STM32-Based IMU Attitude Estimation - Revision history

Liding: /* References */

2025-04-06T12:06:35Z

References

← Older revision		Revision as of 20:06, 6 April 2025
Line 310:		Line 310:
	https://www.st.com/en/microcontrollers-microprocessors/stm32f103c8.html		https://www.st.com/en/microcontrollers-microprocessors/stm32f103c8.html

	* [3] '''~~Complementary Filter vs Kalman Filter~~''', ~~Robotics Stack Exchange Discussion~~.		* [3] '''How to Interface the MPU6050 (GY-521) with STM32''', Controllerstech.
	https://~~robotics.stackexchange~~.com/~~questions/21476/complementary~~-~~filter~~-vs-~~kalman~~-~~filter~~		https://controllerstech.com/how-to-interface-mpu6050-gy-521-with-stm32/

	* [4] '''MPU6050 ~~with STM32~~ (~~I2C + Kalman + OLED~~) ~~– Tutorial~~''', ~~Controllerstech~~.		* [4] '''STM32 I2C for MPU6050 (e.g., GY-521) Get Angle''', GitHub Repository by creative-apple.
	https://~~controllerstech~~.com/~~mpu6050-with~~-~~stm32~~/		https://github.com/creative-apple/STM32_MPU6050_ComplementartFilter

Liding: /* References */

2025-04-06T12:05:12Z

References

← Older revision		Revision as of 20:05, 6 April 2025
Line 304:		Line 304:
	== References ==		== References ==

	* [1] MPU6050 Register Map and Datasheet, InvenSense.		* [1] '''MPU6050 Register Map and Datasheet''', InvenSense.
	https://invensense.tdk.com/products/motion-tracking/6-axis/mpu-6050/		https://invensense.tdk.com/products/motion-tracking/6-axis/mpu-6050/

	* [2] STM32F103C8T6 Datasheet, STMicroelectronics.		* [2] '''STM32F103C8T6 Datasheet''', STMicroelectronics.
	https://www.st.com/en/microcontrollers-microprocessors/stm32f103c8.html		https://www.st.com/en/microcontrollers-microprocessors/stm32f103c8.html

			* [3] '''Complementary Filter vs Kalman Filter''', Robotics Stack Exchange Discussion.
			https://robotics.stackexchange.com/questions/21476/complementary-filter-vs-kalman-filter

			* [4] '''MPU6050 with STM32 (I2C + Kalman + OLED) – Tutorial''', Controllerstech.
			https://controllerstech.com/mpu6050-with-stm32/

Liding: /* 7. Discussion */

2025-04-06T12:04:20Z

7. Discussion

← Older revision		Revision as of 20:04, 6 April 2025
Line 266:		Line 266:
	Overall, the complementary filter worked effectively for basic attitude estimation using only the MPU6050 sensor, especially for roll and pitch. The system demonstrates reliable performance for short-term orientation tracking without external reference.		Overall, the complementary filter worked effectively for basic attitude estimation using only the MPU6050 sensor, especially for roll and pitch. The system demonstrates reliable performance for short-term orientation tracking without external reference.

	== 7. Discussion ==		== Discussion ==

	=== Accuracy and Stability of Roll and Pitch ===		=== Accuracy and Stability of Roll and Pitch ===
	The complementary filter ~~performed~~ well in estimating roll and pitch angles. These two values ~~remained~~ stable when the sensor was ~~held still~~, and ~~they~~ responded ~~consistently to manual tilting~~. The accelerometer provides ~~reliable~~ long-term reference ~~based on~~ gravity, while the gyroscope ~~helps improve~~ short-term responsiveness. Minor fluctuations were observed ~~due to~~ hand tremors ~~and environmental~~ vibrations, but ~~they were within acceptable range~~ for basic ~~applications~~.
			The complementary filter worked well in estimating roll and pitch angles. These two values stayed stable when the sensor was stationary, and responded predictably when the module was manually tilted. The accelerometer provides a consistent long-term reference by measuring gravity, while the gyroscope enhances short-term responsiveness due to its ability to capture quick changes in angular velocity. Minor fluctuations (within ±1°) were observed, likely caused by hand tremors or ambient vibrations, but overall the performance was satisfactory for basic real-time tracking.

