Machine Learning Integration for Real-Time Naval Sonar Data Analysis

During the RIMPAC 2026 exercise off the coast of Hawaii in summer 2026, Lockheed Martin, the United States Navy, and FUSE Inc. evaluated an artificial intelligence system for airborne antisubmarine warfare. The collaborative framework integrates real-time acoustic signal processing with secure over-the-air software distribution to enhance digital infrastructure in naval operations.

Challenge and Operational Context
Subsurface threat detection relies on analyzing complex underwater acoustic spectral energy transmitted by deployed sonobuoys. Legacy systems constrain operators to four simultaneous sonobuoy streams, while variable ocean acoustics require continuous adjustments to signal classification models without interrupting airborne missions.

Technical Architecture and Partner Responsibilities
The deployed architecture processes multi-sensor acoustic feeds and streams processing data to helicopter aircrews and ground stations. Lockheed Martin developed the underlying machine learning classification system, which enables operators to monitor up to eight simultaneous sonobuoy surveillance feeds. FUSE Inc. provided the encrypted data link capability to ensure resilient over-the-air transmission of software and model packages to active platforms.

Closed-Loop Model Retraining and Deployment
The workflow operates as a closed-loop sequence comprising signal capture, target labeling, model retraining, and wireless deployment. When a new acoustic signature is detected, ground operators label the noise pattern, retrain the classifier on the new data, and transmit a small update packet directly to the aircraft. The complete retraining iteration requires less than five minutes.

System Integration and Performance Impact
The solution was integrated and tested on two MH-60R helicopters during anti-submarine missions in summer 2026. By enabling inflight updates via encrypted wireless links, the system maintains continuous situational awareness without requiring physical software access or mission interruption.

Edited by Evgeny Churilov, Induportals Media - Adapted by AI.

www.lockheedmartin.com