[Official 2006 Game Design] Autonomy And Other Technology Discussions

[dlavery]

This thread is a spin-off of this discussion, and has been started to focus on suggestions for autonomous elements of the game, and other new technologies that could be introduced into the game or kit of parts. While autonomy need not be a part of a specific game, creative uses of autonomous components in any game are sought. For example, a discussion may be presented that proposes no dedicated autonomous time period during the game, but may require that a robot complete a certain function during the course of the game autonomously while other robots on the field are being controlled by their drivers. Alternately, ideas about new drive technologies (anyone know of a source for inexpensive CVTs?) or inter-robot communications may be reviewed.

-dave

[Eugenia Gabrielov]

I have a few thoughts, as a spectator.

The concept of inter-robot communication fascinates me, but I have one or two concerns. First of all, how would this affect teams who don't have that kind of funding, or is it just kind of an extra thing?

I really enjoy watching the autonomous mode improve from all teams over the course of the season, so I support keeping a 15 second mode. I think learning to refine that is great technique, and gives the match a focus. That extra hanging tetra in autonomous was sometimes what it took to get the match.

I think new drive technologies could be helpful, but it really depends on the context of the game. I felt that many of this year's were chosen appropriately to the game, but they really may not apply from year to year. The CMU wouldn't have helped much last year, unless you wanted to seek out that big yellow ball.

[sanddrag]

Ever since the autonomous mode started in 2003, there has been a bigger and bigger push to get teams to complete tasks autonomously. However, there has not been a huge leap in the number of teams that are able to complete complex autonomous tasks very well.

The game must have some provision for huge points scoring for completing a task in autonomous mode. There needs to be so much incentive to do it that no one will not try. It needs to be easy enough for everyone to do, but hard enough to make a lot of effort go into it. Also, this huge amount of points should be available only in autonomous mode. So basically, I'm just saying more points for auto mode.

As for hardware, the gear tooth sensors were a step in the right direction but many teams find it easier to work with shaft encoders. I know many of the kit parts are donated, but perhaps the switch could be made somehow.

Also, I would say to include a couple good potentiometers along with all the instructions on how to use them.

I don't think we need to give up on the camera. But if we are to keep it we should let everyone know that as soon as possible. I would say include it again, but make the task a little easier (like pushing/pulling open a colored door or climbing up a colored ramp) and worth more points. Let teams be aware of what the sensing device (camera, IR, etc) will be and give it to them in a preseason kit so they can try it out sooner.


What would be really interesting would be if you had an autonomous mode that could be actuated at any time during the match by the opponent team/alliance. But I'm afraid that for fear of messing up the robot many teams would elect to have no autonomous mode. You would still need the really big incentive, and it would probably be only effective on a flat field with few obstacles.

[Conor Ryan]

I love autonomous, It just shows where the technology is headed, look at programs like DARPA, and all the NASA (as in the space robots not the sponsored ones) the demand for autonomous is there and growing. If anything, it should be growing (Cameras were a big step) But one thing may help autonomous, make it simple enough to learn, that every rookie team could use it. Like the Kit Transmission this year.

Continously Variable Transmissions (CVTs) are great, snowmobiles have them. You can do just about anything with them.

Interrobot Communication is great, there is just soo many possibilities of what can happen.

Thinking outside of the box is the only way First is going to get better.

[Mike]

I say definitely a bigger reward for those who complete the autonomous goal. For example, this year capping a vision tetra on the center goal was the "big thing". No teams did it, making it one of the hardest things to do in FIRST history. Yet if we were to do it, we would receive only 9 points.

I support sanddrag's view on keeping the camera, but disagree on making the task easier. Maybe you should make an easy task, worth a small amount of points, and also a larger task. The larger task would be worth 30 points or some high amount like that. This method lets rookie teams learn about the camera and autonomy, and also lets veteran teams expand on their knowledge.

Autonomous mode should definitely be a bigger portion of the game, as that is surely where robotics is going in the future. Maybe having robots interact with humans during autonomous?

[Meandmyself]

Originally Posted by sanddrag

What would be really interesting would be if you had an autonomous mode that could be actuated at any time during the match by the opponent team/alliance. But I'm afraid that for fear of messing up the robot many teams would elect to have no autonomous mode. You would still need the really big incentive, and it would probably be only effective on a flat field with few obstacles.

I would love to see an autonomous mode sometime in the middle of a match, or at the end. It would be at some set time, not activated randomly, for example the last 15 seconds would be autonomous. If the points for autonomous tasks were high enough, few teams would elect not to have autonomous. It would give a big advantage to teams who used sensors to effectively track robot position, but teams who didn't have time/money/weren't ingenious enough could still pull it off by driving the robot during manual operation to some sort of starting point.

[Donut]

The autonomous has to use another tracking sensor (like IR and the CMU cam). Maybe we can do something with sonar sensors? I've been wanting to try those out.

Autonomous should have more point incentive to get more teams to attempt it, but I think it should also include a tactical advantage. For example, there could be a gate giving your robot easy access to scoring objects or goals, but the gate is only open during autonomous mode.

There should be multiple tasks, some harder than others, to allow both veteran and rookie teams a shot to do something in autonomous. If the amount of the robots on the field continues stays at 6 or gets any bigger, there needs to be a way to tell where other robots are on the field (if robots are going to be moving alot in autonomous, we don't want them crashing into each other). Alliance robots and opposing robots could also be distinguished between, so teams could attempt to implement defensive autonomous modes.

