inverted-pendulum-bot

Firmware for a self-balancing inverted pendulum robot, written in C++20 with strict real-time and memory constraints. The repository was bootstrapped from embedded-pro/embedded-scaffold and carries its structure, build system, dev container, CI, and documentation scaffolding, so the project starts from a working, testable baseline.

Status: the structure is in place; the example_* components and cli are the scaffold's worked examples and are expected to be replaced by the robot's own components (sensing, control, actuation).

Overview

The project separates portable logic from hardware through one key seam: an abstract platform::Platform interface in core/platform_abstraction/. Application logic in core/ depends only on that interface; each board provides a concrete PlatformImpl in targets/platform_implementations/. The same application is therefore compiled once and runs on the host (for testing) and on the microcontroller.

The current headline example is cli: blink a status LED and serve a tiny UART command-line interface (ping, id). It is written once against the interface and builds for the host and the ST NUCLEO-WB55RG.

Features

• Platform abstraction: application logic depends on the platform::Platform interface, not on an MCU — so it is built once and unit-tested on the host against a PlatformMock, then run on real hardware.
• Real firmware output: cli builds to a flashable .elf/.hex for the NUCLEO-WB55RG (LED blink + UART CLI).
• No heap allocation in runtime/embedded code — bounded containers from infra/embedded-infra-lib (infra::BoundedVector, infra::BoundedString).
• Worked examples in every top-level folder showing the conventions to follow.
• Unit tests (GoogleTest) and BDD integration tests (cucumber-cpp / Gherkin).
• Dev container with the full toolchain (CMake, Ninja, ccache, ARM GCC, Qt6).
• CI for build, linting/formatting, static analysis, documentation and requirements validation, and release-please versioning.

Getting Started

Prerequisites

• Docker (the project is developed inside a Dev Container) or a local toolchain with CMake ≥ 3.24, Ninja, a C++20 compiler, and (for embedded) ARM GCC.
• VS Code with the Dev Containers extension is the recommended workflow.

Quick Start

1. Clone with submodules:

   git clone --recursive https://github.com/embedded-pro/e-Inverted-pendulum-bot.git
   cd e-Inverted-pendulum-bot

2. Configure & build for the host:

   cmake --preset host
   cmake --build --preset host-Debug

3. Run unit + integration tests:

   ctest --preset host

4. Run the example tool / app on the host:

   ./build/host/bin/Debug/inverted_pendulum_bot.tool.example 2 3 4   # -> total = 9
   ./build/host/bin/Debug/inverted_pendulum_bot.example_app          # -> accumulator total = 5

5. Build the cli firmware for the board (produces .elf/.hex):

   cmake --preset NUCLEO-WB55RG          # ST Nucleo-68
   cmake --build --preset NUCLEO-WB55RG-Debug

   On hardware the status LED (LD2, green, PB0) blinks and a UART command-line interface (115200 8N1) accepts ping and id. The CLI is on USART1 (PB6 TX / PB7 RX), which the NUCLEO-WB55RG routes to the on-board ST-LINK virtual COM port — no USB-UART adapter needed.

All presets are defined in CMakePresets.json (host, host-single-Debug, windows, coverage, NUCLEO-WB55RG).

Project Structure

e-Inverted-pendulum-bot/
├── core/                      # Reusable libraries only — no entry points
│   ├── platform_abstraction/  #   platform::Platform interface (+ mock) — the seam
│   ├── cli/            #   portable app: LED blink + UART CLI (+ unit test)
│   └── example_component/     #   trivial interfaces/ + implementations/ (+ unit test)
├── targets/                   # Application entry points + platform implementations
│   ├── cli/            #   one Main.cpp reused across host/st
│   ├── example_app/           #   a trivial host-only entry point
│   └── platform_implementations/
│       ├── host/              #   PlatformImpl: stubs + loopback serial (host build)
│       └── st/                #   PlatformImpl: NUCLEO-WB55RG (LED PB0, USART1)
├── tools/                     # Host-side developer tools
│   └── example_tool/          #   a CLI reusing a core library
├── integration_tests/         # BDD integration tests (cucumber-cpp / Gherkin)
├── infra/                     # Infrastructure submodules
│   ├── embedded-infra-lib/    #   bounded containers, build helpers, toolchains
│   ├── numerical-toolbox/     #   PID, filters, fixed-point algorithms
│   └── hal/st/                #   STM32 hardware abstraction layer
├── documentation/             # Architecture/design/theory/requirements (+ templates)
├── scripts/                   # Build and utility scripts
└── build/                     # Build artifacts (generated, not committed)

Each top-level folder has its own README.md describing its conventions and how to add new components.

Key Design Principles

• No heap in runtime code: all memory statically allocated; bounded containers instead of STL containers. Host tools and tests may use the heap.
• Interface-driven design + dependency injection: hardware is injected via the platform::Platform interface through constructors, never global state.
• Documentation-first: update the relevant documentation/ doc before or alongside behavioural changes. Diagrams use Mermaid or ASCII art only.
• SOLID / DRY: reuse infra/numerical-toolbox/ algorithms; do not duplicate.

Documentation

Document	Description
System Architecture	Components, interfaces and cross-cutting concerns of the robot
Use Cases	Actors, flows and the Gherkin scenarios every requirement traces to
Requirements	Requirement sets, validated against documentation/tools/requirement.schema.json
Safety Supervisor	Mode state machine: self-test, calibration, arming and the disarm conditions
Balance Control	The selectable control-strategy interface and setpoint handling
Attitude Estimation	Pitch estimation with a selectable complementary or Kalman filter
Control Loop	Sample-driven pacing of estimation, balance and outer loops, and loop timing
Motion Actuation	Motor driver, effort mapping and the bridge disable states
Wheel Odometry Design	Wrap-safe count accumulation, wheel velocity and chassis motion
Inertial Sensing Design	MPU9250 wiring, body frame, gyroscope bias calibration and sample validity
BLE Service	GATT layout for teleoperation, telemetry and tuning
Platform Abstraction	The peripheral roles each board must supply
Pendulum Dynamics	Equations of motion, linearisation and the fall time constant
Control Laws	Cascaded PID and LQR derived from the shared plant model
Pitch Estimation Theory	Complementary and Kalman formulations, drift and noise
Wheel Odometry Theory	Quadrature decoding and differential-drive kinematics
Documentation Templates	Starting points for new architecture/design/theory/requirements docs
Performance Optimization Guide	Embedded performance techniques, assembly analysis, cycle budgets
AI Agent Instructions	Development guidelines, patterns, and constraints

License

This project is licensed under the terms in the LICENSE file.

Third-party components and submodules keep their own license terms as documented in their respective directories.