Among many hyped agriculture technologies under industry 4.0 umbrella I’ve ever seen, this Microsoft solution, namely FarmBeats, appropriately addresses Indonesia’s geographical challenges.
As a solution architect, what I like the most from the presentation is that how they synthesized the problem and designed a solution for it.
Here are the summaries of the challenges:
1) limited connectivity from the farm (sensors) to farmer’s home/office (few miles away, multipath propagation due to hill/trees). Solution: extend WiFi signal using empty UHF/VHF channels which have much better propagation & are abundant in rural areas
2) limited resources (sensors) to portrait the entire field. Solution: combine readings from only a few sensors and images from a flying moving object into a machine learning model’s inference to predict unmeasured points in order to build a complete precision map
3) high-cost aerial imagery. Solution: instead of an expensive drone, “helium” balloon equipped with smartphone/power bank attached to moving object (such as walk, motor, car, bike, tractor) is used
4) limited Internet of farmer’s home/office. Solution: edge approach. The trained model stays local to infer real-time data, data stored locally and sent in a compressed format to the cloud for model improvement (retrain), web server instance/cache replicated from one in the cloud stays local to serve multiple services, such as irrigation, pest infection, fertilizing, etc.
Here is the full video from Microsoft.