Autonomous Environmental Sensor Performance Prediction Tool for Multi-Static Active and Passive Anti-Submarine Warfare (ASW) Systems

SSC proposes to improve multistatic Air ASW by incorporating environmental information into the P-8A Likelihood Ratio Tracker (LRT) software to provide better classification and localization performance. The overall objective is to create an LRT with an Environmental Probability of Detection (Pd ) Model (EPM) included. This LRT-EPM effort seeks to improve Air ASW performance and mature technology for transition into the P-8A Engineering Change Process.The LRT is an automated Bayesian multi-static detection and tracking fusion engine. This Bayesian processor processes MAC (or DICASS) contact reports generated by a signal processor and fuses related contacts. The LRT processing is a proven method to improve the detection and classification and characteristics of active Air ASW systems. LRT detection and tracking algorithms rely on statistical models of the environment and sensor characteristics to help classify and localize submarine targets. One of the critical LRT models is the Pd associated with a submarine target. This Pd varies as a function of target location and velocity relative to the sonobuoy locations and the waveform being transmitted.Under this second SBIR Phase II effort, archival and in-situ environmental information will be used to provide more accurate Pd of target echoes. SSC will use a combination of physics based acoustic propagation methods as well as machine learning techniques to predict the sonar equation-based probability of target detection associated with every possible target contact generated by the Multistatic Air Coherent (MAC) system. L3Harris Technologies will support SSC in evaluating different Navy MAC acoustic propagation models to find the right combination of accuracy and speed for predicting echo Pd.SSC will evaluate several ocean propagation models and use machine learning technology to regress the physics-based model predictions to actual at-sea detection events. We will also investigate machine learning methods to reduce the computational load of calculating multistatic signal excess, the fundamental quantity needed to calculate Pd.  By having more accurate Pd forecasts, LRT and Air ASW performance is improved.