unwanted variable change

Projects Programming
1

Hi all,

Im encountering a change of a variable i do not understand, probably due to a lack of programming experience.
For a ball on a beam project im trying to build a kalman filter in using Arduino Playground - MatrixMath, most is working fine, until i get to this piece of code:

Serial.println(a[0],10);

x_kalman = X00[0];
v_kalman = X00[1];

Serial.println(a[0],10);

in line 2 and 3 im assigning the result the previous code to x_kalman and v_kalman coming from a array X00[2].
this seems to work great, however in these two lines the array a[1] appears to change it's value.

the piece of code results in:

0.0018310546
0.0003591835

What is happening here? the first value is the correct value, how can a[0] change value?

Thanks,

2

What is happening here? the first value is the correct value, how can a[0] change value?

Without seeing your code, we'd only be guessing, but I'm guess that you are writing beyond the end of an array.

3

PaulS,

How would i check, if that is the case?

Thanks,

4

How would i check, if that is the case?

Check that every write to an array is writing within bounds. Post your code, like you were supposed to, and we'll help.

5

This is what i have so-far (total code is to large to paste):

void loop (){

a[0] = 2.000;//(float)analogRead(angle_read);
float x_meas = 15.0000;
//a[0] = calib_apow(a[0],3)+calib_bpow(a[0],2)+calib_c*a[0]+calib_d;

// Kalman filter----------------------------------------

// X01[1] = F[1]X11[1]+F[2]X11[2]+B[1]a;
// X01[2] = F[3]X11[1]+F[4]X11[2]+B[1]a;
Matrix.Print((float)H,1,N,"H");
Matrix.Print((float)F,N,N,"F");
Matrix.Print((float)B,N,1,"B");
Matrix.Print((float)Q,N,N,"Q");
Matrix.Print((float)X11,N,1,"X11");
Matrix.Print((float)P11,N,N,"P11");
// X01------------------------------------------------------
Matrix.Multiply((float*)F,(float*)X11,N,N,1,(float*)temp2);
Matrix.Print((float*)temp2,N,N,"temp2");
Matrix.Multiply((float*)B,(float*)a,N,1,1,(float*)X01);
Matrix.Add((float*) temp2, (float*) X01, N, 1, (float*) X01);
//Matrix.Print((float*)temp2,N,N,"temp2");
Matrix.Print((float*)X01,N,1,"X01");
Matrix.Print((float*)temp2,N,N,"temp2");

// P01------------------------------------------------------
Matrix.Multiply((float*)P11,(float*)F,N,N,N,(float*)temp33);
//Matrix.Print((float*)temp22,N,N,"temp22");
//Matrix.Print((float*)temp2,N,N,"temp2");
Matrix.Multiply((float*)temp33,(float*)F1,N,N,N,(float*)P01);
//Matrix.Print((float*)P01,N,N,"P01");
Matrix.Add((float*) Q, (float*) P01, N, N, (float*) P01);

Matrix.Print((float*)P01,N,N,"P01");

float temp2[N] = {0,0};
float temp[1] = {0};

//K--------------------------------------------------------
Matrix.Print((float*)temp2,N,N,"temp2");
Matrix.Multiply((float*)H,(float*)P01,1,N,N,(float*)temp2);
Matrix.Print((float*)temp2,1,N,"P01H");
Matrix.Print((float)H,1,N,"H");
Matrix.Multiply((float*)temp2,(float*)H,1,2,1,(float*)temp);
Matrix.Print((float*)temp,1,1,"P01HH'");
Matrix.Add((float*) temp, (float*) R, 1, 1, (float*) temp);
Matrix.Print((float*)temp,1,1,"P01HH'+r");
Matrix.Invert((float*)temp,1);
Matrix.Print((float*)temp,1,1,"inv P01HH'+r");
Matrix.Multiply((float*)temp,(float*)H,1,1,N,(float*)temp2);
Matrix.Multiply((float*)P01,(float*)temp2,N,N,1,(float*)K);
Matrix.Print((float*)K,N,1,"K");

//X00-----------------------------------------------------
//float temp2[N] = {0,0};
temp[0]=0;
temp2[0]=0;
temp2[1]=0;

Matrix.Multiply((float*)H,(float*)X01,1,N,1,(float*)temp);
Matrix.Print((float*)temp,1,1,"temp");
temp[0]=x_meas-temp[0];
Matrix.Print((float*)temp,1,1,"temp");
Matrix.Print((float*)K,1,N,"K");
K2[0] = K[0]temp[0];
K2[1] = K[1]temp[0];
Matrix.Print((float)K2,1,N,"K'");
Matrix.Add((float) K2, (float*) X01, 1, N, (float*) X00);
Matrix.Print((float*)X00,N,1,"X00");

//P00----------------------------------------------------
temp[0]=0;
temp2[0]=0;
temp2[1]=0;
temp33[0][0] = 0;
temp33[0][1] = 0;
temp33[1][0] = 0;
temp33[1][1] = 0;

Matrix.Multiply((float*)K,(float*)H,N,1,N,(float*)temp33);
Matrix.Print((float*)temp33,N,N,"temp33");
Matrix.Print((float*)I,N,N,"I");
Matrix.Subtract((float*) I, (float*)temp33,N,N, (float*)temp33);
Matrix.Print((float*)temp33,N,N,"temp33");
Matrix.Multiply((float*)temp33,(float*)P01,N,N,N,(float*)P00);
Matrix.Print((float*)P00,N,N,"P00");
x_kalman = X00[0];
v_kalman = X00[1];

;
delay (1000000);
// -----------------------------------------------------

dt=millis()-t0;
}

the error occurs @ "x_kalman = X00[0];" . Here a[0] changes for some reason

6

You can post your full code as an attachment. Please do so.

How to use this forum

7 
 float temp[1]  = {0};
...
  Matrix.Invert((float*)temp,1);

What is the point of an array with only one element? How do you Matrix.Invert a single float? I thought a matrix was ... well ... more than one thing.

8

Nick,

You are completely right, in this case useless.
I was hoping to end up with a more general form, in some cases this matrix is larger than 1 element.
But got a bit diverted from this goal trying to understand the MatrixMath code (and Arduino coding in general).
Thanks for reading it that thoroughly

Attached is the complete file

kalman.ino (6.61 KB)

9 
float a[0];
...
  a[0] = 2.000;//(float)analogRead(angle_read);

You are declaring the variable "a" to have no dimensions (effectively, it occupies no memory) and then you assign a value 2 to element 0, which is out of range.

Plus, this is confusing:

float temp[0];   // no elements
float temp2[N];

...

 float temp2[N] = {0,0};
 float temp[1]  = {0};

This is redefining a local variable "temp" which shadows the global variable "temp". The global variable has no elements.

10

PaulS:

What is happening here? the first value is the correct value, how can a[0] change value?

Without seeing your code, we'd only be guessing, but I'm guess that you are writing beyond the end of an array.

PaulS guessed correctly, it seems. :slight_smile:

11

That seems to be the source of my problems.
Got a bit confused in pointing to the correct position in the array (0= first position) and defining the length of the array ( 0 is none)

Thanks for the help!

12

mathematically 1x1 matrix is perfectly valid, inversion is same as reciprocal in this case.
A matrix library needs to support them in fact as the product of a 1xN row and Nx1
column matrix is a 1x1 matrix.

13

MarkT,

As you mention the code (inversion) off course still has a function.
In this specific case where the size of the array is hard coded to be 1, it could be done simpler (what i assume was the point of Nick)