Help building standalone exe from slx model that uses TCP/IP Receive Block from Instrument Control Toolbox

High level question up front:

Should I be using the TCP/IP Receive block from the Instrument Control Toolbox when building a standalone app from a Simulink model to communicate over ‘localhost’ (127.0.0.1) with another app built in C#/Unity or are there other alternatives I should consider? I only need my Simulink app to act as the TCP client and receive data only.

Details:

I’ve run into a confusing issue when trying to build a standalone app from a Simulink model that uses the TCP/IP Receive block from the Instrument Control Toolbox. I have successfully built a standalone executable using the “mcc” command and I can run the executable without issue on my (first) PC which has a full Matlab R2023a licensed install plus Matlab runtime installed. Now, the issue comes when I try to run the executable on another (second) PC that has the correct Matlab runtime version R2023a (9.14) but does not have matlab installed. I have been able to run standalone executables on the second PC (that don’t have the TCP/IP Receive block) with the same setup for a few months now so I’m confident the issue is with trying to run a standalone executable containing the TCP/IP Receive block.

On the second PC, I get the error when trying to run the executable as shown in the attached image. I tried to capture the error with a try/catch ME statement and print the ME text and got the text shown in the second attached image.

I inspected the C/C++ Code Generation documentation for the TCP/IP Receive block which states the following:

C/C++ Code Generation Generate C and C++ code using Simulink® Coder™. This block generates platform-specific code for the host machine's platform only (Windows®, macOS, Linux®). Set the Device vendor and Device type in Model Settings > Hardware Implementation. You must also select Support long long under Device details. You can only use a single conversion specifier for the ASCII format string parameter.

I didn’t know how to properly set the Device vendor and Device type but did try selecting “Support long long” but that didn’t fix the issue.

I’m also wondering if the second PC had a full matlab install if the executable would work? But why wouldn’t it work with just Matlab Runtime?

Lastly, I was trying to implement in Simulink a Matlab function block that calls tcpclient but was running into an error that “The current BuildConfig has not been set”.

[Brad]

For more context...

I am required by my employer to produce a standalone executable that acts as a TCP/IP client that receives data packets from a TCP/IP server sent as a uint8 vector from an external app built in C#/Unity. Both my simulink app executable and the C#/Unity app need to both run at the same time on a windows PC where they communicate over address 127.0.0.1 or localhost. The combined software package needs to be distributed to multiple customers around the globe. The C#/Unity app runs fine when users just drag and drop that package into the target windows PC I assume because it includes all required packages/libraries.

For more context we are designing a flight simulator where the aero/flight physics equations are done in simulink and we already have it setup to communicate with the C#/Unity app (which handles visualization of an aircraft) over UDP connection in a way where it works as a standalone app. The feature we are trying to add is to handle user triggered “events” over TCP such as when the user starts a session, pauses, resumes, restarts, and exits a session. I think as a workaround we can handle events over udp but in the long term for our system to be distributed to many customers and have a reliable event system, we would prefer to use TCP driven events.

I had also looked into using a matlab function block that calls the ‘tcpclient’ function but got an error when trying to run my model in simulink that states “The current BuildConfig has not been set”. Any idea how to address that?

Answer 1

[akshatsood]

Hi @Brad,

I understand that you are facing issues in executing the standalone application on your system. Prior to reflecting thoughts on the high-level question "Should I be using the TCP/IP Receive block from the Instrument Control Toolbox ...", it would be beneficial to address the message "Rapid accelerator process crashed while simulating the model" when trying to execute the standalone app on your second PC.

Debugging "Rapid accelerator process crashed while simulating the model"

First take on this message would be disabling the anti-virus software on your system to check whether it is blocking the process or not. Depending upon the anti-virus software in your system, you can either put it off and run the executable or add the executable to its permission list (if any for your anti-virus software) so that it is not excluded.

In reference to the screeshot attached, in where you are catching the MException, the "stack" property of "MException" object captures the Stack Trace information which if you could please share would be helpful in having a more clear view on the issue.

Selecting Device Type and Device Vendor

In respose to your confusion pertaining to selecting the appropraite Device Type and Device Vendo, please refer to the following documentation pages for a brief overview on the available options.

https://in.mathworks.com/help/simulink/gui/device-type.html

https://in.mathworks.com/help/simulink/gui/device-vendor.html

Will installing MATLAB fix the issue?