[Pat Fairbank]

I would say that there would be a lot of benefit in having two radically different autonomous goals each year: a "super hard" goal worth a lot of points, and a "super easy" one worth fewer points but much more easily attainable.

The hard goal would require actuation of an arm or other appendage as well as the drive-train and would necessarily rely on sensors like the camera and not soley on dead reckoning, just like capping with the vision tetras this year.

The easy goal would simply be to have the robot move itself to a predefined location, for example onto a coloured 4' by 3.5' rectangle. If the robot were completely within the area (i.e. not touching the carpet outside) at the end of autonomous mode, the alliance would get bonus points added to their final score.

Although this task would be laughably simple in comparison to past games, I think it would encourage most of the less programatically-advanced teams to attempt an autonomous mode. As a result the autonomous period would be much more interesting to the majority of spectators who aren't programmers, who I'm sure think it's a lot more fun to see six uncontrolled robots charging forward in unison than five robots sitting idly while one attempts to manipulate a game object.

[Pat McCarthy]

More memory on the RC would be great!
We ran a three wheel omni/automotive drive with each wheel being independently powered and each turret rotating independent of the others. We only had 10% of memory left after our drive program was put on! 10% wasn't really enough space for a complex autonomous like we had hoped to pull off. A memory expansion chip or something like a thumb drive would be cool for those who like to do really amazing things with code!

[ColleenShaver]

I think autonomous mode was an innovative addition to the games the last few years and is worth keeping in some regard (beginning, middle, end of match, whatever).

However, I think the auto options, like the entire game itself, gets caught up in trying to force teams down a specific path of what they should do rather than keeping options open to do lots of things. For instance, this year there was a huge focus on the vision tetras and have many have said, the payoff wasn't worth the effort. Some of the coolest, most unique auto modes came from teams that didn't even worry about the vision tetras (ala using freebies to knock and cap, etc.).

I actually don't think there needs to be any auto function worth specific points per say. The 2k4 game I think is a good example of this. Although the push seemed to be for teams to knock the 10pt ball in auto, it wasn't as popular as one would expect. However, there were tons teams could do: Knock the 10 pt ball, go for a yellow ball, grab a mobile goal, jockey for position, hang, etc. More advanced teams could go bigger, while less advanced teams could just drive a try and push a mobile goal or get in a better position for the match.

My point is... include whatever technology FIRST can get donated, and then make sure teams have the accessibility and the budget allowances to get better or different sensors if they like. Then stress to teams that autonomous isn't necessarily about achieving the biggest, hardest task possible, but maybe about having a lot of options to make yourself a better alliance partner. Too much technology can be overwhelming and teams give up rather than just trying it simply (some great auto mode have been done with dead reckoning and a lot of fine tuning).

[Billfred]

Inter-robot communication? Did somebody just say laser tag?

Seriously, though, I'm thinking of the real world. You've got some things that are strictly human-controlled and some things that are strictly autonomous. But then you've got some things that have some human input, but do a lot of the stuff on their own.

So imagine this. Autonomous mode ends, and your operators take over, operating four push-button switches to handle the entire robot operation for the next N seconds. Teams can program their robots to react to those switches however they choose, but they can only use those four switches; joysticks and other switches are still disabled. I assume this would either take some trickery on IFI's part, or FIRST ordering a lot more penalty flags.

If this actually works, I will kindly ask that all rocks, baseball bats, torches, and pitchforks be kept away from me. Thank you for your cooperation.

[Mike]

I'm definitely liking the idea of autonomous at the end of a match, rather than at the beginning.

[StephLee]

Originally Posted by Meandmyself

I would love to see an autonomous mode sometime in the middle of a match, or at the end. It would be at some set time, not activated randomly, for example the last 15 seconds would be autonomous. If the points for autonomous tasks were high enough, few teams would elect not to have autonomous.

If the autonomous mode was the last 15 seconds, teams might not be willing to not have one even if the points weren't that high. Think about it: the end of the match is determined with no driver control at all. It's like a basketball coach backing off and not coaching for the last 30 seconds of a tight game. If FIRST went with this format, they would probably have even more over stressed drive teams on their hands than they have now. Vision would be really important in a format like this. The drivers would otherwise have to be extremely careful to place the robot in exactly the right spot so the autonomous mode works correctly. Of course, having 6 robots(assuming FIRST keeps the 3v3 format) all running autonomously with vision systems at the same time might cause some damage to a few bots...But I do think having the autonomous mode somewhere other than the beginning would be cool.

[Conor Ryan]

Would a RC thats based more like a computer be a possibility?

A faster processor for the RC would be great to, there are some ideas that I've had that could call for one.

[dhitchco]

I totally agree with sanddrag who said

Ever since the autonomous mode started in 2003, there has been a bigger and bigger push to get teams to complete tasks autonomously. However, there has not been a huge leap in the number of teams that are able to complete complex autonomous tasks very well.

Having an autonomous mode IS CRITICAL to the game. However, I believe it should remain at the BEGINNING of the game, and human driver mode really becomes the BACKUP plan for auto mode.

I would like to see the autonomous mode last for more than 15 seconds, but not shrink the 120 second human mode.

Each year, we're asked to push the technology more and more in autonomous, which is a good thing. Some teams did get the camera working, other teams did "work-arounds" in auto mode (using dead reckoning) while other teams chose to not participate in auto mode. All are good flexible strategies.

I'd like to see a return toa bit more focus on drive train technologies by re-introducing a three-dimensional aspect which would invlove steps or ramps....