Hacker News Reader

Comment by incognito124 | Hacker News Reader

This was an invaluable resource 5 years ago when I was working on a summer research project making a ray tracing-based optical levitation simulator - initially it felt a bit insane to try to deeply understand this obscure bit of maths to implement rotations, but once it clicked it clicked. Quaternions ended up being a super neat formalism for writing and computing rotational equations of motion.

https://github.com/jdranczewski/optical-levitation-raytracin... for my repo, and https://onlinelibrary.wiley.com/doi/abs/10.1002/nme.5165 for the rotational dynamics with quaternions.

Visualizing quaternions (2018) - https://news.ycombinator.com/item?id=38043644 - Oct 2023 (42 comments)

Visualizing quaternions: an explorable video series (2018) - https://news.ycombinator.com/item?id=31083042 - April 2022 (15 comments)

Visualizing quaternions: An explorable video series - https://news.ycombinator.com/item?id=18310788 - Oct 2018 (32 comments)

FYI - if you're like me, you just looked at the page thinking it had some links to 3blue1brown videos. Not obvious that those are thumbnails link to interactive video apps. Very cool.

wow love the fact that you can interact with the videos

That's what's most valuable to me too. If I remember correctly, Ben Eater (8bit CPU from scratch guy) and Grant Sanderson (3b1b guy) tried to do a little experiment on education with this guided-yet-interactive learning material back in 2018 (!!). As someone working in EDU sector of tech, I'm constantly amazed by what people come up with in order to explain things to other people (or even themselves). Whatever the experiment was, I'd say it was a success, and I hope we get to see more such material on other topics in future.

I still gotta grasp this for IMUs

In general, the better sensor fusion algorithms incorporate kalman filtering.

https://www.olliw.eu/2013/imu-data-fusing/#chapter23

Best of luck =3

Yeah something else I need to do. lazy question, if I have an IMU that is swaying around as I try to move in a linear direction (e.g. Forward) is that something this kind of filtering would be used for? Regarding displacement estimation.

Edit: I get fusion is regarding multiple sensors

Normally, there are several types of sensors available, but most use 9DOF packages (3-axis gyroscope, accelerometer, and magnetometer)

gyroscope: fast over-sampled low-pass filter, but slowly drifts compounding heading errors

accelerometer: relatively stable, but dead-reckoning errors compound quickly

magnetometer: best stability, but low-sample rate and vulnerable to metal/magnets fooling/blinding the sensors

The fusion algorithms usually weights which data is consistent with the motion path, and attenuates the estimated pose errors.

Notably, not all sensors are equal quality, but there are probably better options now. =3

What a beautiful and high quality presentations. All the praises to the author.

Reply on news.ycombinator.com