The documentaiton page attached below clearly states that "MATLAB® Runtime contains the libraries needed to run compiled MATLAB applications on a target system without a licensed copy of MATLAB." Hence, installing MATLAB might not be the one-shot solution to the problem.

https://in.mathworks.com/help/compiler/install-the-matlab-runtime.html

“The current BuildConfig has not been set” in MATLAB Function block

When you attempt to extract output from the MATLAB Function block, it participates in size and type propagation during which it does not provide the "BuildConfig", which includes settings that are sometimes crucial for code generation for "tcpclient", hence the error.

A possible workaround which might help address this could be by assigning all the output variables with their appropriate types before calling "tcpclient". For instance, if you have two ouptuts 'A' and 'B' from the MATLAB Function block, then before initiating a call to "tcpclient", make sure to define 'A' and 'B' with the appropriate data types. An example code snippet is as follows

% earlier implementation

function y = fcn()
    persistent client
    if isempty(client)
        client = tcpclient("127.0.0.1", 4000);
    end
    y=1;
end

% proposed implementation

function y = fcn()
    y = 0; % defining the correct type  
    persistent client  
    if isempty(client)
        client = tcpclient("127.0.0.1", 4000);
    end
    y=1;
end

I hope this helps.

[Brad]

@akshatsood thanks for your thorough response.

I recently came across this post: https://www.mathworks.com/matlabcentral/answers/1979839-how-should-i-handle-inputs-and-outputs-for-my-simulink-compiler-model

which suggests I use these functions for TCP communication within a matlab script:

• simulink.compiler.setExternalInputsFcn
• simulink.compiler.setExternalOutputsFcn

I'm most interested in designing my deployed app where it sets up a TCP client and only needs to handle inputs, not outputs. So I checked out the documentation for simulink.compiler.setExternalInputsFcn and in my opinion, it is severely lacking. Do you have any examples of how I would use this function in a matlab function that does the following:

• Creates a Simulink.SimulationInput object
• Configure for deployement with simulink.compiler.configureForDeployment
• Establishes TCP client on port 8200
• Reads 22 bytes of data over TCP every simulation frame
• Process this data where the 22 bytes needs to be processed with the 'typecast' function into separate vars of varying data types:

            Var_1: int8  (1 byte)
            Var_2: int32 (4 bytes)
            Var_3: int32 (4 bytes)
            Var_4: int32 (4 bytes)
            Var_5: double (8 bytes)
            Var_6: int8 (1 byte)

• Pass these 6 vars into my model, let's assume the model has 6 inputs
• Script/function needs to be compatible with 'mcc' command to get an executable

Now, I'm hung up on how to pass these 6 vars into my model with the simulink.compiler.setExternalInputsFcn function. Any examples of how to do this?

Otherwise, answering your other questions directly:

Debugging "Rapid accelerator process crashed while simulating the model"

Don't think antivirus is the issue because I can run the executable.

The "stack" trace is already printed out above below where it says "ME.stack is...". Not a very helpful error message :(

Selecting Device Type and Device Vendor

I've heard elsewhere from Mathworks support and Reddit that the TCP/IP Receive block from the instrument control toolbox isn't supported for code generation at all. Is this accurate?

“The current BuildConfig has not been set” in MATLAB Function block

This is an interesting approach, I will give it a shot and let you know!

[Brad]

@akshatsood - I tried implementing your suggested code that uses tcpclient inside a Matlab function block within my simulink model and I'm still running into the error "The current BuildConfig has not been set."

Here is my script titled tcp_client_script.m:

function data = tcp_client_script()
    
    % Pre-allocate data size and type to avoid error related to not
    % providing a current BuildConfig
    data = uint8(zeros(1, 22));
        % Sized for 22 bytes assuming the incoming data structure is:
        % packetType (int8) = 1 byte
        % packetVersion (int32) = 4 bytes
        % id (int32) = 4 bytes
        % sessionId (int32) = 4 bytes
        % timeEpoch (double) = 8 bytes
        % eventID (int8) = 1 byte

    persistent client

    if isempty(client)
        % Construct a tcpclient object assuming there is a TCP server at port 8200
        client = tcpclient("localhost", 8200);
    end
    
    % Read 22 bytes of data from the host
    data = read(client, 22, "uint8");
end

Here is how I call the tcp_client_script() function with a Matlab function block:

and inside the Matlab function block it looks like this:

function data = tcp_client_sim()

data = tcp_client_script();

If it matters, the simulink model was made as a new model with default settings and then I changed the solver settings to:

• Start time: 0
• Stop time: 5
• Fixed-step solver with auto solver selection
• Fixed-step size = 0.005

Now, when I save everything, then in the simulink model, an update diagram (ctrl+d) or try to run the sim, I get the following error printout:

Error:The current BuildConfig has not been set.
    
