best technical feature at GLR
 

imho:

Wildstang's guidance system. I watched a match where in autonomous mode the robot hit a wall that it wasn't supposed to hit, corrected itself, climbed the ramp, knocked over some bins, and got into a position where it could defend a human player stack. Three things were obvious to me:

1. This robot knew where it was
2. This robot knew where it wanted to go
3. This robot had the means to get there

wow

Ken

yes, i agree.

wildstang's robot = wow.

Re: best technical feature at GLR
 

Does anyone have video of this match? i'd LOVE to see it

Tom

StangPS
 

Thanks for the compliment and thanks for noticing! I'm so glad that someone not on our team realized what happened during that match. The students and adults that worked on our positioning system (called "Stang Positioning System" aka "StangPS") thought that was great but we figured that no one else would appreciate it.

Your three observations above are exactly correct.
1)The robot knows where it is by using our custom circuit (with a Motorola HC08 processor) that reads a homemade encoder wheel, the gyroscope and the wheel direction pot. The pot tells the custom circuit where the "crab" wheels are pointed relative to the frame of the robot. The gyro is integrated to tell us exactly where the robot is pointing on the field (like a compass). With that information, as the wheel moves, we can map our current location to an X,Y coordinate on the field. So, the custom circuit always knows where the robot is on the field (within about an inch...although position is reported to the RC in 2" increments).

2/3)The robot location is passed from the custom circuit to the RC, which is programmed up with a list of X,Y "waypoints" on the field. Each loop, the RC uses trigonometry (TAN) to calculate the angle the that the "crab" wheels need to be pointed so they are always aimed at the next target waypoint (what a great real world application of trigonometry!) Anyway, the robot can be pushed off course or twisted but we will continue to try and get to our next waypoint.

We also have some features programmed into the RC where we will determine the most efficient direction to aim the drive wheels to get to our next waypoint. I can go into details, but the idea is we don't program in how to drive the wheels, we program in waypoints and the RC figures out the most efficient way to get to that waypoint based on its current location on the field.

One other nice feature is that the robot is constantly correcting its angle (or compass reading) on the field so that it will hit the bins exactly square with the wings every time. The match you were talking about, we ended up getting turned so much by the wall that we couldn't correct fast enough, but in a normal run we will hit the bins perfectly straight every time. Even if we are hit and twisted, or a bin is placed in the way we will constantly correct.

There were numerous WildStang members at GLR that were taking video of the matches for soap108.com, so I'll make sure that someone posts a link to the video of this match.

I'll also have someone post screen shots of the computer program we have that allows you to click on points on the field and this program will output the waypoints that are programmed into the RC. This allows us to write new programs or make modification while at a competition.

We have a few students that wrote the majority of the software logic in the RC (with help from the engineers) to implement this autonomous operation and I am sure they would be happy to explain it in more detail either at the Midwest Regional or on this thread so feel free to post any questions you might have!

PS: Here is why we hit the ramp wall in the first place...Just prior to that match we had a problem with the chain on our wings and we were working like mad to fix it prior to the match. We were in the pits and actually saw the video feed of our match being setup on the field. We ran down the hallway and onto the field. In the rush, we didn't setup our wheels to be aimed in the correct direction. Since our wheels only have 360 degrees of rotation (not continuous turn) the RC ended up calculating that our only way to get the top of the ramp was to reverse the direction of the drive wheels (instead of the usually smooth arching turn we usually use) and this caused us to hit the ramp wall...for a bit!

Truely Awesome
 

Wow, that is really impressive. My team currently wants the same, rather, out mentor does. And as the team programmer, It hasn't been easy so far. I hope to have something only minutly close to what you have by houston, but wow! Good job. My hat off to you guys!

stangPS sounds nuts, I can't wait to see it at nats.
how fast do ya'll ussually get up the ramp?

Wildstang was nuts out there, especially how it showed up at the end of autonomous ready to defend the stack. Wow, what a pain it was to try and get around them too. We ended up stuck on them for almost the entire match (Never thought of the robot going up that steep of an incline, the back bottomed out)

I think the 'only' logical way to beat Wildstang was to double team them from two angles, since they are so mobile with the steering, then like Truck Town in the finals, lock them in between the ramps.

