Efficient Aerial Coverage Search in Road Networks

Abstract

Wide-area coverage is well suited for small unmanned aircraft, however common coverage algorithms are particularly inefficient in many common environments such as urban areas. Classical open-loop spiral or lawn-mower patterns typically cannot represent large sub-regions that are unlikely to contain an object of interest that are thus wastefully covered. More general algorithms can maintain arbitrary likelihood distributions but fundamentally solve a complex continuous-space trajectory optimization problem, requiring a trade-off between look-ahead and computational complexity while providing at best asymptotic coverage guarantees. This paper considers generating global coverage patterns in the space of the coverage area, with specific focus on road networks, and mapping these to UAV actions. A new method is proposed for producing and following such patterns with dynamics-constrained vehicles, compared with existing coverage strategies, and shown to be advantageous in many realistic environments through simulations of real-world aircraft. Results suggest environment density as a metric for algorithm selection given a lack of dominance by any one strategy across all environments.