Whisk wins the Mobot Mini Challenge!

[Somchaya]

Hey! I just recently completed building the chassis of my robot Whisk. Whisk is designed to take part in the Mobot competition, so the design choices were made with that in mind.

Basically, I'm using two Hitec servos to power the differential drive, and I'm using an omni wheel as the caster wheel. My original prototype used a regular swivel wheel, but it didn't work out too well and would cause some issues when the robot tried to go straight.

Here's the picture of the current version of Whisk:

So, Whisk's chassis is just a base that carries the laptop armed with a webcam, with wheels to navigate the Mobot track. The wheels that I'm using are really big (6" in diameter) because the servos I originally used had around 60 rpm, so to get decent speed I needed big wheels.

This is an early picture of Whisk, with just the bottom base and servos attached:

Here's a picture with all the electronics put in:

And without further ado, here are the videos:

[youtube=425,350]jYjw8aNFUNU[/youtube]

This is the first version of Whisk, which basically worked, but had a couple of issues. It was really difficult to take out the laptop and make changes to the code, since the laptop was securely fastened on the inside of the chassis. Also, the servo mounts I was using weren't designed for the heavy weight I think, so the wheels would slowly splay outwards as time went by. Also, the software, heh heh, didn't work basically.. So the robot just didn't like to follow the line.

[youtube=425,350]abVYDMyxt5E[/youtube]

This is the present (second) version of Whisk. Essentially, I made the servo mounts a little more sturdy (I hope), and moved all the electronics below. This way, I can remove the laptop and make changes as I need to, without having to dismantle half the robot each time. Also, although it looks like Whisk is semi-intelligently navigating the decision points, there's actually no code right now that does that so it was kind of a fluke that it survived that area.

Yep, that's basically the state of Whisk right now. I'm still working on the software part. I think I've fixed the bug that caused it to fail twice on the second video, but I haven't tested it out. Right now, I'm thinking of running some learning algorithms on Whisk so that it can drive itself without explicit code. No idea if that'll work though!

What do you guys think of Whisk? Any suggestions/comments would be greatly appreciated!

[sdk32285]

Hi
Good Luck with Mobot. ill probably go and watch some of it.

Something that might help is if you access the laptop using ssh so you don't need to pull it off every time you want to change something or check its status.

[Tsukubadaisei]

Good job!

I think the problem with the program not working is not your fault. Instead of using white for the lines try using a color that has more contrast with the floor, like red or black. Camera robots have problem with lightning specially in open areas like that place.

[JesseWelling]

but that *IS* the challenge of mobot.

[Tsukubadaisei]

I see... I did not know anything about this mobot competition.

[Somchaya]

Yeah..  I found out that the second run was kind of a fluke too, cos it was overcast that day. When I ran it on a day with sunlight, the light and shadows killed the program too!

[youtube=425,350]OaRsph7jvdI[/youtube]

This is a run that did worse than the second, mainly due to shadows, and also some bug in the code that refused to go away or something.

[youtube=425,350]ZSzbsrmSLjQ[/youtube]

This is a run from yesterday, where it does better in sunlight, but poorly in the noisy areas where the ground is speckled and the contrast of the line isn't so good.

I kind of gave up doing the learning algorithm, since it's pretty difficult to get that to work Right now I'm working on detecting the lines accurately, which is a pretty tough problem on its own!

[Admin]

Mini-Slalom Challenge:
       1st place ($99): Whisk--1:16.31
       2nd place (no prize money): Plan B--1:32.47

impressive! ive never been able to win the mini . . .

i see the fertilizer is still a problem, hope it doesnt gunk up my sensors again . . .

as for the bolts and nuts coming loose, use locknuts, or a glue called lock tite.

just a side note, you should mention how much you speed up the video just so it isnt misleading.

ill see you on raceday . . . ill probably start building my robot on monday . . . oh man im in for some long nights . . .

[Somchaya]

Oh.. the speed-up is mentioned before the videos are played, something like "Note: Video is playing at 2x speed" and the like..

I have 2 new videos online now too One of which is the mini challenge run (unedited)..

[youtube=425,350]1yIvvLYOf1Y[/youtube]

This is version 4 of Whisk, tried playing around with sun shields, but 2 weeks of trying didn't get much progress..

[youtube=425,350]8hw70QGwiRg[/youtube]

This is the first run Whisk made during the mini challenge. It's unedited, so all the sounds (like the strong wind) is in there. The video is also playing at normal speed for this one

Hopefully, some time this weekend, I'll put the 3 runs Whisk made into a composite video and upload it..

[Somchaya]

Mini-Slalom Challenge:
       1st place ($99): Whisk--1:16.31
       2nd place (no prize money): Plan B--1:32.47

impressive! ive never been able to win the mini . . .

i see the fertilizer is still a problem, hope it doesnt gunk up my sensors again . . .

as for the bolts and nuts coming loose, use locknuts, or a glue called lock tite.

just a side note, you should mention how much you speed up the video just so it isnt misleading.

ill see you on raceday . . . ill probably start building my robot on monday . . . oh man im in for some long nights . . .

Hehe yeah the fertilizer is still there, grass is starting to grow though, but I think bits and pieces of fertilizer is still going to fly around..

