Mapping Orchards for Autonomous Navigation

Abstract

Autonomous orchard vehicles have been shown to increase worker efficiency in tasks including pruning, thinning, tree maintenance, and pheromone placing. To cover entire blocks, they must be able to repeatedly exit an orchard row, turn, and enter the next. The authors’ experience deploying autonomous vehicles for five years in commercial and research orchards shows that, when the map of the block is available, navigation errors decrease and the vehicle’s probability of finding the next row increases. From a cost perspective, it is important that the map be built without expensive surveying equipment, and preferably using the same sensors used for vehicle guidance. Here, we present a landmark-based method that creates a local map of the block in which the vehicle operates using the laser scanner already present on the vehicle. We describe the mapping procedure and present results obtained in commercial orchards.