[Official 2005 Game Design] Autonomy Discussions

[dlavery]

This thread is a spin-off of this discussion, and has been started to focus on suggestions for autonomous operations for the FRC. This is a thread to present new ideas for autonomous elements in the game. While autonomy need not be a part of a specific game, creative uses of autonomy components in any game are sought. For example, a discussion may be presented that proposes the autonomous portion of the game be ______ (and we look forward to the many variations of filling in the blank).

-dave

[mtrawls]

Hrm ... maybe I'm crazy and this is just a tad bit radical, but here goes nothing!

I would like to see a return of infrared. Maybe a miniture robot could start moving through the field randomly (or controlled actively by FIRST officials) once some task is performed by a robot. This "mini bot" would have an IR beacon on it, and its release would kick the robots into autonomous mode. The goal would be to somehow capture the mini bot, possibly having to place it in some recepticle. Control could then return to the operators, or maybe there could be some other task to perform.

Now programming a robot to follow another robot and capture it wouldn't be easy, and might present a barrier for some teams, I admit (although I can't think of anything funner than trying to do this!) The twist that I like to this idea, though, and what helps to eliminate that barrier is the activation of auto mode. Maybe a robot has to enter into some zone on the playing field, or knock off a ball similar to this year. A robot which didn't have an auto mode could be designed defensively to stop the activation of auto mode. It would also be interesting strategically ... the best teams would be able to start in auto mode from anywhere, while those still stubborn enough to rely on dead reckoning would want to be sure they were at a certain spot when auto mode is activated (presuming there's some other minor task beside capturing the robot ... although I suppose a dead reckoning bot might also randomly go about trying to capture the mini bot, but then the starting position wouldn't be so important).

Well, I told you it was crazy

[Tom Schindler]

Autonomous mode has to be a task that is generally easy to complete-- both this year, and last year satisfied that requirement. Knocking a ball off of the post, or hitting a giant stack of bins is "relatively" easy to get a robot to do autonomously.

The game needs to keep this "ease" of play, but add a new element. What if, the alliances were randomly selected by the computer, so you did not know if you were red or blue until the match started. The field was also in a sort of "random" setup. The robots could have a "dead reckoning program", but it might knock off the wrong ball, or hit the wrong stack.

If a game used two different colored balls, use of sensors could allow a robot to detect which color balls you are touching, would make the autonomous task quite a bit more interesting. This would also still allow for rookie teams to design dead reckoning programs that just suck up any color ball at random, and the drivers can later decide what to do with them.


Also--

What about moving autonomous to the end of the match? That way there is no way of knowing where you are starting at the beginning of autonomous. This could still allow rookie teams to use "dead reckoning", but their drivers would have to be disciplined enough to get back to a certain spot on the field before the start of autonomous.

Just throwing a few ideas on the table..

Tom Schindler

[Joe Matt]

My only big recomendation for autonomous is that you think about it happening at the end of the game. Think about having an infared on the bar from this year. The bar would start off at 10 feet, but then after auton begins, it lowers to 7 feet. Not only would that give an advantage to attach during auton, but also requires the robots to attach to the bar higher if they do it during regular play. I think implementing that idea, not specificly, but in the same context, would be awsome. Any field features that move, crash, or are tall are cool.

[Ben.V.293]

I think that for now the automonous mode should be as it has been for the past 2 years, a simple goal at the beginning of the match. Before making it a more critical part of the game i think we should let it develop a little more so teams can become more comfortable with it. There were a lot of teams i saw this year that didn't have any automonous at all. If it becomes too complicated too fast then many teams will be left behind and the teams with experianced programmers will have a huge advantage. I also think that the majority of the match should remain under human controll. As a spectator it is much more fun to see competative human driven robots compete than slow computer controlled ones. I do really like mtrawls idea of having a mini bot on the feild that teams would have to catch. Mayby something like 100 points for catching it during automonous and 50 for getting it under human controll. Although as i driver i would prefer the whole match to be under human controll. I just think that 2 minutes of competitive gameplay is more fun than 1 minute and 45 seconds of it.

Ben Van Selous

[Yan Wang]

Autonomous mode usually consists of:
A) Robots doing nothing.
B) One robot doing a lot very quickly.
C) The majority of the time consisting of nothing, with very few exceptions.

Thus, either shorten the time for the autonomous mode or lengthen the driver control period back to the 2min length prior to 2003.

In addition, large targets like the wall of bins in 2003 were fun, but it is more interesting for there to be a smaller target (10pt ball in 2004) or perhaps a target that moves???

[Doug G]

Originally Posted by Yan Wang

Autonomous mode usually consists of:
A) Robots doing nothing.
B) One robot doing a lot very quickly.
C) The majority of the time consisting of nothing, with very few exceptions.

Many teams this past year, disabled their autonomous or ended up keeping it simple, like drive forward and push the moveable goal (or robot) to the other side. Many teams actually didn't want the balls to fall in Auto mode, since it was often more advantagous to move the moveable goal under the ball drop area and score an easy 4-5 balls. This year we worked hard on an auto program that included wheel counting and a yaw rate sensor - but found that it didn't really give us an edge and just kept it disabled for most matches. It seemed to heavily favor those few experienced teams who could accomplish several key tasks in auto mode.

In contrast, I thought the autonomous mode in 2003 was very successful since even the very inexperienced teams could write a simple auto program for dead reckoning the ramp and it proved to be very important which way that stack fell.

