Integral in Simulink fcn Block

Hey,

frist: I'm pretty new to Matlab / Simulink. I'm wrote an algorithm to generate a trajectory in Matlab and wanted to integrate my working Matlab code into a Simulink Project by using a Fcn Block.

Within my algorithm I have to integrate a cos / sin function. My code looks like:

ds = .1;
length = 1;
phi = 0;
R=5.5;
A=sqrt(R*length);

i=1;
int_x1=@(l) cos((l.^2/(2*A.^2)) + phi);
int_y1=@(l) sin((l.^2/(2*A.^2)) + phi);

x=zeros(1,(length / ds)+1);
y=zeros(1,(length / ds)+1);

for l=0:ds:length
    x(i)=integral(int_x1,0,l);
    y(i)=integral(int_y1,0,l);
    i=i+1;
end

This generates x and y coords in ds steps. It builds a Clothoid for a Radius of 5.5 within a length of 1.

My Problem with Simulink is, that the integral function does not work.

"Function 'integral' not supported for code generation."

In the next step I tried to fix this by the use of:

coder.extrinsic('integral');

which leads to:

"An error occurred while running the simulation and the simulation was terminated

Caused by:

• First input argument must be a function handle. "

Is there any possibility to numerical solving of integrals within a function block?

[Jim Riggs]

I'm not sure what you are trying to do in Simulink. Can you post an image of your Simulink model, and specify exactly what code is in the Fcn block?

[Björn Meyer]

Unfortunately I can't post an image of the Model. In general my part is to create a parking in / out trajectory (x,y coordinates for example).

In the actual fcn I'm getting Inputs from my model like gps coordinates of the car and where to park in / out.

In Matlab it looks like this, the trajectory is starting at X1000.9363 Y1062.6396 in a parking spot and ending in X1009.6008 Y1051.8693 on the road.

What I'm calculating in the code snippet above is the first bluish line from the start.

Answer 1

[Fangjun Jiang]

You need to understand the difference between MATLAB and Simulink.

This Simulink block diagram will give you the curve. Set the values of "A" and "phi" in base workspace.

[Björn Meyer]

Thank you for your answer, I'm trying to. My problem is that I need a combination of a straight, a curve, a radius, a doubled curve, another radius a a curve and a spline. The A depends on the Radius and the length of the curve. Phi depends on the Radius and A and is necessary for the start of the next curve or radius.

My problem is that i wrote the whole code in a Matlab function, because I thought I could just paste it in a Simulink Fcn Block. I did not know that Simulink does not support Matlab built in functions like integral(fun,a,b).

[Sam Chak]

The approach proposed by Jiang is a typical standard way to carry out the integration. If there are multiple dependent variables, then you will need to create the feedback loops manually and using the Mux block at the input port of the Fcn block.

Another way is to use advanced level method, the S-function.

[Fangjun Jiang]

That is right. This problem seems to be a little harder than this diagram illustrates. But one of the fundermetal point is that Simulink will do the "integration" for you, rather than in MATLAB, you have to specify the time, step and call integral() function and do for-loops.

Formulate your problem and construct the algorithm using blocks, let Simulink take care of the time, step and integration.

[Björn Meyer]

Thank you for your answers.

I'm still not sure if my problem can be solved by this way. Let me give you a little bit more details on my project and problem.

Right now we have two Simulink Models.

One for "offline / off-car" simulation, to evaluate problems and test new functions. This model is fed with data from a measurement recorded in the vehicle. The Simulation time is fixed to for example 30s, depending on how long the measurement record was.

The other model is used to control a car. For this, the model is calculated in C code and passed to a dSpace autobox. The model is then executed on the autobox and fed with live data from the vehicle bus system. The incoming data is manipulated and sent back to the bus.

Back to my problem:

Is it possible to use the simulink integration independent from the simulation time?

note: If I'm building a Model like you said with the parameters: start time 0.0, stop time 1, fixed step, step size .1 I'm getting the same output as within my matlab algorithm. Meaning an Integral from 0 to 1 with a step size of 0.1 between data points. But I have no idea what to do if I cannot change the start time, stop time or even a step size.

[Björn Meyer]

Added Information:

I cannot use the "interpreted MATLAB FCN" block, because it does not allow C-Code generation.

Do you have any other ideas how I could solve my problem?