PDA

View Full Version : Next Game Autonomous Mode


MaxMax161
06-02-2010, 09:11 AM
While watching a NOVA special about a completely autonomous robot race it got me on the line of thinking that robots in the future will be developed with the goal of minimizing the amount of human input by giving the robot the ability to make dynamic decisions in real time. From what I can gather many if not all FRC robots are, in over simplified terms, the merely most awesome RC cars on the planet and do not incorporate this idea.

Dose anyone think that next years FRC game should/will incorporate a larger focus on programing and building the robots to be more independent? Maybe a longer autonomous mode, or maybe a bonus for the robot completing the endgame without human input?

RyanCahoon
06-02-2010, 11:37 AM
Dose anyone think that next years FRC game should/will incorporate a larger focus on programing and building the robots to be more independent?

Should they? Absolutely. Will they? Probably not; the usual argument given is that the GDC likes to make sure that all the games are at least playable to some degree by rookie teams who may not have a lot of experience. May this would be possible if the default code included a useful default autonomous routine, the same way the default code includes basic driving code by default. They have actually done similar things in the past, such as including auto-targeting code for the years that have included vision targets, but the level of adoption by teams is still fairly low, as you point out. A lot of the challenge here is finding a way to provide code that will work on all designs of robots.

You allude to this, but an interesting way to put more importance on autonomous function is to make the end game autonomous. Seattle Robotics Society is hosting a "FIRST-like" game (http://www.firstwa.org/DoubleCross/tabid/252/Default.aspx) that includes this feature. I'm very interested to see what kinds of entries are generated.

--Ryan

lineskier
06-02-2010, 10:06 PM
From what I can gather many if not all FRC robots are, in over simplified terms, the merely most awesome RC cars on the planet and do not incorporate this idea.

Dose anyone think that next years FRC game should/will incorporate a larger focus on programing and building the robots to be more independent? Maybe a longer autonomous mode, or maybe a bonus for the robot completing the endgame without human input?

I don't think FIRST needs to provide a bonus, but rather I encourage programmers to each get a test chassis from their teams, and create robots that can auto aim, lead aim, and perform obstacle detection.

The teams that have auto aiming tend to do the best in many competitions. 469 has been praised for their loop hole robot, but I think one of the most impressive aspects, was the robots ability to actually play the game. Almost entirely due to one of the fastest auto aims I've seen.
http://www.youtube.com/watch?v=FA0FSXlvf_Q

I know my team, 319, had our auto aim to thank for the success in 2006.

And trust me, drivers being proud and confident in their abilities, will always claim to not need the auto aim. At the end of the day, if your robot can auto aim they will use it.

Autonomous used to be a much more exciting time period, and based on FIRST's choices for autonomous post-2006 I don't think we can count on a exciting one. If you look at 2003, 2004, and 2006 they were all very impressive autonomous years with huge advantages for robots with a good autonomous. Interesting too, they were exciting not only because of points, but the spectator aspect. Spectators love to see robots score, they get bored fast with low scoring matches. So who knows, maybe we're overdue for a strong autonomous incentive again.

I sure hope so...

davidthefat
06-03-2010, 11:16 PM
http://www.chiefdelphi.com/forums/showthread.php?t=84797

I am doing it, no matter the game, I am doing full autonomous mode. Im just hoping the game next year won't be super complex...

Tom Line
06-04-2010, 01:10 AM
I don't think FIRST needs to provide a bonus, but rather I encourage programmers to each get a test chassis from their teams, and create robots that can auto aim, lead aim, and perform obstacle detection.

The teams that have auto aiming tend to do the best in many competitions. 469 has been praised for their loop hole robot, but I think one of the most impressive aspects, was the robots ability to actually play the game. Almost entirely due to one of the fastest auto aims I've seen.
http://www.youtube.com/watch?v=FA0FSXlvf_Q

I know my team, 319, had our auto aim to thank for the success in 2006.

And trust me, drivers being proud and confident in their abilities, will always claim to not need the auto aim. At the end of the day, if your robot can auto aim they will use it.

Autonomous used to be a much more exciting time period, and based on FIRST's choices for autonomous post-2006 I don't think we can count on a exciting one. If you look at 2003, 2004, and 2006 they were all very impressive autonomous years with huge advantages for robots with a good autonomous. Interesting too, they were exciting not only because of points, but the spectator aspect. Spectators love to see robots score, they get bored fast with low scoring matches. So who knows, maybe we're overdue for a strong autonomous incentive again.

I sure hope so...

I was not aware that 469 was using auto aim during the non-autonomous portion of the game. Are you sure that they were?

