ai-robot

What it is

RK3506 Voice Robot is a pure-C embedded application intended to run on the Rockchip RK3506 development board. It records audio via ALSA, uploads it over WiFi to a remote AI agent, and renders the received reply on an LCD framebuffer. The project emphasizes static memory usage with no dynamic allocation at runtime and a single-threaded event loop.

How it works

The software uses a 50ms event loop in main.c to coordinate modules:

• Audio capture via ALSA PCM
• HTTP client to upload audio (multipart/form-data) to an AI server
• WiFi management via a WiFi manager
• Display rendering via a framebuffer with FreeType for text
• State machine with 5 states (IDLE, RECORDING, PROCESSING, RESULT, ERROR) handling transitions and errors

Key components include: audio_capture, wifi_manager, http_client, display, font_renderer, and state_machine. The design aims for 100% static memory allocation and no malloc/free at runtime.

Getting started

The README provides these steps:

• Clone the repository and prepare dependencies

  git clone https://github.com/yourname/rk3506-voice-robot.git
  cd rk3506-voice-robot
  

  # Debian/Ubuntu
  sudo apt install libasound2-dev libcurl4-openssl-dev \
                   libfreetype6-dev libconfig-dev cmake
  # Also install the aarch64 cross-compiler
  sudo apt install gcc-aarch64-linux-gnu
  

• Configure runtime in config/app_config.ini (WiFi credentials at minimum)

  [wifi]
  ssid = "YourWiFiSSID"
  psk  = "YourWiFiPassword"
  

• Cross-compile

  mkdir build && cd build
  cmake .. -DTOOLCHAIN_PATH=/usr -DCMAKE_BUILD_TYPE=Release
  make -j$(nproc)
  

• Deploy and run on RK3506 board

  # On the target device
  ./scripts/start_wifi.sh
  ./scripts/run_voice_robot.sh
  

Recent releases

• Releases: latest 0 listed as none; repository shows no tagged releases in the provided data.

Traction

• Stars: 67

License

• License: none listed