few meters and reoriented, then steered properly around the van. We dinged up the front of the Suburban slightly - nothing the grinder couldn't fix. We finished out that run clearing 4 out of 5 obstacles and a good score.
The van was in a difficult position for us - the course had you coming around a turn and there the van was before you got straightened back out and made the geometry a little weird, at least for us and our algorithms.
The hardest obstacle I thought was the tank trap. Our LIDAR intersected it at the criss-cross level and made the obstacle appear smaller than it really is, thus we tend to understeer when avoiding it.
Here you can see some low quality video of one of our runs here (taken with a digital still camera with MPG capability):
http://www.insightracing.org/NQE-R2R2-A.mov
This has us entering the hay bail funnel before the tunnel, going through the tunnel, around the first parked rental car, on to the rough terrain section, down the long striaght and turning toward the van and clearing it this run (too far away to see clearly, though), then back around and finally to the final stretch where we clear the remaining obstacles and cross the finish. The mountain pass section is not shown because you couldn't see it very well from where the video was shot.
Our main obstacle sensor is a SICK LMS-291 LIDAR mounted at front bumper level. It works very well, has a range of about 40 meters. I wrote the obstacle avoidance code - it works pretty well most of the time, but is not perfect. The LIDAR can be dazzled by bright sunlight resulting in blind spots, but other than that, it is a fantastic sensor.
We were highly constrained by our budget so we really couldn't afford more LIDARs which are about $6000 each. SICK sold them to GC teams for $3000 each so at least we did not have to pay full price. We do have a second one mounted on top to sense the terrain profile in front of the truck, but we had some problems with it during the NQE and it was not running during these runs, so the front LIDAR was the only obstacle sensor we had. Had we had more budget, we probably would have had several mounted on top each at a slightly different angle to get a better terrain and obstacle picture. We saw many teams that used that configuration.
-Brian
--
Brian Dean
ATmega128 based MAVRIC controllers
  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Brian Dean wrote:

Hey Brian. Thanks for all the info regarding the obstacles. Looks like your vehicles needed to be working well just to get through that. Good job, for going 26 miles too. Oops, my little sumo just ran into the wall - blind as a bat :O).
BTW, what do you think might work for a moose detector? Especially, for a moving moose.
- dan ===========

Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Mon, Oct 17, 2005 at 08:33:21PM -0700, dan michaels wrote:

Thanks!
That's a tough one. I thought you were joking when you first posted that, but the more I think about it, I think it would be a great idea. I was driving home at that twilight time in the evening the other day when deer are rustling about and was intensely alert, not that my reaction time still would have been fast enough to avoid one since they come out of nowhere like a bullet. What a great idea to have a sensor to give you some forewarning.
What about an infrared camera - look for its body heat which should be substantial. CMUCam/AVRcam for IR. One should be able to differentiate a moose or other animal (deer, cow, person) from other heat sources like oncoming cars due to temp difference between animals and a car engine. Also, size of the animal could be estimated as well - do you care about squirrels? If the IR cameras see warm bodies nearby, sound the alert.
-Brian
--
Brian Dean
ATmega128 based MAVRIC controllers
  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Brian Dean wrote:

Moose and dear cause tons of trouble.

I had one miss my truck by inches, almost twilight, into the setting sun, Probably would have totaled the van, as I was doing 70 or so. I'd pay serious money and I think others would as well.

Don't care at all about anything house cat sized on down.
How do you tell a car far away from a deer/cow up close? Do you have range as well? or triangulate from two or more sensors?
Maybe combine IR with motion and/or visual?
--
Pat



Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Tue, Oct 18, 2005 at 01:01:12AM -0400, Pat Farrell wrote:

I was going on the idea that the IR sensor can actually sense temperature. A car engine should be hotter than an animal's body temp and it can thus be filtered out. Only look for things in the temp range of a large animal.

