May 4, 2007

Steve Hassenplug's FLL Robot

400 points, 55 seconds. 'Nough said?

Steve Hassenplug has been coaching FLL teams for a long time, and always likes building an FLL robot of his own (hey, you can't expect a guy like Steve to just watch everyone else do a challenge...). Well, he's done a wonderful job of both building a Nano-challenge robot, as well as describing the how and why of the design. If you like videos of a good FLL run, take a look at this:



Steve has another video up on his webpage for FLL 2006 that I'd also recommend. But for any of you really interested in how folks think about these challenges, Steve's FLL NXTlog post, with a complete description of the robot & strategy, is really enlightening. This is a great example of a robot built to a specific goal, with a careful analysis of just what that goal is.

--
Brian Davis

15 comments:

Jonathan Daudelin said...

... Wow... that is absolutely amazing! How consistent is it?

-Jonathan

Steve said...

There are two things that are not 100%. The dirt doesn't always go down into the dirt trap, so it can lose 2 (or 4) points there. And, occasionally, it won't get three atoms off the table.

But, it very consistantly gets a score of at least 388.

Steve

David Levy said...

Steve ,
That's a great jig on the dirt trap delivery! It looks like it's attached somehow and releases when the dirt is dumped. It also looks like there are no moveable parts and that there is some sort of hook that fits around the dump lever on the way in.

I've seen a couple of other vehicles with front fender assemblies that force the vehicle to be perfectly aligned to the field model simply by crashing into it. ( question: Is the vehicle grabbing the model on the molecular motor as well ? )

Now I am very inspired to develop curriculum in this area. Last season I spent a lot of time covering jig creation for alignment from home base. Your technique and the others I've seen should make a nice addition to those lessons.

On the "individual atoms" ... You should get a bonus for knocking them clear off the table ;)

Thanks for a great video!

- David

Steve said...

Once I realized I could use the wall and the fabric tester to align the robot, it became much easier to place the dirt trap in the correct location (100% of the time)

The dirt trap attachment is NOT attached to the robot. The robot just pushes it out. Then the robot backs away and releases the lever, causing the spring-loaded dumper to activate.

David Levy said...

"spring loaded dumper"?
I thought one had to physically reach around the back of the model to pull the dumper. It's not clear from the video how it is being activated. That's way I thought the jig was attached and hooking around the dumper arm.

Steve said...

Check out the dumper pictures on NXTLog. Maybe that will explain it better.

Steve

Maniac of Mayhem said...

I looked at the program and didn't see any blocks indicating the use of sensors...


...the consistency of the robot pretty much smashed my beliefs about dead reckoning:)

Dean Hystad said...

Thanks for setting a very bad example Steve. We're trying to get the kids not to depend on odometry and aiming. Remember?

Luffe said...

Great work Steve!

Brian Davis said...

Dean, while I agree that teaching how to use sensors etc. is very important... it may not be for the missions. If FLL really wants to students to use sensors, then the FLL challenges should force the use of such. These don't... as this example has just beautifully shown.

Take-home here? Not that sensors are worthless or dead-reckoning is the way to go. This teaches a very good lesson IMO: accomplish the task at hand with the simplest feasible approach, but no simplier.

Dean Hystad said...

Excuse me, my tongue appears to be stuck in my cheek here. Ahhh, that's better.

Yes Brian, the challenge this year didn't do much to promote sensor use. Lots of lines, but not many in any useful places. And none of the challenges were, well, very challenging. So even a very simple robot like Steve's could score lots of points. There goes that blasted tongue in the cheek again!

But seriously, a really nice job. The stain resistant fabric solution is inspired. My team had a tough time with that one because the dirt trap and their arm for tripping the stain release threw off the robot CG and messed up odometry. They used a Y shaped guide that correced for small errors, but your wall following wedge is a superior solution. All you need to add is a chute to direct the dirt to the trap so it can't get caught up in the fabric. I was impressed and delighted when I saw a team come up with that solution. And then I saw it again at the world festival (in a video, alas I was not invited to the ball). But it does kind of defeat the purpose of testing the stain resistant fabric if you don't allow the dirt to touch the fabric.

I am suprised you don't have problems with knocking red atoms of the table. Our solutions are similar and my robot knocks a red atom off 1 in 4 tries. I kept an eye open for the Eureka solution to that mission, but have yet to see one.

Steve said...

Dean,

I read your first comment as you intended. :)

I played around with a couple ideas for the fabric tester, and when I came up with the "wedge" idea, I knew it was the way to go. If I make any changes to the robot, I will be adding a ramp, so the dirt will not make contact with the fabric, on the "Fabric Tester" mission... !?!

Unfortunately, I don't think it's possible (in the FLL format) to create a challenge that requires the use of sensors.

Dean Hystad said...

How about a randomly placed obstacle. The City Sights challenge had a random component in which one of the bus flags you needed to raise. Many teams solved this by programming three different flag missions and chosing the correct mission at the table (sometimes using a touch sensor as a click counter). A few teams used a light sensor to find the correct flag(it was different color).

But a random obstacle would be very difficult to pre-program around because it could affect your solution to almost every mission.

Another possibility is to place a terrain obstacle between the base and one of the finesse missions. Terrain obstacles have been done in the past with bridges and snow fields. The terrain throws off odometry and you are required to establish your robot's position again. You could always do the square against the wall thing, but that would lose appeal if a conveniently placed visual marker were available.

FLL should have us coaches help design the challenge. We are used to thinking up fiendishly difficult problems for our teams to solve, and aware of the deviously clever solutions they will try. It would be so much fun if I could torment 6000 teams instead of just one.

Robolab 2.9 said...

Very amazing! Great job Steve!

Dean,

Just so you know, it was No Limits that the random flag, not City Sights. We solved the problem by using the light sensor to find it.

This is my own thinking, but I think this year was out of the norm easy was because that FLL was expecting a lot of new teams because of the launch of the NXT, and they wanted it a little easier for them. I guess they forgot about us more expierenced teams.

Again, great job Steve. I look forward to seeing more from you.

Robolab 2.9

Brian Davis said...

Dean, while you remove your tongue from your cheek, I'll admit to take my very large foot out of my oversized mouth. Sorry 'bout that, I was clearly not well-enough rested to filter for irony that evening.

As to challeges that require (or at least encourage) sensor usage, I'd agree - movable or unpredictable locations, or actions, should be relatively easy to do. For instance:

1) And obstacle that is in a new location each run
2) A ball that must be dropped in the proper box, with which box should be selected determined by a giant on-the-playfield slider that is set just *after* the start of the round.
3) A sensor conditional: you can only grab the ball if the colored spot under the ball is red, not blue, or an assembly you need to set or trigger only if it is not already set or triggered (with the state either determined in some way the team can not see, or better yet by the moment-by-moment performance on the neighboring playfield.

--
Brian Davis

Related Posts Plugin for WordPress, Blogger...