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 lawnmower 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.

Optimization approaches for civil applications of unmanned aerial vehicles (UAVs) or aerial drones: A survey

“…First of all, since drones perform observations from the air, there is a trade‐off between taking pictures from a higher altitude with a larger camera footprint, but lower resolution and higher energy consumption, and taking pictures from a lower altitude with a smaller camera footprint, but higher resolution and lower energy consumption. Secondly, cameras and sensors may be attached to a drone at different orientations (eg, side‐aimed cameras vs. directly‐downward facing cameras) , so that different shapes and positions of the camera footprint relative to the drone are possible. Moreover, more sophisticated sensors can change their orientation during flight.…”

“…Preferred paths of drones have to avoid unproductive movements, such as repeated surveillance of the same points in or “idle” flights to the next zone of operation. Common objective functions aim to minimize some distance‐related cost function , completion time (in case of several drones or setup times between drone flights) , or energy consumption (because the drone's energy consumption depends on the altitude of flight and the direction of acceleration) . Furthermore, since each turn of the drone requires additional time and energy (see Section 3), a cost‐minimizing path will generally contain fewer turns.…”

“…Independent strategic drones may negotiate a solution for partitioning into individual areas of responsibility according to some specified auction protocol . Coverage of a road network is examined in Dille and Singh , Oh et al , and Oh et al .…”

Design Challenges of Multi-UAV Systems in Cyber-Physical Applications: A Comprehensive Survey and Future Directions

“…This situation resulted in the issue of designing, powerful, flexible, and lightweight UAVs while emphasizing the practicality of target mobility predictions. Another challenge is that sensors and cameras may be placed in UAVs at various positions [102]. Therefore, numerous structures and situations of the camera coverage area relative to the UAVs are likely.…”

Drone arc routing problems

“…This is solved with a greedy insertion heuristic that models drone travel distance between the road components as a Dubins path, as for a fixed wing drone. Dille and Singh also model drone travel using Dubins paths for a setting where the drone has a sensor with a radius of coverage. The authors convert the road (edge) covering problem to a TSP by discretizing the road network into a set of coverage points designed so that visiting each point will ensure that all parts of the road network are within the sensor range of a drone.…”