Also, you forgot to mention 2008. That year the good teams usually owned the game by the end of autonomous (see Simbotics that year).

kramarczyk
06-04-2010, 06:59 AM
I was not aware that 469 was using auto aim during the non-autonomous portion of the game. Are you sure that they were?

I asked their coach, Dan, that very question toward the end of the Troy event. His response was that they were not using the camera to target. The aiming skill during teleop is practice, not programming.

Chris Hibner
06-04-2010, 08:24 AM
I was not aware that 469 was using auto aim during the non-autonomous portion of the game. Are you sure that they were?

Also, you forgot to mention 2008. That year the good teams usually owned the game by the end of autonomous (see Simbotics that year).

As a side note, there were quite a few robots that used auto-aim in 2006. I'm pretty sure 469 was one of them that year (but not so much this year).

lineskier
06-05-2010, 12:33 PM
I asked their coach, Dan, that very question toward the end of the Troy event. His response was that they were not using the camera to target. The aiming skill during teleop is practice, not programming.

Thanks for clarifying, I watched the one YouTube clip and the accuracy was awesome from midfield. Also with the shots during autonomous, the way the robot rotates I thought it may be auto aiming. Definitely impressive driving from a team that spent most of their time sitting.

I still think auto aim and drive to target code are something you can expect in many autonomous periods, so having this work already done will put you in a good place for next year.

Also, you forgot to mention 2008. That year the good teams usually owned the game by the end of autonomous (see Simbotics that year).
Totally forgot about that one, probably one of the best autonomous modes I've ever seen. I mean one full lap 2 track balls. Crazy....

Does anyone know of clips of riot crew in 2004, they were able to go out grab a movable goal uncap it and bring it back below the corral all in autonomous.

oddjob
08-18-2010, 12:15 PM
The current game structure has it about right. A short autonomous only period, followed by human control with no restriction on autonomous play in that period. I'd hate to see a fully autonomous game. It would dramatically lower the appeal of the FIRST competition to the casual audience. Only a few technosavants would pay to watch a fully autonomous NASCAR race or two robots playing tennis.

If a team were to develop a really good fully autonomous robot, it's legal. If this actually happened I think it would be a problem for FIRST. I doubt they want a fully autonomous game but there is no need for a rule now to prevent it because it is thought to be out of reach. So while it would be an awesome robot, it might force FIRST to change the rules and outlaw it.

JesseK
08-18-2010, 12:47 PM
From a derived conclusion from a customer perspective of my "Value Proposition for an Optionally-Piloted Vertical Takeoff and Landing Platform" grad school project for the Blackhawk this summer:
The time you would spend piloting the helicopter is not better spent worrying about why the helicopter is doing what it is doing.

In >90% of today's mobile platforms that perform any useful function, full autonomy does not save any time, money, or resources due to the exorbitant cost of development to reach the market in any shortened time frame. They simply add capability that previously was not there and usually increase cost by a factor of more than 2. Long-term, full autonomy is an option since the cost is so spread out, yet that is 25-30 years away for a full-scale system.

lemiant
09-11-2010, 01:24 PM
How about a game, where autonomous allowed for a specific strategy. For example:

The field is divided into two sides by a curtain running width-wise (the short distance) across the middle of the field. There are a couple of ways to go through (over, around, under) the curtain. However, none of these passages provide line of sight from the driver stations to the opposite side.

Each side has 3 goals in it. The goals are shaped like barrels a couple of feet in diameter. The goals are positioned against the wall of the driver station (as far from the curtain as possible); one on the far left, one on the far right and one in the middle.

The game could be split in two parts, 1:30 (1 minute 30 seconds) with the curtain down, and 1:00 with the curtain up (motors could lift it mid-game). During the first half, any ball scored in the goals would be worth double their normal value.

Since attacking drivers have no line of sight (you are attacking the other teams goals... on the other side of the curtain), the attacking robots would have to be able to identify which goals were unguarded and then be able to target them and score. Drivers could help as much as they want, but due to the lack of visibility it would be very hard to attack without at least some autonomous routines.

Teams that couldn't manage to create a good enough autonomous, could defend their goals, or store up balls, wait to the end, and hope that they could make up for their balls being worth less by scoring more of them.


What do you think?

Lemiant

EricH
09-12-2010, 02:18 AM
I'll answer with some reactions.

Rookies: "We have to do what??? We don't even know what autonomous is!"

Veterans: "So, after the lights in 2006 and 2007, and the targets in 2009 and 2010, we get to deal with zero visibility for half the game. Sounds like fun. Let's see if we can steal Wildstang's code for that."

