Winning Entry of the 2025 Munich EDTH Quantum Systems Challenge for aireal search path optimization for drones. The objective is to find flight paths that:
-
are hard to predict
-
minimize flight time
-
maximize chance to detect targets
- Terrain feature extraction and analysis of feature opacity from satellite images, as well as edge detection to determine from which side to view buildings and forests
- Drone simulator capable of calculating overall detection coverage probability, based on terrain features and modeling drone camera behavior.
- Initial path guess based on physical dynamics and brownian motion.
- Optimization drone flight paths based on cumulative detection probability across an entire area using state-of-the-art global optimization algorithms.
Here are some example animations and plots.
extracted map information including terrain information, terrain accessibility, edge detection
Left: Naive drone search path implementation with a lawnmower pattern.
Right: Optimized drone search path. Compared to naive approach, by our benchmarks, this path minimizes expected target detection time by 23%, while mainting same path length
Left: simulation of the drones overall detection cone. Right: Cumulative detection probability over the drones flight path