AI-Enhanced Subsurface Infrastructure Mapping on AWS

This article explores an innovative approach to subsurface infrastructure mapping using AI and advanced magnetic imaging technologies deployed on AWS infrastructure.

• Researchers from S2 Labs, Empact AI, and Kraken Robotics developed a physics-informed deep learning solution for detecting underground structures
• The AI method uses U-Net neural networks to translate magnetic field measurements into detailed underground images
• Training involved generating 202,000 3D susceptibility models using Amazon EC2, S3, and AWS Batch
• The workflow includes four stages: data acquisition, survey domain preparation, parallel processing, and AI-based inference
• Successfully demonstrated in a case study detecting underwater oil well conductors buried 35-45 meters deep in the Gulf of Mexico

The technology promises transformative capabilities in infrastructure mapping, offering unprecedented accuracy in detecting buried structures across multiple industries.