Hey all! We are trying to decide weather it is worth it to do a complex vision system for our robot, or rely on the FMS for the autonomous mode this year, as it gives us a ton of data. Any thoughts?
Think of it as this: you have two inputs as to which switch/scale side to go for. You have the vision targets (red/blue LEDs), or you have the FMS code. There are two factors in choosing which to go for.
**Time: **
You only have 15 seconds to do place your game piece(s), make the most of it. Why wait for your robot to process the vision data (1-2s) when the FMS tells you at the beginning of auto (instant)?
Reliability
What could go wrong in interpreting both inputs? The vision could be skewed with lighting, whereas FMS code could only fail if it glitches (probably rare).
I’m not a programmer, but if it’s not hard to interpret the FMS code sent to your robot, I wouldn’t even consider using vision.
My vote is for FMS, if you aren’t trying a multi-cube autonomous.
Just go thru FMS for auto .I reccomend some type of vision to see the cubes behind the switch while picking up cube because visibility might be hard to manual see if their a cube in your mechanism
FMS tell you which side you need to move to. Your robot will still need to know where it is in relation to the switch or scale to be able to score. If you don’t use vision, you’ll have to rely on dead-reckoning to be successful.
Problem I see is you really need to test code under real life conditions, to know if you have anything.
This year that is problematic, as you really won’t see FMS data until your first event. The timing of when you receive this data, will really be unknown. (aka https://www.chiefdelphi.com/forums/showpost.php?p=1363106&postcount=48 2014 auto timing issues)
If it is your plant to use the FMS data, I would also plan to be be successful if this data shows up a couple of seconds delayed.
Vision with the lights is also problematic, as I doubt team will have resources to do the vision with actual field lights, and the lexan reflections to accurately mimic the field.
It think you will be more successful with fms as it is defined as part of the game. If fms acts up and does not send this data, that is the definition of a field fault and on FIRST to fix. If you do vision, it is on you to fix and make work, which is pretty tough if you don’t have an accurate representation of a field. If you don’t play week 1, you can probably take your vision testing on practice day, and acquire images of a real field for testing.
For testing purposes, see the screenshot below so you can test the FMS sending you the data.
In Build Season, as well as at other times, there’s another way to ask this question - rather than decide between an easy and more difficult option, think about this: if we choose the easy option, what else can we accomplish in the remaining time? If we choose the more complex option, what are we giving up in the time it will take?
Absolutely process the FMS data and use it to make the primary decision as to where you’ll place your first auto cube! (reasons given above)
If you use the vision, it should be to direct you to the targets (switch, scale, spare cubes, whatever) that you want to drive to, or (if you’re really ambitious) avoid collisions.