The wizardry of improved weed control
Michigan’s cabbage fields rarely make headlines but they quietly carry economic weight (and mounting weed control strain) across the state.
In 2024 about 3,800 acres of cabbage were harvested in Michigan – a crop that is small in acreage but significant in value. Growers face a persistent profit pinch because cabbage returns only slightly exceed production costs, leaving little cushion for rising inputs and labor challenges. Weed management alone can average about $500/acre, largely driven by hand weeding after herbicide misses or cultivation breakdowns.
Rising labor costs, herbicide resistance and weather volatility further complicate dependable weed control, pushing researchers and growers toward automated alternatives. At Michigan State University, graduate research assistant Akane Takenaka and Daniel Brainard, Ph.D., evaluated a compact robotic mechanical weeder as a potential solution for cabbage systems.
Their study aimed to measure how an autonomous cultivator influences weed density and cabbage yield under conventional management conditions. They hypothesized that robotic weeding would improve suppression and yield, particularly when pre- or post-emergent herbicides failed.
Field trials ran in 2024 and 2025 at MSU’s Horticulture Teaching & Research Center in Holt on sandy loam soil. Plots tested robotic cultivation intensity using the Naïo OZ robot equipped with sweeps, finger weeders and flex-tine rakes. Treatment levels included no robotic passes; low intensity with two to three passes; and high intensity with four to five passes. Tool combinations and their settings were based on timing – rakes were only used toward the end of cultivation.
Four herbicide programs were included: no herbicide; pre-emergent (metolachlor); post-emergent (oxyfluorfen plus clethodim); and a combined pre- and post- program.
Cabbage was transplanted in early July with standard fertility, irrigation and insect management practices across all treatments.
In 2024, robotic cultivation consistently reduced weed density and increased yield across herbicide regimens compared with untreated controls. Improvements appeared whether pre- or post-emergent herbicides were used or omitted, indicating a strong standalone performance of the system.
In 2025, benefits were more conditional, with significant gains mainly when post-emergent herbicide failure scenarios were simulated. Under these conditions robotic cultivation improved yield and reduced weeds in control and pre-emergent only treatments.
Few differences emerged between low and high intensity passes, suggesting diminishing returns beyond moderate cultivation frequency. Operational limits were noted, including slow travel speeds (below 1.25 mph) and the occasional need for human intervention.
These findings highlight robotic cultivation as a promising tool for integrated weed management in cabbage production systems. The technology is most valuable when herbicide programs fail or when environmental conditions reduce chemical effectiveness. However, current limitations in speed, reliability and field capacity prevent it from replacing conventional methods.
Takenaka and Brainard emphasized that improvements in navigation, tool design and operational efficiency are necessary for broader adoption. Robotic weeding currently appears best positioned as a complementary strategy that strengthens – but does not replace – existing weed control systems.
Future research will focus on refining tool configurations and adjusting robotic passes to optimize crop safety, weed suppression and efficiency. Additional studies will also explore interactions between robotic cultivation and reduced herbicide programs under varying environmental conditions.
Researchers hope these advances will help move autonomous weeding from experimental trials toward practical adoption in vegetable production.
Improving robustness under real farm conditions will be as important as refining weed detection or mechanical design in future iterations.
by Enrico Villamaino
