PDA

View Full Version : Fully autonomous game


Hypnotoad
07-30-2013, 12:10 PM
By defenition, a robot is "A machine capable of carrying out a complex series of actions automatically." (with google as my witness). Now, Traditionally, FRC "robots" operate on their own for fifteen seconds and are controlled by human operators for the rest of the match. We know that autonomous mode is not often very complex in robots, and the complex task of moving to the feeder station, loading frisbees, finding a suitable firing position, aiming, firing, and later climbing, are not being performed automatically.

What I am getting at is that the machines we build are not "robots", but are instead really fancy RC cars. Now, that doesn't make them any less awesome (these are some REALLY fancy RC cars), but it does make them less... well... robotic.

This is not a problem on the mechanical side of things as whether or not the robot is fully autonomous or not does not affect their learning of CAD, power tools, milling, etc. This does, however, limit the programming because there simply isn't anything to code that the driver and operator won't do better. Why make a PID loop if humans are already the best PID controllers you could have.? Why make vision tracking code when humans are faster than code at lining up shots and firing?

This could be just a personal complaint that only I have, but I want to know from you guys.

So what is keeping us from making our robots fully autonomous as things are right now? If people really want to make autonomous robots why not just make them? Absolute position tracking, or rather lack thereof. The robot cannot on it's own know its exact position. GPS deprived navigation is not even a fully developed military technology yet, let alone tech that our robots can use. If the robot were to know its actual position on the field, both rebound rumble and ultimate ascent would be perfectly playable as is by a fully autonomous robot.

The actual methods of doing this can be discussed on the technical forum if the idea catches on. I suspect a localized gps style system would do the trick quite easily, but there are always multiple ways of doing something.

AdamHeard
07-30-2013, 12:18 PM
I would argue there is a high level of useful automation that can be done. Human beings are certainly not better PID controllers than the robot in a lot of cases. Can they control multiple system at once? How would they control flywheel speeds? etc...

In terms of positioning, if FIRST really wanted teams to solve that problem a great solution (stolen from robocup) is an overhead camera feed that all teams have access too. Robots would each be required to have a distinctive shape somewhere on the top with a distinctive color pattern so that each could be identified.

Andrew Lawrence
07-30-2013, 12:18 PM
-Robot needs to move autonomously to be a robot
-FRC has a 15 second autonomous period where robots move autonomously
-Therefore what we build = robot.

Basically, your definition says a robot is "capable" of doing tasks autonomously. Ours are certainly capable of that, they do it for 15 seconds each match. That qualifies them as a robot, no matter how much teleoperation comes after it.

Hypnotoad
07-30-2013, 12:31 PM
-Robot needs to move autonomously to be a robot
-FRC has a 15 second autonomous period where robots move autonomously
-Therefore what we build = robot.

Basically, your definition says a robot is "capable" of doing tasks autonomously. Ours are certainly capable of that, they do it for 15 seconds each match. That qualifies them as a robot, no matter how much teleoperation comes after it.

My post explicitly mentioned "complex series of actions". I do not feel that firing, driving forward while running a pickup motor, driving backwards the same amount and firing again qualifies as complex. There is no decision making. If it did qualify as "complex" you could call wind up toys robots too.

Al Skierkiewicz
07-30-2013, 12:36 PM
So what is keeping us from making our robots fully autonomous as things are right now?

Inspiration! With a fully auto robot, there is no need for drivers, strategy or data observers or interaction with other teams.

Joe G.
07-30-2013, 12:45 PM
For FRC to achieve its culture changing goals, the competitions must be exciting and accessible for the general public. This means that someone who is not well-versed in the technologies involved in the robots should be able to walk into a competition and be immediately impressed by the task. Of course, one wants to preserve the opportunity for students to learn advanced engineering concepts. But raising the minimum acceptable performance by creating autonomous games would result in many more "failed robots" than you see today, and lowering the bar enough to make the game more autonomous friendly would ruin the spectacle.