	=== Yaw Angle Limitations ===		=== Yaw Angle Limitations ===
	Unlike pitch and roll, yaw cannot be ~~derived from~~ the accelerometer ~~alone~~. In this project, yaw was ~~computed~~ by integrating ~~gyroscope data around~~ the Z-axis. ~~This approach worked~~ for short-term ~~motion but exhibited noticeable~~ drift over time~~, as expected~~. Without a magnetometer ~~to provide an external reference~~ (e.g., Earth's magnetic field), yaw estimation lacks long-term ~~accuracy. This is a known limitation of using MPU6050 without additional sensors~~.
			Yaw angle is significantly more difficult to estimate. Unlike pitch and roll, which rely on the gravity vector, yaw involves rotation around the vertical (Z) axis, and cannot be observed by the accelerometer. In this project, yaw was estimated by integrating the gyroscope’s Z-axis angular velocity. While this method works for short-term orientation changes, it introduces drift over time because small sensor biases accumulate with each integration step. Without a magnetometer (e.g., to detect Earth's magnetic field), yaw estimation lacks an absolute reference, making it unreliable in the long term.

	=== Display and Refresh Rate ===		=== Display and Refresh Rate ===
	~~Displaying angles on the~~ OLED screen in real time (~100 Hz~~) allowed for intuitive observation of system performance~~. However, the ~~OLED display has~~ limited resolution and ~~update~~ speed ~~compared to serial terminals~~ or ~~PC-based plots. Some rapid~~ changes may not be ~~fully captured~~ visually.
			The OLED screen displays the estimated angles in real time at roughly 100 Hz. This update rate was sufficient to observe smooth changes during hand-held motion. However, due to the limited resolution and refresh speed of the OLED, rapid or subtle angle changes may not be visually apparent. For more precise analysis, data could be logged via serial communication and plotted on a PC.

	=== Potential Improvements ===		=== Potential Improvements ===
	~~To further~~ improve the system:
	* A magnetometer (e.g., HMC5883L) ~~can be added~~ to correct yaw drift.		Several enhancements could improve the accuracy and expand the capabilities of this system:
	* Kalman filter or Madgwick filter could replace the complementary filter for ~~higher accuracy~~.
	* Initial calibration routines can be added to correct ~~zero-bias~~ errors ~~in accelerometer and gyroscope~~.		* Adding a '''magnetometer''' (e.g., HMC5883L) to correct yaw drift and provide an absolute heading.
	* ~~Angle output could also be sent~~ to PC or mobile ~~app via~~ serial/Bluetooth for ~~better~~ visualization.		* Implementing a more advanced fusion algorithm such as a '''Kalman filter''' or '''Madgwick filter''' for better dynamic performance.
			* Including a '''calibration routine''' at startup to correct sensor biases and scaling errors.
			* Sending angle data to a PC or mobile device over serial or Bluetooth for richer visualization and analysis.

			These changes would bring the system closer to what is used in real-world IMU applications, especially in drones and robotics.

	=== Applications and Extensions ===		=== Applications and Extensions ===
	~~This basic attitude estimation system could be extended for use in:~~
	* Drone orientation feedback
	* Self-balancing robots
	* Gesture-based control systems
	* VR/AR head tracking

	~~The understanding gained from~~ building this ~~project~~ from ~~scratch—both in hardware wiring~~ and ~~software logic—provides~~ a ~~strong foundation for further exploration~~ in embedded ~~sensing~~ systems.		Although this is a small-scale project, the techniques used here reflect real principles behind many motion-sensing systems. Possible applications include:

			* Orientation feedback in drones or quadcopters
			* Balancing control in two-wheeled robots
			* Gesture recognition in wearable or handheld devices
			* Head tracking in VR or AR environments

			By building this system from scratch—including raw register access, filtering, and angle estimation—we’ve gained a practical understanding of how physical motion is sensed and processed in embedded systems. This hands-on experience ties directly into the themes of the Physics of Sensors course.

	== References ==		== References ==

Liding: /* Theory & Angle Estimation */

2025-04-06T11:58:46Z

Theory & Angle Estimation

← Older revision		Revision as of 19:58, 6 April 2025
Line 96:		Line 96:

	== Theory & Angle Estimation ==		== Theory & Angle Estimation ==
	To estimate the orientation (attitude) of an object using the MPU6050 ~~sensor~~, we need to understand ~~both~~ the physical meaning of the ~~sensor~~ outputs ~~and how they~~ can ~~be processed to obtain~~ rotation ~~angles~~.
			To estimate the orientation (attitude) of an object using the MPU6050, we need to understand the physical meaning of the quantities it measures. The accelerometer responds to linear acceleration (including gravity), while the gyroscope senses angular velocity. By interpreting and combining these outputs, we can reconstruct the rotation of the sensor in space.

