Drones Navigate Without GPS Using AI Vision

Moscow Aviation Institute (MAI) is creating an innovative autonomous navigation system designed to enable drones to operate reliably even when GPS signals are unavailable. This advanced technology integrates algorithms, cameras, and sensors for spatial navigation, complemented by a neural network that analyzes video streams to detect people.

The system employs visual-inertial navigation based on SLAM, allowing drones to construct their own maps and maintain accurate positioning through complex maneuvers and GPS loss. Developed by a team of MAI students, the core innovation lies in machine learning algorithms that not only facilitate navigation but also enable real-time human detection.

This MAI solution offers distinct advantages over existing systems, which often require expensive hardware or operate as separate, less integrated units. The MAI project embeds navigation directly into a custom flight controller for enhanced stability and reliability. The technology is designed to be scalable, adaptable for various unmanned aircraft systems and tasks, including monitoring remote locations and surveying infrastructure.

Field trials are currently underway, with ongoing efforts to enhance functionality and data accuracy. Future developments include expanding human detection capabilities for autonomous object tracking and trajectory adjustments. The project is slated for completion in the second quarter of 2026, with a focus on improving performance in adverse conditions like low light or smoke.