Ranveer Chandra spent much of his childhood on a family farm, and he uses that experience as part of Microsoft's FarmBeats team. FarmBeats is a part of the company's AI for Earth initiatives and one of its larger research programs. The problem statement seems dire - food production needs to be increased significantly by 2050, farmable land is at a premium if not peak, and the world holds dangerously low levels of usable water. FarmBeats goal is to help farmers use data for better farming with a full system approach.
FarmBeats wants a large scale deployment, with early testing on farms lasting over six months at each location and collecting more than ten million sensor measurements and one million camera images. Algorithms applied to drone flight patterns can increase area coverage by up to thirty percent. The platform used television broadcast channels to send high bandwidth farming data to a central hub.
Moisture sensors, pH sensors, stationary and drone cameras all send data back to the IoT base. The base station controller manages the data between the system and the cloud, and then uses that data to help the farmer predict trends and schedule tasks. Farmers can control the functions from a phone app and access data from other farms.
Two major functions of FarmBeats jump out to me as the most interesting and helpful - the idea that all of these sensors are commercially available but being used in new ways, and the work that's being done with AI to make the drones more efficient. The paper FarmBeats: An IoT Platform for Data-Driven Agriculture gives a dry technical overview of the work being done and has great sections on drone coverage and later discusses how wind data is taken into account for more efficient drone operations. As food production becomes a larger problem in the next few decades we'll need several solutions, and FarmBeats looks to be a good tool in making farms more productive.
