DON26BZ01-NV025 — Leveraging Machine Learning for Advanced Passive Sonar Tracking

Award Maximum: $140,000 (Base) / $100,000 (Option) Period of Performance: 6 months (Base) + 6 months (Option) Phase Type: Phase I

OBJECTIVE: Develop advanced automation to detect, locate, classify, and correlate contacts across multiple sonar sensors and multiple display surfaces.

DESCRIPTION: Passive sonar systems employ a standardized signal processing pipeline to track, classify, and localize underwater contacts. This automated process, often referred to as "automation," begins after front-end processing generates visual displays for sonar operator analysis and automated processing. Existing algorithms that track energy signatures on these displays typically include Kalman filters, probabilistic multi-hypothesis trackers, and particle filters. However, these traditional tracking methods, as implemented in current operational systems, often fail to fully leverage the potential of modern machine learning techniques. This SBIR topic seeks to incorporate cutting-edge machine learning technologies into passive sonar processing to significantly improve tracking, classification, fusion, and localization of current anti-submarine warfare passive sonar systems.

Targeted Improvement metrics include: Tracking — Increase Hold Time Ratio (Threshold: 10%, Objective: 20%); Tracking — Reduce Time to Detect (Threshold: 10%, Objective: 20%); Classification — Increase Probability of Correct Classification (Threshold: 10%, Objective: 15%); Classification — Reduce Probability of False Alerts (Threshold: 10%, Objective: 15%); Track Fusion — Increase Probability of Correct Association (Threshold: 15%, Objective: 20%); Localization — Reduce Area of Uncertainty (Threshold: 15%, Objective: 20%).

Work produced in Phase II may become classified.

PHASE I: Develop algorithms that improve sonar automation for tracking, localization, classification, and multi-sensor fusion. The approach will reduce the burden of operators to maintain and promote tracks and be supported by theory.

PHASE II: Implement the proposed approach in a simulated environment (e.g., MATLAB) and demonstrate stated performance using government-provided data from a Navy sonar system. Important metrics will be, but not limited to, probability of correct association, hold time ratio, time to track, and probability of correct classification. It is probable that the work under this effort will be classified under Phase II.

PHASE III DUAL USE APPLICATIONS: Support transition to Navy use. This effort is anticipated to have dual-use applications in commercial surveillance systems with towed arrays or ISR uncrewed aerial vehicles.

KEYWORDS: Multi-sensor data fusion, operator workload reduction, advanced automation