I would definately like to know more about this positioning system, maybe after nationals you guys could explain it a little more in detail to all of us. We partnered with you guys a few times, and even though Raul said it would eventually find its way, I never realized it had that much intelligence. Every year I am amazed at the quality and content in the 111 robots. The machining and build is top notch, fantastic robot Wildstang.

Thanks for all the compliments, everyone! I think Steve did a great job in the overall description of StangPS. We will definitely welcome everyone to come to our pits at Midwest and Nationals for a demonstration. It's been quite a rewarding accomplishment, and there's still more to come.

In regards to RogerR's question about how fast we can get to the top of the ramp, I think I can safely say that we can do it right around 4 seconds, give or take. After Great Lakes, we saw some great machines race to the top, and I think we are right around those times. Although, Baxter Bomb Squad (16), who we saw in St. Louis, is very fast!! I fear (and look forward) to competing against them!

We need to talk. I have been wondering about attempting the same thing on a few robotics projects of mine and It could help if I talk to you.

Re: StangPS
 

Based on what you said above, if your robot is up against a hard barrier and is spinning it's wheels, will it throw off all of your calculations because your encoders would be still spinning? Or is your encoder non-powered so that it just follows the terrain and doesn't spin when against a barrier. I would definately like to know more about this 'Wild' system.

You are correct that "if" the wheels started spinning it would throw off our position calculations, but the wheels have a gummy belting on them that doesn't slip easily. When we run up against a wall on the carpet, we end up stalling the motors.

Early in design, stalling the motors became a major concern for us in autonomous mode so we also created a board that is capable of measuring current draw on our motors so we could detect a stall condition. We shipped the board with our robot but we haven't spent any time actually integrating it into the StangPS system. Our current monitoring is almost exactly what we used last year ("StangSense") to report motor current back to a Palm that was connected to the OI. We hope to work on integrating those two pieces together and have it working before Houston, but we haven't found that it as necessary as we expected during early design since we haven't run into a stall condition for more than a few seconds.

BTW...your robot was my (and my dad's) absolute favorite at the GLR hands down. I was very disappointed when you didn't make it further. I just wish I made is over to your pit to check it out in person. Maybe I'll make it to Houston and check you out there!!

Thanks for the inspiration to create something as awesome as this next year!

Good Luck!!!

This is a system that is very similar to what we have. We have what is called INS, Inertial Navigation System. Not being a part of the electrical design team, I do not know the fine details of it all, but I can give a brief overview, and a detailed whitepaper is in the works.

We spent very much time debating how we should approach the autonomous time. After some debate, it was determined that dead reckoning and line tracking would not be very accurate or very quick, respectively. An encoding wheel was discussed, but again... if we are up against a solid stop and the wheels spin on the carpet, our calculated position is no longer the same as our actual position.

Soooo, using a gyro chip and dual-axis accelerometer, we are able to determine our exact position on the field. Our gyro operates in very much the same way as 111's, securing us with "heading hold" to ensure that we are perpendicular (or parallel depending on how you look at it) to the ramp, and if we get spun around will take the most efficient wheel path to correct itself. The dual-axis accelerometer reads the X-Y acceleration of the robot on the field. We take the double integral of these values to read X-Y position. By converting these "INS units" to real-life units, we are able to tell the robot where to move on the field, using knobs on our human control station; we can chose X and Y position -- the ability to choose which side of the ramp to go up, and how far to go along the ramp. We also account for the angle of going up the ramp, because our acceleration is now being miscalculated because of the 32 ft/s downward acceleration which now becomes a component of the X-Y coordinates. Possibly under development is way to detect if we are flipped during autonomous which would allow us to self-right ourselves autonomously!

I am not too sure of WildStang's drive system, but it too seems as if it may be similar to ours. Our robot name this year is "King CrabGoat 2k3." Perpendicular to our main drive wheels, we have a fully redundant driveline that is known as the "crab system." These 4 kicker wheels drop down in 0.2 seconds and lift the entire robot with an over-center mechanism, so we are able to almost instaneously change our direction of movement 90 degrees.

Up to this point, I had not heard of any other team using an actual "guidance system" such as we have done. I give congrats out to WildStang, and would definitely love to see it in action.

