One of TurbineOne's most popular prototypes was federated learning across drones. Or as we like to call it; Drone Dreaming. This technological invention creates a new capability for drones that improves with each and every mission. It maximizes the resources available to a drone without requiring additional storage, weight or deployment infrastructure. It creates drones that learn.
The US military and intelligence drone services deploy where they are needed and into locations that rarely coincide with a large internet backhaul. Additionally, as technology progresses - cutting-edge machine learning relies on the power of the cloud, rendering this technology nearly impossible to use in a disconnected “drone-edge” environment. Further the amount of data collected by our aircraft is measured in petabytes - outstripping even high-bandwidth fiber networks.
TurbineOne is composed of accomplished engineers from Google, Amazon, Twitter, Databricks, Salesforce, Cloudera, Toyota, Kindred.ai, Mozilla, and many others. TurbineOne’s co-founders have contributed to over 90 patents and have built transformative technologies on top of data ranging from earth satellites to sensors at the deepest depths of the ocean. We approach every problem through a “Rapid Evaluation Methodology” pioneered to find fast and effective solutions to all types of technical issues. Leveraging this depth of experience we’ve created a solution that ensures a drone fleet has access to the techniques and technologies that use just-in-time learning.
Drone Dreaming enables drones-at-rest to coordinate their on-board resources amongst their physically co-located peers to tackle huge data-processing tasks. All without operator intervention. The components of Drone Dreaming contain all those found in rich-cloud environments, but enhanced with solutions to limited or isolated connectivity with potential access to at-rest field experts.