Shipped: Hybrid Quantum-Classical AI — Drone Gas Leak Detection

An open-source AI research proof of concept investigating how hybrid quantum-classical optimization can improve autonomous drone coordination for industrial gas leak detection.

The framework combines machine learning for gas leak severity estimation, quantum-inspired Quadratic Unconstrained Binary Optimization (QUBO) for drone deployment and Voronoi partitioning, and classical 2-Opt Traveling Salesperson Problem (TSP) heuristics for route planning.

Using synthetic hazard environments and the GPLA-12 acoustic gas leakage dataset as a real-world data proxy, the system prioritizes high-risk areas and minimizes the time required for a drone fleet to reach potential leaks.

In a simulated seven-drone experiment, the hybrid approach achieved a mean time-to-detection of 10.0 seconds, compared with 43.9 seconds for an intensity-weighted K-Means baseline. The repository also includes hazard heatmaps, drone partitioning visualizations, and an animated mission demonstration.

The project explores applications in AI-assisted infrastructure monitoring, autonomous inspection, environmental safety, and intelligent emergency response. All quantum components currently run through classical simulation; this is a research POC rather than a field-validated or quantum-hardware deployment.