This is a companion repository for an article about the Kalman filter.
In the article we identify a discrete linear state-space system by using System Identification app in Matlab from input.csv file:
% Predict
y = Cx + Du
% Update state
x = Ax + Bu
Sample code in this repository then demonstrates how to use these two equations in practice to simulate the system.
The code is provided in Matlab (.m) and C++ (.cpp). Since Matlab is essentially a Python REPL, it should be easy to convert the Matlab script to Python.
Launch Matlab, set current directory to this repository, then execute simulate.m simply by typing simulate into Matlab prompt and pressing ENTER.
The C++ example can be built and executed with the following steps.
- Download Eigen 3.4
- Extract and change into the folder:
cd eigen-3.4.0 - Create a build folder:
mkdir build && cd build - Setup CMake:
cmake .. - Install:
make install
Note: C++ 11 is required because matrix initialization shorthands are only supported by Eigen library on this standard. This is why
CMakeLists.txtincludes this line:set(CMAKE_CXX_STANDARD 11).
- Create a build folder:
mkdir build && cd build - Setup CMake:
cmake .. - Build:
make - This should create
simulateexecutable inbuildfolder
Run the demo from the current directory:
./build/simulate
This will read inputs from input.csv and write filtered result into output.csv.
Re-configure the build to add symbols and launch with LLDB debugger:
- Setup CMake from
buildfolder:cmake .. -DCMAKE_BUILD_TYPE=Debug - Build:
make - Open this folder in VSCode
- Switch to Debug tab
- Click Debug on (lldb) Launch configuration
The following steps were used to configure debugging in VSCode:
- Switch to Debug tab
- Click create a launch.json file link
- Select C++ (GDB/LLDB)
- Click Add Configuration on bottom right
- Select C/C++ (lldb) Launch
- A new entry should have been added to
launch.json - Set
programto"${workspaceFolder}/kalman"