AltitudeEstimation

A hardware-independent, header-only, C++ arduino library to perform vertical position estimation using an IMU-Barometer system via a two-step Kalman/Complementary filter.

This work is an implementation of the algorithm explained in this paper written in 2016 by Jung Keun Lee. Although the original is in Korean you can find an English version of it here thanks to Simon D. Levy.

Setup

Hardware

The current setup used for testing consists in the Butterfly STM32L433 Development Board plus the MPU9250 Mini Add-On connected through USB to the computer.

Yo can see it in the images below:

Usage

This library is extremely easy to use. Go to the releases page of the library and download the latest release. Once downloaded, uncompress it in your Arduino/libraries folder. Alternatively, clone this respository in your Arduino/libraries folder by opening a terminal session there and typing:

git clone https://github.com/juangallostra/AltitudeEstimation.git

Clonning/Downloading the library in your Arduino/libraries folder is the only step that has to be performed before getting into the actual coding. To be able to use the library in your code add the following line at the beginning of your file:

#include "altitude.h"

We are ready now to instantiate the estimator where needed. This can be done by:

// Altitude estimator
AltitudeEstimator altitude = AltitudeEstimator(0.0005, 	// sigma Accel
                                               0.0005, 	// sigma Gyro
                                               0.018,   // sigma Baro
                                               0.5, 	// ca
                                               0.1);	// accelThreshold

Note that here we are specifying the value of the parameters that will be used to perform the estimations. These can be tuned to achieve higher accuracy. A more in-detail description of the parameters is provided below.

Once we have our estimator the only thing we have to do to estimate the current vertical position, velocity and acceleration is to call its estimate method. Since one of the aims of this library is to be hardware-independent, this method expects to receive the latest acceleremoter, gyrometer and barometer readings as well as the timestamp they were obtained as parameters. Jump to the Available methods section if you wish to know more about these parameters and their expected values.

altitude.estimate(accelData, gyroData, baroHeight, timestamp);

The call to the previous method will not return anything. Instead, it will perform the estimation and store the results. To access the results of the latest estimation you just have to call the following methods:

altitude.getAltitude()	// get estimated altitude in meters
altitude.getVerticalVelocity() // get estimated vertical velocity in meters per second
altitude.getVerticalAcceleration() // get estimated vertical acceleration in m/s^2

A fully working example can be found at AltitudeEstimation.ino under examples/AltitudeEstimation. The provided code assumes that the hardware used is an MPU9250 IMU and a MS5637 barometer. Arduino libraries for both them are available here (MPU9250) and here (MS5637).

There is another example thanks to Simon D. Levy, making use of the EM7180 SENtral sensor hub. The sketch is SENtralAltitude.ino and can be fund under examples/SENtralAltitude. The drivers for the EM7180 can be downloaded from here.

Available methods

estimate

Each time this method is called the estimation of the vertical position, velocity and acceleration will be updated. To update the estimation you must provide the latest available readings from the accelerometer (in g-s), the gyrometer (in rad/s) and the baro (in meters) as well as the timestamp in which the readings were obtained. Currently the library assumes the timestamp is measured in microseconds.

Calling this method will update the estimations and store the results internally but it will not return anything. There are specific methods provided to get the estimated values.

Method signature:

void estimate(float accel[3], float gyro[3], float baroHeight, uint32_t timestamp)

Parameters:

• float accel[3]: length 3 array of floats with acceleration readings in g-s. The order in which the estimator expects the readings is ax, ay and az.

• float gyro[3]: length 3 array of floats with gyrometer readings in rad/s. The order in which the estimator expects the readings is gx, gy and gz.

• float baroHeight: vertical height in meters as estimated from the barometer signal. The conversion from pressure to height can be easily achieved with a small helper function. This is the formula I used in the provided example to achieve so (see lines 29-33). Since the algorithm requires the altitude estimated from the baro and not the pressure reading itself I prefer to let the user choose freely how he wants to map pressure to altitude.

• uint32_t timestamp: The timestamp at which the readings were obtained. Currently the library expects it to be in microseconds.

getAltitude

This method can be called to obtain the latest vertical position estimation. The estimated altitude is measured in meters.

Method signature:

float getAltitude()

getVerticalVelocity

This method can be called to obtain the latest vertical velocity estimation. The estimated velocity is measured in meters per second.

Method signature:

float getVerticalVelocity()

getVerticalAcceleration

This method can be called to obtain the latest vertical acceleration estimation. The estimated vertical acceleration is measured in meters per second^2.

Method signature:

float getVerticalAcceleration()

Parameter tunning

There are a few parameters that can be tuned to try to achieve higher accuracy. These paremeters are:

1. sigmaAccel: standard deviation of the accelerometer
2. sigmaGyro: standard deviation of the gyroscope
3. sigmaBaro: standard deviation of the barometer
4. ca: constant value for the markov chain acceleration model: a(k) = ca * a(k-1) + e
5. accelThreshold: vertical acceleration threshold. If 12 consecutive vertical acceleration values are below the threshold the vertical velocity will be set to 0.

The value of this parameters must be specified when calling the constructor of AltitudeEstimator. The order is the one listed above. Below one can see the signature of the constructor:

AltitudeEstimator(float sigmaAccel, float sigmaGyro, float sigmaBaro, float ca, float accelThreshold)

Results

Legend:

• Yellow: Estimated vertical acceleration (m/s^2).
• Green: Estimated vertical velocity (m/s).
• Blue: Estimated altitude from barometer pressure readings (m).
• Red: Two step Kalman/Complementary filter estimated altitude (m).

Test 1

Setup at rest the whole time.

Test 2

Lift to full cable length, hold, and lower it back to the initial position.

Extras

There is a Python implementation of the algorithm under the folder extras. You can read about it here.