Intro to Odometry in FTC | The Clueless

The Clueless FTC #11212, based in San Diego, CA, provide an introductory explanation to odometry within FTC.
Desmos Graph: www.desmos.com/calculator/osc...
Useful resources:
gm0.org/en/latest/docs/common...
• Intro to Odometry, Part 1
• What is Odometry? | An...
Tips For FTC Odometry
1) Build your odometry pods as sturdy as possible
- The mechanical design and components used can heavily influence your odometry's accuracy and performance. To mechanically minimize the amount of error, it is encouraged that your odometry modules are fabricated to precision, thus reducing the need for post-fabrication (calibrate your 3d printer/ CNC well). Post-fabrication, such as sanding down edges for your bearings to fit, can typically lead to wiggle room and increase the amount of potential accuracy error.
- Minimizing the parts used can often help decrease movement within your hardware components. For example, use one 3d printed encasing for your encoder + non-powered wheel instead of 2 plates + standoffs sandwiching your rotating wheel.
- Adding elastic compliance is another method to increase the robustness within your system. By using rubber bands or springs, you are able to ensure that your odometry module will always be in contact with the floor. This increases your odometry's data accuracy to your robot's movement.
2) Make sure they can spin freely and are not obstructed
- Minimize any friction between your non-powered odo wheel, encoder, and encasing, especially if the dead wheels are close to the center or in corners of the robot.
- When the dead wheels are located in the corners there is a component of force on them that points into the encoder, the rollers should help take care of this but when moving quickly or if sprung too hard this can result in these rollers locking up resulting in forces that can increase friction on the encoder.
3) Increase loop speeds & use a good approximation
- The faster the loop speed the more close the delta is to the differential.
You can increase loop speeds in a few ways:
- Only call a hardware component 1 timer per loop/don’t call it if you don’t use the data.
- Use BulkReads to call multiple devices at the same time gm0.org/en/latest/docs/softwa...
- Use analog and digital equivalents to I2C sensors (especially avoid the REV 2m Distance Sensor).
- Only update motors when they need to be updated
- Generally a constant velocity arc approximation is good enough for the 30 seconds of auto (In a good system expect 0.5 in of error). Road runner already uses this naturally.
- Try to minimize strafing, it makes your auto faster and increases accuracy.
4) Measure the Pod's Positions Accurately
- For the equations discussed in the video, multiple odometry values are needed such as the x / y distance between your horizontal or vertical odometry pod and the center of your robot. In order to get precise measurements, we suggest measuring in your robot's CAD software as a first pass. If your team does not use CAD software, we would suggest using precise forms of manual measurements such as calipers. If this doesn’t work just use a measuring tape.
- After you get a preliminary measurement try rotating the robot 10 times in place. Then you will need to adjust your track width by multiplying the left and right odometry wheels by 3600/numDegrees.
- We highly recommend that you follow these steps and find the experimental positions as opposed to just trusting your CAD because an inaccuracy as small as a few tenths of an inch can severely decrease accuracy when turning as the heading calculations will be inaccurate.
5) Know the Limits of your Odometry
- Although odometry provides a very accurate measurement for the location of your robot on the field, theoretically, error will always exist. It is best to design other subsystems in order to minimize the preciseness of odometry accuracy required.
- It is also useful to plan knowing the limits of odometry. Even the best odo systems expect about +- 0.5 in of error. If your design requires high levels of positioning to function autonomously consider adjusting this to be more forgiving.

[@pjwetherell9414]

Hey everybody. After looking more into the math behind constant acceleration arcs I don't think that I can make a tutorial that is understandable and concise. It is especially important to note that there is almost no benefit to a constant acceleration arcs tutorial over a constant velocity arc when you have fast enough loops (50 hz), because they already contribute such small errors in comparison with the physical build of the odometry wheels and the tuning of their positions.
Thanks for watching the video and I hope you enjoyed it, but this will probably be the last in the series.

[@user-em9ir8ek5b]

Considering that the loops are run that fast, I think that also applies to linear odometry. I think if you call the loops fast enough it still can be very accurate

[@velocityspider635]

dam i wish i had this earlier. best explanation of odometry and particularly constant velocity assumption ive heard. Great video

[@superarash6594]

Bro this is so good. My team was just about to start using 3 wheel odo for the first time this season and this was really helpful for understanding the basics
7:25 I am also very interested in the constant acceleration arcs tutorial

[@fasvi1285]

Great Video

[@ThatTutorGuy1]

This is great! It’s so awesome to get some advice from someone who’s done this a lot, most tutorials are too basic. This is our second year using odometry pods and roadrunner, and we have some pretty large errors still. Like we’re off 4 inches crossing the field and back just once. Reading your helpful tips in your videos description, is it possible that slow loop times are the issue? (We have very nice pods, using Loony’s design, so I don’t think it’s a mechanical problem.)
Any chance you could make a video about some common errors in loop speed? (You have some tips in the description, but they are a little over my head.) We already have 50pt auto and get some very high scores, but we are now ready to start cycling stacked pixels during auto, and it’s going to kill us if we always have to use the April tags. Thanks!

[@andeboyer]

The music is very distracting. And almost completely drowns out your narration in places. Please remove it.

[@mahjsa]

You should not have used the background music. It is extremely distracting. Moreover it is somewhat depressing, as if odometry is really difficult.