A ROS filtering package

This repo is about a filtering package for various sensors used in the AAU ASWP project. Currently the package supports any number of singals, with a varying window lenghts and two basic filtering methods.

Features

As of right now (20-02-2019) the package only supports singular value signals, such as an axis of acceleration from an IMU. However, it is possible to use multitude of axis, applying a different window to each.

Simple Moving Average (SMA)

This feature extraction is based on this methodology from Wikipedia. Moving average - Wikipedia

Exponential Moving Average (EMA)

Same applies to the exponential moving average. Exponential Moving average - Wikipedia

The exponential moving average only uses the previous average sample rather than calculating the average for the lenght of the window each time. It also uses a custom ALPHA_WEIGHT to determine how much should the function wiegh in the past results. This method is a lot faster computionally, but if the weight is not set according to the application's needs, it could undervalue or overvalue the past, resulting in noise average.

Getting started

To run the package do: $ rosrun sensor_filter_kit sensor_filter_kit_node

It is neccessary to pass the lenght of the window, sensor readings and the desired filter.

// Window length
const uint window_size = 100;

// number of sensors as a constant
const uint SENSOR_NUMBER =  3;

// The index of the sensors as an enum, here X_DDOT=1, Y_DDOT=2 etc.
uint sensors[SENSOR_NUMBER] = {X_DDOT, Y_DDOT, Z_DDOT};

// The readings array, same length as the number of sensors.
float sensor_readings[SENSOR_NUMBER];

// Class constructor, passing the number of sensors and the window size
FilterKit filter_kit(SENSOR_NUMBER, window_size);

Usage

To pass the sensor readings to the filter class as such:

  // Storing the features for each window declare a double type std::vector
  std::vector<double> features;
  int counter = 0;

  while(ros::ok())
  {
    ros::spinOnce();
    
    // Populating the sensor readings array
    sensor_readings[0] = imu_data.x;
    sensor_readings[1] = imu_data.y;
    sensor_readings[2] = imu_data.z;
    
    // Passing the values to the filterkit window. SMA = Simple Moving Average
    filter_kit.window(sensor_readings, sensors, SMA);

    // The features for each window are returned as a vector double
    features = filter_kit.get_features();
    std::cout << counter << " -X_DDOT: - " << imu_data.x << " - " << features.at(0) << \
                            " -Y_DDOT: - " << imu_data.y << " - " << features.at(1) << \
                            " -Z_DDOT: - " << imu_data.z << " - " << features.at(2) << std::endl;

    loop_rate.sleep();
    counter++;

To use the EMA, simply pass the EMA as the 3rd parameter to the window as such: filter_kit.window(sensor_readings, sensors, EMA); The resulting features will no be calculated with the exponential moving average.

Authors

This package and interface has been developed by David Michalik for the use on the project AAU ASWP.

License

This project is licensed under the MIT License - see the LICENSE.md file for details.

How to implement a moving average in C without a buffer? - Signal Processing Stack Exchange