A good example of what I mean is Lunacy. It's not a very popular game. I think a big reason why is that it crippled FRC robots' potential. Here we have these sophisticated machines, with industrial control systems and drive trains capable of precision maneuvering, and then we cripple them by making them slide around out of control on a slick surface and slick wheels. Those of us who bury ourselves neck-deep in the robots know all about the intricate details of the challenge, and the impressive things top teams came up with to operate as well as they did, like fans and follower wheels. But to the casual observer, a lunacy robot manuvering was simply not an impressive thing to watch, especially when compared to robots that climb a pyramid or shoot basketballs or launch a track ball.


Sidenote: Lunacy was also the last FRC game to encourage autonomous interaction with your opponents. In theory it sounded like it should have worked like it did in the game animation, with robots chasing each other using vision code. In practice, it resulted in 6-way collisions at the center of the field 90% of the time, with the occasional robot dead reckoning over to a loading station, and practically no reliable autonomous scorers. Even with no interaction from 2010-on, autonomous mode has been much more exciting since the mid-field lines started appearing in the rulebook.


The simple truth is, FRC consists of complex robots built by teams of high schoolers in six week periods, and plenty of teams struggle with it already the way it is. Demanding that these robots match the performance of cutting edge professional robots may increase the technical learning opportunities for some, but would push many away from the sport, and make getting interested in it even more difficult.

Pault
07-30-2013, 01:00 PM
I would argue that there are some robots who fulfill this definition, from this year some automatic cimbers (as in I push a button and all of the sudden my robot is a the top of the pyramid) and floor pickups preformed complex series of actions throughout the match.

But if we went fully autonomous, it would take the excitement out of everything. There is so much more uncertainty when a human element is in play. Changing stratgies, defense, and imperfect driving all add to what distinguishes FIRST as a competition instead of a science fair. If it wasn't for the drivers, coach, and human player, people could easily predict every match and be right about it 95% of the time.

Jon Stratis
07-30-2013, 02:09 PM
I would challenge your definition of a robot. What I get from dictionary.com:
a machine that resembles a human and does mechanical, routine tasks on command.

Our robots may not always resemble humans, but they DO perform routine tasks on command - simply driving forward on command could be considered to meet this part of the definition, even if that's all the robot does.

For your point of trying to go fully autonomous... Attempting to create something fully autonomous that can react appropriately to random events (ie other robots) in its surroundings is hard. Really hard. I don't care how good of a programming team you have, there's pretty much no way you'll beat a driver with just 6 weeks of design and programming.

That said, I always encourage teams to automate as much as is reasonable. For example, this year my team's climbing mechanism was as close to fully autonomous as you could get - manual lining up and triggering, and a safety release to stop it in mid-motion if needed, but once it started lifting there was no human feedback to keep it going from one level to the next. This was really only possible because we could be sure of there being no "random" interference from other robots during the process!

And localized GPS isn't a perfect solution, either. A couple of years ago I watched a college/post-college autonomous snow plow competition, and many of the competitors utilized localized GPS to help guide their robots. Despite this (and the fact that they had much longer to work on the robots AND much less complex robots - ie a drivetrain with a blade in front - than we do), no competitor performed perfectly. They all missed some area of the pre-planned course. They all wandered off course a bit and had to be manually reset.

Also, consider years where there is no way to know ahead of time where the game pieces will be at any point in the match. They aren't always fed in from a slot, whose location and orientation is known. Sometimes (quite often, actually) they're bouncing around the field and have to be tracked down. robot position for that doesn't matter - vision tracking of moving objects is. In fact, the only real constant we've had (except for Lunacy) in the past few years or so is goals - they seem to always have some sort of vision target available, which really is all a robot should need to line itself up to score appropriately. But even doing that much is incredibly difficult to do in the code.

ebarker
07-30-2013, 02:20 PM
This looks like a thread from a few years ago all over again. Back then, If memory serves me well, Dave Lavery piped in and presented the definition of a 'robot'. Bottom line is that robots are not necessarily autonomous.

gblake
07-30-2013, 02:28 PM
...
So what is keeping us from making our robots fully autonomous as things are right now?
...
Absolutely nothing. Go right ahead and make one. I'm serious.