Function 'tcp_client_script.m' (#36.617.645), line 18, column 18:
"tcpclient("localhost", 8200)"
Launch diagnostic report.

Error:Function call failed.
    
Function 'MATLAB Function' (#63.41.60), line 3, column 8:
"tcp_client_script()"
Launch diagnostic report.

Error:Errors occurred during parsing of MATLAB function 'tcp_simple_client/MATLAB Function'

Error:Simulink cannot determine sizes and/or types of the outputs for block 'tcp_simple_client/MATLAB Function' due to errors in the block body, or limitations of the underlying analysis. The errors might be inaccurate. Fix the indicated errors, or explicitly specify sizes and/or types for all block outputs.

Error:Simulink cannot determine sizes and/or types of the outputs for block 'tcp_simple_client/MATLAB Function' due to errors in the block body, or limitations of the underlying analysis. The errors might be inaccurate. Fix the indicated errors, or explicitly specify sizes and/or types for all block outputs.

What do you recommend we try next to resolve this?

[akshatsood]

Hi @Brad,

I see that you have tried implementing the following in the MATLAB Function block,

function data = tcp_client_sim()
data = tcp_client_script();

The workaround, I mentioned in my answer focused on assigning the output variable inside the MATLAB Function block and not in a script. Could you please test it out if this works fine.

function data = tcp_client_sim()

data = %define data
data = tcp_client_script();

If the issue still persists, I consider this to be an issue with the "tcpclient" function in MATLAB Function block. To work around this for TCP communication from Desktop, I would recommend using the "TCP/IP Send" and "TCP/IP Receive" block from Instrument Control Toolbox.

https://www.mathworks.com/help/instrument/tcpipreceive.html

https://in.mathworks.com/help/instrument/tcpipsend.html

Note that for hardware targets (except for Raspberry Pi), using the "tcpclient" function in a MATLAB Function block will not work. Make sure to leverage the "TCP Send" and "TCP Receive" block from the hardware-specific library.

[Brad]

@akshatsood - I think we're getting closer here with the approach using tcpclient in an embedded matlab function block. So I followed your advice now I'm getting size mismatch errors when I try to compile the model. Here's what I tried...

Assuming my simulink model looks like this:

Inside the Matlab function block, the code now looks like this:

function data = tcp_client_sim()

% Pre-allocate data size and type to avoid error related to not
% providing a current BuildConfig
data = uint8(zeros(1, 22));

persistent client

if isempty(client)
    % Construct a tcpclient object assuming there is a TCP server at port 8200
    client = tcpclient("localhost", 8200);
end

% Read 22 bytes of data from the host
data = read(client, 22, "uint8");

Now when I try to compile with ctrl+d, I get error:

=== Update Diagram (Elapsed: 2 sec) ===
    Error:'data' is inferred as a variable-size matrix, but its size is specified as inherited or fixed. Verify 'data' is defined in terms of non-tunable parameters, or select the 'Variable Size' check box and specify the upper bounds in the Size box.
    Error:Errors occurred during parsing of MATLAB function 'tcp_simple_client3/MATLAB Function'
    Error:Errors occurred during parsing of MATLAB function 'tcp_simple_client3/MATLAB Function'

The 'data' in the Diagnostic viewer is clickable so I click it and it takes me to the Model Explorer with the signal highlighted:

The next thing I think to try is to change the size of 'data' in the Model Explorer to [1, 22] like so:

Then I attempt to build the model again with ctrl + d and I end up with the same error in the Diagnostic Viewer:

=== Update Diagram (Elapsed: 2 sec) ===
    Error:'data' is inferred as a variable-size matrix, but its size is specified as inherited or fixed. Verify 'data' is defined in terms of non-tunable parameters, or select the 'Variable Size' check box and specify the upper bounds in the Size box.
    Error:Errors occurred during parsing of MATLAB function 'tcp_simple_client3/MATLAB Function'
    Error:Errors occurred during parsing of MATLAB function 'tcp_simple_client3/MATLAB Function'

Ok so then I try to reall follow what the error message is saying...the part about "Verify 'data' is defined in terms of non-tunable parameters" so I choose to uncheck the "Tunable" box for 'data' in the Model Explorer

Then retry compiling model with ctrl + d and get the same error again in the Diagnostic viewer. Ok so the other part of the error message says "select the 'Variable Size' check box and specify the upper bounds in the Size box."

