Detecting Adversary Military Assets on Autonomous Edge Devices using Deep Learning
Open Access DepositedThe research in this praxis explores the potential use of deep learning AI on small, unmanned aircraft systems (sUAS) to detect adversary military assets in electronic warfare (EW) disrupted environments. This praxis identifies several impacts of EW to the command and control of sUAS in military applications. Several research questions used in this praxis investigate various constraints that hardware and environmental conditions have on AI deployed on edge devices. This praxis implements a convolutional neural network (CNN) using a medium size version of the 8th generation of the You Only Look Once model architecture. This model is trained on images of modern Russian military assets including main battle tanks and infantry fighting vehicles. This praxis also uses a dedicated test set of images as well as simulated images generated in an Unreal Engine test environment. The experiments in this praxis are conducted on an NVIDIA Jetson Orin Nano Super Developer Kit to simulate the limitations found on edge devices. The results of this praxis demonstrate the potential application of a lightweight CNN to detect adversary military vehicles from an edge device to mitigate the disruption of EW on sUAS. This research contributes to the future development of artificial intelligence equipped edge devices in the defense sector, particularly with efforts like the U.S. Department of Defense (DOD) Replicator Initiative. Future research may focus on robust dataset generation, model refinement, integration with autonomous flight control systems, and integration with military battlefield intelligence systems.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.