[Official 2007 Game Design] Autonomy And Other Technology Discussions

[GaryVoshol]

Originally Posted by Andrew Blair

The beautiful thing about autonomous this year, and in years prior, is that you didn't *have* to use the camera, or other sensor if you didn't want to or didn't have the resources.
You could always dead-reckon it if you had to, and I think that needs to stay. I love the advanced sytem autonomous points, but we should keep an easy couple of points in there for the rest of the teams.

In Triple Play, you could use dead reckoning to get the hanging tetra and/or use the one tetra the alliance was given to cap a goal. Yes, getting the vision tetra involved, well, vision.

The problem with auton in Triple Play was that the rewards were not enough. 26 points were theoretically possible - drop both hanging tetras, get the vision tetra on the goal in the middle of the field, and cap the center home-row goal to generate a row. But how many vision tetra cappings were there all season? In reality, your alliance was doing good to get 5 points in auton, a rather insignificant number in the whole game.

Aim High had a very meaningful auton reward - not only the 10 point bonus, but playing defense first. Winning auton put you in a very good position to win the match. Maybe the pendulum swung too far this year, but it certainly made auton an important feature in the game. And there were multiple things to do - high goal via either camera or dead reckoning, low goal, play defense.

We need an autonomous mode with real rewards for the teams that master it. There also has to be something for those with lesser programming skills to do. IMO, this year's game provided more of those kinds of options than Triple Play did. Hopefully future games will continue the trend.

[Peter Matteson]

I would like to see two to three completely different possible scoring methods for auton.

One that any team can easily do with dead reckoning. This would be the low hanging fruit that any team should be able toreach and worth the least points.

One that requires a level of sensor feedback to accomplish worth a moderate amount of points. This should be attainable with software out of the box so that teams who put the effort in can accomplish this.

The third should be a pie in the sky real challenge that has a corresponding bonus to make it worth the effort for the teams to do this.

I know that first has tried to do some of this over the last few years but I like the idea of challenging the veterans but still keeping things accessible to the rookies and mid-level teams. This tiered objective system would allow teams to work towards a goal if they saw it recur over a few years. Also you could illuminate each with a different color vision target like we saw on Einstien this year.

[Dave Scheck]

Adding on to what Pete brought up...what if there were 3 seperate start times? Maybe give the hard task 20 seconds to work with, the middle 10 seconds, and the easy 5 seconds? You could even penalize for interacting with the other scoring opportunities before your task is complete.

[Donut]

Originally Posted by Dave Scheck

Adding on to what Pete brought up...what if there were 3 seperate start times? Maybe give the hard task 20 seconds to work with, the middle 10 seconds, and the easy 5 seconds? You could even penalize for interacting with the other scoring opportunities before your task is complete.

Why not do it in reverse, and force the teams going after the harder task to do it in less time? Sounds like a better challenge to me.

I think FIRST has hit a good point with autonomous this year; namely that it matters in the game. Autonomous could help you get a few extra points the past few years, but truthfully 90%+ of all matches were completely unaffected by what happened in autonomous mode. Games have gone from "oh, we don't need to worry about autonomous, we won't be down by more than 3 points" to "we need someone to try to block them or we'll be starting down 15 points and they'll get to start on defense". Autonomous may have actually had too large a bonus attached to it this year, but I think it's much easier to lower the bonus a little than to continue trying to increase the importance of it and hoping it will finally affect the outcome of the game.

Autonomous is forcing teams to look at even more of an all around robot; you need good drivers, a well built robot, good programming, and some good alliance partners to win.

[Mike Shaul]

Originally Posted by dlavery

Uhhh, says who? .....

I think that limiting it only to students could hurt FIRST. Some teams have different resources. Some schools don't have any programing classes or students interested in programing. Our team has a limited interest in programing (most people want to build, its more exciting than sitting at the computer) but we use the opportunity to teach the interested students about the basics of programing, why its important, etc. Our students do write code for the robot but autonomous would be difficult without engineer involvement. Not to mention, it gives the engineers an opportunity to show why they went to college, what they have learned and how it applies to our jobs.

