search

TASRobotics / RaidZero-FRC-2019

Competition code for 2019 season
1 stars 0 forks source link

Klein (Limelight Vision) #6

Open qsctr opened 5 years ago

qsctr commented 5 years ago

Are we using raspberry pi or phone or both?

bayntun commented 5 years ago

We're not sure yet, we're currently checking which method is fastest or easiest.

On Sat., Jan. 19, 2019, 23:54 Bretton, notifications@github.com wrote:

Are we using raspberry pi or phone or both?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/TASRobotics/RaidZero-FRC-2019/issues/6, or mute the thread https://github.com/notifications/unsubscribe-auth/AQTbPaPioZ2cXQz13z0DtXIc2IubLfXqks5vEz_PgaJpZM4aJMvQ .

Gediminas19 commented 5 years ago

Vision for retroreflective tape works, issue now is communication with roborio with networktables.

Gediminas19 commented 5 years ago

Vision for ball works in GRIP, able to get both position of largest ball and size now even when multiple balls overlap. This GRIP pipeline has not been tested on limelight yet though.

As with limelight in general, networktables doesn't work yet.

Gediminas19 commented 5 years ago

Vision for ball works on limelight now.

Networktables are even stranger; roborio creates a duplicate “limelight” table instead of accessing the existing one as seen in Shuffleboard.

Gediminas19 commented 5 years ago

Networktables works after reflashing limelight and matching hostname with network table name. Limelight info accessible from RoboRIO now.

qsctr commented 5 years ago

Yayyyyyyyyyyyyyyyyy

bayntun commented 5 years ago

Issue to be renamed to Klein now On Wed, Jan 23, 2019 at 11:59 PM Bretton notifications@github.com wrote:

Yayyyyyyyyyyyyyyyyy

— You are receiving this because you commented.

Reply to this email directly, view it on GitHub https://github.com/TASRobotics/RaidZero-FRC-2019/issues/6#issuecomment-456857638, or mute the thread https://github.com/notifications/unsubscribe-auth/AQTbPU04pCvYf6m0vG_LMBSwo0NzU9TNks5vGIblgaJpZM4aJMvQ .

qsctr commented 5 years ago

what i don't get it what's klein

bayntun commented 5 years ago

Shin yi renamed the limelight kaluza.

Kaluza and Klein are evil dogs to shin yi On Thu, Jan 24, 2019 at 12:37 AM Bretton notifications@github.com wrote:

what i don't get it what's klein

— You are receiving this because you commented.

Reply to this email directly, view it on GitHub https://github.com/TASRobotics/RaidZero-FRC-2019/issues/6#issuecomment-456872809, or mute the thread https://github.com/notifications/unsubscribe-auth/AQTbPVOeObWHL-NAkJ5tSM9k-3-QiSAeks5vGI_lgaJpZM4aJMvQ .

qsctr commented 5 years ago

lol ok

Gediminas19 commented 5 years ago

Given gyro angle, robot can quite accurately determine relative x y coordinates to vision target.

Gediminas19 commented 5 years ago

Further testing driving the robot around shows that it can very accurately determine its relative coordinates to a vision target given the Limelight info and the Pigeon angle.

qsctr commented 5 years ago

Moved the phone vision part into #15. This issue will be for limelight vision.

Gediminas19 commented 5 years ago

The robot is able to hermite spline to a orange ball with decent accuracy.

Gediminas19 commented 5 years ago

The robot is able to hermite spline to retroreflective tape targets quite accurately.

usymmij commented 5 years ago

im confused what does the limelight have to do with string theory

Gediminas19 commented 5 years ago

@qsctr @Kof223 What should I do if I made the Vision branch too early so there are no pathgen classes inside yet? I need to make methods outputting Point objects, but it would be weird if I just copied them in from outside.

EDIT: Question answered, I realized you can merge between branches.

bayntun commented 5 years ago

FYI, Klein had very little to directly do with string theory.

He was a mathematician with strong ties to extra dimensional geometry. Hence Kaluza-Klein theories and Klein bottles. The relevance to the limelight is that the limelight is learning about higher dimensions through its limited scope. Ie it's observing its two dimensional world and making 3 dimensional predictions based on our models.

Maybe we're all three dimensional limelights being controlled and programmed by for dimensional beings.

On Fri., Feb. 1, 2019, 17:13 Karakorum91, notifications@github.com wrote:

@qsctr https://github.com/qsctr @Kof223 https://github.com/Kof223 What should I do if I made the Vision branch too early so there are no pathgen classes inside yet? I need to make methods outputting Point objects, but it would be weird if I just copied them in from outside.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/TASRobotics/RaidZero-FRC-2019/issues/6#issuecomment-459656253, or mute the thread https://github.com/notifications/unsubscribe-auth/AQTbPVgHuaFfFxLujcCscXaHxTijkDo2ks5vJAUjgaJpZM4aJMvQ .

Gediminas19 commented 5 years ago

I'm getting nightmares from vision. I'm trying to see if SolvePNP with pose estimation can be better.

Gediminas19 commented 5 years ago

It appears people online have been using SolvePNP with some mild success. I will be attempting it soon. See this thread for more info.

Gediminas19 commented 5 years ago

I have tried SolvePNP. It seems ok for determining position but not amazing.

Gediminas19 commented 5 years ago

image

The limelight interface isn't good enough for good SolvePNP at the moment, so I am writing my own pipeline. By the look of things they should be upgrading their thing soon.

Gediminas19 commented 5 years ago

Further testing shows that SolvePNP can get position in camera coordinates well, but not world coordinates because rotation vector is inaccurate.

If these coordinates are used in conjunction with gyroscope angle, more accurate results can be obtained.

Gediminas19 commented 5 years ago

Oh my Limelight released a new image with onboard SolvePNP, so all my Java CV code for SolvePNP on the Roborio is no longer necessary.

At least the onboard code will be a LOT faster than running SolvePNP on the RoboRIO.

Gediminas19 commented 5 years ago

Also ball tracking is no longer using spline, but a crude PID system directly using vision coordinates of ball. This is working pretty well to be honest.

Gediminas19 commented 5 years ago

I have tested onboard SolvePNP on the Limelight, and it seems to work decently well.

Gediminas19 commented 5 years ago

Limelight has released an interesting way to approach the vision tape targets that seems more reliable than spline. I will test it soon.

I think this may even work for auto since we know the absolute position of the cargo ships, so we can continue splining afterwards with no qualms.

qsctr commented 5 years ago

@Karakorum91 are you still working on the vision branch?

Gediminas19 commented 5 years ago

Oh yea, I didn't commit my new stuff, oops.

qsctr commented 5 years ago

Ok, I was just making sure

Gediminas19 commented 5 years ago

Ball tracking works with the 2019 base.

LinDylan commented 5 years ago

someone create branch for me to upload pls <3

On Tue, Feb 26, 2019 at 11:32 PM Shin-Yi notifications@github.com wrote:

Ball tracking works with the 2019 base.

— You are receiving this because you were assigned. Reply to this email directly, view it on GitHub https://github.com/TASRobotics/RaidZero-FRC-2019/issues/6#issuecomment-467485049, or mute the thread https://github.com/notifications/unsubscribe-auth/Ae-4AJ7BDytV9lPsXMt93BbOIUwE17V8ks5vRVOSgaJpZM4aJMvQ .

qsctr commented 5 years ago

Make it yourself, it literally takes 3 seconds

LouisAsanaka commented 5 years ago

@Karakorum91 Could you please upload your new code?