Your OP focus on wanting a field-based nav system like the GPS would simplify things, but isn't necessary.

So, go get busy; and be sure to post a picture of all of the accolades the end product earns for the team of innovators who create it. It will be an achievement to be proud of.

Blake

Rangel
07-30-2013, 03:18 PM
FRC I think is so fun because it usually features fast paced, back and forth games. It is very spectator friendly because it is mostly teleoperated. I would venture to say that the reason we don't have full autonomous is because a great many teams struggle with 15 seconds as it is. Now, there are definetely some teams out there that have the abillity to go that extra mile to fully autonomous. That is why there are plenty of other competitions besides FIRST out there such as:

Robosub- http://www.auvsifoundation.org/foundation/competitions/robosub/
IARC- http://iarc.angel-strike.com/
RoboBoat- http://www.auvsifoundation.org/foundation/competitions/roboboat/

Having recently competed with my team at the Robosub competition, I can say that it is a lot of fun and definetely very challenging. The best part too is the super cheap entry fees. Robosub for example was only a mere $500 for the weeklong competition it is.

RyanCahoon
07-30-2013, 08:53 PM
Inspiration! With a fully auto robot, there is no need for drivers, strategy or data observers or interaction with other teams.

For FRC to achieve its culture changing goals, the competitions must be exciting and accessible for the general public. This means that someone who is not well-versed in the technologies involved in the robots should be able to walk into a competition and be immediately impressed by the task.

Along these lines, I think having humans in the game also makes the game more interesting in a direct manner: humans like watching humans (this is why professional sports are so popular). I think if we removed or distanced humans from the visible part of the competition, it would make the game much less attractive to the general public, but also to students who may not have "drunk the koolaid" completely yet.

When you come down to it, FIRST is about people inspiring other people; STEM and the robots are just the subject matter.

It's been commonly pointed out that automobile racing is one of the closest other sports to FIRST. I think there's something to be said for why so much of the presentation of those sports focuses on the drivers, when it could very easily have a much greater focus on the cars. Much of the technical information presented serves as much to illustrate how impressive the feats are that the drivers are performing as it does to laud the performance of the cars.

--

That said, I'd love to see more autonomous in FIRST games. I'm personally in favor of an autonomous zone, rather than a time period.

qzrrbz
07-30-2013, 10:31 PM
Pretty sure that the underwater sub challenge that 842 just competed in could be called "fully autonomous".

Water game!!! :p

N7UJJ
07-31-2013, 07:47 PM
Pretty sure that the underwater sub challenge that 842 just competed in could be called "fully autonomous".

Water game!!! :p

and they came in 8th overall. The second highest ranking high school in the international competition. Thus they can claim to be a world champion high school in an international autonomous robot competition besting university teams from the U.S., China, Russia, etc.

Congratulations to the 2013 Winners!


1st Place Cornell University, $8,000
2nd Place University of Florida, $4,000
3rd Place Far Eastern Federal University, $3,000
4th Place University of Maryland, $1,000
5th Place Harbin Engineering University, $1,000
6th Place Amador Valley High School
7th Place National University of Singapore
8th Place Falcon Robotics

Special Awards:
Best New Entry: National University of Singapore, $500
Bang for the Buck: Daytona Beach Area Homeschool, $500
Best Paper: Cornell University, $500
Best Static: Ecole de Technoligical University, $500
Outreach (3-way tie, $500 each):
•Embry-Riddle Aeronautical University
•Carl Hayden High School
•North Carolina State