[Dave Scheck]

Originally Posted by Donut

Why not do it in reverse, and force the teams going after the harder task to do it in less time? Sounds like a better challenge to me.

I disagree...Let's apply some examples of a staggered start system to Triple Play.

I would make the easy task scoring the hanging tetra (either knocking it down or placing it on top). This is a relatively easy task that doesn't need a whole lot of time to accomplish. Look back to when the game was played, and you'll remember that there was quite a bit of downtime after teams hit the hanging tetra.

I would make the medium task be pick up off the autoloader (1 or 2) and score on the side goal. This is more complex than the previous task, but it can be done without fany vision/sonar/etc... systems. Everything you're trying to accomplish is at known positions on the field.

I would make the hard task to score the randomly placed vision tetra on the center goal. This would require some form of sensory system to seek out the vision tetra. Once it collects the tetra, the robot is in a semi-known position. The goal is in a known absolute position, but that would have to be combined with the current position, to determine where to drive next. This makes it more complex.

I would allot 20 second to the hard, 15 seconds to the medium and 5 (maybe 10) seconds to the easy. I think this cuts down on the amount of downtime in the match. Maybe you could go so far as to skip the first 5 seconds if nobody's attempting the hard task.

If you flip that around, you would have a robot that was halfway to picking up the vision tetra at the end of autonomous, and a robot that knocked the hanging tetra down and sat there for 15 seconds. I definitely think that the harder skill would need to be allocated more time.

I think FIRST has hit a good point with autonomous this year; namely that it matters in the game.

I definitely agree with this. Not only that, but large movement was required to accomplish the task, keeping things interesting. Triple Play's hanging tetras required such little movement to accomplish that sometimes you couldn't even tell that the robots even moved.

[Donut]

I don't think downtime is going to really be removed by that; sure some robots will start later, but you still have 15 or 20 seconds for the whole autonomous to run (and forcing teams to announce what they will be doing before autonomous will give away their goal, not something I want my opponent to know for their defensive autonomous).

If you keep them the same time then the difficult tasks are more difficult because it's more to do in the same time; by increasing the time for harder tasks you remove a large part of the difficulty from them (time is usually the biggest problem in autonomous mode).

[Dave Scheck]

Originally Posted by Donut

...and forcing teams to announce what they will be doing before autonomous will give away their goal, not something I want my opponent to know for their defensive autonomous.

Who said there was defense? The only reason that defense was so prominent this year is because the field was wide open and it was possible to have robot interaction. 2005 didn't have defense because robots couldn't interact (with maybe the exception of somebody blocking a middle row goal). 2004 didn't have defense because the task at hand was inherently offensive. 2003 was the same way...the only real way to defend other than to drive under the bar or chase stacks was to get to the stack first.

If you keep them the same time then the difficult tasks are more difficult because it's more to do in the same time; by increasing the time for harder tasks you remove a large part of the difficulty from them (time is usually the biggest problem in autonomous mode).

I agree that time is usually the problem, but why not allot just enough time for teams to actually succeed at the task at hand? In 2005, if there were another 2 or 3 seconds, I think that there would have been some center goal capping. I would much rather watch a game where teams are successful at doing something autonomously than watching a game where they almost get it done.

[Jherbie53]

This seems like the right area to talk about this, so here it goes. It would be nice to have more and different sensors in the KOP. I think if there are certain motors that are known to get a little warm if put under the wrong strain, then it would be nice to have heat sensors that you can attach to them for monitoring them. I'm not sure if hot motors are a big problem, but they do heat up pretty quickly when they are pushed to far. Also there could be a heat source in stead of a light for tracking during the game.

This is also a stretch, what about a small LCD screen that is used on the controls or something. This could replace some of the warning lights on the control panel. If theres enough room for memory, it could also be used for other things, like a heads up display with specific information about your robot, real time scoring, using a camera to look around, and other things I'm not thinking of.