I'm envisioning one of those infrared cameras that detect body heat. Used by law enforcement to find people at night. Or do those just exist in Hollywood?
A signal image wouldn't provide depth, I think you'd need stereo for that. Given the infrared image(s), standard video processing techniques should work to do the rest - find the moose, sound the alarm.
-Brian
--
Brian Dean
ATmega128 based MAVRIC controllers
  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Brian Dean wrote:

They exist, more or less; firefighters use similar equipment to evaluate buildings for residual heat sources. Unfortunately, they tend to be expensive. The low energies involved in thermal radiation require high sensitivity for measurement, and the receiving element must be shielded from background radiation (often by actively cooling the camera)
- Daniel
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
D Herring wrote:

Hmmm, the obvious place to put sensors is on the front of the car. Right near the screaming hot iron engine, radiators, A/C condensers, etc.'
--
Pat



Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Brian Dean wrote:

Actually, I was being semi-serious. As noted in the original post, there are 100s of 1000s of deer collisions every year, so just the cost of the auto repairs, and possible personal injuries, can be high. And I read that in Maine, approx 1 moose is hit every day. How many people even live in Maine? 6, maybe ;-). A 1500# moose will put a dent in the hood.
Down southwest of Denver, they opened up a new ring corridor 10 years or so ago, and there were deer collisions every single day, until they put up about 15 miles of tall fences the deer couldn't jump over on both sides of the road.
The moose/deer sensor would have to:
(a) detect the animal, and maybe note its size (b) track if it's moving (c) compute its speed (d) predict whether it will intercept the path of vehicle
A good engineering problem.
- dan michaels www.oricomtech.com ===================

Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
"Brian Dean" wrote

Listening to an interview of Montmerlo (spelling?), the team leader of Stanford team, he said it was quite amusing to see Stanley trying to swerve when a bird crossed flew a couple meters in front of Stanley bumper.
Padu
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
"dan michaels" wrote in message <snip>

Nothing to do with robotics, but two days ago at WalMart I saw this "Deer avoidance" device.
It is a set of two little devices you install on your front bumper and they are "wind activated". From what I saw it is nothing more than a whistle set to a certain frequency that must bother that kind of animal. How effective it is I really don't know.
Padu
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Brian Dean wrote:

So just what were some of the approaches, and what accounts for the big improvement ? Is there real ooh-la-la AI involved here ? Or just sophisticated ad hoc adaptations ?
Congratulations to you and everyone involved, BTW. I can hardly imagine how one would go about doing this, which is why I'm so curious.
I read an account of one team which qualified ( Syracuse ? ) and they just talked about computers catching fire and vehicle problems.
Lew Mammel, Jr.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Randy M. Dumse wrote:

Did you see the Blue Team's self-balancing motorcycle? The three-unit snake-like articulated vehicle? There were a few unusual entries that worked.
There was a straightforward qualification process. Your vehicle had to actually work. First, you sent in a video of your vehicle working. Everybody who had some indication of having something working got a site visit.
At the site visit, two DARPA inspectors watched your vehicle run through a very simple course they'd specified. You had to be able to avoid two trash cans. That brought it down to 43 entries for the National Qualification Event.
At the NQE, at the California Motor Speedway, each team got five tries at a 2-mile obstacle course. Everybody who had one clean run went to Primm for the main event. That brought it down to 23 teams.
Of those 23 teams, five finished the 132 miles.
Of those five teams, Stanford had the best time.
I'm not complaining, and I was the team leader of one of the losing teams. The elimination process was tough, but fair.
While one can argue about the funding of the CMU and Mitre teams, the Stanford team was funded by Volkswagen and by Mohr Davidson Ventures. Not tax money.

It sure beats dealing with NASA, which sucks robotics researchers into decade-long projects that go nowhere.
                    John Nagle                     Team Overbot
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Polytechforum.com is a website by engineers for engineers. It is not affiliated with any of manufacturers or vendors discussed here. All logos and trade names are the property of their respective owners.