Co-robotic intra-row weed control system

• Published in: Biosystems engineering 2014-10, Vol.126, p.45-55
• Main Authors: Pérez-Ruíz, Manuel, Slaughter, David C., Fathallah, Fadi A., Gliever, Chris J., Miller, Brandon J.
• Format: Article
• Language: English
• Subjects: Actuation; Agricultural machinery and engineering; Agronomy. Soil science and plant productions; Assessments; Automation; Biological and medical sciences; Co-robot; Crops; Fundamental and applied biological sciences. Psychology; Generalities. Biometrics, experimentation. Remote sensing; Odometry; Parasitic plants. Weeds; Phytopathology. Animal pests. Plant and forest protection; Precision farming; Real time; Robots; Tomato; Weed control; Weeds
• ISSN: 1537-5110; 1537-5129
• DOI: 10.1016/j.biosystemseng.2014.07.009

The automation of intra-row weed control in row crop production systems is very challenging. This work describes the development and in-field assessment of an automatic intra-row, hoe-based weeding co-robot system with real-time pneumatic hoe actuation based on an accurate odometry sensing technique. The US National Science Foundation has identified a need for robots (called co-robots) that serve as co-workers and work beside, or cooperatively with, people. These co-robots have a symbiotic relationship with a human partner, where, as a team, they combine their relative strengths to jointly perform a task. Such co-robots should be relatively inexpensive and easy to use. In this work, mechanical weed control was achieved by a co-robot actuator that automatically positioned a pair of miniature hoes into the intra-row zone between crop plants. The design was tested in a precision transplanted row crop and may also be suitable for direct seeded row crops. Co-robot cost was minimised by limiting the system to a single, simple odometry sensor. Co-robot hoe actuation was controlled using pre-programmed knowledge of the crop planting pattern and real-time odometry data as the control input for hoe positioning. Low-frequency drift in the odometry control points relative to the actual plant locations was corrected occasionally as needed in real-time by a human partner monitoring system performance. The co-robot was evaluated in an experimental trial conducted on the UC Davis campus farm. Assessment was based upon the follow-up hand hoeing required after automated intra-row weeding in comparison to the labour required to manually hoe a control plot. The mean person hours required for hand hoeing weeds in the control were 0.241 h for the 100 m2 plot, while only 0.102 h 100 m−2 were required in follow-up labour to complete the weed removal in the plots weeded by the co-robot. This represents a 57.5% reduction in hand labour requirements for intra-row weed control and indicates that the co-robot could help reduce traditional hand hoeing labour requirements with mechanised weed control in intra-row areas between the crop plants. •An automatic intra-row, weeding co-robot system was successfully demonstrated.•A co-robot actuator automatically positioned a pair of miniature hoes to kill weeds.•A 57.5% reduction in hand labour for intra-row weed control was achieved.•Weeding co-robot system design was suited for all precision planted row crops.