AUTOMATING ORCHARDS: A SYSTEM OF AUTONOMOUS TRACTORS FOR ORCHARD MAINTENANCE

� Abstract— This paper presents results from an extended field test of an autonomous multi-tractor system that performs mowing and spraying operations in a citrus orchard in Southern Florida. The system includes two autonomous tractors and uses a remote, human supervisor to assign tasks and help when needed. Each autonomous tractor detects obstacles with a perception system using ladar and cameras. The perception system makes use of both a geometric-base d detector and an appearance-ba sed classifier to detect hazards in the cluttered orchard environment and guide the tractors down the center of the tree rows. A mission planner uses a map of the orchard to produce optimized paths that cover the area requested by the supervisor. The paper also describes how the autonomous tractors fit into the existing orchard operations and how current practices used during manual spraying can improve the overall safeguarding of the autonomous tractor system by restricting access to areas of operation. The autonomous tractors have shown significant productivity improvements and have driven over 1,500 km, mowing and spraying, during field tests in the orchard.