	=== Sensor Principles ===		=== Sensor Principles ===
	~~The MPU6050 is a 6-axis MEMS-based Inertial Measurement Unit (IMU). It consists of:~~

	~~A 3~~-axis ~~accelerometer – measures linear acceleration in x, y, z directions~~ (~~in units of g~~)		The MPU6050 is a MEMS-based 6-axis Inertial Measurement Unit (IMU). Internally, it contains:

	A 3-axis gyroscope – measures angular velocity around x, y, z axes (in °/s)		* '''3-axis accelerometer''' – measures linear acceleration along the X, Y, and Z axes, in units of ''g''.
			* '''3-axis gyroscope''' – measures angular velocity around the X, Y, and Z axes, in ''°/s''.

	These ~~sensors give raw~~ physical ~~measurements~~:		These two types of data correspond to physical forces acting on the sensor:

	The accelerometer reflects the direction of gravity ~~and motion~~		* The accelerometer reflects the direction and magnitude of gravity (when stationary) or combined forces (when moving).
			* The gyroscope provides information about how fast the object is rotating around each axis.

	~~The gyroscope measures how fast~~ the ~~object rotates~~		By combining both types of data, we can estimate the orientation of the sensor relative to a fixed reference frame.

	~~By analyzing both, we can estimate~~ the ~~orientation (roll, pitch, yaw) of the object.~~		[[File:MPU6050.png\|center\|thumb\|300px\|MPU6050 breakout board showing the chip and I²C interface]]

	~~[[File~~:~~MPU6050.png\|center\|thumb\|300px\|MPU6050 module (breakout board)~~, ~~showing chip and I2C pins]]~~		=== Attitude Angles: Roll, Pitch, Yaw ===

	~~=== Attitude Angles: Roll, Pitch, Yaw ===~~		Attitude in 3D space is commonly expressed using three angles:
	Attitude ~~refers to the rotation of an object~~ in 3D space~~. It~~ is commonly expressed using three angles:
			{\| class="wikitable"
			! Angle !! Axis !! Description
			\|-
			\| Roll \|\| X-axis \|\| Tilting left and right
			\|-
			\| Pitch \|\| Y-axis \|\| Tilting forward and backward
			\|-
			\| Yaw \|\| Z-axis \|\| Rotating around the vertical axis
			\|}

	~~Angle Axis Description~~		In our project, we focus on roll and pitch. Yaw is difficult to measure accurately without a magnetometer, because the accelerometer does not respond to rotation around the vertical axis, and gyroscope-based yaw accumulates drift over time.
	~~Roll X-axis Tilting side-to-side~~
	~~Pitch Y-axis Tilting forward/backward~~
	~~Yaw Z-axis Rotating left/right~~
	In ~~this~~ project, we ~~mainly estimate~~ roll and pitch. Yaw is ~~harder~~ to ~~determine~~ accurately without a magnetometer.

	[[File:roll_pitch_yaw_axes.png\|center\|thumb\|~~200px~~\|~~Illustration of roll~~ (X~~-axis~~), ~~pitch~~ (Y~~-axis~~), and ~~yaw~~ (Z~~-axis~~) in a 3D ~~coordinate system.~~]]		[[File:roll_pitch_yaw_axes.png\|center\|thumb\|240px\|Roll (X), Pitch (Y), and Yaw (Z) in 3D space]]

	=== Using the Accelerometer ===		=== Using the Accelerometer ===
	~~When the object is still or moving slowly, the accelerometer can be used to estimate roll and pitch based on the gravity vector. The equations are:~~

	<math>pitch = \arctan{\left(\frac{a_x}{\sqrt{a_y^2 + a_z^2}}\right)}</math>		When the sensor is at rest or moving slowly, the accelerometer mainly senses the Earth's gravity. We can use the direction of this gravity vector to estimate the tilt angles with respect to the horizontal plane. Using basic trigonometry:
	<math>roll = \arctan{\left(\frac{a_y}{\sqrt{a_x^2 + a_z^2}}\right)}</math>
	These ~~formulas give a rough estimate of~~ the ~~angle,~~ but ~~are sensitive to~~ motion ~~and vibration~~, ~~which cause noise~~.		<math>
			\text{pitch} = \arctan\left( \frac{a_x}{\sqrt{a_y^2 + a_z^2}} \right)
			</math>

			<math>
			\text{roll} = \arctan\left( \frac{a_y}{\sqrt{a_x^2 + a_z^2}} \right)
			</math>

			These estimates are relatively stable when the sensor is static but become noisy during motion, since linear acceleration disturbs the gravity vector.

