Differences

This shows you the differences between two versions of the page.

--- projects:robots:inertial_navigation_system [2024/04/25 14:53] – jhagstrand
+++ projects:robots:inertial_navigation_system [2024/04/29 04:51] (current) – [videos] jhagstrand
@@ Line 1: / Line 1: @@
 ====== Inertial Navigation System (INS) ======
+spreadsheet of data:\\
+https://docs.google.com/spreadsheets/d/1SMg0OuFvwoGWqhNJKjKyEFaudhpbxwleUFlkU9-4UDw/edit?usp=sharing
 INS Intertial Navigation System - Sensors, processors, and software to estimate position.
@@ Line 21: / Line 25: @@
   * magnetometer - measures the strength of the magnetic fields operating on a vehicle.
-  * barometer - measures the air pressure
+  * barometer/altimeter - measures air pressure, which can be used to calculate altitude above sea level.
-  * altimeter - measures the height of a vehicle above sea level, based on the barometer reading
+=== Inertial Measurement Unit (IMU) ===
-Inertial Measurement Unit
+A collection of 3 accelerometers and 3 gyroscopes.
+This device gives us six numbers: linear motion along each of the x,y,z axes,
+and rotational motion around each of the x,y,z axes.
+This gives us the complete picture of inertial motion of the vehicle.
-movement relative to the earth's magetic field
-o
+It is not enough to measure motion.
-Sensors
+We also need to know our current position, or orientation, or attitude.
-motion sensors
+Imagine we experienced an acceleration along the x-axis for two seconds.
+During that two seconds we went from standing still to moving upwards at 8 meters per second squared.
+Now where are we
-accelerometer
+???? does the accelerometer measure acceleration or motion ?
-gyroscope
+types of accelerometer: mems, piezo electric, piezo resistive
+https://youtu.be/To7JagpPDwY
-IMU - Inertial Measurement Unit -
+velocity vs acceleration
+per Steppe School youtubes:
+bias, aka offset, must be removed from each measurement
+  * accelerometer
+    * measures linear acceleration, not velocity
+    * unit: mg, 1g = 9.8 m/s2, 1mg = 1 thousandth of a g
+    * has bias, but not so critical as gyroscope and magnetometer
+    * gravity is an acceleration of 1g or 1000mg.  This acceleration is always acting on the sensor.
+    * the upright and at-rest measurement should be x:0, y:0, z:1000.
+    * as the sensor is rotated the acceleration due to gravity is distributed among the x,y,z axes.
+  * gyroscope
+    * gyroscope measures angular velocity, not acceleration
+    * gyroscope unit: degrees per second
+    * at rest, gyroscope returns the bias
+    * by subtracting the bias from a readingd, the at rest measurement is x:0, y:0, z:1000
+    * calibration is done at rest for some seconds, the average of readings is the bias
+    * 20948 chip has "offset" registers that we can set with the 3 bias values, each divided by 4
+  * magnetometer
+    * no registers, so we must keep variables and do the subtraction from each measurement
+    * calibration is done during 3D rotation for some seconds
+    * ((max - min) / 2) should be zero, so the actual value during calibration is the bias
+    * in addition to the bias, we also have an adjustment for declination, to convert magnetic north to true north
+localization = changing reference system from the body of the sensor to the earth
+  * sensor readings give data relative to the body, the vehicle
+  * ahrs values give data relative to the earth
+units:
+  * linear velocity: meters per second
+  * linear acceleration: meters per second squared
+  * angular velocity: degrees per second, radians per second
+  * angular acceleration: ?
+  * rotation velocity: hertz, cycles per second
+  * 10**-6  one millionth, μ, micro
+  * 10**-3  one thousandth, m, milli
+accelerometer and gyroscope measure motion
+we know that the z-axis is pointing straight up, but the x and y-axes are relative to some random or arbitrary starting point
+by adding a magnetometer, we can fix the x axis relative to north-south, and the y-axis relative to east-west
+the magnetometer measures the strength of magnetic fields acting on the sensor
+the most significant magnetic field is that of the earth
+naming absurdities:
+  * IMU inertial measurement unit
+    * does not measure inertia. It's components measure accleration, velocity, and magnetic field strenth.
+    * is not a unit of measurement
+  * 6-axis imu or 9-axis imu.  There are only three axes: x,y,z.
+  * 6dof or 9dof.  A sensor does not have degrees of freedom.
@@ Line 81: / Line 139: @@
 obsolete, replace with ICM-20948
 https://youtu.be/Oen3HqUbctM?si=3UnWGo-XOvs3ZwCZ
-TDK Invensense
-makes both MPU9250 and ICM20948
-EOL in 2018
-registers are different, therefore different driver required
-https://invensense.tdk.com/wp-content/uploads/2018/10/AN-000146-v2.0-TDK_Migration_MPU_9250toICM-20948.pdf
-breakout boards from adafruit and sparkfun
-adafruit ICM20948
-use Adafruit ICM20 library for arduino
-https://github.com/adafruit/Adafruit_ICM20X
-provides GPIO pins but also STEMMA QT connector
-no video demo
-see Paul McWhorter youtubes
-: Adafruit 9-DOF Absolute Orientation IMU Fusion Breakout - BNO055
-https://www.adafruit.com/product/2472
-video demo, no heading mentioned, euler and quat, quat more accurate, too slow for vehicle
-https://www.youtube.com/watch?v=EsgKAawwT9A&t=731s
-Bosch Sensortec BNO055
-"not recommended for new designs"
-"You could use BHI260AP in new design", but has no magnetometer.
