Scaleout implemented decentralized AI training for military drones

Startup Scaleout has adapted decentralized machine learning technologies to work on board unmanned aerial vehicles. The solution allows drones to autonomously identify targets and make decisions in conditions of limited communication with command centers.

The technology is based on the use of compact AI models that can function directly on the drones' hardware. Unlike centralized systems that require data transmission to the cloud, the company's approach provides local information processing in real-time. This reduces delays in object recognition and increases the system's resistance to electronic warfare means.

The development has received support from NATO structures and is aimed at integration into modern military protocols. The system uses federated learning methods, which allows updating target recognition algorithms based on data obtained during real missions, without the need to transmit raw video streams to servers. This ensures the preservation of tactical data confidentiality and optimizes the use of communication channel bandwidth.

The implementation of such solutions on the battlefield changes the paradigm of drone management, transforming them from operator-controlled devices to autonomous tactical units. Details of the project implementation and technical aspects of interaction with military bases are described in the Ars Technica material, which also considers security and ethics issues of applying autonomous systems in conflict zones.