Perhaps going back to an auto mode where what is done in autonomous somewhat determines how the match is played. Perhaps a ramp or platform in the center of the field where balls (scoring objects) could be dropped onto at the end of autonomous. If your robot can get up there (or fight for position) during auto mode, it could catch the balls. If it didn't, more of the match would be spent gathering them up off the floor. I know the idea is not that original - but somehow get auto mode to be important (pointwise) and have it change how the rest of the match is played. That way, the matches are always different and fun to play and watch.

[Astronouth7303]

Originally Posted by Doug G

... If your robot can get up there (or fight for position) during auto mode, it could catch the balls. If it didn't, more of the match would be spent gathering them up off the floor. I know the idea is not that original - but somehow get auto mode to be important (pointwise) and have it change how the rest of the match is played. That way, the matches are always different and fun to play and watch.

Some teams did this, including Digital Fusion (a rookie). And some quite effectively (33, I believe, had a good one).

[Billfred]

Originally Posted by Astronouth7303

Some teams did this, including Digital Fusion (a rookie). And some quite effectively (33, I believe, had a good one).

If you're talking about getting right under the ball dump, then I don't think you're talking about 33. They had a devilishly effective ball collector, though. I'd hate to go against it.

On the other hand, I can think of one team that consistently got right under the ball dump and made all of our collective lives interesting...(lame pun alert)...now who was that bbat?

[Andy A.]

Lose automode.

What ever it adds to the game is more then blown away by the gap that is quickly developing between the teams that can manage to build two bots at once and those who can't. I realize that this is perhaps not the best place for this to be brought up, but I honestly feel that autonomy is a bad thing for the game.

I have yet to find any auto period exciting. Most of them are plain boring. It's like watching blind rats try to feel their way across the field. Usally they just end up rammed up against a barrier or accomplishing absoultly nothing. More so, the most effective moves I saw were simple preprogrammed moves. Perhaps this is because of the completely inane restrictions that FIRST puts on additional electronics, but I think its more a matter of teams not having the time to properly program much more advanced moves (with out a second identical bot, that is). In anycase, usally nothing is accomplished, except by the teams that have the resources to replicate their 'bot and perfect the program after the build season is over.

Let auto mode die.

-Andy A.

[Billfred]

Originally Posted by Andy A.

Usally they just end up rammed up against a barrier or accomplishing absoultly nothing.

Well, they do do things...there was at least one burnout done on the Archimedes carpet. (A team got hung up on the platform, and one wheel -just- touched the floor. Love that dead reckoning.)

Actually, that post made me take a long, hard look at autonomous mode--and I do have to say, it does seem kinda long. I mean, a good amount of teams did go autonomousless this year, and those that did were usually done at the ten-second mark.

So for those die-hards who want their autonomous mode, how about we make it a ten-second mode? If we're gonna trigger some ball dumps, then let's do it with speed, dangit!

[ZACH P.]

Originally Posted by Andy A.

Lose automode.

Without autonomous mode, they aren't robots. They are, as I have heard people degradingly describe them, "big remote control cars". Robots are by nature devices that do things autonomously, weither that task is a function of logic, or preprogrammed is irrevelant.

I think we should keep autonomous mode, and we should keep it in the begining of the match. Having it at the end is a real buzzkiller and takes away from the game. And having it in the middle of the game would be just plain frustrating. I think it should start at the begining, last for a specific amount of time, ten seconds sounds good, and then enter control mode. Additionally I like the idea of having the length of the autonomous mode optionaly extendable, with a point incentive.

[mtrawls]

Originally Posted by Andy A.

In anycase, usally nothing is accomplished, except by the teams that have the resources to replicate their 'bot and perfect the program after the build season is over.

I agree with you to a point. It is hard, near impossible to perfect an auto mode during the build season if you rely on your bot being done early enough. But we have Robovision (I really don't think that name is going to stick). Also, for veteran teams you can use past year's bots. And there's always Thursday (practice day) to work it all out! Regardless, just because it has proven difficult doesn't mean it should be thrown out ... rather, something should be designed to allow the teams climb to the challenge. I say it would be more beneficial to give teams the chance to learn how to do this than to throw it away because some teams won't. What's the chance of, after the typical build season, giving an extra week to the programmers to work on software design? Maybe just a more capable Edu-bot (err.., Robovision) would be enough? Sure, auto mode could stand to be improved ... but that's not enough of an argument to get rid of it, for me at least.

[Adam Y.]

We could figure uses for ultrasonic sensors, touch sensors, and force sensors, if they were readily available and didn't blow the electronics budget.

Actually there are sensors out there that are gaurenteed not to blow the budget and are fairly easy to build. You just have to do a little research. So far I have found four sensors that are very easy to build. The most expensive of which involves buying a C compiler for PIC's.

[Andrew]

Another important thing keeping us from true autonomy is the inability to determine who our alliance partner is and who our opponents are.

I'm all for unstructured environments, but, three 130 lb moving obstacles makes things a little too interesting.

Classic things to do in autonomy...
1. wander around and explore
2. look for things and cluster
3. avoid obstacles
4. map

I still think a maze would be the most interesting thing that FIRST could do to make autonomy interesting.

For instance, you might have to drive out of a simple maze to start the competition. The maze would always be the same and would have a stripe to follow. Actually, it would just be a few turns, not really a maze.