US 10,893,107 B1
Techniques for managing processing resources
Francesco Giuseppe Callari, Seattle, WA (US); Jean-Guillaume Dominique Durand, Seattle, WA (US); Pradeep Krishna Krishna Yarlagadda, Issaquah, WA (US); and Tatiana Glozman, Kirkland, WA (US)
Assigned to Amazon Technologies, Inc., Seattle, WA (US)
Filed by Amazon Technologies, Inc., Seattle, WA (US)
Filed on Jul. 25, 2019, as Appl. No. 16/522,570.
Int. Cl. G06F 13/00 (2006.01); H04L 29/08 (2006.01); G06K 9/62 (2006.01); G06F 8/658 (2018.01); G06N 20/00 (2019.01); B64C 39/02 (2006.01)
CPC H04L 67/125 (2013.01) [G06F 8/658 (2018.02); G06K 9/6231 (2013.01); G06K 9/6256 (2013.01); G06N 20/00 (2019.01); H04L 67/34 (2013.01); B64C 39/024 (2013.01)] 19 Claims
OG exemplary drawing
 
1. A computer-implemented method, comprising:
obtaining, by a computing system, a training data set comprising historical sensor data associated with a plurality of unmanned aerial vehicles (UAVs);
managing, by the computing system, a set of UAVs configured to perform tasks as instructed by the computing system, the computing system managing a distributed computing cluster comprising at least two UAVs of the set of UAVs;
identifying a first subset of UAVs from the set based at least in part on identifying that UAVs of the first subset are in an idle state;
transmitting, by the computing system to the first subset of UAVs of the distributed computing cluster, instructions associated with training a machine-learning model based at least in part on the training data set;
receiving, by the computing system, additional sensor data from an unmanned aerial vehicle (UAV) of the set of UAVs;
transmitting, by the computing system to a second subset of UAVs of the distributed computing cluster, instructions associated with calculating an incremental update to the machine-learning model based at least in part on the additional sensor data;
receiving, by the computing system, model parameter updates from at least one UAV of the distributed computing cluster;
updating the machine-learning model based at least in part on the model parameter updates; and
deploying the machine-learning model as updated to one or more UAVs of the set of UAVs.