Intro to Odometry, Part 1

Рет қаралды 13,672

Күн бұрын

Intro to Odometry, Part 1
In this video, we systematically derive and develop intuition for general odometry processes, as well as the linear odometry assumption.
Song: gymnopedie-no-1 - Erik Satie

Пікірлер: 37

@neowang9033 4 жыл бұрын

Grant Sanderson is that you??

@punchster289 4 жыл бұрын

I wish

@joeym9169 3 жыл бұрын

Me 3 minutes in: This looks easy enough 6 minutes in: ok this is getting complicated 9 minutes in: I'm just gonna guess the robots distance

@EXPLODINGETDOOD 2 жыл бұрын

Really insightful video... Good job, love the graphics Edit: So where's the other 3 videos... I understand this must have been very difficult and time consuming, but this video was amazing. Very high quality...

@Comedy4 Жыл бұрын

Great video!?!!! I loved learning how to rotate things

@lemrabottabdelaziz4357 3 жыл бұрын

Amazing !!! Please do the other parts

@harshchaudhary743 3 жыл бұрын

It will be great if you make whole series on odometry

@yotamdubiner2545 3 жыл бұрын

gymnopedie no 1 is sooo good.

@vedantneema 3 жыл бұрын

Part 2 pls this is too high quality

@Thatguy-fg4tc 3 жыл бұрын

Love you ❤️

@theclueless11212 4 жыл бұрын

Thanks for this, it's really well-produced, goes through the content clearly and in a good progression. If you have the voice I couldn't tell it apart from 3blue1brown. I do have one question about 8:23, the video says "Rotate the approximation (of the displacement vector) to the global heading" does it mean to rotate the displacement vector by the change in the global heading? In either case, from the animation (which is really nice btw) I can't figure out what controls the rotation of the displacement vector. Especially this move at 8:58, the robot's heading barely changes, yet there is a huge change in the displacement vector's direction.

@punchster289 4 жыл бұрын

To rotate to the global heading is to undo the effect of the robot's overall rotation, not its change in rotation. Thanks for pointing that out. I should've been more clear when I said "changes in these values", as to only refer to the translation values.

@theclueless11212 4 жыл бұрын

@@punchster289 Thanks, the math is confusing but after studying the animation I think I get it. I'll definitely try out this approach with the real robot!

@theclueless11212 4 жыл бұрын

@@punchster289 I've just tried it and it seems quite accurate as long as the robot's final destination doesn't have a different orientation. However when the robot does have a different final heading, there is some error, I think because the odometry wheels aren't always at the center of rotation. We're running a 3 wheel odometry setup like in the video, so if the robot turns in place 90 degrees, the strafe/horizontal odometer returns a positive reading (while the forward/backward odometers cancel out). As the robot's global heading changes, the translational vectors rotate from 0 to 90 degrees, and the robot ends up thinking it shifted 45 degrees diagonally. Do you have a solution for this?

@punchster289 4 жыл бұрын

@@theclueless11212 Did you subtract off the rotation component from the strafe wheel? If you don't, it will think a rotation is a strafe.

@punchster289 4 жыл бұрын

@@theclueless11212 Also, the center of rotation doesn't have to be the center of the odometry wheels. I drew it like that for simplicity sake, but the only thing that matters is that the left and right wheels are parallel, and that the strafe wheel is perpendicular. In the case where they're not centered on the center of rotation, you can still test for accuracy by moving the robot such that it ends up in its starting pose, and observing how close to 0 the measurements are. To adjust its center, simply add a vector representing the offset between the centers in the local heading rotated to the global heading. You can find this offset by measuring your robot directly. Keep in mind that the position you want to increment onto is the one corresponding to the center of the odometry wheels.