moving_average.h

#pragma once
#include <vector>
 
/**
 * Interface for filters
 * Use add_entry to supply data and get_value to retrieve the filtered value
 */
class Filter {
 public:
  virtual void add_entry(double n) = 0;
  virtual double get_value() const = 0;
};
 
/**
 * MovingAverage
 *
 * A moving average is a way of smoothing out noisy data. For many sensor readings, the noise is
 * roughly symmetric around the actual value. This means that if you collect enough samples those
 * that are too high are cancelled out by the samples that are too low leaving the real value.
 *
 * The MovingAverage class provides a simple interface to do this smoothing from our noisy sensor
 * values.
 *
 * WARNING: because we need a lot of samples to get the actual value, the value given by the
 * MovingAverage will 'lag' behind the actual value that the sensor is reading. Using a
 * MovingAverage is thus a tradeoff between accuracy and lag time (more samples) vs. less accuracy
 * and faster updating (less samples).
 *
 */
class MovingAverage : public Filter {
 public:
  /**
   * Create a moving average calculator with 0 as the default value
   *
   * @param buffer_size    The size of the buffer. The number of samples that constitute a valid
   * reading
   */
  MovingAverage(int buffer_size);
  /**
   * Create a moving average calculator with a specified default value
   * @param buffer_size    The size of the buffer. The number of samples that constitute a valid
   * reading
   * @param starting_value The value that the average will be before any data is added
   */
  MovingAverage(int buffer_size, double starting_value);
 
  /**
   * Add a reading to the buffer
   * Before:
   * [ 1 1 2 2 3 3] => 2
   *   ^
   * After:
   * [ 2 1 2 2 3 3] => 2.16
   *     ^
   * @param n  the sample that will be added to the moving average.
   */
  void add_entry(double n) override;
 
  /**
   * Returns the average based off of all the samples collected so far
   * @return the calculated average. sum(samples)/numsamples
   */
  double get_value() const override;
 
  /**
   * How many samples the average is made from
   * @return the number of samples used to calculate this average
   */
  int get_size() const;
 
 private:
  int buffer_index;            // index of the next value to be overridden
  std::vector<double> buffer;  // all current data readings we've taken
  double current_avg;          // the current value of the data
};
 
/**
 * ExponentialMovingAverage
 *
 * An exponential moving average is a way of smoothing out noisy data. For many sensor readings, the
 * noise is roughly symmetric around the actual value. This means that if you collect enough samples
 * those that are too high are cancelled out by the samples that are too low leaving the real value.
 *
 * A simple mobing average lags significantly with time as it has to counteract old samples. An
 * exponential moving average keeps more up to date by weighting newer readings higher than older
 * readings so it is more up to date while also still smoothed.
 *
 * The ExponentialMovingAverage class provides an simple interface to do this smoothing from our
 * noisy sensor values.
 *
 */
class ExponentialMovingAverage : public Filter {
 public:
  /**
   * Create a moving average calculator with 0 as the default value
   *
   * @param buffer_size    The size of the buffer. The number of samples that constitute a valid
   * reading
   */
  ExponentialMovingAverage(int buffer_size);
  /**
   * Create a moving average calculator with a specified default value
   * @param buffer_size    The size of the buffer. The number of samples that constitute a valid
   * reading
   * @param starting_value The value that the average will be before any data is added
   */
  ExponentialMovingAverage(int buffer_size, double starting_value);
 
  /**
   * Add a reading to the buffer
   * Before:
   * [ 1 1 2 2 3 3] => 2
   *   ^
   * After:
   * [ 2 1 2 2 3 3] => 2.16
   *     ^
   * @param n  the sample that will be added to the moving average.
   */
  void add_entry(double n) override;
 
  /**
   * Returns the average based off of all the samples collected so far
   * @return the calculated average. sum(samples)/numsamples
   */
  double get_value() const override;
 
  /**
   * How many samples the average is made from
   * @return the number of samples used to calculate this average
   */
  int get_size();
 
 private:
  int buffer_index;            // index of the next value to be overridden
  std::vector<double> buffer;  // all current data readings we've taken
  double current_avg;          // the current value of the data
};