PAWS
Personal Autonomous Weather Station — From Sensors to Local Intelligence
Preface
PAWS (Personal Autonomous Weather Station) is a low-powered ESP32 weather station with SD card logging, RTC timekeeping, phone-based data gateway, time-series database, online dashboard, plant watering decisions, and local weather forecasting.
A learning-oriented project covering the full stack: embedded firmware, solar power, asynchronous data pipeline, and machine learning. This document is the single authoritative reference for the end-to-end design — iterate quickly, validate early, add complexity only when the foundation is solid.
Related project — 3D-PAWS (3D-Printed Automatic Weather Station, UCAR/COMET): an open-source initiative for low-cost weather stations using 3D-printed sensor housings. Their sensor designs (tipping bucket rain gauge, anemometer) are a useful reference for the DIY hardware phases of this project. See 3dpaws.comet.ucar.edu.
How to read this document
| Goal | Chapter |
|---|---|
| Understand why this project exists | Motivation & Goals |
| See the big picture and data flow | System Architecture |
| Know what hardware to use | Hardware |
| Understand the embedded firmware | Firmware |
| Set up the server and dashboard | Backend |
| Understand the ML approach | Forecasting & Decisions |
| Know what to build in what order | ROADMAP.md |
Build philosophy
Each phase of this project must be independently useful — you can stop at any phase and have a working, valuable system. Hardware and software investments made in one phase are never thrown away. The architecture (FSM, data format, connector interfaces) is frozen from Phase 1 onward. Only capabilities are added.
Document status
Each chapter carries a status badge:
- 🟢 Stable — design frozen and validated
- 🟡 Draft — design in progress, may evolve
- 🔴 Planned — not yet started
The architecture (FSM, deep-sleep model, data format, connector interfaces) is considered frozen from Phase 1 onward. Hardware and software can evolve; the architecture does not.