COTS-Based Framework for Zero-trust Autonomous Drone Swarming

Overhead Intelligence proposes to develop a framework for intelligent autonomous drone swarming, where operators can queue complex multi-drone tasks such as formation flight, investigation of areas of interest, and geolocation of electromagnetic signals of opportunity. The autonomous drone network will delegate tasks to available UAVs and coordinate the completion of all tasks in the queue with minimal human intervention.  The research will create a product designed around scenarios like the following: three UAVs are launched by a team of two operators. One of the drones is designated as “master,” supporting satcom, LTE, EM spectrum analyzer, and EM direction finder. The other drones are “subordinate,” carrying the same equipment except for satcom, LTE, and EM equipment. All three drones remain in standby, loitering in place or orbiting safely with respect to one another. Either operator can issue a task to the master drone through LTE, IP radio, or satcom link. The master drone autonomously coordinates instructions for itself and the subordinate drones and issues those instructions. For instance, the master drone can be issued a formation flight task to an indicated waypoint. The flight path of the master drone is generated using that waypoint, but the master drone must calculate paths for the subordinate drones relative to its own before instructing them to follow those paths. If the master drone is tasked with locating the emission source of a given EM frequency, it coordinates its scans with those performed by the subordinate drones to instantaneously geolocate the EM source. If multiple tasks are issued to the master drone, a task queue is formed. The master drone monitors the progress of each task and issues new instructions once a task is complete. While creating new tasks and managing the queue will primarily be performed through a user interface, a voice command interface will also exist to allow operators to more easily instruct drones to complete common tasks. As all of the technologies proposed employ an “offline-first” approach, the speech-to-text machine learning model will be run locally on the drone itself to prevent Internet dependence. UAVs will communicate amongst themselves, with nearby operators, and with individuals anywhere on the globe through IP-based networks.The security of the network will be ensured through expiring encryption keys, facilitated by a blockchain-based administration infrastructure. Administrators can approve or deny requests by credentialed users for access to specific drones and for limited spans of time. The proposed research leverages the architecture constructed by Overhead Intelligence in Phase I of Open-source Autonomous Vehicle Control over IP-based Communication Infrastructure research and development.