For version specific documentation for MATLAB R2022b to R2023b and see README_R2023b.md instead. For releases newer than R2023b continue with this document.
Note Currently this example fails in R2024a, due to the inability to change the signal name for
VelocityRef. This is under investigation, a newer release is recommended.
- MATLAB
- Simulink
- MATLAB® Compiler™
- MATLAB® Compiler SDK™
- Simulink® Compiler™
- Simscape™
- Powertrain Blockset™
- Vehicle Dynamics Blockset™
- Vehicle Network Blockset
Note: This example can only be compiled on a Linux® system to produce a binary that will run on Linux, as used by the Databricks® cluster. Running MATLAB itself on Databricks is one way to achieve this.
For details of running MATLAB on Databricks see: https://github.com/mathworks-ref-arch/matlab-on-databricks This uses the Linux version of MATLAB.
Before running the example, copy it to a temporary directory by running:
workDir = fullfile(tempdir, "Simscape");
copyfile(databricksRoot("examples", "Simscape"), workDir)
cd(workDir)The pwd argument creates it in the current folder, the one moved to in the
previous step.
autoblkEvStart(pwd)This will automatically open the freshly created project, and open the example
model, ConfiguredEv1EmVirtualVehicle.
It will also switch the directory to
Simscape/VirtualVehicle/VirtualVehicle/System/ConfiguredVirtualVehicle.
At the top level, add an Inport block, and connect it to a Goto block
(with the tag DRIVECYCLESOURCE) as follows:
Make sure the Inport block has a unit of m/s set.
Go down in the scenarios subsystem
And then double-click the generator
This brings you to the subsystem
ConfiguredEv1EmVirtualVehicle/Scenarios/Reference Generator/Drive Cycle.
Remove the DriveCycleSource block (highlighted in red)
Now insert a From block, with the same tag as in the Goto block (DRIVECYCLESOURCE)
Open the model settings (CTRL+E), and verify that logging of output signals is selected.
First, move back up to the example working directory.
cd(workDir);
We are now at a point where we can simulate the model with external data.
Note: If this is the first time the simulation is run, it will take some time, as the models must first be compiled.
>> outputs = runEv1Em(makeCycleTable(100))
outputs =
20071x13 table
Time AccelFdbk BattCurr BattSoc BattVolt DecelFdbk EMSpd EMTrq GearFdbk ax ay az xdot
__________ _________ ________ _______ ________ _________ __________ ______ ________ ___________ __ _________ __________
0 0 0 60 363.71 0 0 0 0 0 0 1.0003 0
1.6177e-05 0 0 60 363.71 0 7.2178e-33 0 0 5.4087e-15 0 1.0003 1.0862e-18
9.7064e-05 0 0 60 363.71 0 1.4159e-25 0 0 1.026e-12 0 1.0001 1.2174e-15
0.0005015 0 0 60 363.71 0 1.3147e-18 0 0 1.098e-10 0 0.99907 8.05e-13
0.0025237 0 0 60 363.71 0 4.2825e-12 0 0 4.3715e-10 0 0.99316 4.1176e-10
0.0045584 0 0 60 363.71 0 1.4051e-10 0 0 -3.6981e-08 0 0.986 3.3135e-09
: : : : : : : : : : : : :
99.98 0.039167 28.964 56.824 359.92 0 510.99 16.346 1 0.0091499 0 0.008112 18.472
99.98 0.039166 28.963 56.824 359.92 0 510.99 16.345 1 0.0091485 0 0.0081125 18.472
99.99 0.039113 28.923 56.824 359.92 0 511.01 16.321 1 0.0091081 0 0.0080563 18.473
99.99 0.039112 28.921 56.824 359.92 0 511.01 16.32 1 0.0091068 0 0.008057 18.473
100 0.039061 28.883 56.824 359.92 0 511.04 16.296 1 0.009068 0 0.0080072 18.474
100 0.03906 28.881 56.824 359.92 0 511.04 16.296 1 0.0090669 0 0.0080106 18.474
Display all 20071 rows.
Note: A warning with respect to visualization blocks in compiled mode may be shown, but this can be ignored.
The input to the simulation is from the function makeCycleData, which simply creates
some trigonometric input data. The argument indicates how many seconds it should run.
In order to compile the model, a few steps must be performed.
First, a schema must be generated. This will create a file runEv1Em.schema,
containing information about the input and output data types for the function.
This is necessary, as it helps the compiler framework generate optimized
marshalling code for the inputs and outputs. As we'll be running this model
on Databricks, data must be converted from Databricks to Compiled MATLAB,
and back again.
generateFunctionSchema("runEv1Em", {makeCycleTable(10)})When this is done, the library can be built. Use the function
buildLibraryEv1Em.
>> PSB = buildLibraryEv1Em()
% Lots of output ...
PSB =
PythonSparkBuilder with properties:
BuildResults: [1x1 compiler.build.Results]
BuildOpts: [1x1 compiler.build.PythonPackageOptions]
PkgName: "sldemo.virtualvehicle"
PkgFolder: []
OutputDir: '/tmp/Simscape/_build_ev1em'
SrcDir: "/tmp/Simscape/_build_ev1em/sldemo/virtualvehicle"
Files: [1x1 compiler.build.spark.PythonFileV2]
GenMatlabDir: "/tmp/Simscape/_build_ev1em/matlab_helpers"
HelperFiles: ["/tmp/Simscape/_build_ev1em/matlab_helpers/runEv1Em_mapPartitions.m" … ] (1x2 string)
ExampleFiles: [1x2 table]
ZipArtifactName: "/tmp/Simscape/_build_ev1em/sldemo.virtualvehicle_Artifact.zip"
To run the model on a cluster (with artificial data), the following steps can be followed:
Upload the wheel file to the cluster, either through the portal, or as follows,
customize the /Volumes path:
f = databricks.Files();
[wf, wp] = PSB.getWheelFile();
f.upload(wf, "/Volumes/main/default/myvolume/MyWheels/" + string(wp));Import the example notebook, customize the location as needed:
matlab.databricks.workspace.import('runEv1Em_notebook.py',"/Users/username@example.com/Examples")Open the notebook in the Databricks portal, attach it to a cluster. The cluster, of course, needs to have the correct MATLAB Runtime installed.
Customize the %pip install command to reflect where the wheel file was uploaded.
Customize the save_name variable, to select where the results should be saved.