Trajectory-synchronized diode laser actuation and dose-response field evaluation for precision weed control using a delta robot

Weed control in Vidalia onion production is constrained by a narrow management window, where delays in intervention can result in yield losses of up to 96% in direct-seeded systems. This challenge often forces growers to rely on costly transplanting practices and labor-intensive hand weeding, while the limited availability of effective post-emergence herbicides further restricts in-season control options. Robotic laser weeding offers a precise, non-chemical alternative by selectively targeting individual weeds while protecting crop health and minimizing environmental impact. This research introduces the application of a delta parallel robot equipped with a 60 W diode laser to achieve photothermal weed control by targeting the leaf emergence point (LEP). The manipulator used a lightweight structure, combining aluminum proximal arms with carbon fiber distal arms to support fast motion with limited vibration. A microcontroller-based control stack executed analytical inverse kinematics and synchronized waypoint sequencing, with transistor-transistor logic (TTL) based laser triggering to ensure precise synchronization between motion and laser actuation. The motion planner excluded targets outside the robot workspace and adaptively reduced speed in regions approaching kinematic singularities to maintain stability and accuracy. Positional accuracy was evaluated in a greenhouse setting across 100 commanded targets, yielding a mean radial error of 0.83 mm. Moreover, field positional validation across five sites yielded a mean radial error of 1.11 mm and a 95 th percentile error of 1.50 mm. Field experiments were conducted using Brassica rapa ‘Twister’, selected for its uniform morphology (canopy width: 10.0 ± 1.5 cm, height: 5.0 ± 1.5 cm), arranged in a completely randomized design. Treatments consisted of laser exposure durations of 0.5, 1, 2, and 3 s at 60 W, with 10 replicates per treatment level. Weed injury was quantified using Chlorophyll Content Index (CCI), visual necrosis, half rosette span, and plant height. Measurements were collected prior to treatment and at 0 (control), 24, 36, 48 and 240 h (10 d) post-treatment. By 10 days after treatment, CCI decreased by 92% and 59% at exposure durations of 3 s and 2 s, respectively. Correspondingly, weed mortality reached 90% at 3 s and 60% at 2 s. In contrast, 1 s and 0.5 s exposures induced only transient injury, followed by recovery, with no mortality observed in controls. Overall, the system delivered repeatable millimeter-scale dosing with deterministic motion-actuation synchronization. Integrating perception and navigation modules could further extend this framework toward fully automated, non-chemical weed control in Vidalia onion production systems.