Well,
Wildstand USED StangPS this past weekend. I was under the impression that the CragGoat still utilized a simple dead-reckoning based code in competition up until this point.

Maybe by Seattle you'll have your INS working?

Great Job Stang!
It's amazing to see your system really working! It's amazing to think that even if your robot hits a wall, it is "smart" enough to correct itself.
Do I "sense" a technical award coming your way?

:D

Thanks John (and everyone else) for the kind & encouraging words. It was exciting to see the robot perform in competition (well, I actually came back home on Friday morning so I only saw it in practice) and hear what others had to say about it. The software/firmware students & engineers have been working on StangPS almost daily since mid-January and we're happy we finally saw it in action somewhere other than our practice field. All the hard work was certainly worth the results.

I don't know the full story about the judges & awards at GLR since I wasn't there, but the story I heard is that the judge who talked to our students didn't want to hear about our custom circuit or autonomous program because he was a mechanical engineer. Oh well, maybe we'll get an EE or CompE to talk to us next time. I just want our students to get the chance to describe to the judges the system we're all so proud of :)

Mike

Similar Systems
 

I always hate tooting my own horn, but I have to on this one.

We have a system that is very similar to WildStang's. We implemented our's slightly differently due to the lack of crab steering on our robot, though.

Unfortunately, we didn't work as well as we could have at the GLR since we had some mechanical issues in the system (our measurement wheel was losing contact with the ground), some electrical issues (an ESD filter caused us to miss some wheel rotations), and some software issues (a systems bug caused us to shift gears at the wrong time in our first 5 matches).

As is typical with our team, we finally got it together just before pit close on Friday night (check out our team record at our 1st regional of every year and you will see that this is what we always do - but we always get much better VERY quickly). Our autonomous mode ran nearly perfectly in our ONLY match on Saturday, which was just too late to matter (esp. considering the match was the 2nd to last match so probably no one was watching).

Anyway, we do a couple of other cool things that I don't even think WildStang does, which we haven't shown yet. I would like to invite everyone at the West Michigan Regional (we're going now!) to come check them out. We know we'll be using them there.

Sorry for the shameless plug (not really),

Chris

I have posted a zip file containing screen shots of WildDraw in our public forums at http://www2.wildstang.org/phpBB/view...hp?p=4050#4050

WildDraw allows us to generate waypoint lists that are downloaded into the robot controller. We can accurately make additions and modifications to our existing lists with a fiew mouse clicks.

WildDraw is a constantly evolving application, so stop by at the Midwest Regional or Houston and we'll be happy to give a demo.

Very Cool
 

Dave,

That's a great system you guys have. We don't do a waypoint map. Since we don't have crab steering, the way we did it seems to make more sense for us.

We did all of our autonomous controller in Matlab/Simulink and it was auto-coded using TargetLink from dSPACE. We developed a mathmatical robot model so that we can simulate the robot path in Simulink before we upload any desired paths to the controller. By doing it this way, we actually have 128 different paths in our controller and we have room for up to 123 more. It usually takes us about 5 minutes to create a path, test it in simulation, and then upload it to the robot controller.

After the simulation is done, we can plot all of the results just like in the attached plot.

I wish we didn't have all of the previously mentioned problems on Thursday and Friday, but we should be ready to rock once we get to West Michigan. I hope people can come check it out.


-Chris

Re: Re: StangPS
 

Matt,
Sorry I missed talking to you between all the times we were in each other's pit. We are using the same system many teams were using for counting rotations. Two light sensors aimed at alternating reflective/non-reflective spots on the wheel. For those not familiar with the system, two encoders placed such that the outputs are in quadrature (i.e. 90 degrees apart) will allow both direction and tach to be determined. Our sensors do have some limitations but the errors are minimal for the required 15 seconds of autonomous mode. In looking at other teams, we are not entirely unique in looking at the problem or solving it. It seems that at least 20 teams at GLR had a good way to determine length traveled and direction. Many will tell you that non-stacking robots were the rule, but those that did stack were engineering marvels and true crowd pleasers. No doubt about it, GLR is a good (and difficult) regional, no wonder it filled up on the first day. Hope to see you all in Chicago for the Midwest. If you have any questions before you come, drop me a line.

