Kalman filter-based sensor fusion

The kalman_combinator sensor platform allows you to filter one or several sensors into one with a reduced error. If using a single sensor as data source, it acts like a exponential_moving_average filter. With multiple sensors, it combines their values based on their respective standard deviation.

The unit_of_measurement, device_class, entity_category, icon, and accuracy_decimals properties are by default inherited from the first sensor. state_class is explicitly not inherited, because total_increasing states could still decrease when multiple sensors are used.

# Example configuration entry
  - platform: kalman_combinator
    name: "Temperature"
    unit_of_measurement: °C
    process_std_dev: 0.001
      - source: temperature_sensor_1
        error: 1.0
      - source: temperature_sensor_2
        error: !lambda |-
          return 0.5 + std::abs(x - 25) * 0.023

Configuration variables:

  • process_std_dev (Required, float): The standard deviation of the measurement’s change per second (e.g. 1/3600 = 0.000277 if the temperature usually changes at most by one Kelvin per hour). A low value here will place high importance on the current state and be slow to respond to changes in the measured samples. A high value will update faster, but also be more noisy.

  • std_dev (Optional, Sensor): A sensor that publishes the current standard deviation of the state with each update.

  • sources (Required, list): A list of sensors to use as source. Each source must have either error or error_function set. These work like the process_std_dev parameter, with low values marking accurate data.

    • source (Required, ID of a Sensor Component): The sensor that is used as sample source

    • error (Required, float, templatable): The standard deviation of the sensor’s measurements. If implemented as a template, the measurement is in parameter x.

  • All other options from Sensor.

See Also