How to build Balancio-Kit
38 steps to build Balancio-Kit, with the source documents.
Balancio-Kit is an ultra-low-cost, self-balancing educational robot that runs a neural network to maintain balance and supports wireless control. The repository provides printable 3D parts, a component list, an electrical diagram, and mechanical assembly instructions.
Software
- Install Arduino IDE 2
Install Arduino IDE 2 by following the linked instructions to program and compile the NodeMCU ESP32.
- Open Preferences
Open “File” → “Preferences” in the Arduino IDE.
- Add ESP32 board URL
Add https://dl.espressif.com/dl/package_esp32_index.json to “Additional Boards Manager URLs”, then click “OK”.
- Open Boards Manager
Go to “Tools” → “Board:” → “Boards Manager…”.
- Install ESP32 board package
Search for “esp32” and install “esp32 by Espressif Systems” version 2.x by clicking “Install”.
- Select NodeMCU-32S board
Go to “Tools” → “Board:” → “ESP32 Arduino”, then select “NodeMCU-32S”.
- Select microcontroller port
Go to “Tools” → “Port” and select the corresponding port to which the microcontroller is connected.
- Open Library Manager
Go to “Sketch” → “Include Library” → “Manage Libraries…”.
- Install MPU6050 library
Install “MPU6050 by Electronic Cats” version 1.0.0.
- Install PS3 Controller library
Install “PS3 Controller Host by Jeffrey van Pernis” version 1.1.0.
- Install EloquentTinyML library
Install “EloquentTinyML by Simone Salerno” version 0.0.7.
- Choose whether to run simulation
Simulation is optional and is not needed to operate or assemble the robot; you can proceed directly to calibration.
- Install Conda tooling
Download and install Miniconda; Anaconda can also be installed.
- Create the balancio environment
In the Miniconda console, run `conda env create -f requirements.yml`, or import the repository's requirements.yml file from the environments tab in the Anaconda GUI.
- Activate the balancio environment
Run `conda activate balancio` in the Miniconda terminal, or click the balancio environment in the environments tab of Anaconda.
- Run the simulation library setup
Run the simulation library's setup.py file inside the activated environment.
- Test the simulation installation
Run the pid.py script to test the simulation installation.
- Import the Balancio application
Enter the MIT App Inventor website and import the Balancio-kit/app/app.aia file.
Calibrate and test
- Open WiFiScan example
Go to “File” → “Examples” → “WiFi” → “WiFiScan” to open the generated test sketch.
- Upload WiFiScan sketch
Press the “Upload” button in the generated sketch.
- Confirm successful upload
Check that the console displays the “Done uploading” message.
- Hold Boot button during upload
If the program cannot be uploaded, try holding the board’s “boot” button during the upload; this should be done only the first time.
- Open the IMU calibration sketch
Open Balancio-kit/Mcu/Src/imu_calibration/imu_calibration.ino with the Arduino IDE.
- Hold the robot level and steady
Place the robot with the IMU parallel to the floor and hold it steady.
- Read the IMU offsets
Upload the program to the board and use the serial monitor to obtain the IMU offsets.
- Set the gyroscope offsets
In balancio-kit/mcu/src/main/config.h, set the X, Y, and Z gyroscope offsets to 181, 77, and 60.
- Account for accelerometer gravity
Gravity must be considered when measuring the accelerometer offsets; measure them only with gravity perpendicular to that direction.
- Open the balance main sketch
Open balancio-kit/mcu/src/main/main.ino in the Arduino IDE.
- Hold the robot at equilibrium
Hold the robot in the equilibrium position.
- Read the equilibrium angle offset
Upload the program to the board and use the serial monitor to obtain the angle offset.
- Set the static equilibrium angle
Set STATIC_ANGLE to -0.04 in balancio-kit/mcu/src/main/config.h.
- Remove the driver's 12V jumper
Remove the 12V jumper from the driver.
- Choose the PID parameters
Choose the PID parameters to use.
- Set the PID constants
In balancio-kit/mcu/src/main/config.h, set KP, KI, and KD to 2000, 22000, and 20.0.
- Test the PID constants
Test the constants; after removing the jumper, testing is possible even with the cable connected. Be careful when doing this.
- Select the PID controller
In config.h, set CONTROL_ALGO to PID to stabilize Balancio with a PID controller.
- Select the reinforcement-learning agent
In config.h, set CONTROL_ALGO to RL to stabilize Balancio with a reinforcement-learning agent.
Use it
- Use the application over Bluetooth
After importing it, the application can be used from a phone over Bluetooth.
Common problems
- A high center of mass and low rotational inertia undermine self-balancing stability #9
- The current motor and wheel combination has unverified self-balancing feasibility #2
- READMEs mislabel calibration as simulation and differ between languages #13 #19
- An incorrect switch image in the blueprint can cause wiring mistakes #18
- Builders lack the battery-protector reset procedure for some battery replacements #15
- Reinforcement-learning training and model saving are cumbersome and hard to reproduce #10
- ESP32 2.x code is incompatible with breaking changes in the 3.x library #16