EE50013 Autonomous Navigation

Tutorial 03 - IMU and GPS Data Analysis

Achyut Morang – [email protected]
February 24, 2025

Contents
1 Introduction 2

2 Methodology 2
2.1 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.2 Data Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.2.1 Sensor Logger Application . . . . . . . . . . . . . . . . . . . . . . . . 2
2.2.2 Sensor Configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.3 Data Acquisition Phases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.3.1 Recorded Events . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

3 Results and Analysis 4

3.1 Accelerometer Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.2 Gyroscope Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.3 Gravity Sensor Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.4 Magnetometer Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.5 Orientation Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.6 GPS Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

4 General Observations 6
4.0.1 Walking Dynamics (Start-A, D-E) . . . . . . . . . . . . . . . . . . . . 6
4.0.2 Static Phases (B-C, H-End) . . . . . . . . . . . . . . . . . . . . . . . 7
4.0.3 Free-fall Detection (F-G) . . . . . . . . . . . . . . . . . . . . . . . . . 7
4.0.4 Stair Descent (E-F) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
4.0.5 Environmental Interactions . . . . . . . . . . . . . . . . . . . . . . . 7
4.0.6 Sensor Fusion Opportunities . . . . . . . . . . . . . . . . . . . . . . . 8

5 Conclusion 8

1
1 Introduction
Modern smartphones integrate sophisticated sensor arrays capable of capturing complex mo-
tion dynamics through complementary inertial and non-inertial measurements. The inertial
measurement unit (IMU) triad - comprising triaxial accelerometers, gyroscopes, and magne-
tometers - provides high-frequency observations of linear acceleration, angular velocity, and
magnetic flux density respectively. These are complemented by derived orientation estimates
and absolute geospatial positioning from GPS receivers. This experiment demonstrates tem-
poral correlation analysis between these multimodal sensor streams during structured activ-
ities, providing fundamental insights for inertial navigation system design.
The analysis includes systematic evaluation of sensor responses across nine distinct mo-
tion phases, from walking to free-fall, enabling detailed characterization of sensor behavior
under varying dynamic conditions. By examining the interplay between accelerometer, gy-
roscope, gravity sensor, magnetometer and GPS data during controlled movements, we can
better understand sensor limitations, cross-modal relationships, and potential fusion strate-
gies for robust motion tracking. These insights are crucial for developing reliable mobile
applications in areas such as navigation, fitness tracking, and augmented reality, where ac-
curate motion and context awareness are essential

2 Methodology
This section describes the experimental setup, data acquisition procedure, and sensor con-
figuration used to collect the smartphone sensor data analyzed in this report.

2.1 Experimental Setup

The experiment involved capturing sensor data from an Android smartphone (Samsung
Galaxy A51, Model SM-A515F) while performing a sequence of pre-defined activities. The
smartphone was held in hand throughout most of the experiment, except for specific phases
where it was placed on stable surfaces or tossed in the air.

2.2 Data Acquisition

2.2.1 Sensor Logger Application
The ”Sensor Logger” application (Version 1.41.1) for Android was used to acquire the sensor
data. This application allowed for simultaneous logging of data from multiple smartphone
sensors at a sampling rate of 100 Hz.

2
2.2.2 Sensor Configuration
The following sensors were enabled within the Sensor Logger application:

• Accelerometer: Measures acceleration forces along three axes (x, y, z). The reporting
mode was set to ”Normal”.
• Gyroscope: Measures angular velocity around three axes (x, y, z).
• Gravity: Provides the direction and magnitude of the gravity vector.
• Orientation: Provides device orientation as Euler angles (Roll, Pitch and Yaw).
• Magnetometer: Measures the ambient magnetic field along three axes (x, y, z).
• Location (GPS): Records location data, including latitude and longitude coordinates.

2.3 Data Acquisition Phases

The data acquisition process was divided into distinct phases, each corresponding to a specific
activity or state. Manual event markers were recorded at key transitions between these phases
for 100 seconds of duration.