-components:
-    Atmel SAM20D processor
-    BMA280 accelerometer
-    BMI055 gyroscope
-    BMM150 magnetometer
-    built-in fusion library, closed source:
-        Euler angles
-        quaternion
-        modi: IMU (motion), compass, M4G, NDOF
 video by very creepy guy using adafruit bno055, but no compass heading, source code on screen but not downloadable
@@ Line 154: / Line 180: @@
 Magnetometer calibration
+hold the sensor level and spin it horizontally
+the graph of the results should be a perfect circle, and the radius of the circle is the strength of the magnetic field
 https://youtu.be/MinV5V1ioWg?si=Bh-e9aDgxxzKsm14
-Scott Lobdell, IMU
+Scott Lobdell, on AHRS calculations, specifically the Madgwick Quaternion Update
+start with 9 measurements from the 3 sensors * 3 axes
+calibrate the magnetometer using a 3D bias to get corrected values of the 3 magnetometer measurements
+begin ahrs
+localize the acceleration vector from [x,y,z] to [north,east,down]
+convert the 9 measurements into a quaternion
+convert the quaternion into a 3x3 rotation matrix
+multiply the acceleration vector times the inverse of the rotation matrix, giving the localized acceleration vector
+remove the acceleration due to gravity [0,0,1] to [0,0,0]
+apply the declination offset
 https://youtu.be/T9jXoG0QYIA?si=HjldywVEFu-QSfJp
+===== videos =====
+Paul McWhorter: 9-axis IMUs with Arduino, youtube \\
+Adafruit Bosch BNO055 \\
+videos, in particular: \\
+#10: Making a tilt-compensated compass \\
+#21: Visualizing 3D rotations in VPython using Quaternions \\
+Scott Lobdell: How to implement an IMU \\
+no specific chip or code \\
+hand-drawn diagrams and verbal description of calibration and AHRS calculations
+MicWro Engr, Michael Wrona: Magnetometer Errors and Calibration \\
+Adafruit Precision NXP 9-DOF Breakout Board - FXOS8700 + FXAS21002 \\
+series of youtubes made while developing his custom "Hummingbird" flight controller
+Magneto calibration software \\
+[[Michael Wrona notes | notes]]
+Steppe School: STM32 and ICM20948 IMU part 1. accelerator and gyroscope
+https://youtu.be/Oen3HqUbctM?si=SeUm8s62af9PdqDz
+videos
+===== Commercial Products =====
+Arduino-compatible, hobbyist IMU products from sparkfun, adafruit, polulu.
+=== MPU9250 ===
+chip: TDK Invensense MPU9250 \\
+date:  ? \\
+status: EOL as of 2018 (supposedly replaced by ICM20948) \\
+comments: users complain that this product is no longer being manufactured and the market is flooded with fakes that don't work. \\
+=== ICM20948 ===
+board: boards available from adafruit and sparkfun \\
+chip: TDK Invensense MPU9250 \\
+date: 2018 \\
+datasheet: https://invensense.tdk.com/wp-content/uploads/2024/03/DS-000189-ICM-20948-v1.6.pdf \\
+status: supposedly replaces MPU9250, but registers are different, therefore different driver required \\
+comment: users complain that the DMP is not finished. DMP output is not available in current libraries
+videos:
+  * https://invensense.tdk.com/wp-content/uploads/2018/10/AN-000146-v2.0-TDK_Migration_MPU_9250toICM-20948.pdf
+  * see Paul McWhorter youtubes
+libraries:
+  * https://github.com/adafruit/Adafruit_ICM20X
+  * https://wolles-elektronikkiste.de/en/icm-20948-9-axis-sensor-part-i
+=== BNO055 ===
+board: Adafruit 9-DOF Absolute Orientation IMU Fusion Breakout, BNO055 \\
+chip: Bosch Sensortec BNO055 \\
+date: 2014 \\
+status: "Not recommended for new designs...You could use BHI260AP...", (but that has no magnetometer). \\
+datasheet: https://www.bosch-sensortec.com/media/boschsensortec/downloads/datasheets/bst-bno055-ds000.pdf \\
+components:
+  * Atmel SAM20D processor
+  * BMA280 accelerometer
+  * BMI055 gyroscope
+  * BMM150 magnetometer
+  * built-in fusion library, closed source:
+      * Euler angles
+      * quaternion
+      * modes: IMU, compass, M4G, NDOF
+libraries:
+  * https://www.arduinolibraries.info/libraries/adafruit-bno055
+videos:
+  * https://www.adafruit.com/product/2472
+  * video demo, no heading mentioned, euler and quat, quat more accurate, too slow for vehicle
+  * https://www.youtube.com/watch?v=EsgKAawwT9A&t=731s
+=== LIS3MDL ===
+board: Polulu #2862 MinIMU-9 v6 Gyro, Accelerometer, and Compass (LSM6DSO and LIS3MDL Carrier)
+chips:  ST LSM6DSO + LIS3MDL, two ST chips combined onto one Polulu board
+datasheets:
+  * https://www.st.com/resource/en/datasheet/lsm6dso.pdf
+  * https://www.st.com/resource/en/datasheet/lis3mdl.pdf
+=== UM7-LT ===
+board: made by Redshift Labs, sold by Polulu
+chip: UM7
+description:
+  * AHRS, includes microcontroller running a Kalman Filter
+  * expensive 160 USD
+  * Redshift Labs appears to be out of business