Chris

That's pretty neat. Simulation is an important tool in engineering. We are working on incorporating some into our system.

How do you feed Matlab the information on where to go? Where does your auto code get loaded? I'm assuming it goes onto a custom circuit. What is your interface to the RC? Does it end up being a dead reckoning system?

Dave

Our system
 

Dave,

Instead of waypoints, we use more of a "path following" technique where we know the shape of the path so we can set the control algorithm up to stay on the path.

The generated code is run on a Motorola HC12 on a custom circuit (all Thursday we struggled with a micro reset issue that we finally tracked down after our 3rd practice). The custom controller sends a 0 - 5 V signal to the RC and the RC just simply reads the signal at the A/D and spits it out to the PWM. We also use the digital inputs to control the arms and other stuff.

Every part of the algorithm is closed loop, so it's not at all dead reckoning. Once during testing we forgot that a stack of chairs was sitting in the path of our right arm; we were able plow through the chairs and track our path without problem.

I was talking to Mike Soukup at the Regional and we were comparing controllers. It seems that they're quite similar (we're using the same measuring techniques as well). We found that one nice thing about have a guidance system is that we can test it on concrete or carpet, good or bad battery, high or low gear, and it really doesn't make much difference to the result. The only real difference is that carpet adds more damping so our settling time is faster on carpet than on concrete.

Anyway, good luck to you guys this week and at Houston.

-Chris

Re: Re: best technical feature at GLR
 

Tom & others who have asked for it, here's the video. It's obviously not our team's finest moment since it wasn't an excellent match, but it's a good showcase for how StangPS can almost always help us recover from bins/walls/robots in our robot's path.

http://www2.wildstang.org/2003/video/gl_062.asf

wow i am so glad we someone else decided to do something other than dead reckoning. Our Machine (yes it happens to be pink) uses the same gyro and counting wheel to determine direction and distance, it can move up the ramp and fight almost intelligently. The only thing we lack is a smart custom circut, all we have is one that reads the holofect sensor on the counting wheel (it uses a magnet to measure passes in 8in intervals) and tells the base control chip that it has gone 8 inches, the chip then counts and we can tell how far we have gone, in the same time the chip calculates that if the robot hasn't moved in so many cylcles and the motors are at full it backs up and trys again, and it fought and spun out a robot @ UCF.

Congrats on your design, i should like to see it in action, you going to nationals???

Yes, we'll be there. We'd love to have you stop by if you have time. I'm looking forward to checking out all these teams who did something other than time-based dead reckoning! As a side note, while our setup may seem overkill, I should point out that it was basically a requirement for us. The way our crab drive works, our orientation often drifts when we try to drive in a certain direction (especially sideways). Because of this drifting, time-based dead reckoning probably wouldn't work at all for us, or at least it would be very inconsistent. I guess I don't know for sure though as we commited to our current method 100% from the beginning. We never even tried a time-based program...

Wow! I didn't know both 308 and 'Stang used a guidence type system!! That is extremely cool!! Never thought of using an NC path type system. Of course all this fancy electronic, Black magic stuff is still a little out of our reach. Our arm driver and the coach, worked on the autonoumous process during the build season and decided Dead reckoning was the best way we could go with without having to create custom circutry. Seems to work pretty well. Our Auton mode is extremmely fast and consistant(when the Controller is functioning properly!) 33,217, 226 and 65 were also very fast and consistant. And those are only the ones I can name off the top of my head, I know there are many fast, consistant machines out there. I believe we are all in the 3.15 to 4.5 second range.

I was in 111's pit several times marvelling at the craftsmenship and awesome engineering invested into 2 complete chasiss!! You guys are amazing!! I love all the press brake and waterjet work done on your machine. Then to top it off you have this really cool tracking system. Maybe you guys should be building the next Terminator! Truly amazing. Team 27 would also like to thank-you for letting us use your drill press this weekend. You guys are truly gracious and it was an honor to play with you this weekend. Our match is one of my two favorite matches this last weekend. The other was with 302 in Qualifying match #102. Good luck in Chicago! Hope to meet you again in Houston. :cool: