A Submission for the World’s Largest Arduino Maker Challenge
Hackster.io has teamed up with Arduino and Microsoft to bring us a contest to promote Arduino's new MKR1000 microcontroller. I decided to enter the contest with a project that's been brewing in my brain for quite some time. Here's my entry, which was accepted:
Without doubt, water is the most precious resource we have on this planet. In agriculture, management of this resource has become extremely important as we recognize the impacts of climate change on our ability to feed everyone equitably. Precision agriculture is an area of study that has rapidly expanded in the last decade to tackle this problem. In the United States, ongoing research in precision agriculture has been effectively applied to large agri-businesses at scale. But what about the small farmer? These operations do not have the capital resources to invest in large scale monitoring systems to save resources and increase crop yield.
In the state of Vermont, we have a long and rich tradition of farming and making do with what you have, or don’t have. Out climate is such that fresh water has always been plentiful, and poorly managed, if managed at all. This has changed dramatically in the last two decades. Water pollution, in the form of excess fertilizer runoff, has and is damaging the ecosystems in our largest body of water, Lake Champlain. A lot of federal, state, and local money has been poured into researching and mitigating the issue. The single largest deterrent to correcting the problem is cost. The average small family farmer cannot afford to implement the mandated solutions.
One piece of this puzzle is maintaining adequate soil moisture for a given crop. Many of these small farms operate in microclimates that defy standard weather prediction models. What if we could bring the predictive learning capability of the cloud right into their fields at low cost? My project aims to develop a mesh network of weather and soil sensors that stream data to the cloud for analysis. The overarching goals for low cost and ease of deployment factor heavily into the choice of hardware and software.
Each sensor station will consist of several sensors tied to a low power microcontroller that is capable of wirelessly streaming data at low bandwidth and long range (hundreds to several thousands of meters). These stations will be solar powered and designed to withstand conditions of a typical growing season in Vermont.
The base station will act as the data collector and gateway to the backhaul network. In addition to the RF hardware needed to communicate with the remote stations, it will include a WiFi connection to an access point with an internet connection.
Remote Station Sensors:Air temperature
Solar radiation (light sensor)
Hardware:Solar panel (size TBD)
Battery and charging circuitry
RF communication (type TBD)
Microcontroller – Arduino Uno, MKR1000, or similar
Gateway:Solar panel (size TBD)
Battery and charging circuitry
RF communication (type TBD to match remote stations)
Headless WiFi enabled Raspberry Pi 2 running Windows IoT Core
The design of the system will attempt to abstract away the RF hardware dependency, so that different RF technologies can be tested. Data collection and analysis will be performed using Microsoft Azure IoT Hub, Stream Analytics, and potentially Machine Learning. Data will be displayed in an Azure hosted ASP.NET Web Site accessible to any device with an internet connection. IF time allows, a Windows 10 UWP application will be developed to run on a WiFi enabled Raspberry Pi 2 with a 7” display. This iteration of the project will be for data collection and analysis only. A future iteration may include incorporation of remote stations to directly control irrigation systems.