Autonomous Strategies

[goodsky]

Due to the severe lack of bonus points in autonomous this year, we've decided to push most of our programming resources to assisting the drivers in teleoperated mode. So Autonomous will just be moving in the most unpredictable paths (away from opposing human players).

[mynameisbob]

We will be lucky if we even have an autonomous mode, therefore we will be happy to move anywhere from where we start.

[Booksy]

Originally Posted by mynameisbob

We will be lucky if we even have an autonomous mode, therefore we will be happy to move anywhere from where we start.

You should at least do the drive straight and crash version, otherwise you're a sitting duck, right at the other team's outpost.

[ExarKun666]

Originally Posted by ComradeNikolai

If we go for "drive straight," I'll hope we get hit so I can try out our angle correction software.

Just straight, if I estimate the angle right you would hit into the outpost so you should throw a couple turns in there, maybe instead of just purposely getting hit just to test software, just use the angle corrector, but use another strategy and if you do other then straight, then you will be hit most assuredly!

[JesseK]

I'm really glad this thread finally evolved into a poll...

Our auton will seek to simply position ourselves in the correct spot for our strategy. As the regional progresses, we'll tweak it further for automatic target tracking. Since we could conceivably start in one of three configurations with one of four objectives, we're simply working on positioning ourselves for those objectives now and actually acquire the objectives later.

[TimCraig]

Originally Posted by PinionTwister

We use Labview, if our alliance partners are using C++ it will be difficult for us to assist in moving in Autonomous as we have no experience! Just my $.02

Systems don't have to be programmed in the same language to cooperate. I don't think you're going to be making major programming changes between matches. Programs written in VB and C++ can get along happily on a single PC and different kinds of computers running different operating systems and programs written in a multitude of languages can get along quite happily on the Internet. You're going to arrive at the competion with a catalog (hopefully) of possible autonomous maneuvers and you'll need to confer with your partners to decide with one is most suitable for that match.

[Tom Line]

If experience serves as any kind of example, the last thing on earth you want to be doing is trying to make any type of major code change between matches.

I suspect you need to come to the table with all your autonomous modes programmed, and work out who will do what from there....

Has anyone else considered training the camera to chase the blue shirts of the judges.... and firing balls at them? We want to impress them with our accuracy .

[Doug Leppard]

In the six years of playing FIRST I have always had a good autonomous mode plan by now. But there are too many unknowns at this point to have one figured out.

Biggest issue is how well do we move and how well can we turn.

Once we know some of that we will work on a good auton modes.

[XXShadowXX]

autonomous we will basically move and shoot. Our movement won't be complex unless somebody else wants it to be, other wise just get away from the human player, and score, 'shrugs', the firing part is hard enough

[Rick Wagner]

It seems to me that there are two competing defensive strategies in autonomous mode. On the regolith, the robots will accelerate at less than a tenth of a g, so a good payload specialist (PS) should be able to get two or three balls into the trailer before it moves out of range. One strategy is to drive straight away, but this leaves the trailer exposed in the best attitude for scoring by the opposing PS. It might be a better strategy to spin on center to put the trailer behind the robot as quickly as possible and then back away from the PS. IF the robot is five feet tall, it will block the trailer from all but the highest arc shots.

Spinning quickly on center is problematic, especially since the trailer wheels will be on carpet in the starting position. This maneuver might best be performed by first pulling the trailer up onto the regolith, and then spinning on center 180 degrees.

[Joshamuffin]

Some sort of randomizer would be nice. Opposing teams would not be able to predict where you might go. However, this randomization might leave you in an unstratiegic spot. I.E. crashed into a wall. A controlled randomizer would be alot better. Maybe some DIP switches could feed input as to where you start at so a properly controlled course could be randomized.

One way for this to be done would be an "on-the-fly" randomizer. This randomizer would determine direction and how long it will go in that direction when it is needed. The program could check approximate postion to where it might turn out to be when it completes the "turn," modifying the values as needed to turn out not to crash into the wall.

Another way would be to plan out so many turns in advance and attempt to take said course. The disadvantage is when it gets bumped it diverts the robot off-course. The on-the-fly version could have an accelerometer to detect when it gets bumped and re-randomize the course as needed After impact.

This, however, is a very intensive algorithm and would take alot longer to debug than a simpler autonomous mode. However, I will try to suggest this to the team and subsequently attempt to program it if they agree.

[Doug Leppard]

Originally Posted by Joshamuffin

Another way would be to plan out so many turns in advance and attempt to take said course. The disadvantage is when it gets bumped it diverts the robot off-course. The on-the-fly version could have an accelerometer to detect when it gets bumped and re-randomize the course as needed After impact.

This is what we will do. We will have various auto modes to select from and use them according to the need, all this planned ahead and decided by drive team.

In my experience 80% of the teams have extensive autonomous plans and only about 10% show up with auto modes that work well, mainly because it takes a working robot and a lot of time to perfect it.

[Jared Russell]

I think that we will see a LOT less autonomous robot scoring than this poll would suggest.

1. Shooting from a moving robot into a moving goal is a HARD problem to solve. It is doable, but expect to spend a lot of time solving this challenge that may have been better spent elsewhere. Many teams will want to sit still as they shoot, making them an easy target for human players.

2. There is no bonus for scoring in autonomous mode versus teleop. Your driver will almost certainly be a better shot (with whatever targeting software you give him) than the robot by itself. The driver knows "when to pull the trigger". The only advantage in autonomous mode is that you can victimize teams with nonfunctional autonomous modes (that just sit there). As the season - and the regionals themselves - carry on, there will be fewer and fewer of these machines.

3. The size of the goal relative to the ball means that even good shots will sometimes not end up scoring.

4. It is far easier to write an evasion program than to write a scoring program.

5. The irregularities of the orbit balls makes consistent shooting a difficult proposition. With poof balls, robots could hit the same square inch of target every time. With orbit balls (many of which will be broken), there's a lot more randomness.

That said, I am 100% sure that some teams will score in autonomous mode, and will do it well. But the plurality of teams responding to this poll, I fear, will discover the issues above.

[Doug Leppard]

In the past when your team alliance partner didn’t have an auto mode it don't hurt you, at least not much.

My fear this year is a teammate that does not have an auto mode will hurt you because for 15 seconds they just sit there and collect balls, maybe all 20.

So you could have a great auto mode but your other teammates don't move and you start the player driven time with up to 40 points scored against you.

[Jared Russell]

Originally Posted by Doug Leppard

In the past when your team alliance partner didn’t have an auto mode it don't hurt you, at least not much.

My fear this year is a teammate that does not have an auto mode will hurt you because for 15 seconds they just sit there and collect balls, maybe all 20.

So you could have a great auto mode but your other teammates don't move and you start the player driven time with up to 40 points scored against you.

Totally agree. I expect that our team, and other veteran ones, will quickly go from team to team after receiving the match list in order to install at least a rudimentary autonomous mode.