Ok so In the Model Explorer, I check the Tunable box again, then I check the "Variable size" box, then click apply:

Then I try to rebuild with ctrl+d and get the error message about the current BuildConfig not being set again!

    === Update Diagram (Elapsed: 0.216 sec) ===
    Error:The current BuildConfig has not been set.
    
    Function 'MATLAB Function' (#213.291.319), line 11, column 14:
    "tcpclient("localhost", 8200)"
    Launch diagnostic report.
    Error:Errors occurred during parsing of MATLAB function 'tcp_simple_client3/MATLAB Function'
    Error:Simulink cannot determine sizes and/or types of the outputs for block 'tcp_simple_client3/MATLAB Function' due to errors in the block body, or limitations of the underlying analysis. The errors might be inaccurate. Fix the indicated errors, or explicitly specify sizes and/or types for all block outputs.
    Error:Simulink cannot determine sizes and/or types of the outputs for block 'tcp_simple_client3/MATLAB Function' due to errors in the block body, or limitations of the underlying analysis. The errors might be inaccurate. Fix the indicated errors, or explicitly specify sizes and/or types for all block outputs.

Ok next I set the DataType for 'data' in the Model Explorer:

retry ctrl+d and get a slightly different error message about BuildConfig not being set but it has a diagnostic report link I can click this time:

=== Update Diagram (Elapsed: 0.275 sec) ===
    Error:The current BuildConfig has not been set.
    
    Function 'MATLAB Function' (#213.291.319), line 11, column 14:
    "tcpclient("localhost", 8200)"
    Launch diagnostic report.
    Error:Errors occurred during parsing of MATLAB function 'tcp_simple_client3/MATLAB Function'
    Error:Simulink cannot determine sizes and/or types of the outputs for block 'tcp_simple_client3/MATLAB Function' due to errors in the block body, or limitations of the underlying analysis. The errors might be inaccurate. Fix the indicated errors, or explicitly specify sizes and/or types for all block outputs.
    Error:Simulink cannot determine sizes and/or types of the outputs for block 'tcp_simple_client3/MATLAB Function' due to errors in the block body, or limitations of the underlying analysis. The errors might be inaccurate. Fix the indicated errors, or explicitly specify sizes and/or types for all block outputs.

Ok so I launch the diagnostic report...and it looks like this:

Ok next, I went back to Model Explorer and turned off the Variable Size box:

And now with a ctrl+d I get another error message in the Diagnostic Viewer:

    === Update Diagram (Elapsed: 2 sec) ===
    Error:Dimension 1 is fixed on the left-hand side but varies on the right ([1 x 22] ~= [:? x :?]).
    More information
    
    Function 'MATLAB Function' (#213.364.368), line 15, column 1:
    "data"
    Launch diagnostic report.
    Error:Errors occurred during parsing of MATLAB function 'tcp_simple_client3/MATLAB Function'
    Error:Errors occurred during parsing of MATLAB function 'tcp_simple_client3/MATLAB Function'

Ok interesting so I click the "More information" link in the Diagnostic viewer which takes me to this page:

Diagnose and Fix Variable-Size Data Errors - MATLAB & Simulink (mathworks.com)

I then think I'm a bit confused about what it wants me to do in terms of using:

    coder.varsize()

or:

    assert

to fix the size of vars.

What do you recommend I try next?

[Brad]

@akshatsood - ok one more google search got me to this post:

How to set Matlab function block output as fixed (non-tunable) variable - MATLAB Answers - MATLAB Central (mathworks.com)

And following their example of having a temporary var then assigning the output of the Matlab function block inside simulink model using indexing, I came up with this code inside my Matlab function block:

function data = tcp_client_sim()

% Pre-allocate data size and type to avoid error related to not
% providing a current BuildConfig
data_temp = uint8(zeros(1, 22));
data = data_temp;

persistent client

if isempty(client)
    % Construct a tcpclient object assuming there is a TCP server at port 8200
    client = tcpclient("localhost", 8200);
end

% Read 22 bytes of data from the host
data_temp = read(client, 22, "uint8");

data = data_temp(1:22);

And now it compiles and runs without any errors!!

Weird ok, now I need to see if I can run it with rapid accelerator mode and then build to an exe with mcc command...

[akshatsood]

HI @Brad,

Thank you for your concise response. I appreciate your prompt action in rectifying the errors. Could you kindly proceed to confirm the system's behavior in rapid accelerator mode, followed by building an executable? Should any issues arise or should you require further clarification on any aspect, please do not hesitate to contact me.

