Improving agricultural robot patch-spraying accuracy and precision through combined error adjustment

Abstract

Site-specific weed management (SSWM) involves precision spraying operations based on the spatial variability of weeds in the field. High spraying precision enables extensive economic and environmental benefits. We developed a high-precision patch-spraying robot system based on weed maps to conduct patch-spraying operations at the individual plant scale. The patch-spraying droplet deposition effect is evaluated by post-spraying droplet deposition images. The agricultural robot mainly consists of a walking chassis, spraying, and control systems. The walking chassis provides walking power and support for the suspended spraying system. Combining the static structural and dynamic error models of the robot improves patch-spraying accuracy and precision. Twelve patch-spraying simulations were carried out on the leveling site. Different correction methods and driving speeds were tested, with driving rates of 0.5, 1.0, and 1.5 m/s. Aerial images of the patch-spraying process were evaluated using pixel point-based discriminant statistics after operation completion. The droplet deposition coverage accuracy ranged from 95.9 to 98.8 %. The center of droplet deposition offset distance was 3.7–7.2 cm. Field results showed that the patch-spraying weed control rate based on individual plant scales (spraying precision ≤ 15 cm) was 94.2–98.1 %. In summary, the developed agricultural robot performs patch-spraying operations at the individual plant scale. Patch-spraying accuracy and precision are improved by modeling corrections for random errors. A general error correction method for the deposition position of fog droplets is explored to achieve precise droplet delivery in the walking state of plant protection operation equipment.