Participating teams:
Amador Valley High School- Pleasanton, CA
Autonomous Underwater Vehicle - University of Arizona (AUVUA)- Tuscon, AZ
BangaloreRobotics- Bangalore, India
Ben-Gurion University of the Negev- Beer Sheva, Israel
Carl Hayden High School- Falcon Robotics- Phoenix, Arizona
Cornell University - Ithica, NY
Daytona Beach Area Homeschoolers - Team S.S. Minnow- Daytona Beach, FL
Delhi Technological University - Delhi, India
Ecole de Technilogie Superieure- Team Sonia-Montreal, Quebec
Embry-Riddle Aeronautical University- Daytona Beach, FL
Far Eastern Federal University- Primorsky Kray, Russia
FuVe- Escuela Tecnica Superior de Ingenieros Navales de Madrid- Madrid, Spain
Gears 2 Robots- Temecula, CA
Harbin Engineering University- Harbin, China
Indian Institute of Technology Bombay - Bombay, India
Istanbul Technical University (AUVTech)-Maslak, Istanbul
Malardalen University- Malardalen, Sweden
Montana State University- Bozeman, MT
Netaji Subhas Institute of Technology- New Delhi, India
North Carolina State University- Raleigh, NC
Prairie View A&M University-Prairie View, TX
San Diego City College- San Diego, CA
St.Georges School Canada- Vancouver, British Columbia
Team Bumblebee - National University of Singapore- Singapore
University of Alberta- ARVP- Edmonton, Alberta
University of Florida- Gainesville, FL
University of Maryland- College Park, MD
University of Toronto- Toronto, Ontario
USC Autonomous Underwater Vehicle Team- Los Angeles, CA
Utah State University Department of Computer Science- Logan, UT
Washington State University - Pullman, WA


The competition is a real challenge, especially for programers, but obviously FIRSTers are capable.

http://www.auvsifoundation.org/foundation/competitions/robosub/

Hypnotoad
07-31-2013, 09:12 PM
:D Now to convince my team to join robosub...

EricH
08-01-2013, 12:36 AM
I wasn't going to jump in on this... but then I remembered back between the 2010 and 2011 offseasons, someone threw out a challenge to make a fully-autonomous FRC robot. A bunch of folks were seriously thinking of taking him up on it.

Now I'd go into the technical details of what even a fairly simple FRC robot would need for full autonomy, but I don't want to take too long. I'll simply point out that my senior design project in college (a robot) was supposed to be fully autonomous, for a simple (single-purpose) task, and even with only 8 sensors (4 drivetrain encoders, 2 location cameras with defined targets, 2 current sensors, IIRC) and a very simple robot (4 wheels, 1-joint arm with a wrist) we couldn't do it in 2 semesters (9 months). Theoretically, it's not that hard to "drive out X distance, lower arm, drive Y, raise arm, locate base and return", but we still couldn't do it autonomously, though the programmers figured that they could probably have finished it given a little more time. (We ended up ditching all the sensors except maybe the current sensors and just driving manually--which code was done pretty quickly.)

Can it be done? I'm not saying it can't, particularly on a "do this when instructed and I don't want to have to control it all" level. But it is going to take some serious commitment, or a game that requires autonomous operation outside of the 15 seconds, to get more than a few dreamers to make the attempt. And in 6 weeks? I think maybe some groundwork gets laid well ahead of time.

scooty199
08-07-2013, 03:36 AM
While I am a CS major and was a programmer for my FRC team and love the idea of autonomous and the challenge behind full autonomous competitions, I just don't see the benefit of implementing it in FRC.

While FRC has to make a challenging platform for its participants, it also has to make some form of an attractive showing for the spectators it does have. That comes in the form of human interaction and driving. As said previously, there's an exhilaration and excitement factor when you add the human element. FRC thrives on it as I've seen.


That said if they ever want to lengthen the time of autonomous to 30 seconds I'd love to see what the future members of the team I'm mentoring can do.