Thanks!

[Brad]

@akshatsood - bad news again. I can run my simulink model interactively in normal mode, interactively in rapid accelerator mode, and can deploy it to a standalone exe that runs on my machine. However, when I share the standalone exe (plus all supporting files) with my coworker (on a separate PC), they get the "Simulink:tools:rapidAccelExeError". See image below:

Our PC setups and matlab installs are as follows:

My PC:

• Windows 10, 64-bit operating system
• Matlab R2023a - installed with "all" toolboxes available
• Matlab Runtime R2023a (9.14)

My coworker's PC:

• Windows 10, 64-bit operating system
• Matlab Runtime R2023a (9.14)

So this actually shows that building a standalone app with mcc from a simulink model with 'tcpclient' used in a Matlab function block or from a simulink model that uses the TCP/IP Receive block from the Instrument Control Toolbox (the original post that started this conversation) produces the same "Simulink:tools:rapidAccelExeError" when the exe is run on a PC that only has Matlab Runtime installed (does not have full Matlab installed).

Lastly, on the topic of Selecting Device Type and Device Vendor - here is the setup for my Simulink model(s) which honestly is just the defaults for any Simulink model I make:

• Device type: x86–64 (Windows64)
• Device vendor: Intel

Lastly, here's the full details from the Configuration Parameters -> Hardware Implementation page of my simulink model:

• I should note, I previously tried checking the box that says "Support long long" as recommended by the documentation for the TCP/IP Receive block in its "C/C++ Code Generation Generate C and C++ code using Simulink® Coder™" guide.
• Checking "Support long long" did not fix the issue with getting the "Simulink:tools:rapidAccelExeError" on my coworker's PC which only has Matlab Runtime installed (not full matlab install).

[Brad]

@akshatsood - and for completeness, here is a screenshot of my latest model:

• Should note I'm using a fixed frame time of dt = 1 second with the automatic fixed step solver selection

and inside the Matlab function block, the code is:

function [packetType, packetVersion, id, sessionId, timeEpoch, eventID] = tcp_client_sim()

% Pre-allocate data size and type to avoid error related to not
% providing a current BuildConfig
data_temp = uint8(zeros(1, 22));
data = data_temp;
packetType = int8(0);
packetVersion = int32(0);
id = int32(0);
sessionId = int32(0);
timeEpoch = 0;
eventID = int8(0);

persistent client

if isempty(client)
    % Construct a tcpclient object assuming there is a TCP server at port 8200
    client = tcpclient("localhost", 8200);
end

% Read 22 bytes of data from the host
data_temp = read(client, 22, "uint8");

data = data_temp(1:22);

packetType = typecast(data(1), 'int8');
packetVersion = typecast(data(2:5), 'int32');
id = typecast(data(6:9), 'int32');
sessionId = typecast(data(10:13), 'int32');
timeEpoch = typecast(data(14:21), 'double');
eventID = typecast(data(22), 'int8');

Then the script I'm using to deploy my app is:

function tcp_simple_client3_app()

    in = Simulink.SimulationInput('tcp_simple_client3');
    in = in.setModelParameter('EnablePacing', 'on');
    in = in.setModelParameter('PacingRate', 1.04);  
        % Through trial and error, discovered a PacingRate of 1.04 gets 
        % the deployed sim to run within 1 percent of real time consistently
    in = in.setModelParameter('StopTime', '5');


    % Configure for deployment
    in = simulink.compiler.configureForDeployment(in);
    try
        out = sim(in);
    catch ME
        disp('ME is...')
        disp(ME)
        disp('ME.stack is...')
        disp(ME.stack)
    end

    if exist("out", "var")
        disp(out)
        disp(out.logsout)     

        figure(1);
        for ii = 1:6
            subplot(6,1,ii)
            plot(out.logsout{ii}.Values)
            grid minor
        end
    end

    disp('Please inspect the plot for expected TCP received data over time.')
    ack = input('Press ctrl+c to exit (will close the plot)', "s");
end

Lastly, the command I'm using to deploy my app in the Matlab Command Window is:

dt = 1; mcc -d tcp_simple_client3 -m tcp_simple_client3_app.m

I was wondering if there's something going on here where maybe I need to pass more optional arguments into the "mcc" command? Like do I need to use the "-a" option to add additional files or directories to be included in the build? For example, do I need to use the -a option to tell it to explicitly compile with a reference to my "tcp_simple_client3.slx" model and/or explicitly compile with a reference to "tcpclient.m" or whatever scripts support the TCP Receive block from instrument control toolbox?