	=== Using the Gyroscope ===		=== Using the Gyroscope ===
	The gyroscope ~~provides~~ angular velocity~~. We~~ can ~~integrate this value~~ over time to ~~get angle~~ changes: <math>\theta = \theta_{\text{prev}} + \omega \cdot \Delta t</math>. However, small measurement errors accumulate over time~~, causing drift~~.
			The gyroscope gives angular velocity ω (in ''°/s''), which can be integrated over time to estimate changes in orientation:

			<math>
			\theta = \theta_{\text{prev}} + \omega \cdot \Delta t
			</math>

			However, this method suffers from drift because small measurement errors accumulate over time.

	=== Complementary Filter ===		=== Complementary Filter ===
	~~To overcome the weaknesses of both sensors, we use a complementary filter, which combines:~~

	The short-term ~~precision of~~ the ~~gyroscope~~		To reduce drift while keeping responsiveness, we apply a complementary filter, which fuses gyroscope and accelerometer data. The gyroscope provides short-term accuracy, while the accelerometer corrects long-term drift.

	~~The long~~-~~term stability of the accelerometer~~		<math>
			\theta = \alpha \cdot (\theta_{\text{prev}} + \omega \cdot \Delta t) + (1 - \alpha) \cdot \theta_{\text{acc}}
			</math>

	~~The formula is:~~
	~~<math>\theta = \alpha \cdot (\theta_{\text{prev}} + \omega \cdot \Delta t) + (1 - \alpha) \cdot \theta_{\text{acc}}</math>.~~
	Where:		Where:
	~~𝜔 is angular velocity (from gyroscope), 𝜃 is the angle from the accelerometer, 𝛼 is a constant (typically 0.95 ~ 0.98).~~

	~~This filter gives~~ a ~~smooth and responsive estimation of the roll~~ and ~~pitch angles~~.		* <math>\omega</math> is the angular velocity from the gyroscope
			* <math>\theta_{\text{acc}}</math> is the angle derived from the accelerometer
			* <math>\alpha</math> is a constant between 0.95 and 0.98

			This gives us smooth and relatively accurate estimates for roll and pitch in real time.

	== Results ==		== Results ==

Liding: /* Software & Code */

2025-04-06T11:55:10Z

Software & Code

@@ Line 74: / Line 74: @@
 === Software & Code ===
-* **Initialize I2C communication** – Set up the I2C peripheral on the STM32 and ensure it can communicate with the MPU6050.
+The firmware was written in C using STM32's standard peripheral libraries. The program runs in a loop, repeatedly collecting data from the MPU6050 and computing the corresponding attitude angles in real time.
-The code runs in a continuous loop, updating at a fixed interval (e.g., every 10ms) for smooth, real-time performance.
+The main software tasks are as follows:
 == Theory & Angle Estimation ==

Liding: /* Wiring / Circuit Overview */

2025-04-06T11:53:50Z

Wiring / Circuit Overview

← Older revision		Revision as of 19:53, 6 April 2025
Line 51:		Line 51:

	=== Wiring / Circuit Overview ===		=== Wiring / Circuit Overview ===
	~~Here's how~~ the ~~components are connected~~:		The sensor communicates with the STM32 via I²C, and the wiring is shown below:

	{\| class="wikitable"		{\| class="wikitable"
	! MPU6050 Pin !! STM32 Pin !! Notes		! MPU6050 Pin !! STM32 Pin !! Notes
	\|-		\|-
	\| VCC \|\| 3.3V \|\| ~~Powered~~ by STM32		\| VCC \|\| 3.3V \|\| Sensor powered by STM32
	\|-		\|-
	\| GND \|\| GND \|\| Common ground		\| GND \|\| GND \|\| Common ground reference
	\|-		\|-
	\| SDA \|\| PB7 \|\| ~~I2C~~ data line		\| SDA \|\| PB7 \|\| I²C data line
	\|-		\|-
	\| SCL \|\| PB6 \|\| ~~I2C~~ clock line		\| SCL \|\| PB6 \|\| I²C clock line
	\|}		\|}

	~~The sensor is connected to the STM32 using the I2C interface, which allows the STM32 to read the sensor's data. I2C is a simple and efficient communication protocol for embedded systems.~~

