Re: [Official 2005 Game Design] Autonomy Discussions
 

Another obstacle to AUTO mode is "vision." I know that Botball uses a CMU-cam. I don't know how expensive or effective that is. But, it might be nice to get one of those and try some stuff.

Re: [Official 2005 Game Design] Autonomy Discussions
 

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.

Re: [Official 2005 Game Design] Autonomy Discussions
 

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

Re: [Official 2005 Game Design] Autonomy Discussions
 

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?

Re: [Official 2005 Game Design] Autonomy Discussions
 

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.

Re: [Official 2005 Game Design] Autonomy Discussions
 

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!

Re: [Official 2005 Game Design] Autonomy Discussions
 

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.

Re: [Official 2005 Game Design] Autonomy Discussions
 

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.

Re: [Official 2005 Game Design] Autonomy Discussions
 

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.

Re: [Official 2005 Game Design] Autonomy Discussions
 

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.

Re: [Official 2005 Game Design] Autonomy Discussions
 

An autonomous maze idea for FIRST...

Have a truly blind maze, occupying say 15'x15' in the center of the playing field. Inside the maze are cameras so that the audience could see what was going on inside the maze.

The maze is a "dungeon" (say with a two foot high ceiling) and on top of it is the part of the playing field that everyone else uses. So, you wouldn't lose valuable playing field real estate.

There are stairs or ramps or elevators to get onto the top of the maze. You could have goals, hanging bars, scoring zones, whatever up there.

The maze would have to require four or five turns to solve it, so that it could not be mapped and completed in one match. It would require multiple matches to solve it.

This would be something that teams would do at the end of the match (say with 20 seconds left) and gradually accumulate knowledge of the maze throughout the tournament. The maze would be different at every tournament. This could be accomplished by constructing the maze modularly with movable panels.

The red team entrance and the blue team entrance (just denoting side of the field, not who can enter) would have a different solution.

At the center of the maze is a blue block and a red block. If you navigate the maze, get your colored block, and return to the outside, you win the match. If you return with the other team's block, they win the match.

This would not be the only element in the competition and would be optional. Most teams would never enter the maze at all.

This would also serve to impose a "high" limit on robot design and maybe a "wide" limit as well. Ie, if you want to try the maze, you must be able to fit inside and be able to maneuver.

Significance of the maze as an autonomous challenge
You could solve the maze through preprogrammed instructions and dead reckoning over many attempts. But, it's not likely and you would lose much competition time in matches (and probably lose a lot of matches).

If you actually use sensors and localization, you might be able to have your robot "learn" the maze a lot quicker (you'd have to store the current state of the learning algorithm at the end of each attempt).

Let's say one team commits the resources to develop a successful maze solver. Once that team "cracks it" at a competition, they cannot be beaten. The only way to beat them is to block the entrance/exit to the maze. And there are two entrances/exits.

If you have the solution to the maze and you can detect and pick up the right colored block (either by knowing its position relative to the entrance or by having a color detector sensor [light sensor]), you could navigate the maze in 30-40 seconds round trip.

A team with a successful maze solver entering eliminations would change the entire complexion and strategy of the tournament in an instant. This would add drama to a tournament as teams track the progress of the "maze solvers" from match to match. You might also add an air of "espionage" as teams attempt to acquire information about the maze from each other.

Although the maze would change for the next competition (and maybe at nationals you could elevate the difficulty of the maze by adding one more turning), once a team solved it at one regional, the "technology race" would be on. Once one team has "done it," every team that wants to be competitive would have to close the gap.

No one could completely ignore autonomy and would have to decide at the beginning of the build phase whether to invest resources in sensors and programming or in the other game tasks. For instance, if you commit to a game task which requires you to be tall, you cannot enter the maze at all.

A rookie team might use the stock drive system and commit all of their resources to autonomy. If they succeed, they could be the 500 lb gorilla at Championships. The veteran team with the mondo drive system and the do-everything manipulator might dominate the early rounds. Then suddenly, the 60 lb weenie-bot solves the maze and becomes the champion.

Audience Interaction
I envision a set of cameras inside the maze with a security cam view around the periphery of the big screen that the audience can watch. The entrance cam would pick up the robot until it made it to the first turning. A couple of other cameras would be focussed on key turnings, and, of course, there would be a top down camera looking at the "trophy room" in the center.

The audience would see a robot disappear into the maze and could watch the camera to the first turning. Then, back to the regular match. If the robot gets lost from this point, the audience will be focused on the regular match and it will be no worse than if a robot tipped over, got disabled on the exterior barrier, etc.

Suddenly, the robot appears on the "last turn" cam. It makes its turn and disappears. Back to the match. Then, the robot appears in the trophy room. Does it get the right block? Now, it's racing back to the entrance. How much time is left? Now you can see the robot on the entrance cam heading back towards the field. Then, bursting out of the maze into general view for the victory.

Re: [Official 2005 Game Design] Autonomy Discussions
 

I really am liking this maze game idea. I think it would be a great idea if you had something on top of this that you could do as well (which would mean teams would have to build under a certain height to fit in this maze.... and I think if the points were setup properly, the maze could be a very interesting twist).

Also, as Andrew said... I'd LOVE to see a changing field. I really enjoyed the dropping balls this year, but I would love to see things on the field change during autonomous everytime. For example, maybe a REALLY valuable object that for points is moved onto the field, but after autonomous it's pulled off and no longer able to be retrieved. Then you could incorporate your whole "IR beacon being forced to be used next year" scenario on this moving object.

