A Physics-Informed Machine Learning Approach to Enhanced Dead Reckoning based Hypersonic Navigation

Global Position System (GPS) aided navigation systems are the gold standard for Position Navigation and Timing (PNT) due to their world-wide, all-weather availability and superior accuracy. However, in contested and Electromagnetic (EM) congested environments GPS may not be available due to environmental conditions and/or adversary Anti-Access Area Denial (A2/AD) technology. In these scenarios our assets may be required to use dead reckoning approaches and inertial measurements, which can cause significant drift over time. This is an especially serious concern for hypersonic systems where communications may not be available for long times. As such, the United States Navy (USN) is seeking advanced dead reckoning navigation systems through the SBIR topic N251-071. To address this need, Spectral Energies (SE) proposes to develop and refine a physics informed Machine Learning (ML) based dead reckoning approach. During Phase I, SE will collect simulated and experimental inertial data and develop the ML algorithms, select the appropriate physical constraints, and demonstrate the feasibility of the approach. During Phase II, SE will refine the algorithms, collect additional data at more relevant conditions, develop a prototype, and create a plan to integrate into a hypersonic test program like MACH-TB.Ý