Robotic weeding's false dawn? Ten requirements for fully autonomous mechanical weed management

摘要

While machines called weeding robots are now commercially available and many more designs are being actively researched, I contend that current machines are not truly robotic weeders, rather they are essentially self-guiding vehicles carrying weeding tools. I consider true robotic weeders to be a far more difficult objective. While advances in robotics have been outstanding, the weeding component often appears to be an afterthought. I contend that the weeding is as complex as the robotics. A genuine weeding robot should be able to: (i) monitor the crop, weeds, weather and soil, (ii) decide when the crop should be weeded, (iii) choose the optimal weeder, (iv) take the weeder to the field, (v) adjust the weeder for optimal performance, (vi) continuously monitor the entire weeder for blockages and mechanical breakages and fix them in the field, (vii) continuously monitor and adjust the weeder's performance, (viii) return the weeder to the farmyard and (ix) clean, maintain and store the weeder, that is replace all human intervention. This ten-point list both defines and is a guide to what is required for completely autonomous robotic weeding. Currently, this list is far beyond current technology and it may be decades before it is realisable. The aim of this study therefore was not to disparage the achievements of agricultural roboticists, rather it is to highlight the complexity and demands of mechanical weeding and therefore describe what is really required to create a true robotic weeder. I therefore hope it will guide and expedite research and lead to more rapid success for robotic weeding.