Can't find where caused this "Unable to compile data properties" error in Simulink

Hello, I was trying to run this code script in Matlab Function in Simulink, however I encountered the Model error and it showed that error in the figure and didn't tell me where it caused. After I preliminary debugging , I think the problems will be in the parts of "path", but the code script runs well in the Matlab Workspace, so I don't know where caused the error. May you help me to check please. Thank you a lot!

Below is my code script in Matlab Function1

function path_adjusted = fcn(waypoints, velocity, window_length, P_N, P_E)

persistent current_index last_index path;

if isempty(current_index)

current_index = 1;

end

if isempty(last_index)

last_index = 0;

end

if isempty(path)

path = zeros(2*window_length, 2);

end

distance_threshold = 30;

poly_degree = 3;

regulator_weight = 0.5;

path_raw = zeros(length(waypoints), 2);

path_adjusted = zeros(2*window_length, 2);

if last_index > 0

last_window_point = waypoints(last_index, 1:2);

distance_to_last_point = norm([P_N, P_E] - last_window_point);

if distance_to_last_point >= distance_threshold

path_adjusted = path;

return;

end

end

start_index = max(current_index, last_index + 1);

end_index = min(last_index + window_length, length(waypoints));

% Check if all waypoints are already processed

if start_index > length(waypoints)

path_adjusted = path;

return;

end

waypoints_window = waypoints(start_index:end_index, :);

velocity_window = velocity(start_index:end_index, :);

time_window = linspace(0, 1, size(waypoints_window, 1));

P_ = zeros((poly_degree+1), (poly_degree+1));

qx = zeros((poly_degree+1), 1);

qy = zeros((poly_degree+1), 1);

for j = 1:size(waypoints_window, 1)

t = time_window(j);

Ti0 = [1; t; t^2; t^3];

Tie0 = Ti0 * Ti0';

Ti1 = [0; 1; 2*t; 3*t^2];

Tie1 = Ti1 * Ti1';

P_ = P_ + Tie0 + Tie1;

qx = qx + -2*waypoints_window(j, 1)*Ti0 + -2*velocity_window(j, 1)*Ti1;

qy = qy + -2*waypoints_window(j, 2)*Ti0 + -2*velocity_window(j, 2)*Ti1;

end

t_last = time_window(end);

t_init = time_window(1);

% Combining the Regular Term

Tir_last = [0 0 0 0; 0 0 0 0; 0 0 4*t_last 6*t_last^2; 0 0 6*t_last^2 12*t_last^3];

Tir_init = [0 0 0 0; 0 0 0 0; 0 0 4*t_init 6*t_init^2; 0 0 6*t_init^2 12*t_init^3];

Tir = Tir_last - Tir_init;

P_ = P_ + window_length*regulator_weight*Tir;

P = 2*P_;

P = 0.5*(P+P');

initial_guess_x = [waypoints(1, 1); 0; 0; 0];

initial_guess_y = [waypoints(1, 2); 0; 0; 0];

% Solve QP

options = optimoptions('quadprog', ...

'Algorithm', 'active-set', ... % Use 'active-set' explicitly

'Display', 'off'); % Suppress output for real-time compatibility

coeff_x = quadprog(P, qx, [], [], [], [], [], [], initial_guess_x, options);

coeff_y = quadprog(P, qy, [], [], [], [], [], [], initial_guess_y, options);

target_pos_poly_x = coeff_x(1) + coeff_x(2)*time_window + coeff_x(3)*time_window.^2 + coeff_x(4)*time_window.^3;

target_pos_poly_y = coeff_y(1) + coeff_y(2)*time_window + coeff_y(3)*time_window.^2 + coeff_y(4)*time_window.^3;

path_raw = [target_pos_poly_x(:), target_pos_poly_y(:)];

if last_index > 0

path = circshift(path,-window_length);

path(window_length+1:end, :) = path_raw;

else

path(window_length+1:end, :) = path_raw;

end

path_adjusted = path;

last_index = end_index;

if end_index < length(waypoints)

current_index = current_index + 1;

end

end

Answer 1

[Githin George]

Hello Cheng,

I suspect that the issue is arising because the size of the output signal of for your “MATLAB Function Block” cannot be determined at compilation time. Notice that “window_length” is an input to your function which means that it is a Simulink signal which may change during the simulation thereby affecting the size of your output signal “path_adjusted”.

See the line:

path_raw = zeros(length(waypoints), 2);

I tried to reproduce the issue with the following setup:

Model Init Code (I‘ve assumed some random data here):

% Define waypoints as an Nx2 matrix, where N is the number of waypoints
waypoints = [
    0, 0;
    10, 5;
    20, 15;
    30, 25;
    40, 35
];

% Define velocities as an Nx2 matrix
velocity = [
    1, 0.5;
    1, 0.5;
    1, 0.5;
    1, 0.5;
    1, 0.5
];

% Define other parameters
window_length = 2; % The number of waypoints to process in each window
P_N = 0; % Current North position
P_E = 0; % Current East position

Model Setup: See ModelSetup.png

Notice that after making some changes to the output variable configuration I was able to make the function work.

Default Output Signal Configuration: See DefaultConf.png

Since, “window_length” is set to 2 and my model has a “Constant” block I know that the Signal Size is [4 2] for the output “path_adjusted”. But the compilation fails with errors if I do not set the “Variable Size” property to “true”.

Modified Output Signal Configuration: See ModifiedConf.png

Note that after turning the “Variable Size” checkbox to “true” there are no compilation errors, and I get the output as shown in image 1. To learn more about Variable Size Signals, refer: https://www.mathworks.com/help/simulink/ug/variable-size-signal-basics.html

Based on this information, you may to want to consider modifying your algorithm, or go about the modelling in a different way assuming that “window_length” may change like a parameter during the simulation.

[Cheng Huan]

Greeting Githin,

Thank you for your answering. After I followed your steps and modified my algorithm, it successfully compilted!!! I really appreciate you! I can move forward my research now. Hope all your researches go successful.

Best wish