Inverted Pendulum

Inverted Pendulum

Self-balancing two-wheeled robot built as the final project for the Control Theory course at the Universidade de Santiago de Compostela (2023/2024).

The robot keeps itself upright on two wheels by reading its tilt with an IMU and driving the motors through a PID controller. It also exposes a small Wi-Fi web interface that can be used to move it around or switch the controller on and off.

How it works

1. Sensing - A MPU6050 IMU is read over I2C. The accelerometer gives an absolute (but noisy) tilt angle, while the gyroscope gives a fast (but drifting) angular rate.
2. Angle estimation - Both measurements are fused with a complementary filter: angle = 0.98 * (angle + gyro_rate * dt) + 0.02 * accelerometer_angle This yields a smooth and stable estimate of the tilt around the X axis.
3. PID control - The error is the difference between the desired angle (upright, 0 deg) and the measured angle. The output is: u = Kp * error + Ki * integral + Kd * derivative and is saturated to the PWM range (-255..255).
4. Actuation - The control signal is applied to both DC motors through a TB6612FNG driver. When the robot leans forward the wheels drive forward, and vice versa, producing the correction torque that brings it back to the vertical position.
5. Tuning - Gains are adjusted in the classic order: first Kp, then Ki, and finally Kd (in this build Kd was left at 0 because it did not improve the response).

The PID loop runs as fast as possible (small delay), since a high sampling rate is essential for balancing.

Repository contents

Hardware

• ESP32 DevKit-C (main microcontroller, 5 V logic, built-in Wi-Fi).
• MPU6050 (GY-521) accelerometer + gyroscope, I2C.
• TB6612FNG dual H-bridge motor driver.
• 2x DC gear motors with wheels (ROB-14450).
• 2x 4xAA battery holders in series (12 V for the motors) and an LM7805 linear regulator (5 V for the ESP32).
• Custom PCB designed in KiCad and a 3D-printed base designed in FreeCAD.

Note: the LM7805 needs at least ~7 V at its input, which is why the two 6 V battery packs are connected in series. The motors receive the full 12 V while the ESP32 is fed with regulated 5 V.

Building and running

Software

1. Install the Arduino IDE and add ESP32 board support through the Boards Manager.
2. Install the MPU6050 and I2Cdev libraries.
3. Open one of the sketches in Programas/ and select the ESP32 Dev Module board and the correct COM port.
4. Upload the sketch.

Recommended order for a first run:

1. Prueba_esp32/ - check the board and serial connection.
2. Programas/Programas_auxiliares/Programa_calibracion_sensor/ - calibrate the IMU offsets with the robot perfectly still.
3. Programas/Programas_auxiliares/Marcha_motores/ - verify motor direction and wiring.
4. Programas/PID/Control_PID/ - test and tune the PID.
5. Programas/PID/PID_posicion/ - final version with Wi-Fi control.

Wi-Fi control

PID_posicion.ino starts the ESP32 as a Wi-Fi access point. Connect to the network shown in the serial monitor and open the IP address (192.168.4.1 by default) in a browser. The page provides forward/backward/left/right buttons, a stop button and buttons to enable or disable the PID.

The AP credentials are defined at the top of the sketch; change them before publishing or deploying the robot.

Authors

• Marcos Fontenlos Ferreiro