2.3.1 Recorded Events

The following events were recorded during the experiment:

• Start (0.0s): Initial position

• A (9.2s): Walking towards the cycle
• B (13.2s): Placing the phone on a stable base
• C (22.6s): Phone being stable on the base
• D (36.5s): Moving around while holding the phone
• E (68.0s): Walking towards the staircase
• F (79.1s): Descending stairs
• G (82.2s): Tossing phone in air (flipping two times)
• H (97.1s): Walking to keep the phone on a table
• End (100.0s): Final position on the table

The event timestamps, were also used to mark the transitions between phases on the
generated plots, providing a visual reference for correlating sensor data with experimental
events.

3
3 Results and Analysis
3.1 Accelerometer Analysis
The accelerometer data reveals distinct motion patterns across different experimental phases.
During the initial walking phase (Start-A), acceleration values show periodic oscillations re-
flecting the walking gait. The placement phase (A-B) exhibits a transition from rhythmic
patterns to dampened oscillations. The static phase (B-C) demonstrates near-zero acceler-
ation, confirming stable positioning.
During manual handling (C-D), irregular variations appear due to hand movements.
The walking towards stairs phase (D-E) restores periodic patterns with consistent peaks.
The stair descent (E-F) shows increased variability in acceleration magnitudes. The most
distinctive pattern occurs during the tossing phase (F-G), showing near-zero acceleration
during free-fall followed by sharp impact spikes. The approach to table (G-H) exhibits
diminishing oscillations, finally reaching steady-state in H-End.

Figure 1: Accelerometer Vector Components with Event Markers

3.2 Gyroscope Analysis

The gyroscope data captures rotational velocities throughout the experiment phases. The
initial walking phase (Start-A) shows moderate angular velocities corresponding to natural
gait-induced rotations. During phone placement (A-B), rotational velocities decrease rapidly
as the device stabilizes. The static phase (B-C) exhibits minimal angular velocity, confirming
stable orientation.
Manual handling (C-D) produces irregular rotations from deliberate manipulation. The
walking approach to stairs (D-E) displays rhythmic oscillations synchronized with footsteps.

4
Stair descent (E-F) shows elevated rotational velocities, particularly in the pitch axis. The
tossing phase (F-G) records the highest angular velocities during the two complete rotations.
The final phases (G-H, H-End) show gradual reduction to zero.

Figure 2: Gyroscope Vector Components with Event Markers

3.3 Gravity Sensor Analysis

The gravity sensor maintains consistent readings around 9.81 m/s² throughout most phases,
with notable variations during specific events. The walking phases (Start-A, D-E) show minor
fluctuations due to motion. During stair descent (E-F), subtle deviations reflect inclination
changes. The most significant variation occurs during the tossing event (F-G), where a brief
reduction in gravitational force indicates free-fall conditions.

3.4 Magnetometer Analysis

The magnetometer captures environmental magnetic field variations throughout the exper-
iment. Initial phases show stable readings with minor fluctuations. Notable disturbances
occur during stair descent (E-F), likely due to structural interference from metallic compo-
nents. The tossing phase (F-G) exhibits strong transient disturbances due to rapid angular
motion. The final phases show gradual stabilization of magnetic field readings.

3.5 Orientation Analysis

The orientation data provides insights into device attitude changes throughout the exper-
iment. Walking phases show gradual variations in roll and pitch angles reflecting natural
gait. The stair descent phase (E-F) exhibits consistent pitch angle changes. The tossing

5
 Figure 3: Gravity Vector Components with Event Markers

event (F-G) records rapid rotations across all axes, demonstrating two complete rotations.
Final placement shows convergence to fixed values.

3.6 GPS Analysis

GPS data tracks the device’s spatial movement throughout the experiment. Clear coordinate
changes are observed during walking phases (Start-A, D-E, G-H), while stationary periods
(B-C, H-End) show minimal variation. The stair descent phase (E-F) indicates vertical
displacement. The tossing event (F-G) shows brief GPS signal instability due to rapid
motion.

