Project Rubus: Are raspberry pi's useful in the field?

[richfitz]

[zkamvar]

It would probably be best to document what we were thinking here. I'm assuming that @richfitz or @sdwfrost will correct me if I'm wrong here.

As @sdwfrost said, this is basically a cheaper version of EWARS in a box with the software embedded to deploy RECON shiny apps, collect and tag data versions, transmit data, and receive data.

The raspberry pi is a fully functional, affordable computer (~30 GBP) that is the size of a playing card. Since it This could be used to host pre-installed versions of RECON packages that can be accessed via shinyapps or htmlwidgets for analysis of data.

The pi's can be adapted to work with paper-based data collection methods using digital cameras and OCR software. These images will then be tagged, compressed, stored, and shipped to other sites via whatever method for data transfer possible (e.g. wireless or satellite phone) so that the work of data entry can be distributed among sites. All of the data provenance will be recorded in the process. Cleaned data will be continuously updated and a button can be added to the raspberry pis for analysts on the ground to push a (physical) button and print useful visualizations such as the epi curve.

[zkamvar]

CODE NAME: Rubus

[zkamvar]

Project Code Name: Rubus

The raspberry pi is a fully functional, affordable computer (~30 GBP) that is the size of a playing card. Since it This could be used to host pre-installed versions of RECON packages that can be accessed via shinyapps or htmlwidgets for analysis of data.

The pi's can be adapted to work with paper-based data collection methods using digital cameras and OCR software. These images will then be tagged, compressed, stored, and shipped to other sites via whatever method for data transfer possible (e.g. wireless or satellite phone) so that the work of data entry can be distributed among sites. All of the data provenance will be recorded in the process. Cleaned data will be continuously updated and a button can be added to the raspberry pis for analysts on the ground to push a (physical) button and print useful visualizations such as the epi curve.

Goals

• Cheap
• Customizable
• Modular
• Performs data hashing and synchronization
• Nearly-automated data collection
• Little to no change in field epidemiologists workflow
• Distributed data entry/analysis
• Real-time epi analysis
  • Easy to extend
  • Easy to develop offshore, test and ship out images (over internet or via sdcard)

Needed equipment

• Raspberry pi
• Printer and enough ink and paper
• Satellite phone or other connection
• Multiple buttons at least one should be big and red
  • Breadboard/arduino for connecting said buttons
• Power supply
• Batteries
• Solar panel for charging
• GSM modem
• Satellite phone modem (rock block)
• Some sort of tray for capturing a sheet of a4 paper really easily (probably just a paper tray)
  • Lighting for this - led flash/torch/etc)
  • Digital camera (the cheap one for the pi)
  • Wireframe and shade
• Waterproof case (for running and/or for transportation)
• More raspberry pi’s for distributed use
• Wifi card(s)
• System disk for each pi
• User-removable data disks for each pi
• HDMI cable
• USB grid for storage
• Keyboard/mouse, perhaps display
• Small touchscreen

Software to include

• R
• Database for indexing documents
  • Document database like couch?
  • SQLite for relational data?
  • What is the memory-lightest server/client SQL database for use across the pi cluster?
• Rstudio server
• Shiny
• Python for lots of the shiny driving bits - probably lighter for driving the camera etc so we’ll have to write some custom code
• Encryption libraries (libsodium, openssl)
• Git for updating code

Things to learn

• How to set up an adhoc network
• How to set up a big red button
• How will the blockchain work?
• Will bittorrent software be applicable in our situation?
  • Can we use a peer-to-peer protocol to transmit data back to HQ over unreliable networks?
• How to generate QR codes that we could encode metadata and what is the resolving resolution of the pi camera?
• Distributed storage over pis in the network for automatic backup and redundancy
• Figure out OCR stuff
  • Handwriting
  • Turning tables into smaller bits (for sharing)

Things it can do

• Turnkey deployment of shiny apps (or html widgets) without needing to install any software on other computers
  • Compute simple epi analyses in the field
  • Display useful graphs
  • Communicate complex analyses from HQ
  • Quiz night
• Automated data collection and sharing
  • Take photos of sheets as they are collected
  • OCR on sheets
  • Use QR codes on sheets for automatic indexing
  • Collect up results over distributed pis and communicate out with the main server pi beast
• Data compression
• Hashing/Encryption
• Data retrieval/updating

Features

• No Power no Pigblock Problem
• Expandable to multiple nodes/data entry stations
• Communication over multiple media, depending on what is available (even serialise over SMS at a pinch)
• Most of the system is cheap enough to leave behind (all but the rock block)
• No installation onto government laptops and no installation headaches
• blockchain compatible
• Buzzword cromulent
• Work with both headless and screen based applications
• For the paper based system, a unique index on each sheet to prevent accidental re-entry

Pitfalls

• Paper-based
• More things for people to learn
• More custom software to maintain
• Only 1GB of RAM so lots of things won’t work as expected (it might be necessary to explore other bigger one board systems for the workhorse node perhaps but then the power issues reappear)
• No excel so higher barrier to use for people. Gnumeric might be enough if that is still a thing?
• theft/loss/damage etc. they’re quite small and not obviously precious.
• Security - encrypted disks might be helpful here but balancing security with ease of use will be hard. If there’s a paper system too though it’s probably moot as that is literally cleartext.
• If we have one everyone will want one

[dirkschumacher]

I do like that you think about blockchain early on. I assume you will interact with the ethereum blockchain?