Locknuts? Hmm.. I've not heard of those.. I'll google around and have a look.. Right now what I do is carry around my box of spare nuts, and tighten the nuts after a couple of runs

Yeah, I'm looking forward to meeting some of the guys from the forum on Friday!

[Tsukubadaisei]

Somchaya, I just inscribed myself in an Robot Contest very similar to this one you are participating. The task is to make the robot carry some fruits over a plataform that must be built over it, follow a path with some obstacles and deliver the fruits to the judge at the other side of the "arena" and then go back. Since I am a student of this subject, I am already experienced with robotics, more precisely with bio-robotics, but it will be the first time ever I will prepare a robot for outdoors activity and I feel like a begginer. By the end of this week, after all the official rules are confirmed (and after I find a team), I will start a thread where I can report my ideas, my progress, ask for help and sugestions and read coments(I have a hp but it is mainly in Russian and Japanese, so no use). Of course I hope you can give me a couple of wise advises, because, from those videos, you are an expert outdoors.

[Somchaya]

Hmm sounds like a very interesting competition! You mentioned that the platform must be built over it? So the robot has to build a platform too? Sounds pretty difficult actually..

I'm not really an expert outdoors So far, the code is kind of working, but I used a couple of cheap tricks to go around some issues, eg praying for good weather, lots of clouds, no rain.. etc.. I'll definitely be more than happy to give some suggestions for your project though!

Working outdoors is really tricky, cos the lighting and weather changes a lot over time, esp in Pittsburgh... I was working on Whisk for a while today and it started raining Hopefully tomorrow it'll be better and I can get some calibration done..

[Somchaya]

And, here's the composite video of Whisk's runs for the Mobot Mini Challenge 2007!

Whisk's second run was the fastest, clocking 1:16+ for Gates 2-8. The first run was really close though, at 1:17+. In between those runs, I tried increasing the speed but apparently it didn't give much improvements in time. The third run failed, but if it worked I think it would have cut the time down by a little more I don't think my code/robot can handle that kind of speed though..

[youtube=425,350]XRbjrvWYdTI[/youtube]

In this video, all the clips are running in real-time, unlike the 2x to 5x speeds of the others. So, it might be a little boring to see the robot move around in circles for a couple of minutes

I actually have the videos of the other team (Plan B) too, but I don't think it's nice for me to post their robot video online until I get their permission to.. Their robot was extremely stable and looks all set to take on the decision gates for the actual competition!

[ed1380]

Congrats.
Great job on the robot.

[Hal9000]

I wish the UK had more competitions like this!

Well dont btw!

[Admin]

but it will be the first time ever I will prepare a robot for outdoors activity and I feel like a begginer

yea, autonomous outdoor robots are waaaaay harder (to the 3rd power) . . . i dont think anyone is an expert at it yet . . .

Somchaya, dont use felt for the sunshield. I tried it for a robot once and like three layers of the stuff still let in 30% of the sunlight . . . I find electrical tape to work VERY well, although its not as flexible . . .

I used a couple of cheap tricks to go around some issues, eg praying for good weather, lots of clouds, no rain

actually, in my tests, i found that rain (wet pavement) actually makes sensing the line much better . . . assuming your robot doesnt fry . . .

[Somchaya]

Somchaya, dont use felt for the sunshield. I tried it for a robot once and like three layers of the stuff still let in 30% of the sunlight . . . I find electrical tape to work VERY well, although its not as flexible . . .

Yeah.. so for the sun shield, what I did eventually was to use black foam as the main shield, and felt only at the bottom, but that didn't work very well, cos the wind would blow the felt into the camera's view, or the felt would just move inwards after a while.. I guess I should have tried black electrical tape..

Ultimately, I just gave up on sun shields Hopefully, it won't be too sunny on Friday and that'll be fine.. hehe.. In theory, a little bit of shadows shouldn't kill the algorithm, but when it's really sunny, the contrast between shadowed and non-shadowed areas becomes too great, and the robot thinks the line is elsewhere, and I don't really blame it When I looked at the camera image it did look as if a line existed there.. haha

actually, in my tests, i found that rain (wet pavement) actually makes sensing the line much better . . . assuming your robot doesnt fry . . .

That's very true.. I ran a test yesterday after the rain, and noticed that the contrast between line and pavement was a lot clearer.. If it's a heavy downpour though.. I probly wouldn't dare to run my robot It's not built for much(read: any) water.. hehe

[nanob0t]

Congrats!   

It looks like a neat competition.  I would love to enter something like that, but I fail to have the funds, or transportation.

Maybe some day I'll see you or Admin there   

Have to get a job                 

[Admin]

Team Whisk won 3rd at the actual Race, too (the mini challenge was a pre-race). His team got a nice check $$$ to go with it
Nice job! Im impressed!

My robot, which I started building two days prior to the competition and used a single photoresistor, didnt fare anywhere near as well . . .
Ill post the design and code up when I can. I used The $50 Robot controller so everyone can directly use what I did for their own bots.

the awards:
http://www.cs.cmu.edu/~mobot/winners.shtml

(I name my robot Pikachu every year - its a famous robot at Mobot )

[Somchaya]

We actually got 2nd

I'll be coming up with a nice summary of the whole project, plus released code, sometime in the next few weeks, once all my work for this semester settles down..