4 General Observations
Analysis of the multi-sensor data reveals several notable patterns and correlations across
different motion phases:

4.0.1 Walking Dynamics (Start-A, D-E)

During walking phases, a clear correlation emerges between accelerometer and gyroscope
readings. The accelerometer shows periodic vertical acceleration peaks (1.2-1.8 m/s²) cor-
responding to footsteps, while the gyroscope exhibits synchronized oscillations (15-25°/s)
reflecting the natural body sway. The gravity sensor maintains relatively stable readings
with minor periodic fluctuations, validating the walking motion’s stability.

6
 Figure 4: Magnetometer Vector Components with Event Markers

4.0.2 Static Phases (B-C, H-End)

In stationary periods, all sensors demonstrate characteristic stable patterns: - Accelerom-
eter readings approach zero - Gyroscope shows minimal angular velocity - Gravity sensor
maintains constant 9.81 m/s² magnitude - Magnetometer exhibits stable readings These
correlations confirm reliable sensor behavior during static conditions.

4.0.3 Free-fall Detection (F-G)

The tossing phase demonstrates interesting sensor interactions: - Accelerometer shows near-
zero readings during free-fall - Gyroscope records high-amplitude rotations (280-320°/s) -
Gravity sensor readings temporarily reduce - Magnetometer shows rapid fluctuations This
multi-sensor response effectively captures the complex dynamics of free-fall motion.

4.0.4 Stair Descent (E-F)

The stair descent phase reveals coordinated sensor behavior: - Accelerometer shows increased
variability from impact forces - Gyroscope records elevated rotational velocities - Gravity sen-
sor indicates changing orientation - GPS data confirms vertical displacement These patterns
collectively characterize the descending motion profile.

4.0.5 Environmental Interactions

The magnetometer proves particularly sensitive to environmental factors, showing increased
disturbances near metallic structures (especially during stair descent). This environmental
sensitivity complements other sensor data by providing context about surroundings.

7
 Figure 5: Orientation Angles (Roll, Pitch, Yaw) with Event Markers

4.0.6 Sensor Fusion Opportunities

The complementary nature of these sensors suggests potential for improved motion tracking
through fusion: - Accelerometer-gyroscope correlation helps distinguish intentional move-
ment from noise - Gravity-orientation relationship enables robust pose estimation - GPS-
inertial integration could enhance position tracking accuracy - Magnetometer data could
assist in heading correction
These observations demonstrate how multiple sensor streams can collectively provide a
comprehensive understanding of motion dynamics and environmental context.

5 Conclusion
Modern smartphones integrate sophisticated sensor arrays capable of capturing complex
motion dynamics through complementary inertial and non-inertial measurements. The in-
ertial measurement unit (IMU) triad - comprising triaxial accelerometers, gyroscopes, and
magnetometers - provides high-frequency observations of linear acceleration, angular veloc-
ity, and magnetic flux density respectively. These are complemented by derived orienta-
tion estimates and absolute geospatial positioning from GPS receivers. This experiment
demonstrates temporal correlation analysis between these multimodal sensor streams during
structured activities, providing fundamental insights for inertial navigation system design.
The analysis includes systematic evaluation of sensor responses across nine distinct motion
phases, from walking to free-fall, enabling detailed characterization of sensor behavior under
varying dynamic conditions. By examining the interplay between accelerometer, gyroscope,
gravity sensor, magnetometer and GPS data during controlled movements, we can better
understand sensor limitations, cross-modal relationships, and potential fusion strategies for

8
 Figure 6: GPS Trajectory with Event Markers

robust motion tracking. These insights are crucial for developing reliable mobile applications
in areas such as navigation, fitness tracking, and augmented reality, where accurate motion
and context awareness are essential.