filters

Several classical filters for IMU(or MARG) sensors.

Required sensors

Filters	Accelerometer(m/s^2)	Gyroscope(rad/s)	Magnetometer(uT)
Integral	☓	◯	☓
Tilt	◯	☓	☓
Complementary Filter	◯	◯	△
Mahony	◯	◯	△
Madgwick	◯	◯	△

Preface

All the units of the measurement inputs are defined in Required sensors, and the unit of the delta time dt is second.

Prerequisites

I have tested the library in Windows 11, but it should be easy to compile in other platforms.

C Compiler

I use MSYS2 to install gcc. Feel free to use your own compiler.
NOTE: remember to update the path of your own compiler in CMakeLists.txt:

set(CMAKE_C_COMPILER "path/to/your/gcc")

Cmake

I write a simple CMakeLists.txt to compile and build all the filters and all the static/dynamic/executable files. I also use MSYS2 to install make and one can make a link between make and mingw32-make.exe for convenience.

Scripts

I write a simple build.bat to run compiling and building. Feel free to write your own scripts.

Usage

In main.c, there is an example of how to use the filters(e.g. integral) in this repository.
The process is as follow:

1. Declare the filter type and initialize it.
2. Set customized configuration.
3. Update the internal state.
4. Get the quaternion from the internal state.

#include "integral.h"
#include <stdio.h>


int main(void) {
  printf("#include test successfully!");

  IntegralT integral;
  integral_init(&integral);
  // Set config
  const IntegralConfigT config = {
    .method = INTEGRAL_METHOD_SERIES,
    .order = 2
  };
  integral_set_config(&integral, &config);
  
  // Read in the IMU(or MARG) sensor measurements and continuously update in the loop
  IntegralInputT input = {
    .dt = 0.01,
    .gyro = {0.123, 1.234, -0.587}
  };
  double quat[4];
  integral_update(&integral, &input);
  integral_get_quat(&integral, quat);
  printf("(qw, qx, qy, qz) = (%lf, %lf, %lf, %lf)\n", quat[0], quat[1], quat[2], quat[3]);
  return 0;
}

Todo

1. Test the correctness and performance between filters
2. Upload an example file and write testing scripts in every filter algorithms.
3. Write Python-C-API instead of adding API to library wrapper manually.

Contributions

1. Integrate several classical filters into an easy-to-use library.

License

Complementary Filter

BSD, following the original implementation

Mahony

MIT, following the original implementation

Note: The implementation of Mahony filter is followed the Madgwick's version which is also provided by xioTechnologies. So here I just declare the license the same as the Madgwick filter.

Madgwick

MIT, following the original implementation

References

Integral

• Quaternion kinematics for the error-state Kalman filter
• Indirect Kalman Filter for 3D Attitude Estimation
• Quaternions

Tilt

• Tilt computation using accelerometer data for inclinometer applications
• Tilt Sensing Using a Three-Axis Accelerometer

Complementary Filter

• Keeping a Good Attitude: A Quaternion-Based Orientation Filter for IMUs and MARGs
• imu_tools

Mahony

• Complementary filter design on the special orthogonal group SO(3)
• Attitude estimation on SO(3) based on direct inertial measurements
• Nonlinear Complementary Filters on the Special Orthogonal Group
• A complementary filter for attitude estimation of a fixed-wing UAV
• xioTechnologies - Open source IMU and AHRS algorithms

Madgwick

• xioTechnologies - Open source IMU and AHRS algorithms