FarmBeats is a Microsoft Research project that builds an end-to-end data system for agriculture, combining sensors, drones, AI, and edge computing so that a farmer can see and act on the variability across a field. It was presented at the USENIX NSDI 2017 conference, with Deepak Vasisht as lead author, by a team drawing on Microsoft, MIT, the University of Washington, and other institutions.
A central obstacle to data-driven farming is that farms often lack reliable internet and power. FarmBeats addressed this with long-range sensor networks built on TV white spaces - the unused spectrum between television channels - to carry data from remote sensors back to a base, with designs that tolerate weather-related outages and power variability. On top of that connectivity, the system uses machine learning to fuse aerial imagery from drones with ground sensor readings, producing precision maps of conditions such as soil moisture and temperature, and applying machine vision to imagery to flag pests, disease, and nutrient deficiencies.
The result is a comprehensive, continuously updated picture of a farm that supports precision-agriculture decisions: where to water, where to fertilize, where a problem is emerging. FarmBeats was deployed in trials across the United States, India, Africa, and China, and its technology later fed into Microsoft’s Azure Data Manager for Agriculture, moving the research toward a commercial cloud product.
Why business readers should care: FarmBeats is a reminder that the hard part of AI in the physical world is often not the model but the plumbing - getting reliable data off a remote farm at all. Solving connectivity and power was the prerequisite that made the machine learning useful.