Field setup: "Uh, the technical crew won't let us go up to put the lifting rig... we can't set this up."

Technical crew: "Hey, don't blame us. We just work here, and your lifting rig interferes with X system that we absolutely have to have there. Why couldn't you guys have a ground-based system?"

Field teardown: "You want to put all these motors and cables into the road cases in what order?"

Audience: "zzzzzz Oh, hey, the robots are scoring like crazy now that the curtain is removed. Oh, match over. Match starting... zzzzzzzz"

I'll let you figure out if any of those reactions is what you're looking for.

Remember, absolutely nothing in any of the 2010 rules (or any ruleset dating back as far as I can remember, for that matter--say about 2005 or earlier) says anything about autonomy not being allowed outside of autonomous mode.

Chris is me
09-12-2010, 03:59 AM
You can say "Rookies will find it hard" with literally any change in the FRC challenge. I'm kind of tired of hearing of it as a blanket "this idea won't work" response, because that applies to any game that doesn't have a role for an unmodified kitbot. Granted, it's been 4 years since the unmodifed kitbot could not score an offensive point (or run empty cells in Lunacy), but still. It's a challenge. Any kind of "it's too hard" response shouldn't be used as a reason not to do something. More like a reason to definitely do something.

gblake
09-12-2010, 12:09 PM
I wonder if the large advantaged gained by successful autonomous operation during FTC matches is part of a devious master plan to trick students (shortly before they join an FRC team) into learning how to do autonomous operations well?

Maybe the value of the autonomous part of FRC matches will be increased in a few years because enough students will have passed through FTC training?

Or maybe not, for any number of reasons - But if there is a master plan, I hope it succeeds.

Blake

EricH
09-12-2010, 01:03 PM
You can say "Rookies will find it hard" with literally any change in the FRC challenge.
Let's take a look at the proposed idea. For simplicity, we'll assume that the barrels are open end up, and the rookie team has infinite materials.

1) Getting objects into goals. This is a standard challenge. I would presume that any rookie could build a robot that would theoretically be able to do this in some way.

2) Going through the curtain. Also somewhat standard, other than entanglement risks. Again, any rookie could build a robot to do this.

3) 1) on the far side of the curtain, where target visibility is nil from the operator's station. In order to be successful, your rookie team needs to, at minimum, have a camera feed going to the driver's station (assuming that that is possible in any given year). How many veteran teams had trouble with that this last year?

4) Apply 3) to semi-autonomous operation. Again, how many veterans had trouble with auto mode last year? Much fewer, but how many used the camera?

As a rookie team, I'd be tempted to either do too much (automode the whole way) or not even attempt it at all. Even a number of veteran teams would probably not attempt blind scoring for a number of reasons. This leads to the audience reaction to that type of game--boredom for the first half except when a robot actually scores on the far side, followed by interest during the second half when the curtain disappears and robots start scoring like crazy. That first reaction alone would probably be enough to take the game down before the GDC finished designing it.

I think Breakaway is about as far as the GDC is willing to go in terms of hidden targets--reversed from the view of the drivers so they have a harder time lining up.

Chris is me
09-12-2010, 01:26 PM
You're assuming that your rookie team in question puts the same amount of emphasis on mechanical versus autonomy that most teams do every year / last year. As in, mechanical tasks are "easy" and programming tasks are "hard". Maybe programming just seems harder because there hasn't been a great emphasis on it? Sure, there was hybrid in 2008, but it wasn't obvious to Random Team XYZ that a winning hybrid was the best strategy of all that year. People had trouble this year because teams decided that the game challenge did not call for an autonomous emphasis moreso than a mechanical emphasis, so they spent 6 weeks building and 3 days programming instead of the other way around.

Why assume a team has more mechanical resources than programming resources rather than the other way around? If anything, an autonomous intensive game rather than a mechanically intensive game would level the playing field; since programming isn't something you use "resources" to do like in the mechanical sense. "Any" team can make good code, unlike a well machined and light mechanical part.

EricH
09-12-2010, 06:42 PM
Chris, how many rookie teams have you seen with an automode? With an automode that works?

I haven't seen any improvements in those numbers over the past few years, either.

In general, automode is not a priority for rookies. This is for a variety of reasons, including the fact that the best automode is useless if the mechanical parts to make it work don't work (which is also why many programmers get very little time).

I get what you're saying--rookies can do well autonomously. But I don't think that it's reasonable for the majority of rookies, or even an alarmingly large number of veterans, to go half the match autonomously. How many teams used robocoach controls in hybrid in 2008?

If the GDC wants autonomy to be a focus at all, 2006 automode was the way to go, with 2007 being a close second. Automode has been almost an afterthought for the last 3 years.

