2006 Autonomous Disappointment

[MikeDubreuil]

Originally Posted by billbo911

There is no way you can track the upper target manually with any real success!

Originally Posted by Rick Thornbro

Much more fun for us programmers.

I agree with both quotes. This year if your robot can't score in the center goal during autonomous I suspect your drivers won't have much luck scoring in there either. I imagine it will be extremely difficult for a driver to line up and score into the center goal. In fact, the best center goal scoring robots will have their ball shooting mechanism running autonomously during the whole match. This makes the software teams job this year extremely important and fun!

[Kevin Watson]

Originally Posted by chris31

How fast can the camera track. Do i need to write some code to make it be able to track better. I havent yet set up the camera so i dont know how fats it really reacts.

You're free to modify/re-write the code any way you see fit. The camera, out of the box, tracks pretty well and can potentially be made faster by adjusting the two proportional gain values that are documented in tracking.h. You might also consider adding some more intelligence to the tracking algorithm. As an example, consider implementing some kind of forward predictor of where the center of the green blob (also known as the centroid) will be in the next t-packet and move the servos ahead of time. If the centroid moves 30 pixels to the left between the last t-packet and the current t-packet, it might be safe to assume that the centroid will move around 30 pixels left in the next frame. With that knowledge, command the servos such that the camera will move, say, half that distance left so that the calculated error on the next t-packet will be much smaller. If you don't understand this, don't worry as it's advanced stuff you'll get later in college if you take a control theory class.

-Kevin

[chris31]

I know exactly what you mean and I am working on ways to make it faster. Also rewriting the search code to make it find the target faster.

[KTorak]

Originally Posted by MikeDubreuil

I agree with both quotes. This year if your robot can't score in the center goal during autonomous I suspect your drivers won't have much luck scoring in there either. I imagine it will be extremely difficult for a driver to line up and score into the center goal. In fact, the best center goal scoring robots will have their ball shooting mechanism running autonomously during the whole match. This makes the software teams job this year extremely important and fun!

You're thinking exactly the same thing I am. This year, programing your robot will be extremely important as well as the use of the CMU Cam. Your robot will more or less run its own "autonomous mode" to score in the center goal becuase it will be eatremely hard to line it up and stay lined up with all of the pushing and shoving I suspect will be occuring in the matches. Without an automatic aim, it is unlikely that you will have an easy time scoring into the center goal. However, I'm sure that some teams will be able to do it, just not as well as anyone else who tries to manually score.

...my $0.02 on the game...

[dm0ney]

Originally Posted by cbolin

Hi,
Coding your bot to drive and shoot in 10 seconds is really challenging...especially if you want to shoot at 100% (10 of 10 balls). This is far more difficult than raising a tetra and sitting it on the center goal like we did a few times last year in autonomous.

Consider this aspect...Software assist during the 120 second drive time. Example:
1. Laptop computer with graphic display connected to Dashboard serial port.
2. Camera x,y data being used to move cross-hairs on laptop.
3. Gunner (human) uses the crosshairs on laptop display to move turrent azimuth and elevation to close proximity. 1 or 2 tracer rounds to allow human to get to the target...and then feed the balls into the target. One team is allowing for loading of over 25 balls.
4. Drive team can be in opposing corner from the robot shooting diagonally while the bot is moving.

Squeezing the most useful information from the sensors with good code is much more challenging this year. Add in X,Y sensors with Yaw rate sensor to dead-reckon is cool.

So, think of a 130 second autonomous mode with driver assist for 120 seconds. :-)

Regards,
ChuckB

This is exactly something I envisioned. The dashboard program can be utilized even further to more autonomous functions into the human mode.

Last year, we had buttons for predetermined arm positions such as 'short goal w. one tetra', 'center goal w. one tetra', and 'fully stowed'.

This year I can see utilizing the camera or a combination of sensors to use a point and shoot dashboard type interface. Line up the crosshairs and fire.

An alternative might be instead of using the camera targeting, use predetermined field positions and different modes of shooting to line up the crosshairs?

I think that this autonomous leaves quite a bit to the imagination and also forces programmers to start thinking about more code in the human mode.

[Astronouth7303]

The holy grail: Your board has one switch, "Fire".

[devicenull]

Originally Posted by Astronouth7303

The holy grail: Your board has one switch, "Fire".

Actually, I forsee two: Fire and "Thats the opponents target, you dumb robot"

[KTorak]

Originally Posted by Astronouth7303

The holy grail: Your board has one switch, "Fire".

...so that's what we can make the big red button do.

[X-Istence]

Originally Posted by devicenull

Actually, I forsee two: Fire and "Thats the opponents target, you dumb robot"

Computers make fast and accurate mistakes, as always, PEBKAC applies here!

[Joohoo]

Just because Auton mode is shortened That doesn't mean you can't have a task done autonomously in the other 2 minutes

[BotLobsta]

What would be really impressive is if your team didnt need any drivers. The robot could acquire the target, shoot, and move all on its own. Then at the end of 120 seconds, it tries to get back to the platform. That would be very cool to watch.

[Donut]

That would be cool, except for the part where it tries to fire the balls it doesn't have, because it's already fired the 10 balls it started with.

And it's no where near impossible to hit the center goal without using the camera at all. The camera is a very good assistance and increases your accuracy humongously; but you can calculate that your robot could hit the goal if it's between x distance and x distance, estimate that distance during the match using field features (and knowing their distance from the goal), and try to line it up so the ball will land in the goal. Alot more room for you to miss, but still possible.

[phrontist]

90 seconds in autonomus? Psssh. Lame! Our driver will just try to amuse the crowd while the robot does it's thing...

[phrontist]

Originally Posted by Donut

That would be cool, except for the part where it tries to fire the balls it doesn't have, because it's already fired the 10 balls it started with.

And it's no where near impossible to hit the center goal without using the camera at all. The camera is a very good assistance and increases your accuracy humongously; but you can calculate that your robot could hit the goal if it's between x distance and x distance, estimate that distance during the match using field features (and knowing their distance from the goal), and try to line it up so the ball will land in the goal. Alot more room for you to miss, but still possible.

That's why you should have sensors to count the balls going in and out of your hopper. FIRST provided some... if you know where to look.

Furthermore, human beings really suck at this sort of thing. I predict no human driver will get an accuracy of above 30% when being harassed by a defensive robot at multiple ranges.

[Donut]

We're already intending on using this sensor, I was just commenting a completely autonomous robot would have great difficulty finding balls to pick up off the floor, and I wouldn't want to rely on my human players trying to throw them in the robot.

If you lower the angle enough (somewhere in the range of 25 to 40 degrees) and fire the ball with the right velocity, you can get it so the ball "peaks" at a height that would still go in. If you make this height very close to the top and do a shallow angle that allows the ball to maintain a height near that for a good lenght of time, you can get a wide range to shoot with (shooting 10 m/s at 32 degrees if your ball leaves your launcher at 54" allows you to hit the goal anywhere from 6 feet to 24 feet away without readjusting, not including air resistance).