/sci/ - Science & Math » Thread #12340374

1MiB, 1836x2568, embedded bullshittery.jpg

View Same Google iqdb SauceNAO

Anonymous Sat 14 Nov 11:27:38 2020 No.12340374 View Reply Original Report

Quoted By: >>12340390 >>12340393 >>12340409 >>12340508 >>12340549 >>12340571

helo /sci/

i've come to you in a time of great need.

i need to solve a linear system of equations, but i am a dum dum.
i have 7 variables and 3 equations that are linearly independent
i need a solution or an aproximation where all variables are between [0-1]
algorithms are fine as long as i can run it inside a 48Mhz ARM0 embedded bullshittery in a reasonable amount of time.

wat do

ps: any other quick method for vectoring thrust with multiple motors is also appreciated

Anonymous

Anonymous Sat 14 Nov 2020 11:34:12 No.12340390 Report

Quoted By: >>12340403

>>12340374
You could probably just use linear algebra to row reduce it. Then parameterize the extra variables and turn each variable into it's own equation.

Anonymous

Anonymous Sat 14 Nov 2020 11:35:23 No.12340393 Report

Quoted By: >>12340403

>>12340374
You can't solve a system of linear equations when you have seven variables and only three equations.

The vector space of all possible solutions will still be four-dimensional.

Anonymous

Anonymous Sat 14 Nov 2020 11:38:06 No.12340403 Report

Quoted By:

>>12340390
thanks anon, indeed i can; but every variable is defined in terms of other variables and i have absolutely zero clue how to constraint these between [0,1]

>>12340393
there are infinetely many solutions, i just need one where all variables are between [0,1]

Anonymous

Anonymous Sat 14 Nov 2020 11:42:16 No.12340409 Report

Quoted By: >>12340413

>>12340374
numerical constrained optimization using some scipy or something. Pray that the local minimum is a decent one

Anonymous

Anonymous Sat 14 Nov 2020 11:45:17 No.12340413 Report

Quoted By: >>12340420 >>12340430

>>12340409
thank you, beautiful approach anon, but isn't elligible for time critical embedded systems, especially ones with no fpu ;_;

Anonymous

Anonymous Sat 14 Nov 2020 11:48:50 No.12340420 Report

Quoted By:

>>12340413
yeah you have to change approach if you want speed and the process is not monitored

Anonymous

Anonymous Sat 14 Nov 2020 11:54:00 No.12340430 Report

Quoted By: >>12340435 >>12340440

>>12340413
just use an interior point method then? your system has really small dimension, you could find an exact solution in like cubic time or whatever.

Anonymous

Anonymous Sat 14 Nov 2020 11:56:43 No.12340435 Report

Quoted By: >>12340440

>>12340430
woops I meant the simplex method not interior point. you can definitely do this.

Anonymous

Anonymous Sat 14 Nov 2020 11:58:52 No.12340440 Report

Quoted By:

>>12340430
>>12340435

thanks very much anon!
will try to implement, hope i can get it to work!

Anonymous

Anonymous Sat 14 Nov 2020 12:42:33 No.12340508 Report

Quoted By: >>12340512

>>12340374
Make your life easier and teach yourself Laplace transforms.

Anonymous

Anonymous Sat 14 Nov 2020 12:45:47 No.12340512 Report

Quoted By:

>>12340508
elaborate pliz.

i already use laplace transforms for control theory, how can i solve this under given constraints using laplace transforms?

Anonymous

Anonymous Sat 14 Nov 2020 13:16:29 No.12340549 Report

Quoted By: >>12340607

>>12340374
get yourself more equations
formulate the solving algorithms yourself
also you could just present us your equations and variables and also explain what meaning they have
>quick method for vectoring thrust with multiple motors
explain what exactly you want to do might me helpful

Anonymous

Anonymous Sat 14 Nov 2020 13:34:51 No.12340571 Report

Quoted By:

>>12340374
try to find estimation correlations in which those unknown vars are used. --> more equations

or

as you know your vars are all between 0-1 you could do a brute force thing.
reduce your equation system to 4 equations and 4 unknown vars and and write the equations implicitly (<all on one side> = 0).
(now if the result would be 0 this equation would be fullfilled)

you have 4 vars between 0 and 1. lets start with [0.0001, 0.1, 0.2, ... 0.9, 0.9999]
so you have 11 versions of the first var and 11 versions of the second and so on.

11x11x11x11 = 14.600 possibilities in total

look for the line where all results are as near as possible to 0.
no need to search on your own, just do a calculation.
summarize all absolute values (which would ideally be 0) and look for the smallest total value.
get the four var values for this case.

if this is not satisfactory you can refine the value range[...] or define one variable and then refine the value range for the rest.

Anonymous

Anonymous Sat 14 Nov 2020 14:00:26 No.12340607 Report

Quoted By:

>>12340549
underwater drone. it has motors that look all kinds of directions. i want to move according to a given vector (ie <1,1,1>, right upward direction) of course i use unit vectors.

how much thrust should i give to each motor to accomplish this? and how to generalize this.

Capcode	All Only User Posts Only Moderator Posts Only Admin Posts Only Developer Posts
Show Posts	All Only With Images Only Without Images
Deleted Posts	All Only Deleted Posts Only Non-Deleted Posts
Ghost Posts	All Only Ghost Posts Only Non-Ghost Posts
Post Type	All Only Sticky Threads Only Opening Posts Only Reply Posts
Results	All Grouped By Threads
Order	Latest Posts First Oldest Posts First

Your latest searches