SOCRAT — Remote Detection of Explosive Objects

SOCRAT is a portable, drone-mountable device that combines pulsating electromagnetic probing with neural-network machine learning to scan large areas for landmines and unexploded ordnance — quickly, remotely, and without putting sappers on contaminated ground.

The problem

Roughly one third of Ukraine’s territory — about 74,000 km² — is contaminated with mines and other explosive remnants of war and requires thorough examination. At the current demining pace of about 64 km² per year, fully clearing this territory would take approximately 70 years. SOCRAT is built to compress that timeline.

How it works

The device emits pulsating electromagnetic probes and captures the raw response from the ground. Neural-network machine-learning models then analyse that raw data to build 3D maps of suspect objects, distinguishing buried munitions from harmless metal clutter. Sappers use the resulting map to enter only the locations that actually warrant inspection.

The current model achieves a detection accuracy of at least 86%, with active development focused on pushing that figure higher.

Key advantages

• Drone-mountable. Mounted on a UAV, the device scans terrain at linear speeds of up to 40 km/h — orders of magnitude faster than ground-based manual sweeps.
• 3D maps, not just hits. Neural-network analysis of the raw electromagnetic data produces a 3-dimensional map of suspect objects, giving sappers precise localisation rather than a flat list of detections.
• Remote and standoff. No contact with the suspect ground is required — humans only enter the field once the map identifies what to inspect.
• Built around real data. The mathematical models and ML are tuned on field-collected sensor data, not laboratory ideal cases.