I'm also a big fan of multipliers in autonomous. I really would like to see either at the beginning or end something similar to the 2001 game... where the sooner you go into autonomous (if at the end) or the longer you stay in it (at beginning) or maybe even a combination of the two (because you could run auto in the beginning, and see what happens during the match to decide when/if to go back into auto at the end) and this can give you valuable points.

I'd definitly like to see some type of system working with the field possibly to help teams position. Dead reckoning is a pain, and line tracking and IR beacon tracking are slow and tedious usually. I'd love to see some type of system on the field that could maybe act kind've like a GPS system where you read the location of these different beacons and can find your position. Like I said with the moving objects, you could use this GPS-like system to move near where this object would be at (I'm guessing things like that would be mid-field) and then you'd have to swap over to IR beacon to actually find the object's location. It would make autonomous more interesting, and with the GPS-like system... programmers could work AFTER the robot is shipped because as long as they know where their sensor is on the robot, they can code it for where to go without a robot, getting rid of some of the disadvantage that teams who have 2 bots have. This all could be incorporated with a maze-like idea as well, or maybe something else. There are some teams who have come up with how to do this already (I'm referring to WildStang's StangPS and I know several other teams have been making similar systems)... but for many teams it's way too difficult to do. I know I was working on a system this year, but with spacing in the robot I had no room to put the gyro wheels and things I needed for this system. Plus, I couldn't use anything in the kit really to do this, and our team's budget didn't allow for these little extra sensors that I'd need, even if we had had the room.

Overall... I really would like to see autonomous have LOTS of value next year and also be setup to where the robots can't just dead reckon or something. As George said, getting the 10 point ball this year was nearly impossible (Only team I know of that did it was Technokats at nationals one match I believe) and strategically it might not be what you want to do. But give that ball some points you can only recieve in autonomous... you'll see many more teams going after that ball.

My final thought is to allow alliance robots to literally "communicate". Perhaps another radio channel, I don't know. I think that if you had a communication between the robots with something like... maybe position or location of that mobile object. So, say the object is randomly moved around each match during autonomous... both can search for it and when one finds it relay to the other "Hey, it's somewhere in this area, here's my location and where I'm looking with my IR beacon" and the other robot can reroute it's autonomous to go to that area. Perhaps even knowing the location of ALL robots, whether on the same alliance or on the opposing. Robots could actually "think" and interact to the movements of the other teams and would have to constantly do a "What do I want to do." and at the same time the other robots will be doing that, so depending on how the robots react, each one will recalculate what to do... removing the "Wow, autonomous is repetitive and boring." scenario some people are talking about.

I know I just put in ALOT that would require a ton of designing and would probably be quite impractical in a money sense, but I've been thinking about it for awhile. Out of all of that, I've noticed with competitions and all my scenarios that a better positioning system is DEFINITLY needed to know absolute location by some type of system. The StangPS is a beautiful system for this, but it's just not practical for some designs. Like I said before, if we could make a like 3 "tower" system of radio beacons or something around the field to act like a GPS system... teams would be able to actually program after their robot is shipped and see their robot do what they told it to at competition... evening out the playing field for teams that can only build 1 robot.

Re: [Official 2005 Game Design] Autonomy Discussions
 

Our Programming team worked on ideas for autonomous ball release while the robot was constructed. By the time we had robot constructed, the team decided that dropping the balls in autonomous was not worth the possiblity of crashing headon into another robot. We decided to use the strategy of moving the goal and waiting for the balls to drop. At regional competition, we noticed that very few robots had reliable autonomous, so we decided to put more emphasis on dropping the balls in auto. We never really got it reliable at the Championships during practice and during qualification matches we only had one match where the strategy worked out with our partner caused us to attempt to release the balls during auto. Contrast this with the 2003 game where there was not a strategy that involved letting your opponent get to the top of the ramp first.

Re: [Official 2005 Game Design] Autonomy Discussions
 

I agree with those that say that currently autonomous mode is generally blah, with the exception of a few teams. To correct that and to make it more viewer friendly, at the very minimum make it worth something. It should be a strategic necessity to have a viable and funcitoning auto mode. Did someone say POINTS for extra effort?

However, in order for those teams that have limited capacity to achieve and effective auto mode, don't make it tremendously difficult.

I did like the idea of 2003 that required the human player to be in place at teh irght time to activate auto mode as well as 2004 mode activating some aspect of the game. Perhaps a combination of some sort. Just a thought.

Remember, FIRST is in a transition period. The game must be challenging enough for veterans and accomedating to rookies. Until there are seperate leauges/divisions there must be the balance. Hey, perhaps a seperate division for regional winners? Now there is and interesting idea for a new twist and discussion...........

Re: [Official 2005 Game Design] Autonomy Discussions
 

I have doubts about autonomy...


For instance this year, to make the game more enthusiastic to the fans and spectators, say a bot removes the ball in auto. as soon as the ball is released have drivers take control. not wait another 3-5 seconds for the balls to fall. Limit the time to 10 seconds. This way there is more excitement for outside viewers.
Not to say this year was an awesome year with teams like WPI(190), The Riot Crew(58), and teams like Buzz(175) had awesome auto. modes that were exciting to watch.

I think FIRST should step back and evaluate the good that auto is doing. Gauranteed It was always a Build/driver robot, and the controls teams usually did'nt have that much of a part in it. However, its been two years, and I have yet to see some major accomplishments in the programming.


just my $.02