Please fix auto

[lemiant]

Dear GDC,

Having kids move from FLL to FTC to FRC is a great idea. At the moment it is executed quite well for builders, with each new level getting progressively harder. But any programmer moving from these two programs, which present a good coding challenge, into FRC would be most unhappy to take a step down into the autonomous mode of FRC. There is so much more capability in the FRC system, but no real value in innovating beyond an encoder and a gyro, or even in getting the hours to properly tune a dead reckoning program. If autonomous required real autonomous thinking instead of dead reckoning, and/or if autonomous was more important, then it would be truly enjoyable.

Please make autonomous fun for us ex-FTC programmers!

Regards,

Lemiant

[Alan Anderson]

Nowhere is it written that you have to limit autonomous robot actions to the "autonomous mode" portion of a match.

Many teams successfully targeted the high goals autonomously in the 2006 Aim High game. Some teams used autonomous tracking in Lunacy to zero in on opponents' trailers during teleoperated mode. There was an autonomous bump traversing mode in the TechnoKats' Breakaway robot that used an inclinometer to deploy and retract outriggers at the appropriate times. Lots of teams this past season used semi-autonomous behavior to position tubes for hanging in Logo Motion.

If you want the challenge of more robot autonomy, just do it.

[lemiant]

True, but in most situations drivers are better than programs. Many teams have incorporated autonomous activities into teleop and even created amazing autos (pink team), but there has been no justification for these skilled teams to really work on their autonomous. Because auto is so simple. Just look how important code is in FTC. And you begin to see a glimpse of what FRC teams could be capable of, if asked.

[Chris is me]

Originally Posted by lemiant

True, but in most situations drivers are better than programs.

Obviously you just haven't written a good enough program yet.

[lemiant]

Originally Posted by Chris is me

Obviously you just haven't written a good enough program yet.

Good point. I'm sure writing an autonomous driver would be a significant challenge

[bduddy]

There always has to be a balance. Even with these comparatively "simple" activities, probably a majority of teams won't or can't (don't have the resources) to build a working autonomous mode. Should this put them at a huge disadvantage?

[Don Wright]

Once we go to the Championship event and see more than 70% (and even that's a weak number) of the robots even just move during auto will I even begin to think that auto mode is too easy for FRC teams.

The FRC GDC has tried "more complex" automodes... In 2005, nobody [in an actual competition match] actually capped the center goal in auto mode. Now whether that was because teams weighed the time and effort verses the benefit and determined it wasn't necessary...or it was a very hard task...is a different argument.

[JesseK]

Here is a paradigm for ya if you're itching to do autonomous software.
Software only exists so that machine can serve man.
Did that just blow your mind? It basically means that if man can't control a machine ... it's practically useless. Since a man/woman can't ever truly make up his/her mind about what he/she REALLY wants to do, autonomous control is only as good insofar as we can alter the behavior on the fly (aka, teleop mode). At that point, why not just relent most, if not all, decision-making high-level control to the human operator? It'll probably make more sense when you get further into a career.

In all seriousness, for the great majority of teams out there the autonomous tasks are already difficult enough to achieve during the build season simply because the mechanical aspects take so many resources. This is just like the real world for any type of sensor-feedback system that MUST work 100% of the time. The physical piece is just as hard as the software piece in 90% of situations (even the ones where all you have to do is build a server from COTS parts...). You didn't experience this in FLL or FTC because the mechanical pieces are all, essentially, puzzle pieces.

For a real challenge of a real-time integrated autonomous software system, look into building a QuadRotor. Then figure out what tasks to give it, and make it do the tasks. They're relatively cheap if you don't break anything: www.diydrones.com. Yet even with them -- the less software, the better!

[sanddrag]

Look at Team 233's 3-tube autonomous mode from this year. Look at Team 330's autonomous stair climb and hang from 2004. Look at 980's autonomous tetra swipe, cap, and grab another from the loading dock in 2005. And there are many more. It's all what you make of it.

[BigJ]

Seconding Alan.

Talk to your drivers. Identify things that they do in the same way, all the time.

For example, our drivers this year always "spit out the tube and slightly overrided the lift down" to hang a tube.

So, the programmers developed an "autoscore" button that the drivers can hit to take half a second to perform the "scoring procedure". The lift/arm/claw operator decided to put the "autoscore" button on the base driver's sticks, and he can just tell him to score once he was snagged.

It worked out great.

[Jared Russell]

I agree that autonomy is what you make of it; every one of our machines since 2003 has had SOME degree of autonomy during teleop. Whether that means having pre-programmed setpoints on an arm, automatic kicker re-cocking, unjamming routines, traction control, etc., depends on the game.

Still, I hear you that the autonomous portion of the game could stand to get a little more attention. Veteran teams have gotten dead reckoning with a gyro and encoders down to a science - give us something new. It would be nice to have a basic challenge with a little reward during autonomous mode, but also a tougher challenge give veterans the incentive to actually use some vision/active sensors that we keep getting in the kit.

For the last 4 years (Lunacy is sort of a special case), the best autonomous modes have been decided more by the quality of the machine (driving straight, controllable turns, reliable game piece acquisition, smooth and consistent scoring motions) and basic state machines than by anything else. Nothing is inherently wrong with this, but I'd like to bring programming more to the forefront in FRC.

[Mike Soukup]

Originally Posted by Jared341

I'd like to bring programming more to the forefront in FRC.

I used to think like you, until the 2005 season when we spent far too many late nights trying to the the one team that scored a tetra on the middle goal. After that, we've closely weighed the effort vs benefits of autonomous and done the bare minimum to excel at the game's challenge. Some years it wasn't worth doing anything fancy (2007, 2009) other years it was critical (2006, 2011) and we adjusted our effort accordingly.

Back to the topic. I agree with Don, once a vast majority of the teams show a capability to complete - or at a bare minimum, attempt - the autonomous challenge then FIRST should consider upping it. Until then, I'll happily take extra nights off after the robot ships instead of staying late to fiddle with autonomous.

[ebarker]

The dumbest technical decision we made this year was to not auto-score. This year's game had "auto-score" written over it. There were teams that used software to really co-exist and autocomplete tube hanging and they did very well indeed.

[EricH]

Originally Posted by sanddrag

Look at Team 330's autonomous stair climb and hang from 2004.

That was FRC190 (up the stairs, around the goal, and up onto the bar).

330 went up the side step and positioned, but never hung.

Don't forget 469's killer automode from 2010--kick 2 and get into the tunnel. 910's "Brother's Keeper" mode from 2007, which they only got fully working well after the season. 111's and 233's loading 2 tetras in 2005.

If you want a more challenging autonomous, go for it!

[lemiant]

Originally Posted by sanddrag

Look at Team 233's 3-tube autonomous mode from this year. Look at Team 330's autonomous stair climb and hang from 2004. Look at 980's autonomous tetra swipe, cap, and grab another from the loading dock in 2005. And there are many more. It's all what you make of it.

And yet even spectacular auto like 233 (sorry I haven't seen the others) is worth very little to the game, it ends up being a challenge they did just to prove they could do it. 233 would get at most 24 extra points for that work in a 150 point game. Its not an equivalency that encourages teams to finish early or put resources into auto.