Reference Number: ARL-SRP-00021
Project Description
A core challenge in modern electromagnetic warfare has adversaries use multiple mobile radio-frequency (RF) emitters for communication, jamming, and interference while continuously changing position and operating within dynamic, hard-to-predict spectrum conditions. The research problem posed here is how to accurately track the locations of these emitters and characterize their effects on the RF environment when only sparse measurements of the battlespace are available.
The project’s objective is to develop a low-latency, self-supervised learning architecture capable of tracking multiple unknown, moving RF emitters using sparse received signal strength (RSS) data and multimodal inputs such as RGB and infrared (IR) imagery. The participant will design, test, and refine this architecture to minimize mean-square error between estimated and actual emitter locations, while also supporting decentralized implementation and edge deployment.
The work will rely on Python and PyTorch for model development, NVIDIA Sionna for ray-tracing-based generation of RF propagation and RSS maps across varying emitter configurations, carrier frequencies, and terrain settings, and joint embedding predictive architectures for multimodal self-supervised learning. Using simulated data, the participant will evaluate system performance and improve robustness under realistic operational conditions.
The expected deliverable is an algorithmic framework for multimodal RF emitter tracking, accompanied by supporting code, software artifacts, and technical documentation.
Technical Skills
- Artificial Intelligence/Machine Learning
- Communication and Networking Technologies
- Computer Modeling/Simulation
- Data Analysis; Data Visualization
- Electrical Engineering
- Statistics
- Python
