← Wiki

Ingredient

Raspberry Pi

Also known as: RPi, Raspberry Pi 5, Raspberry Pi 4

ARM-based single-board computer — full Linux on a credit-card-sized board. Runs Debian-derived Raspberry Pi OS, Ubuntu, or any ARM64 Linux distribution; capable of hosting web services, running ROS2 nodes, processing camera feeds, training small ML models, and acting as the always-on brain of a farm-automation system. The right ingredient when you need a real OS, network services, a filesystem, or compute that exceeds what an MCU can deliver. Models: Pi 5 (4-core ARM Cortex-A76, up to 16 GB RAM, ~$80), Pi 4 (4-core A72, up to 8 GB, ~$45), Pi Zero 2 W ($15, low-power applications), Compute Module (industrial-form-factor, for embedded production).

Inputs / outputs

  • GPIO: 40-pin header — 28 GPIO, 3.3V logic; multiple I²C, SPI, UART; PWM via software or DMA.
  • USB: 4× USB-A (Pi 4/5: 2× USB 3.0, 2× USB 2.0).
  • Network: Gigabit Ethernet + WiFi 5/6, BLE.
  • Display/camera: HDMI, dedicated CSI camera ribbon connector, DSI display connector.
  • Storage: microSD primary; Pi 5 supports NVMe SSDs via a HAT.
  • Power: 5V/3A USB-C (Pi 5 wants 5V/5A under load); ~5–10 W typical.

Solves / unlocks

  • Always-on farm-automation hub (web dashboard, scheduler, MQTT broker)
  • Computer vision at the edge (camera + OpenCV/YOLO nano)
  • ROS2 control node for small robots and stationary actuator-arrays
  • Local LLM inference (Pi 5 + 8 GB can run small quantized models like Phi-3 mini)
  • [[home-assistant|Home Assistant]] or OpenHAB self-hosted automation server
  • [[node-red|Node-RED]] visual flow orchestration

Constraints

  • No real-time guarantees — Linux scheduling is not deterministic; pair with an MCU for tight timing loops.
  • Microsd reliability — frequent writes destroy SD cards; use SSD or read-only root for production.
  • GPIO weakness — single GPIO can sink/source ~16 mA; never drive motors or solenoids directly.

Source

See also

Auto-generated from this entry’s typed relations: frontmatter, grouped by relation type so the editorial signal isn’t flattened.

  • Parallels: [[nvidia-jetson]] · [[beaglebone-black]]
  • Member of: [[ingredient]]
  • Combines with: [[ros2]] · [[opencv]] · [[node-red]] · [[home-assistant]] · [[mqtt]] · [[rgbd-camera]]

What links here, and how

Inbound connections from across the wiki, grouped by lens and by relationship. These appear automatically — every entity page declares what it links to, and that data populates here on the targets.

Practical

parallels

  • BeagleBone Black BBB competes for industrial-control roles where its real-time PRUs matter; RPi wins on cost and ecosystem
  • NVIDIA Jetson Jetson is the GPU-accelerated step up from RPi when ML/CV is the bottleneck

contains

combines with

  • FarmBot Genesis FarmBot OS runs on a Pi inside the FarmBot's electronics box
  • Home Assistant Home Assistant OS images directly to a Pi SD card; the standard self-hosted automation stack
  • iNaturalist API Pi 5 + camera + iNaturalist CV API = field species ID without needing local model
  • LiDAR rangefinder RPLiDAR A1 + RPi 4 = entry-level autonomous-rover perception stack
  • Modbus / RS-485 USB-RS485 adapters or HAT; Python pymodbus is mature
  • MQTT Mosquitto MQTT broker runs trivially on Pi as a service
  • Node-RED Node-RED runs as a service on the Pi; visual orchestration of farm IoT
  • OpenCV Pi 5 with Camera Module 3 = real-time OpenCV pipelines for crop monitoring
  • PlantVillage dataset Pi 5 with quantized model (TFLite) = lightweight in-field disease screening
  • RGB-D camera RPi 5 can stream RealSense, but heavy processing should move to Jetson
  • ROS 2 ROS2 runs on Pi 4/5 with Ubuntu ARM64; standard low-cost ROS2 control node
  • Servo motor needs hardware PWM HAT for jitter-free control (PCA9685 is standard)
  • YOLO (object detection) Pi 5 with YOLOv8-nano + TFLite delegate: real-time at ~10 FPS for low-frame-rate work

combines

19 inbound links · 9 outbound