Autonomous robot navigation for precision horticulture

Certain experimental horticultural operations require an autonomous vehicle capable of navigating through a field of crop plants. Since the vehicle must be able to operate in many fields to be economic, artificial navigation beacons and detailed prior maps are disadvantageous. A novel navigation scheme has been devised, which allows the crop rows themselves to be used as a navigation aid. The commanded path of the vehicle through the field is expressed by path curvature as a function of forward distance; parts of the path are marked as being aligned with crop rows. Data from image analysis is combined with that from a solid state compass and dead reckoning using an extended Kalman filter (EKF). Rather than position in an arbitrary Cartesian coordinate frame, the EKF estimates position and orientation error from the commanded path, which may be used directly for feedback control of vehicle motion. The method has been implemented on a small horticultural vehicle, allowing it to follow crop rows and turn at the end of the rows fully autonomously. The system has been tested in the field, and results showing the accuracy of guidance along crop rows are presented.