[akshatsood]

Hi @Brad,

After gathering and reviewing all the details you have provided, and conducting a comprehensive investigation, I have arrived at the following conclusions:

The Simulink Compiler currently does not support the Instrument Control Toolbox, which explains the error triggered by the inclusion of the TCP/IP Receive block in the model. To view which toolboxes are compatible with the Simulink Compiler, please consult the list available at this link

https://in.mathworks.com/help/releases/R2023b/slcompiler/gs/prerequisites-for-simulink-compiler.html

To resolve the issue at hand, it would be beneficial to utilize the `simulink.compiler.setExternalInputFcn` function, which you have previously mentioned. There is a specific article from MathWorks Support that provides an in-depth explanation of how to use this function.

https://in.mathworks.com/matlabcentral/answers/1979839-how-should-i-handle-inputs-and-outputs-for-my-simulink-compiler-model

For the use case, you have mentioned, you can try out the following,

model = 'your_model_name_here';

% creating a SimulationInput object for your model
simIn = Simulink.SimulationInput(model);

% configuring for deployment
simulink.compiler.configureForDeployment(simIn);

% establish the TCP client connection on port 8200
tcpipClient = tcpclient('localhost', 8200);

% define the external input function
function extInputs = getExternalInputs(~)
    
    % reading 22 bytes of data from the TCP client
    data = read(tcpipClient, 22, 'uint8');
    
    % processing the data using typecast
    Var_1 = typecast(data(1), 'int8');
    Var_2 = typecast(data(2:5),'int32');
    Var_3 = typecast(data(6:9),'int32');
    Var_4 = typecast(data(10:13),'int32');
    Var_5 = typecast(data(14:21),'double');
    Var_6 = typecast(data(22),'int8');
    
    % passing the processed variables into the model
    extInputs = {Var_1, Var_2, Var_3, Var_4, Var_5, Var_6};
end

simulink.compiler.setExternalInputsFcn(simIn, @getExternalInputs);

% mcc -m <YOUR_SCRIPT_NAME>.m

I hope this helps.

[Brad]

Thanks will try it out!

[Brad]

@akshatsood - there a way to make the tcpclient in the above code work where “blocking mode” is disabled (like the option we get with the instrument control toolbox TCP/IP receive block)? I was thinking doing a check on the NumBytesAvailable and if it’s greater than 0 then read TCP data, else keep my sim stepping. Would that work?

[akshatsood]

Hi @Brad,

Could you please check this documentation page for further reference for reading the data when "NumBytesAvailable" becomes positive. As correctly pointed by you, I believe it should work on similar lines as that of "blocking mode".

https://in.mathworks.com/help/matlab/ref/tcpclient.html?searchHighlight=NumBytesAvailable&s_tid=srchtitle_support_results_5_NumBytesAvailable

I hope this helps.

[Brad]

@akshatsood

Trying this code out, but first question is on this line:

extInputs = {Var_1, Var_2, Var_3, Var_4, Var_5, Var_6};

Why did you opt to define extInputs as a cell array?

Trying this code out, I had to move the function definition to the bottom of the file, define a fixed step size dt that I use in my model, then run the deploy command as:

dt = 0.005; mcc -d tcp_app -m tcp_app.m

Then I startup an external Python script to make a TCP server that sends 22 bytes of data every 1 second. Then I run tcp_app.exe and get error:

Error using tcp_app
The current simulation mode is unknown or not supported in deployment.  Please set SimulationMode model parameter to 'Rapid', using Simulink.SimulationInput object.
simulinkcompiler:runtime:UnsupportedSimulationMode

Ok then I fix the error by adding the following line to tcp_app.m:

% set rapid accelerator mode
simIn = simIn.setModelParameter('SimulationMode', 'Rapid');

then deployed again and ran and got different error:

For deployment, set the RapidAcceleratorUpToDateCheck model parameter to 'off', using Simulink.SimulationInput object.

simulinkcompiler:runtime:RapidAcceleratorMustBeUpToDateCheckOff

So then fixed that error by adding the line to tcp_app.m:

simIn = simIn.setModelParameter('RapidAcceleratorUpToDateCheck', 'off');

and it runs now.