I don't know how much these would add to the cost of the KOP, but they would be cool to have. It might be a couple of years before they are cost effective for FIRST to put in the KOP, but I will just have to wait.

[sanddrag]

It would be neat to have stereoscopic vision and advanced pixel comparison and image processing capability. Unless at intersections, you won't really find a nice green light out in the real world.

[Donut]

Originally Posted by sanddrag

It would be neat to have stereoscopic vision and advanced pixel comparison and image processing capability. Unless at intersections, you won't really find a nice green light out in the real world.

Some how I see a camera like that a little too expensive for the Kit of Parts. Not to mention since half or less of teams got the camera working with just a green light, it would probably be even worse dealing with a more complicated camera.

On the note of heat sensors, why not bring back the current sensors? I'm pretty sure they were included in the 2004 KOP. LCDs would be nice, but they'd need something so we could output effectively to it.

Defense was actually something I liked about autonomous this year. I want to watch the "superb" autonomous modes out there not only score effectively, but compensate for the interference of another robot and still score effectively.

In 2005 they should have increased the overall time a little, but that was by far the most difficult task for autonomous (this year had a stationary target). This year was good though. I think as long as a reasonable time is found for the task every year then things are good for difficulty.

[ahecht]

I want to see some easy way for teams to determine their exact position on the field. It would greatly expand the possabilities in autonomous mode for those teams that don't want to build an INS.

There are lots of options: a commercial Local Positioning System, fixed ultrasonic emitters around the field (used for triangulation), overhead camera data fed to the robots (ala RoboCup Soccer), a gradient on the carpet (ala RoboCup Jr), etc.

[Billfred]

Originally Posted by Donut

On the note of heat sensors, why not bring back the current sensors? I'm pretty sure they were included in the 2004 KOP.

While I also wouldn't mind seeing a set included in the KOP, I don't recall seeing them in the 2004 KOP.

This is going to sound a little silly, but knowing how many teams (mine included) like to take the "Who needs sensors?" approach to autonomous for whatever reason, would it be possible to get a smallish patch of carpet in the kit to test how the robot handles in that oh-so-critical first few feet of the program? The difference between standard-issue school tile floor and FIRST field carpet is somewhat significant.

[Lil' Lavery]

Keep autonomous big. Keep it REALLY big. This year was good, but it could be MORE. And KEEP robot interaction in auto.
While rookies and some teams may suffer from big auto modes, with auto modes that enable and even promote interaction by having big rewards and close starting positions you can create simpler auto codes that can alter the autonomous portion of the game. A simple "drive straight" code could change whether or not the opponent managed to hit a bunch of shots in the center, or none. It allows for basically a lesser function autonomous to acheive an important goal. You don't even need an advanced manipulator to accomplish it.
Another cool idea would to be follow FVC's example. Many people have talked about having auto portions of the field. FVC had two seperate games, autonomous and operator controlled. Your rankings in both games were averaged to form you overall seed, then you picked your alliance partners and played the operator controlled for the elimination rounds. The auto game also had slightly different rules (the field was divided into 4 sections, one for each team, center goal was worth 2 points, no ownership of goals, and only 30 seconds instead of 2 minutes). I would have liked to see it have the same rules though, with the exception of maybe staying 30 seconds instead of 2 minutes. Interaction and alliance partners would have made interesting strategy discussion, along with deciding which goals to pursue, as you couldn't adjust to the opponent's strategy afterwards. Scouting and pre-game strategy would become incredibly important.

[Donut]

I'm not sure what year the current sensors are from, I just know we have 2 current sensors sitting around from either 2004 or 2005.

How did you get an INS accurate enough to not be off by 2 meters in 15 seconds (this was what I was expecting to potentially get from the high end Analog Devices accelerometers)? You know, they could do that nice field position thing if they bring back those IR beacons from '04.