In addition, I believe the competition that does it best in terms of autonomous might be VEX U, 1 whole minute for autonomous allows a lot of freedom and capabilities out there.

evanperryg
08-09-2013, 04:00 PM
Although driving and a few functions of the robot are human-controlled, many features are autonomous. On my team's robot, these include shooter angle adjustment (PID), shooter angle brake(allows us to fix our shooter in a single position, very useful), full court alignment, shot timing(robot won't fire unless the angle and wheel speed are right). Some teams made their climb autonomous, just drive up and press a button and the robot does it all itself.

Invictus3593
10-21-2013, 08:01 AM
...if they ever want to lengthen the time of autonomous to 30 seconds I'd love to see what the future members of the team I'm mentoring can do.

I think making autonomous 30 seconds would require more revision that FIRST wants to do right now. With that kind of time and these same sort of games, robots would be sitting idle for about 15 extra seconds, including the ones who went and grabbed other frisbees for a 7-throw autonomous.

yash101
11-01-2013, 10:12 PM
By defenition, a robot is "A machine capable of carrying out a complex series of actions automatically." (with google as my witness). Now, Traditionally, FRC "robots" operate on their own for fifteen seconds and are controlled by human operators for the rest of the match. We know that autonomous mode is not often very complex in robots, and the complex task of moving to the feeder station, loading frisbees, finding a suitable firing position, aiming, firing, and later climbing, are not being performed automatically.

What I am getting at is that the machines we build are not "robots", but are instead really fancy RC cars. Now, that doesn't make them any less awesome (these are some REALLY fancy RC cars), but it does make them less... well... robotic.

This is not a problem on the mechanical side of things as whether or not the robot is fully autonomous or not does not affect their learning of CAD, power tools, milling, etc. This does, however, limit the programming because there simply isn't anything to code that the driver and operator won't do better. Why make a PID loop if humans are already the best PID controllers you could have.? Why make vision tracking code when humans are faster than code at lining up shots and firing?

This could be just a personal complaint that only I have, but I want to know from you guys.

So what is keeping us from making our robots fully autonomous as things are right now? If people really want to make autonomous robots why not just make them? Absolute position tracking, or rather lack thereof. The robot cannot on it's own know its exact position. GPS deprived navigation is not even a fully developed military technology yet, let alone tech that our robots can use. If the robot were to know its actual position on the field, both rebound rumble and ultimate ascent would be perfectly playable as is by a fully autonomous robot.

The actual methods of doing this can be discussed on the technical forum if the idea catches on. I suspect a localized gps style system would do the trick quite easily, but there are always multiple ways of doing something.

Yes. I agree!
Please do not feel offended by this, anyone. Our "robots" are not robots. They are more-or-less remote controlled cars meant to carry out a special function

EricH
11-01-2013, 10:23 PM
Yes. I agree!
Please do not feel offended by this, anyone. Our "robots" are not robots. They are more-or-less remote controlled cars meant to carry out a special function

We've discussed this before, on a couple of occasions. Consensus says:

They ARE robots--but they are a special class of robots known as "telerobots". They fall into a similar category as bomb disposal robots, possibly some drones, and any robot operating under a "learning" mode that is being directly controlled by a human, as well as Mars rovers. These robots are mostly under operator control, but are also capable of running autonomously (and to one degree or another, they all have some autonomy).

Tristan Lall
11-01-2013, 11:46 PM
By defenition, a robot is "A machine capable of carrying out a complex series of actions automatically." (with google as my witness). Now, Traditionally, FRC "robots" operate on their own for fifteen seconds and are controlled by human operators for the rest of the match. We know that autonomous mode is not often very complex in robots, and the complex task of moving to the feeder station, loading frisbees, finding a suitable firing position, aiming, firing, and later climbing, are not being performed automatically.