I added some plotting logic to check that I'm actually getting values input to my model that get recorded by signal logging. I also want to slow execution down and only run for 10 seconds so I added these lines to tcp_app.m:

% Set pacing rate and stop time
simIn = simIn.setModelParameter('EnablePacing', 'on');
simIn = simIn.setModelParameter('PacingRate', 1.04); 
simIn = simIn.setModelParameter('StopTime','10');

The code for tcp_app.m now looks like:

% clear; close all; clc

function tcp_app()

    dt = 0.005;
    
    model = 'tcpClient_model';
    
    % creating a SimulationInput object for your model
    simIn = Simulink.SimulationInput(model);
    
    % configuring for deployment
    simulink.compiler.configureForDeployment(simIn);

    % set rapid accel mode
    simIn = simIn.setModelParameter('SimulationMode', 'Rapid');
    simIn = simIn.setModelParameter('RapidAcceleratorUpToDateCheck', 'off');
    
    % Set pacing rate and stop time
    simIn = simIn.setModelParameter('EnablePacing', 'on');
    simIn = simIn.setModelParameter('PacingRate', 1.04); 
    simIn = simIn.setModelParameter('StopTime','10');

    % establish the TCP client connection on port 8200
    tcpipClient = tcpclient('localhost', 8200);
    
    % Drive external inputs to model with function
    simulink.compiler.setExternalInputsFcn(simIn, @getExternalInputs);
    
    % Simulate model
    simOut = sim(simIn);

    save("simOut")

    figure(1);
    for ii = 1:7
        subplot(7,1,ii)
        plot(simOut.logsout{ii}.Values)
        grid minor
    end

    saveas(figure(1), 'plotOut', 'fig')
    
    % define the external input function
    function extInputs = getExternalInputs(~)
        
        % reading 22 bytes of data from the TCP client
        data = read(tcpipClient, 22, 'uint8');
        
        % processing the data using typecast
        Var_1 = typecast(data(1), 'int8');
        Var_2 = typecast(data(2:5),'int32');
        Var_3 = typecast(data(6:9),'int32');
        Var_4 = typecast(data(10:13),'int32');
        Var_5 = typecast(data(14:21),'double');
        Var_6 = typecast(data(22),'int8');
        
        % passing the processed variables into the model
        extInputs = {Var_1, Var_2, Var_3, Var_4, Var_5, Var_6};
    end

end

After running "mcc", then running tcp_app.exe, the plot I get is showing I'm not actually getting any data driving my inputs 1 through 6 (as shown by grids 2-7 being constant at zero) . Note I'm plotting "sin_wave" as a control value just to show my sim is stepping.

The expected behavior from what my Python TCP server is sending is that values will change every 1 second as follows:

• packetType: remains fixed at 0
• packetVertsion: remains fixed at 1
• id: starts at 0 but increments by 1 every second
• sessionId: stays fixed at 2
• timeEpoch: a float that increments by 1 every second
• eventID: starts at 0 but increments by 1 every second

Below I have attached my models if you wanted to try it out for yourself.

It's not allowing me to attach my python script as a .py so I've attached it as "tcp_server_send_data.py.m" - to get that to run, would have to remove the ".m" part of the file name and run it with a python version 3.10 or newer. Alternatively, could drive the TCP server output with another matlab session.

[Brad]

I'm also finding when I change this line:

simulink.compiler.setExternalInputsFcn(simIn, @getExternalInputs);

to this:

simIn = simulink.compiler.setExternalInputsFcn(simIn, @getExternalInputs);

assuming I'm defining extInputs as a cell array:

extInputs = {Var_1, Var_2, Var_3, Var_4, Var_5, Var_6};

or a normal array:

extInputs = [Var_1, Var_2, Var_3, Var_4, Var_5, Var_6];

that I get the following error when running tcp_app.exe:

Error using tcp_app
Error occurred while invoking ExternalInputsFcn.
Caused by:
    Error using rapid_accel_target_utils
    Too many input arguments. - Show complete stack trace

SimulinkExecution:SimulationService:InputFcnError

Ok so running with syntax:

simulink.compiler.setExternalInputsFcn(simIn, @getExternalInputs);

results in no errors but I get zeros for the values of all inputs then running with syntax:

simIn = simulink.compiler.setExternalInputsFcn(simIn, @getExternalInputs);

results in an error.

What other methods might there be to have getExternalInputs() pass out multiple signals in order to drive multiple inputs to the model with simulink.compiler.setExternalInputsFcn() ??

