Team 254 Presents: Skyfire

[Jay Meldrum]

Originally Posted by JABot67

My opinion on vision targeting is that it's something to do if your programmers have nothing else to do. Even so, most years it's a waste of time. This year I was so sure that vision targeting would be necessary, and it turns out it is not.

I would not say it is necessary, but to reach a high shot accuracy, camera auto aim is a must.

We tried having our driver (who is quite good) aim it himself. It took much more time, and yielded much less accurate results then using the auto aim. This could be very bot specific, but this is how it has worked out for us.

Looking forward for the reveal video guys!

[Travis Hoffman]

Originally Posted by jsmeldru

I would not say it is necessary, but to reach a high shot accuracy, camera auto aim is a must.

We tried having our driver (who is quite good) aim it himself. It took much more time, and yielded much less accurate results then using the auto aim. This could be very bot specific, but this is how it has worked out for us.

You don't need a camera to have extremely high accuracy (but it certainly helps increase confidence in ramping up throughput).

If you have a repeatable shooter, then pick your hot spots on the field, line up, and execute the shot the same way every. single. time. Practice makes perfect.

Every shot our robot took in the Wisconsin finals was manually aimed, with no camera present on the robot. We hit every shot we took and made at least 5 shots in each match. Tyler, our pilot, found an aiming method in the semis that is very intuitive for him, one he is eager to continue using. Now he just has to collect ammo better/faster so we can fire more salvos. I'm actually impressed by what he can do with our *piddly* little single-wide inside the frame perimeter collection method, but I know we can do better - guess what we're working on this week?

A skilled and practiced drive team (I can't imagine how much practice time 254's crew has been getting at home!) operating a machine with the potential capability like 254's could do even more damage in full manual due to faster/easier collecting and shorter times between salvos (provided the ammo is there to collect). I have no doubt that 254's machine can be killer when manually aimed.

Now develop this kind of manual driver skill and then add a layer of automatic aiming control on top of it, with the ability to quickly bypass the system during any malfunctions? Then you are truly deadly.

Think of it as Top Gun for robots *flashbacks to 2004 IRI talent show*. Don't rely upon missiles only. Learn how to dogfight and maneuver for a shot.

[Jay Meldrum]

Originally Posted by Travis Hoffman

Now develop this kind of manual driver skill and then add a layer of automatic aiming control on top of it, with the ability to quickly bypass the system during any malfunctions? Then you are truly deadly.

100% agree with this statement

Originally Posted by Chickenonastick

254 isn't just good at building their robot. They're also good at taking pictures of it.

Totally thought it was a rendering as well. Crazy.

[Tom Bottiglieri]

I can give a bit more insight into how our software process works.

For the most part, we try to find mechanical solutions to everything. Once we get our robot to the point where extra mechanics and driver practice starts to result in diminishing returns, we start adding automation on top. We try not to do anything in software that the driver can't do on their own, in case of failures. It would be terrible for our camera to come unplugged during finals and be a sitting duck because of it. Sure, the software will make it faster, but that's just gravy once you have a properly trained operation crew.

We do have auto aim on our robot. We use it in autonomous and the driver has a button to activate it in teleop. To tell you the truth though, he hasn't really needed to use it. Between intuition, practice, and a live feed from the camera back to the DS, auto aim doesn't save enough time to justify the risk of practicing with it. We plan on improving this part of the code as time goes on and maybe our strategy will change, but for now it makes sense to give more practice cycles to the drivers than the programmers.

[trilogy2826]

Originally Posted by jsmeldru

I would not say it is necessary, but to reach a high shot accuracy, camera auto aim is a must.

Myself and another mentor on 2826 argued this on several occasions leading up to the Wisconsin regional. Since on our practice field, vision targeting was working perfectly, it seemed a moot point.

Upon arriving at WI, we immediately had issues with our vision and the FMS interfacing in an unknown manner, forcing us to abandon vision tracking. Using the Axis camera and a well calibrated crosshair on the DS, we were able to score an average of 11 3 pt shots from the key through elims with a max of 13 made of 16 shot (last 3 missed accounted to driver pulling away as shots were fired).

Basically, I second that vision as an augmentation can be a benefit, but a well trained drive crew can pull off some amazing things when push comes to shove.

Good luck with Skyfire. The poofs are undoubtedly a role model for us all!