What I am getting at is that the machines we build are not "robots", but are instead really fancy RC cars. Now, that doesn't make them any less awesome (these are some REALLY fancy RC cars), but it does make them less... well... robotic.
Yes. I agree!
Please do not feel offended by this, anyone. Our "robots" are not robots. They are more-or-less remote controlled cars meant to carry out a special function
We've discussed this before, on a couple of occasions. Consensus says:

They ARE robots--but they are a special class of robots known as "telerobots". They fall into a similar category as bomb disposal robots, possibly some drones, and any robot operating under a "learning" mode that is being directly controlled by a human, as well as Mars rovers. These robots are mostly under operator control, but are also capable of running autonomously (and to one degree or another, they all have some autonomy).

Eric is correct—and unfortunately, you two are not. The Wikipedia article (https://en.wikipedia.org/wiki/Robot) describes the nuances much better than the simple dictionary definition that forms the premise of this thread.

If that's not good enough, consider that autonomy can be expressed in both large and small scales. Even a feedback control system like a PID loop is to some degree autonomous, even if the overall direction is being provided by a human.


As for making FRC robots fully autonomous in a game similar to past ones, it could be done—at the cost of most of the teams after the first year, and a dramatically lower level of competition. Barring a much more capable basic software and hardware platform, situationally-aware robots are really hard to build. One of the most important characteristics of that awareness would be a grasp of strategy, but that may well require knowledge of the state of the game in real time—something which isn't yet passed to the robots by the field, and which is very hard to sense directly.

Another way to solve it is to play a different kind of game. Perhaps something like FLL, where there is no opposition to add randomness to the game. I don't think I'd like that.

Hypnotoad
01-05-2014, 09:50 PM
I guess the game creators disagree with me, as evidenced by this even simpler (programming wise) game.

Ernest314
02-18-2014, 04:03 PM
I wouldn't go as far as to want a fully-autonomous game--but the few seconds we do have seems rather pathetic (no offense). I really don't know what the GDC should do, but the current situation seems inadequate. Sure, we need "the human element" and excitement; the fact that FRC is a spectator sport is a good thing. But to accomplish FIRST's mission:

"[...] to inspire young people to be science and technology leaders, by engaging them in exciting mentor-based programs that build science, engineering and technology skills, that inspire innovation, and that foster well-rounded life capabilities including self-confidence, communication, and leadership"

requires more than making students and the public feel like robotics is "cool." There needs to be something more. I get that many teams already struggle with the challenge of the game, but that doesn't mean we should set a limit on how much students can accomplish. Instead, there should be easy tasks as well as difficult ones--I think FIRST is doing a great job with this, for example having part of the autonomous be simply "drive forward".

Back to the definition of "robotics." Notice how the mission statement doesn't even mention that word? FIRST is interested in STEM, not robots in specific. "Robotics" is simply the medium (if you will) that they chose to spread their message. I do agree with OP on the definition of robotics though: even though there are ways to argue for what is and isn't a robot (and teleoperated robots are robots, kinda), current research into robotics is mostly in the autonomous realm. Just take a look at the articles on IEEE Spectrum! So autonomous operation of robots definitely should be emphasized, not just swept under the rug. This could be a selling point; imagine FIRST advertising that they were inspiring students to build robots that performed advanced autonomous tasks. FIRST really can't say that right now.

How should this be done? I don't know. I'm sure the GDC can come up with something though. Give robots more difficult tasks, along with easy ones. Vision targeting is cool, but that hasn't changed from year to year. And the algorithms that students have to use is very simple, compared to some of the mechanical things that students have to accomplish. (Maybe I'm overstating this because I'm a programmer.) Whatever happens, students shouldn't ever be limited by the program.

That was a long first post.

(On a side-note: I realize what people mean when they say that PID is an automated thing, but come on...)

wilhitern1
02-19-2014, 07:19 AM
Sorry I haven't read the whole thread. Anyway, here's my view. Perhaps 30% of teams have an effective autonomous at the regionals that I've been to. You'd basically be eliminating 2/3 of the teams. They might as well stop coming.

Instead, how about having a separate contest each year. Teams could sign up for it. unopposed scoring. Highest score wins.

Neal