[Brad]

Alright I figured out how to pass simTime into setExternalInputsFcn and getExternalInputs. See my modified code below:

% clear; close all; clc

function tcp_app()

    dt = 0.005;
    
    model = 'tcpClient_model';
    
    % creating a SimulationInput object for your model
    simIn = Simulink.SimulationInput(model);
    
    % configuring for deployment
    simulink.compiler.configureForDeployment(simIn);

    % set rapid accel mode
    simIn = simIn.setModelParameter('SimulationMode', 'Rapid');
    simIn = simIn.setModelParameter('RapidAcceleratorUpToDateCheck', 'off');
    
    % Set pacing rate and stop time
    simIn = simIn.setModelParameter('EnablePacing', 'on');
    simIn = simIn.setModelParameter('PacingRate', 1.04); 
    simIn = simIn.setModelParameter('StopTime','0.1');

    % establish the TCP client connection on port 8200
    tcpipClient = tcpclient('localhost', 8200);
    
    % Drive external inputs to model with function
    simIn = simulink.compiler.setExternalInputsFcn(simIn, @(pn, simTime) getExternalInputs(pn, simTime));
    
    % Simulate model
    simOut = sim(simIn);

    save("simOut")

    figure(1);
    for ii = 1:2
        subplot(2,1,ii)
        plot(simOut.logsout{ii}.Values)
        grid minor
    end

    saveas(figure(1), 'plotOut', 'fig')
    
    % define the external input function
    function extInputs = getExternalInputs(~, simTime)
        
        if tcpipClient.NumBytesAvailable > 0
            % reading 22 bytes of data from the TCP client
            data = read(tcpipClient, 22, 'uint8');
            
            % processing the data using typecast
            Var_1 = typecast(data(1), 'int8');
            Var_2 = typecast(data(2:5),'int32');
            Var_3 = typecast(data(6:9),'int32');
            Var_4 = typecast(data(10:13),'int32');
            Var_5 = typecast(data(14:21),'double');
            Var_6 = typecast(data(22),'int8');
        else
            Var_1 = int8(0);
            Var_2 = int32(0);
            Var_3 = int32(0);
            Var_4 = int32(0);
            Var_5 = double(0);
            Var_6 = int8(0);
        end
        
        % passing the processed variables into the model
        % extInputs = {Var_1, Var_2, Var_3, Var_4, Var_5, Var_6};
        % extInputs = [Var_1, Var_6];
        extInputs = Var_3;
        % disp(extInputs)
    end

end

I'm just having extInputs be of size 1 for now.

Compiling with "mcc", the exe runs and I can push data through the one input (see screenshot for simplified model) but it's clearly waiting to step the entire model until it receives a TCP packet of data (which the Python script is pushing every 1 second of real-time). This means the TCP client is running in "blocking mode" still.

Oh and we can tell my sim is running much slower than clock time because I set in my script:

simIn = simIn.setModelParameter('StopTime','0.1');

This means my sim runs for 0.1 seconds of simulation time but the printout of the following property of the simOut object indicates it's taking 2.3 seconds of "ExecutionElapsedWallTime" to actually compute 0.1 seconds of simulation time:

>> simOut.SimulationMetadata.TimingInfo

ans = 

struct with fields:

WallClockTimestampStart: '2024-1-29 13:18:57'
WallClockTimestampStop: '2024-1-29 13:18:59'
InitializationElapsedWallTime: 0.3040
ExecutionElapsedWallTime: 2.3270
TerminationElapsedWallTime: 0.0090
TotalElapsedWallTime: 2.6400

[Brad]

Now, trying with multiple inputs, I changed extInputs to be:

extInputs = {Var_1, Var_2, Var_3, Var_4, Var_5, Var_6};

with model looking like:

Running mcc, then running exe throws error:

Error using tcp_app
The dimensions, data type or complexity of the return value from the ExternalInputsFcn for port 1 are incorrect.

SimulinkExecution:SimulationService:InvalidDataFromExternalInputs

Changing it to be defined as an array throws the same error but complains about port 2:

extInputs = [Var_1, Var_2, Var_3, Var_4, Var_5, Var_6];

error:

Error using tcp_app
The dimensions, data type or complexity of the return value from the ExternalInputsFcn for port 2 are incorrect.

SimulinkExecution:SimulationService:InvalidDataFromExternalInputs

I had received a separate email from Mathworks support that provided a script example of using simulink.compiler.setExternalInputsFcn and their example recommended each input be set to a fixed Execution -> Sample time so I double clicked on each input and changed that setting like shown below:

Didn't seem to resolve the issue though...