# ADR 000: Arduino Uno Prototype Platform

- HTML version: https://robbiepalmer.me/projects/autonomic-satellite-swarm/adrs/000-arduino-uno-prototype-platform
- Project: Autonomic Satellite Swarm (https://robbiepalmer.me/projects/autonomic-satellite-swarm.md)
- Status: Accepted
- Date: 2016-07-01

# Context

The research needs several independent, programmable agents for a physical CubeSat swarm
demonstration. Each agent must read sensors, exchange messages, calculate a candidacy score, and
control a simple actuator. The prototype has to fit a summer internship budget and schedule.

An [Arduino Uno](https://docs.arduino.cc/hardware/uno-rev3) provides enough processing and I/O for
that job without a general-purpose operating system. It is widely stocked, documented, replaceable,
and can run from USB or a battery. Its modest processor, memory, and power needs resemble the
constraints that make embedded spacecraft software interesting.

Arduino also has relevant precedent. The purchasable
[ArduSat Space Kit](https://web.archive.org/web/20160725193115/http://store.ardusat.com/products/ardusat-space-kit)
includes an Arduino and the same kinds of sensors sent into orbit on Spire CubeSats. ArduSat put
Arduino-based CubeSats into orbit in 2013, as recorded by both the
[Arduino project](https://blog.arduino.cc/2013/08/12/ardusat-successfully-launched-in-space/)
and a [2014 NASA comparison of Arduino and Raspberry Pi hardware](https://ntrs.nasa.gov/api/citations/20140017620/downloads/20140017620.pdf).
This is useful precedent for a research prototype, though it does not make an off-the-shelf Uno
flight-qualified.

ArduSat described itself as the first "open satellite platform." That matters beyond hardware
choice. An accessible platform lets other people program experiments, inspect how they work,
and build on the results. It follows my [build-in-public
philosophy](/projects?tab=philosophy#build-in-public), creates [short feedback
loops](/projects?tab=philosophy#short-feedback-loops) through external use and scrutiny, and starts a
[flywheel](/projects?tab=philosophy#build-flywheels) in which shared tools and findings make each
later experiment cheaper. Choosing a common, open-source board gives this project the same
possibility instead of making the prototype a closed artifact that only its original author can
reproduce.

A [Raspberry Pi 3 Model B](https://www.raspberrypi.com/products/raspberry-pi-3-model-b/) has much
more computing power, memory, storage, and networking. The demonstration does not use that extra
capacity. A Pi also adds an operating system, removable storage, a longer boot path, and a much larger
power supply requirement. The smaller microcontroller is a closer match for an always-on agent with
a bounded job.

# Decision

Build each prototype agent around an Arduino Uno. Keep the workload within its 32 KB of flash, 2 KB
of SRAM, and 16 MHz ATmega328P rather than selecting a more capable computer pre-emptively.

# Consequences

The boards are inexpensive, easy to replace, and quick to program. The demonstration can use common
sensors and actuator modules without designing a processor board first.

The Uno's limits are real. Software must conserve memory, avoid operating-system assumptions, and
keep computation bounded. The prototype does not establish radiation tolerance, thermal fitness,
component reliability, or flight readiness.

---

Markdown index of this site: https://robbiepalmer.me/llms.txt