How to use 4d rotors

The most common way to represent goemetric algebra multivectors (including rotors) in a computer is via an array of coefficients, one for each canonical form algebra basis element (canonical basis blade) ie. for a 4D basis space you will have a 2^4 dimensional algebra and have 2^4 dimensional array of coefficients. An alternate but probably faster way to represent them is with a dynamically resizing list with each element containing an index to a blade and the coefficient of the associated blade. In this case the multiplication of two multivectors will only use non zero basis blades and so should be algorithmically cheaper and lighter on memory usage.

In terms of practical usage I found the easiest place to get started with playing around with geometric algebra is probably in python with https://github.com/pygae/clifford . Full disclaimer I use this library daily and contribute to it extensively. This library uses the flat array of coefficients approach. With this python library you can apply 4D rotors via the sandwich product and do reversion (inversion of a rotor) via the tilde operator:

# Create a 4D geometric algebra with euclidean metric
from clifford.g4 import *

# Create a rotor
R = layout.randomRotor()

# Create a vector to rotate
V = layout.randomV()

# Apply the rotor to the vector
V2 = R*V*~R

The specific definition of the geometric product and reverse for multivectors from an N-dimensional geometric algebra can be found in Chapter 4 of Geometric algebra for Physicists by Chris Doran and Anthony Lasenby.

A good C++ GA reference implementation for N-dimensional GAs using the list of elements approach can be found in Leo Dorst's book Geometric Algebra for Physicists or on his website: http://www.geometricalgebra.net/code.html . In general this is a great resource for GA, especially the conformal model and numerical implementations and concerns.

I was able to use Rotors after learning more on the subject thanks to this youtube series about Geometric Algebra : https://www.youtube.com/watch?v=PNlgMPzj-7Q&list=PLpzmRsG7u_gqaTo_vEseQ7U8KFvtiJY4K

It's really well explained and I recommend it to whoever wants use geometric algebra.

If you already know about Quaternions multiplication, Rotor multiplication won't be any different, and the i, j, k units of quaternions are analog to the basis bivectors of Geometric Algebra : e12, e13, e23 (or e01, e02, e12)

So a Rotor in 4D will be (A + B*e12 + C*e13 + D*e14 + E*e23 + F*e24 + G*e34 + H*e1234).

A table showing how to multiply those units can be found on this page : http://www.euclideanspace.com/maths/algebra/clifford/d4/arithmetic/index.htm

To get the gist of it, consider the 2D Rotors.

They're of the form: R = A + B*e12

Now, if we compute product between 2 arbitrary rotors R_1 and R_2, we get:

R_1*R_2 = (
  R_1.a     * R_2.a
+ R_1.a     * R_2.b*e12
+ R_1.b*e12 * R_2.a
+ R_1.b*e12 * R_2.b*e12 )
// but: e12*e12 = e1e2e1e2 = -e1e2e2e1= -e1e1 = -1
// this is confirmed by the computation rules I linked above
=
( (R_1.a * R_1.a - R_2.b * R_2.b)
+ (R_1.a * R_2.b + R_1.b * R_2.a) * e12 )

So in code, you would do something like :

R_3.a = R_1.a * R_2.a - R_1.b * R_2.b
R_3.b = R_1.a * R_2.b + R_1.b * R_2.a

Now it's only a matter of doing the same with those big 4D rotors, applying the multiplication rules for dimension 4 as linked above.

Here is the resulting code (using e0, e1, e2, e3 as basis vectors):

e: self.e*other.e - self.e01*other.e01 - self.e02*other.e02 - self.e03*other.e03 - self.e12*other.e12 - self.e13*other.e13 - self.e23*other.e23 + self.e0123*other.e0123,
e01: self.e*other.e01 + self.e01*other.e - self.e02*other.e12 - self.e03*other.e13 + self.e12*other.e02 + self.e13*other.e03 - self.e23*other.e0123 - self.e0123*other.e23,
e02: self.e*other.e02 + self.e01*other.e12 + self.e02*other.e - self.e03*other.e23 - self.e12*other.e01 + self.e13*other.e0123 + self.e23*other.e03 + self.e0123*other.e13,
e03: self.e*other.e03 + self.e01*other.e13 + self.e02*other.e23 + self.e03*other.e - self.e12*other.e0123 - self.e13*other.e01 - self.e23*other.e02 - self.e0123*other.e12,
e12: self.e*other.e12 - self.e01*other.e02 + self.e02*other.e01 - self.e03*other.e0123 + self.e12*other.e - self.e13*other.e23 + self.e23*other.e13 - self.e0123*other.e03,
e13: self.e*other.e13 - self.e01*other.e03 + self.e02*other.e0123 + self.e03*other.e01 + self.e12*other.e23 + self.e13*other.e - self.e23*other.e12 + self.e0123*other.e02,
e23: self.e*other.e23 - self.e01*other.e0123 - self.e02*other.e03 + self.e03*other.e02 - self.e12*other.e13 + self.e13*other.e12 + self.e23*other.e - self.e0123*other.e01,
e0123: self.e*other.e0123 + self.e01*other.e23 - self.e02*other.e13 + self.e03*other.e12 + self.e12*other.e03 - self.e13*other.e02 + self.e23*other.e01 + self.e0123*other.e,