Self-Balancing Robot

A two-wheeled self-balancing robot built around the Nordic nRF52840 and Zephyr RTOS, using nRF Connect SDK 2.8.0. The project includes the embedded control system, custom electronics and mechanical designs, a BLE desktop application, over-the-air firmware updates, and tools for modelling and controller development.

Documentation

• Project overview
• Environment setup
• Building and flashing
• BLE protocol
• Control-loop structure
• Application state machine
• Mathematical model

Demonstration

https://github.com/user-attachments/assets/bdc0d3e3-bf6b-4125-bade-0684ad42b085

🎥 See the robot in action! Check out the Lapun Embedded Control Lab on YouTube for more demonstrations.

Features

Robot firmware

• Cascaded control loop for balancing, driving, and rotation
• PID controllers with user-adjustable setpoints and parameters
• Optional LQR balance controller
• Trajectory generation for linear motion and rotation
• MPU6050 inertial measurement and quadrature encoder feedback
• PWM motor control through a dual-channel TB6612FNG driver
• Battery-voltage monitoring
• Safety angle validation and controlled motor shutdown
• Binary BLE protocol for commands, telemetry, and diagnostic logs
• MCUboot-based device firmware update over BLE

Control and development tools

• PyQt6 desktop application for BLE control, live telemetry, PID tuning, and DFU
• Offline model identification and LQR tuning tools
• MATLAB/Simulink models of the inverted-pendulum system
• Optional Raspberry Pi Zero 2 W web interface with camera streaming
• KiCad schematics, PCB layout, Gerber files, and mechanical CAD

[!NOTE] The default firmware configuration uses PID control. The LQR controller and on-device model-identification mode are implemented but disabled by default.

The Raspberry Pi web interface and the nRF7002 DK controller are legacy or experimental components and may require adaptation to the current binary BLE protocol.

Control architecture

The firmware uses a cascaded control structure. Distance and linear-speed controllers generate the balance-angle setpoint, while the balance and rotation controllers determine the wheel-speed setpoints. Separate wheel controllers then generate the motor PWM outputs.

Hardware

The robot is based on:

• Nordic nRF52840 Dongle
• MPU6050 IMU
• Two DC gearmotors with quadrature encoders
• TB6612FNG dual motor driver
• 2-cell, 7.4 V LiPo battery
• Custom PCB and 3D-printed mechanical structure
• Optional Raspberry Pi Zero 2 W and CSI camera

Hardware design files are available in Schematic/.

Custom PCB

Getting started

Prerequisites

• nRF Connect SDK 2.8.0
• nRF Connect toolchain 2.8.0
• West and the Zephyr development environment
• An external SWD debugger for the initial firmware installation

Follow the environment setup guide before building the project.

Build the robot firmware

From the repository root:

cd Sources/applications/app
west build -b nrf52840dongle_nrf52840 -p always

A successful build generates:

Sources/applications/app/build/merged.hex
Sources/applications/app/build/dfu_application.zip

The firmware can also be built through nRF Connect for VS Code. See Building and flashing for the complete GUI, terminal, SWD, and BLE DFU workflows.

Flash the initial firmware

After connecting an external SWD debugger:

cd Sources/applications/app
west flash --erase --skip-rebuild

Subsequent firmware versions can be installed over BLE using the desktop application or the command-line DFU tool.

Run the desktop application

The desktop application requires Python 3, PyQt6, Bleak, and PyQtGraph:

python3 -m pip install PyQt6 bleak pyqtgraph
python3 run_controller_app.py

The DFU workflow additionally requires smpmgr. A Windows binary is included in 3rdParty/smpmgr/; on Linux, smpmgr must be available on PATH.

Repository structure

.
├── Sources/applications/app/          # Main nRF52840 robot firmware
├── Sources/applications/controller/   # Experimental nRF7002 DK controller
├── Sources/applications/controller_rpi/
│                                      # Raspberry Pi web interface
├── Sources/drivers/                   # IMU, encoders, motors, BLE, DFU, battery
├── Sources/robot_control/             # Control loop, PID, LQR, state machines
├── robot_control_app/                 # PyQt6 BLE desktop application
├── mathematical_model/                # Models, identification, and LQR tooling
├── Schematic/                         # PCB, schematics, Gerbers, and CAD
├── docs/                              # Setup and build documentation
└── dts/bindings/                      # Custom Zephyr devicetree bindings

Roadmap

• Swing-up procedure
• Neural and other intelligent controllers
• Vision-based autonomous navigation

Licensing

• Software and firmware: Apache License 2.0
• Hardware designs (schematics, PCB, and CAD): CERN-OHL-P-2.0

Commercial use is permitted under the terms of the respective licenses.