	~~Optional~~: ~~A USB to TTL converter is used for programming the STM32~~ and ~~debugging through the serial port. This allows us to print the sensor data or estimated angles to the terminal.~~		Below are two views of the hardware setup: a breadboard diagram and an abstract schematic:

	<div style="display: flex; gap: 10px; align-items: center;">		<div style="display: flex; gap: 10px; align-items: center;">
	[[File:Imu circuit diagram.png\|center\|thumb\|500px\|~~imu_wiring_diagram~~]]		[[File:Imu circuit diagram.png\|center\|thumb\|500px\|Breadboard wiring diagram]]
	[[File:Schematic diagram of electrical connections.jpeg\|center\|thumb\|500px\|Schematic ~~diagram~~ of ~~electrical~~ connections]]		[[File:Schematic diagram of electrical connections.jpeg\|center\|thumb\|500px\|Schematic representation of connections]]
	</div>		</div>

Liding: /* Equipment */

2025-04-06T11:53:14Z

Equipment

← Older revision		Revision as of 19:53, 6 April 2025
Line 48:		Line 48:
	The MPU6050 is connected to the STM32 using the I²C protocol. I²C is widely used in sensor interfacing due to its simplicity and ability to handle multiple devices over just two lines (SDA and SCL).		The MPU6050 is connected to the STM32 using the I²C protocol. I²C is widely used in sensor interfacing due to its simplicity and ability to handle multiple devices over just two lines (SDA and SCL).

	~~```wiki~~
	[[File:Stm32c8t6.jpg\|center\|thumb\|300px\|STM32F103C8T6 "Blue Pill" development board]]		[[File:Stm32c8t6.jpg\|center\|thumb\|300px\|STM32F103C8T6 "Blue Pill" development board]]

Liding: /* Equipment */

2025-04-06T11:53:03Z

Equipment

← Older revision		Revision as of 19:53, 6 April 2025
Line 35:		Line 35:

	== Setup ==		== Setup ==

	=== Equipment ===		=== Equipment ===
	~~The sensor communicates with the STM32 via I²C, and the wiring is shown below:~~

	~~{\| class=~~"~~wikitable~~"		The following components were used to build our attitude estimation system:
	! MPU6050 ~~Pin !!~~ STM32 ~~Pin !! Notes~~
	\|-		* STM32F103C8T6 ("Blue Pill") – A low-cost 32-bit microcontroller used here to handle I²C communication, filtering, and OLED output.
	~~\| VCC \|\|~~ 3.3V ~~\|\| Sensor powered~~ by STM32		* MPU6050 – A 6-axis IMU that combines a 3-axis accelerometer (for sensing gravity and linear motion) and a 3-axis gyroscope (for measuring angular velocity).
	\|-		* USB to TTL Serial Converter – Used mainly during development to upload programs to the STM32 and read data from the serial port.
	~~\| GND \|\| GND \|\| Common ground reference~~		* Breadboard and jumper wires – For making temporary connections during prototyping.
	\|-		* 3.3V Power Supply – The MPU6050 operates at 3.3V, which is provided by the STM32 board itself.
	~~\| SDA \|\| PB7 \|\| I²C data line~~
	\|-
	~~\| SCL \|\| PB6 \|\| I²C clock line~~
	\|}

	~~Below are~~ two ~~views of the hardware setup: a breadboard diagram~~ and ~~an abstract schematic:~~		The MPU6050 is connected to the STM32 using the I²C protocol. I²C is widely used in sensor interfacing due to its simplicity and ability to handle multiple devices over just two lines (SDA and SCL).

	~~<div style="display: flex; gap: 10px; align-items: center;">~~		```wiki
	[[File:~~Imu circuit diagram~~.~~png~~\|center\|thumb\|~~500px~~\|~~Breadboard wiring diagram~~]]		[[File:Stm32c8t6.jpg\|center\|thumb\|300px\|STM32F103C8T6 "Blue Pill" development board]]
	~~[[File:Schematic diagram of electrical connections.jpeg\|center\|thumb\|500px\|Schematic representation of connections]]~~
	~~</div>~~

	=== Wiring / Circuit Overview ===		=== Wiring / Circuit Overview ===

Liding: /* Equipment */

2025-04-06T11:52:34Z

Equipment

← Older revision		Revision as of 19:52, 6 April 2025
Line 36:		Line 36:
	== Setup ==		== Setup ==
	=== Equipment ===		=== Equipment ===
	~~Here are~~ the ~~main components used in this project~~:		The sensor communicates with the STM32 via I²C, and the wiring is shown below:

	* STM32F103C8T6 ("~~Blue Pill~~"~~) – A versatile microcontroller for controlling the system.~~		{\| class="wikitable"
	~~* ~~MPU6050 – A 6-axis IMU sensor that includes both a 3-axis accelerometer and a 3-axis gyroscope.		! MPU6050 Pin !! STM32 Pin !! Notes
	* USB to TTL Serial Converter – Used for programming the STM32 ~~and communicating with the serial terminal.~~		\|-
	* Jumper wires and breadboard – For making connections between the STM32 and MPU6050.		\| VCC \|\| 3.3V \|\| Sensor powered by STM32
	* Power supply (3.3V~~) – Provided~~ by ~~the~~ STM32 ~~or through an external 3.3V regulator.~~		\|-
			\| GND \|\| GND \|\| Common ground reference
			\|-
			\| SDA \|\| PB7 \|\| I²C data line
			\|-
			\| SCL \|\| PB6 \|\| I²C clock line
			\|}

	~~The MPU6050 module is powered with 3.3V,~~ and ~~it communicates with the STM32 over the I2C bus.~~		Below are two views of the hardware setup: a breadboard diagram and an abstract schematic:

	[[File:~~Stm32c8t6~~.~~jpg~~\|center\|thumb\|~~300px~~\|~~STM32F103C8T6 "Blue Pill" development board~~]]		<div style="display: flex; gap: 10px; align-items: center;">
			[[File:Imu circuit diagram.png\|center\|thumb\|500px\|Breadboard wiring diagram]]
			[[File:Schematic diagram of electrical connections.jpeg\|center\|thumb\|500px\|Schematic representation of connections]]
			</div>

	=== Wiring / Circuit Overview ===		=== Wiring / Circuit Overview ===

Liding: /* Project Overview */

2025-04-06T11:50:05Z

Project Overview

← Older revision		Revision as of 19:50, 6 April 2025
Line 24:		Line 24:
	FAN XUTING		FAN XUTING

	== Project Overview ==		== Project Overview ==

	The idea of this project is to estimate the orientation (attitude) of an object using sensor data. We used the MPU6050 module, which ~~combines~~ a gyroscope ~~and an accelerometer~~, to measure ~~how the object is moving~~ and ~~rotating in space~~. ~~The sensor data is collected using the STM32F103C8T6 microcontroller through I2C communication~~.		The core idea of this project is to estimate the orientation (attitude) of an object in space using inertial sensor data. We used the MPU6050 module, which integrates a 3-axis accelerometer and a 3-axis gyroscope, to measure two different physical quantities: linear acceleration and angular velocity. These are fundamental mechanical observables when analyzing motion.

	~~Since the raw~~ data from the sensor is ~~not very stable or accurate by itself~~, we applied ~~a filtering method (in our case,~~ a complementary filter) to combine the gyroscope ~~and~~ accelerometer ~~readings~~. This ~~allows~~ us to reduce noise and ~~drift, and get smoother and more reliable angle values over time~~.		The data from the sensor is acquired through I²C communication using the STM32F103C8T6 microcontroller. However, the raw measurements—especially from the gyroscope—tend to drift over time, and the accelerometer is sensitive to external vibrations. Therefore, we applied a complementary filter to combine the short-term precision of the gyroscope with the long-term stability of the accelerometer. This approach helps us reduce noise and improve the reliability of the estimated angles.

	~~The final goal~~ is to ~~calculate~~ the roll, pitch, and yaw angles of the ~~system~~ in real time. ~~This kind of setup~~ is commonly ~~used~~ in ~~things like self-~~balancing robots~~, drones~~, camera ~~stabilizers~~, or even ~~mobile phones~~ for ~~detecting screen~~ orientation. ~~Our~~ project ~~is a small-scale version~~ of such systems~~, built from scratch~~ to ~~better~~ understand how ~~attitude estimation actually works~~.		Our objective is to continuously compute the roll, pitch, and yaw angles of the sensor in real time. In practice, only roll and pitch can be estimated reliably without a magnetometer, since yaw depends on integrating rotational velocity around the vertical axis and is subject to drift.

			Systems like this are commonly found in drones, balancing robots, camera gimbals, and even smartphones for orientation detection. In our project, we replicated the basic functionality of such systems on a smaller scale in order to understand how physical quantities—like gravity and rotation—can be sensed, processed, and interpreted through embedded electronics.

	== Setup ==		== Setup ==