Virtual Reality for Multi-INT Deep Learning (VR-MDL)


TECHNOLOGY AREA(S): INFO SYSTEMS OBJECTIVE: Develop a new high-fidelity modeling and simulation (M&S) framework that addresses the need for voluminous and high-quality Multi-INT training data for deep learning networks that would be too expensive and infeasible based on costly field experiments. The focus of this effort is on multiphysics-based modeling of radio frequency (RF) signals in realistic physical and contested environments. Effective training methods for deep learning systems are essential for improving the performance of autonomous systems. DESCRIPTION:A key enabler for improved system performance while reducing operator workload is autonomy powered by next generation machine intelligence. However, for such deep learning systems to be effective, massive amounts of realistic and relevant “training data” is required [1]. Given the highly sensitive and variable nature of RF collection and exploitation, it is simply not possible to conduct field experiments capable of meeting these requirements. Fortunately, commercially available next generation high-fidelity physics-based M&S tools have been developed that can form the basis for an RF “virtual reality” Multi-INT Deep Learning (VR-MDL) environment (see for example [2]). Thus, the goal of this project is to develop an M&S environment to address the robust training needs of deep learning networks and other machine intelligence and cognitive systems such as DARPA/AFRL KASSPER and CoFAR projects [3-5]. The VR-MDL environment should be capable of supporting all physical elements of the RF collection process, from raw multichannel, multiplatform in-phase and quadrature (I&Q) signals, through the various RF processing chain (e.g., mixing, amplifiers, analog-to-digital conversion, etc.). The output of this effort should be a general-purpose M&S environment that is agnostic to the particular machine learning algorithm or architecture. PHASE I: Develop a baseline design for a VR-MDL environment. The design should have the potential of achieving the aforementioned goals of producing massive amounts of high-fidelity, physics-based, RF training data generated via realistic CONOPS. Quantitative analyses and experiments shall be conducted that establish the scalability of the proposed VR-MDL approach. Basic M&S examples shall be conducted in Phase I that establish the viability of the training data generated by comparing it with actual collected data with the same type of sensor and scenario being modeled by the VR-MDL M&S tool. PHASE II: Further refine and develop the VR-MDL design from Phase I, and enhance the complexity and sophistication of the VR scenarios conducted. The output of Phase II should be a VR-MDL tool that is ready to enter low-rate initial production (LRIP) at the beginning of Phase III. PHASE III: The proposer will identify potential commercial and dual use applications such as non-military applications of deep learning techniques. These could include training for autonomous systems such as self-driving cars and unmanned air systems (UAS) operating in civilian airspace. REFERENCES: 1.X.-W. Chen and X. Lin, "Big data deep learning: challenges and perspectives," IEEE access, vol. 2, pp. 514-525, 2014.;2.RFView(TM). Available:;3.R. Guerci, R. M. Guerci, M. Ranagaswamy, J. S. Bergin, and M. C. Wicks, "CoFAR: Cognitive fully adaptive radar," presented at the IEEE Radar Conference, Cincinnati, OH, 2014.;4L. Bell, C. J. Baker, G. E. Smith, J. T. Johnson, and M. Rangaswamy, "Cognitive radar framework for target detection and tracking," IEEE Journal of Selected Topics in Signal Processing, 9(8), 1427-1439, 2015. KEYWORDS:Multi-INT, Deep Learning, Autonomous Systems, Cognitive Systems, Sensors, Electronics, Modeling and Simulation CONTACT(S):DanStevens AFRL/RIGC 3153302416

Agency Micro-sites

SBA logo
Department of Agriculture logo
Department of Commerce logo
Department of Defense logo
Department of Education logo
Department of Energy logo
Department of Health and Human Services logo
Department of Homeland Security logo
Department of Transportation logo
Environmental Protection Agency logo
National Aeronautics and Space Administration logo
National Science Foundation logo
US Flag An Official Website of the United States Government