Dave Scheck
09-14-2010, 03:50 PM
As a side note, there were quite a few robots that used auto-aim in 2006. I'm pretty sure 469 was one of them that year (but not so much this year).We also had an auto-aiming turret. The driver would get the light within the field of view of the camera and a light would turn on to let him know that auto-tracking was ready. He could then hold down a button and the turret would follow the light. It also determined the distance to the goal (using the height of the light in the frame with some trig) and determined the optimal tilt of the shooter (when you're close you have a high trajectory, moving back lowers the trajectory to a point, then you have to start aiming higher again to play the arc)

If a quick maneuver was made or if we were pushed quickly, the turret would turn in the direction that the light should have been based on the way the light left the field of view. This approach let the driver have control of how the turret behaved based on the situation.

You can see the tracking in autonomous (both pan and tilt) in this video (http://www.thebluealliance.net/tbatv/match/2006new_qf1m2). There used to be a video from Wisconsin where we got spun and the turret self-corrected, but I can't seem to find it.

We positioned the turret by hand so that it would pick up the light when it started trying to track. The same section of code was reused between both modes, so we had nice consistent behavior.

In my opinion, this is how vision targets should be used. Teams could really use it to their advantage throughout the game. Every other attempt that FIRST has used has failed. In 2007 some teams used the light in auto, but it wasn't worth it to use it during driver control. In 2009 there were a handful of teams that used the vision targets in auto and in driver control (40 and 494 come to mind), but using that target didn't make teams leaps and bounds better. 2010 was worthless because everything was in a known location in autonomous and in driver control it was relatively easy to score knowing a robot's capabilities. What's the takeaway here? Vision targets are best used when a task has a hard (not impossible) difficulty to do by hand, but easy when the target is used. It is also important to note that the illuminated targets yielded much better results than color pattern and image detection.

If anything, an autonomous intensive game rather than a mechanically intensive game would level the playing field; since programming isn't something you use "resources" to do like in the mechanical sense. "Any" team can make good code, unlike a well machined and light mechanical part.I can't disagree with this more. An autonomous intensive game will increase the gap in the playing field. Teams that have experience will always have less of a hurdle to clear and will be able to raise the bar even more. You say "any" team can make good code, but how many teams know how to? Yes, programming basic FIRST robots is getting easier every year with libraries and tutorials out there, but I am still amazed at the number of teams that struggle with simple autonomous. For example, look at 2008. All you had to do was drive across that first line to get points. How many teams didn't even have that at their first regional? Don't take this as an insult to rookie teams. I know there are well supported rookies that hold their own every year, but they are the minority. In general, rookies will always have major hurdles to clear which sets them back. Once they get over those hurdles though, great things can happen.

sithmonkey13
12-03-2010, 07:30 PM
About the game idea, couldn't a team just build a bot to stay on the side of the curtain they can see and just push all the robots trying to do autonomous out the way? Or you could just use a camera if allowed. A few strategies could become game winning if you simply ignored trying to program your own autonomous and focus on using your brain (able to quickly react to a change) versus a robot (that just follows a predetermined set of instructions) to disrupt the other alliance. My bet would be on the drivers controlling the robots preventing the autonomous bots from scoring a significant amount of points.

Also, many teams try to find jobs for all the students, like driving or scouting, pit crew, etc. Eliminating the drivers by making an all autonomous game would prevent some people from being able to participate, and at least in my experience, everyone wants to drive the robot, and would be disappointed if told the robot was all autonomous (if they even understood what it meant).

Finally, many teams take their robots to events outside of FIRST to promote the competition, their team, etc. A lot of times, this requires doing something the robot wasn't exactly designed to do in competition (such as occasionally trying to manipulate something other than what it was built to manipulate). An automous robot would probably require a large area, have to have a fence or wall to keep it in, and you cannot control it to send it out to interact with the crowd.

Just my thoughts, and if you have a counter argument against any point, or think I am trying to bring back an old thread, I am just trying to spark a healthy debate on something that many people just keep bringing the same point on.

EricH
12-03-2010, 08:37 PM
About the game idea, couldn't a team just build a bot to stay on the side of the curtain they can see and just push all the robots trying to do autonomous out the way? Or you could just use a camera if allowed.
Too easy to prevent from the GDC's perspective. Robot starting position is on the far side of the curtain, and the bottom of the curtain has a bar that is difficult for a robot to move (ideally, attached to the lifting mechanism). Then all you have to do is either ban cameras or let the drivers deal with camera lag. Defense level goes down pretty dramatically with just two tweaks...