China is developing an ultra-fast torpedo with artificial intelligence capable of identifying real submarines amid decoys and sophisticated countermeasures.
The technology, presented by Chinese military researchers, achieved a 92.2% success rate in distinguishing real targets from fake ones, surpassing previous systems.
This advancement is crucial in the face of modern underwater warfare strategies, which include the use of acoustic holograms, false bubble trails, and decoys that mimic submarine sounds and movements.
To tackle these challenges, researchers used a combination of physical models and advanced neural networks. With limited real combat data available, they simulated decoys based on bubble collapse patterns and turbulence. These simulations fed a generative adversarial network, composed of two AIs competing against each other: one creates acoustic forgeries while the other tries to detect them. The result is a system trained with thousands of artificial profiles that continuously improves its ability to accurately identify real targets even in noisy and chaotic environments.
The system works like “sonic image recognition,” converting sonar signals into spectral representations and using neural network layers to detect distortions. The technology represents a significant step toward more autonomous and effective torpedoes capable of operating without external communication in intense combat situations.
The project is part of a global race for intelligent underwater weapons, with countries like Russia and the United States also investing in supercavitating torpedoes that still struggle with precise target identification.
Source: South China Morning Post | Photo: Freepik | This content was created with the help of AI and reviewed by the editorial team
