Multi-UAS path planning for non-uniform data collection in precision agriculture

Abstract

This paper presents an augmented path-planning technique for unmanned aerial systems to generate focused trajectories about one or more areas of interest for non-uniform sensor data collection. The technique described in this paper uses a coordinate transformation that augments the work space with a temporary, virtual space in which existing path-planning and control algorithms can be used to provide uniform coverage. Transforming back to the original work space forces the planned trajectories to focus on regions of interest. We illustrate the application to precision farming, where regions of interest in a crop field correspond to stressed crop health. When collecting aerial survey data, we seek to have a higher density of sensor data in areas of interest (e.g., RGB images, multispectral images, etc.). The technique presented in this paper offers a method for concentrating sensor measurements around these regions of stressed crop health for one or more vehicles. In agricultural domains with multiple regions of interest, a Voronoi partitioning algorithm partitions the operating area into individual regions in which the augmented path-planning technique is applied. The path-planning in each region takes into account the resources available — i.e., vehicles with larger sensor footprints are assigned to larger regions and execute trajectories that are more broadly spread as compared to vehicles with smaller sensor footprints. Theoretical results are applied to commercial off-the-shelf unmanned systems, both in simulation and in a fully realized precision agriculture demonstration field experiment.