Apresentação IMU (2).pdf

System of
Inertial Navigation
for indoor positioning
Jhonatan Brandel de Souza
Machine Translated by Google

Introduction
starting point in latitude and longitude, by processing data
provided by gyroscopes, accelerometers, and magnetometers.
Inertial Navigation System (INS):
• Allows obtaining and determining the position in relation to the
Source: www.decea.gov.br/sirius

Introduction
• INS was widely used in the Pre-GPS era, by airplanes,
ships and missiles.
• It is still used as the main means of controlling the attitude of
spacecraft and satellites.

Job offer
• Create handheld device to estimate the position of a
individual walking, without the aid of other positioning
systems (such as GPS and AGPS).
• Determine the length of each user's step, to correct the
distance covered, acting as an optimized pedometer.

The INS device also
records the data, storing it in a
CSV file.
The portable INS device is
placed in the user's backpack,
and through a WLAN connection,
data is collected and displayed.
in real time on the server
side.
Project

The solution was developed based
on the MPU9250 module, which
has an accelerometer, gyroscope
and magnetometer, all triaxial and
with a resolution of 14 bits for the
magnetometer, and 16 bits for the
other 2 sensors.
All MEM devices, integrated.
With I²C interface.
Solution development

The Raspberry Pi 2 B was the
board chosen to collect and process
the readings.
MPU9250 by I²C, and connecting to
the remote server
A library was developed
Communicating with the module
via WLAN network.
in Python to communicate with the
module, as well as a script to estimate
the position.
Solution development

Operation
1. Self-Calibration of accelerometers and gyroscopes;
2. Step detection; 3. Determination of azimuth; 4.
Determining the step length 5. Integration and domain
transformation; 6. Data display.

calibration
1. It is performed with the device stopped, in the initial position.
2. It serves to correct the accelerometer and gyroscope polarization.
3. Performs the standard deviation from a set of measurements, to
4. The standard deviation ends up characterizing the mechanical noise present.
determine a threshold, used as a safe margin, ensuring that the new data is not
just a random error, but a valid new data.
5. At the end of this routine, the system is ready to start readings.

step detection
1. The pitch detection routine monitors vertical (Z) axis acceleration.
3. The detector is a trigger mechanism, when a threshold is passed a new
step is computed, at the same time the trigger function is inactive for a
period of time, not to count multiple steps
The threshold was found empirically, and it is 11 m/s².
2. When a step is taken, there is a spike in acceleration.
mistakenly.
The trigger dead time is 400 ms.

azimuth determination
2. From this angle, the gyroscope's angular velocity readings are read, and
integrated to obtain the new azimuth angle.
1. At the initial moment, the azimuth variable is updated with the data of the
4. To combat drift, every time the magnetometer indicates north
5. The magnetometer is not used alone to determine azimuth as it is sensitive
to field distortions, so the azimuth value is only validated by the
magnetometer, if the field module is compatible.
magnetômetro.
3. As there is an error associated with the angular velocity, when integrating the
error, there is an accumulation of error in the angle variable as time passes.
This phenomenon is called drift.
magnetic, the azimuth value is set to zero. This eliminates the drift caused
by the gyroscope.

Determination of azimuth by means of the gyroscope.
Determination of azimuth by means of magnetometer.

stride length
At each step taken, the time between the given step and the previous step is
evaluated, thus calculating the step period, and this is the independent variable
of the function, which returns the step length.
There is a correspondence between the step length, and the time period
between steps. This correspondence was modeled in this project through a
first degree polynomial.
The premise used was: fast steps are short, and long steps take a longer period
of time.

a=0.3
b=0.8

Getting the position by line integral
Every time a step is taken, the user's position is recalculated. Being that the
position is mapped in a Cartesian plane, having as origin the initial position.
To update the position, we take a vector, with a length equal to the
calculated step length, and an angle equal to the azimuth, decomposing this
vector in the x and y directions, obtaining its components, and adding the
previous position value .

RESULTS

300m
It was the distance traveled within an error of less than 7m.

<5%
Despite the trajectory errors, the distance traveled had an error of less than 5%.

Sources of systematic errors
The magnetic field has a slope,
which creates a distortion when
calculating the azimuth angle
through the arc tangent of the x
and y components, parallel to the
ground.

Sources of systematic errors
A suggestion for future work is
to compensate for the magnetic
tilt, decomposing the tilted
component in relation to the
device's reference plane,
determining the tilt of the device
through the accelerometers.

Now it's time to
see it working!

jhonatan.brandel@engenharia.ufjf.br
Thanks!
Jhonatan Brandel de Souza

Apresentação IMU (2).pdf

More Related Content

Similar to Apresentação IMU (2).pdf (20)

Recently uploaded (20)

Apresentação IMU (2).pdf