cs.ROOct 13, 2026

MARTS: A Modular Agricultural Robot for Targeted Spraying of Diseased Tomato Plants in Greenhouses

Authors: Rajmeet Singh, Manveen Kaur, Yusuf Adamu, Lochan Kshetrimayum, Irfan Hussain

Organizations: Mohamed bin Zayed University of Artificial Intelligence Abu Dhabi, UAE · Khalifa University Center for Autonomous Robotic Systems (KUCARS) Khalifa University, Abu Dhabi, UAE · Department of Mechanical Engineering University of Windsor ON, Canada · Kobe University Japan

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Broadcast spraying remains the dominant method of disease control in greenhouse tomato production, yet it treats healthy and diseased tissue alike, exposes workers to hazardous aerosols, and depends on operator skill. This paper presents MARTS (Modular Agricultural Robot for Targeted Spraying), a low-cost robotic platform that closes the loop from plant-level disease detection to leaf-localised spray delivery in commercial greenhouses. MARTS integrates a Clearpath Husky A200 mobile base, a UR5 manipulator, a novel modular spraying end-effector built around a single 12 V micro-diaphragm pump, and a wrist-mounted Intel RealSense D435 RGB-D camera. Disease detection reuses our previously published Tomato Leaf Disease Vision Language Model (TLDVLM; Grounding DINO, SAM-2 and LoRA-tuned BLIP-2, 97.27% accuracy over 10 classes); this work focuses on actuation, control and system integration. A ROS pipeline lifts detections into 3D targets, generates collision-aware nozzle poses at a 120 mm stand-off using fitted leaf normals, and scales pump pulse duration to lesion area. Water-sensitive paper calibration showed a linear pulse-volume relationship (R^2 = 0.994). In MuJoCo simulation, an adaptive base-repositioning strategy achieved 100% coverage of 18 waypoints with a maximum orientation error of 1.1 deg. Deployed at the SILAL commercial greenhouse (Al Ain, UAE) on cherry tomatoes under natural infection, MARTS treated 312 diseased leaves in 12 autonomous trials with 92.4% targeting success, 18.7 mm mean spray-centre error, 6.1% adjacent-leaf overspray, and a 71% reduction in pesticide volume versus broadcast spraying. Mean cycle time was 4.8 s per leaf, and the spraying subsystem costs USD 412, 6-10x lower than comparable systems, showing that accurate leaf-level targeting is achievable on commodity hardware.

Figures & tables

Figure 1: Representative conventional bulk-spraying practices in greenhouse production. (a) Hose reel trolley sprayer with 200 L tank for manual broadcast application along the row. (b) Self-propelled compact sprayer for ornamental and vegetable greenhouses. (c) Operator pushed cart sprayer with vertical nozzle boom for uniform canopy treatment. (d) Pesticide applicator in full PPE chemical resistant coverall, respirator, face shield, and gloves.
Figure 2: System architecture of the MARTS platform. The four subsystems: Clearpath Husky A200 mobile base, Universal Robots UR5 manipulator, modular spraying end-effector, and on-board perception stack with the Intel RealSense D455 RGB-D camera communicate through a ROS graph running on a single on-board PC. The dashed green target indicates the diseased tomato leaf: the perception stack observes it through the VLM disease detector, and the end-effector delivers the leaf-localised spray pulse, closing the detection-to-spray loop.
Figure 3: Selective spraying end-effector designed for integration with the UR5 manipulator: (a) isometric CAD rendering showing the key components, including the Intel RealSense RGB-D camera mounted on top for target localization, the spraying nozzle at the front, the UR5 tool-flange mount, and the hosepipe connector at the base; (b) sectional cutaway view illustrating the internal routing of the brass threaded fitting that channels the spray fluid from the hosepipe connector to the nozzle outlet; (c) 3D-printed prototype mounted on the UR5 end-effector during field deployment in the greenhouse.
ParameterValue
Pump type12 V DC diaphragm
Free-flow rate4.0 L/min
Max pressure0.55 MPa
Reservoir capacity10.0 L (food-grade)
Tubing inner diameter4.0 mm (polyurethane)
Nozzle typeHollow-cone (default)
Table 1: Modular spraying end-effector specifications.
Figure 4: MARS integrated hardware-software architecture for autonomous disease detection and selective impulse spraying. The system orchestrates perception (GroundingDINO, SAM-2, BLIP-2 LoRA), spray decision logic, and mobile manipulation via ROS 2.
Figure 5: System architecture of the precision spraying platform: the on-board ROS-based PC communicates via impulse mechanism with an Arduino-controlled pump driver (5 V input) to command the micro-diaphragm pump (12 V supply). The pump draws liquid from the spray tank and delivers pressurized fluid to the proposed end-effector, which atomizes the pesticide formulation for selective spot-spraying application.
Figure 6: MuJoCo-based simulation framework for validating trajectory planning and mobile-base positioning strategies prior to hardware deployment.
Figure 7: Predefined waypoint layout and trajectory design for plant scanning, spanning the full plant height with 18 discrete inspection poses.
MetricCase ICase IICase IIIUnit
Average PTE3.2±1.13.2\pm 1.16.8±2.46.8\pm 2.42.1±0.82.1\pm 0.8mm
Average OTE0.34±0.180.34\pm 0.180.78±0.420.78\pm 0.420.41±0.190.41\pm 0.19deg
Max Cumulative OTE2.88.41.1deg
CJM142.3189.7156.2rad
Cycle Time156142168s
Workspace Coverage94.4100.0100.0%
Table 2: Comparative simulation results for three mobile base positioning strategies.
Figure 8: MARTS experimental setup in a real greenhouse environment. The annotated frame shows the complete system deployed in a tomato crop aisle, with key components labelled: the Husky A200 mobile platform, UR5 robot manipulator, front module (RGB-D camera and spray nozzle assembly), on-board PC, and surrounding tomato crop rows. The greenhouse aisle width ( ≈1.2\approx 1.2 m) constrains the mobile base footprint and manipulator reach envelope.
Figure 9: MARTS real-greenhouse experiment workflow illustrating the four-stage trial protocol: (1) greenhouse row preparation and environmental assessment; (2) MARTS deployment and aisle alignment; (3) autonomous plant interaction with front module operating near leaves and tomato clusters; (4) post-trial field evaluation using video evidence and treatment logs.
Figure 10: Time sequence of MARTS experimentation in the greenhouse row. Six representative frames captured at (a) t=0t=0 s: initial deployment with the platform at the row entrance; (b) t=20t=20 s: front module approaches the crop canopy; (c) t=40t=40 s: row traversal with coordinated mobile manipulation; (d) t=65t=65 s: close-range positioning at the tomato canopy ( ≈0.15\approx 0.15 m working distance); (e) t=100t=100 s: mid-trial repositioning in the aisle; (f) t=130t=130 s: in-canopy interaction with dense foliage and fruit clusters.
Figure 11: Close-range robot plant interaction during MARTS field testing: (a) full view showing the front module positioned within the tomato canopy between upper foliage and lower fruit clusters; (b) zoomed view of the module canopy interface, with the front module housing and RealSense D435 camera operating at 0.20–0.25 m working distance from the leaf surface; (c) zoomed view of a green tomato fruit cluster with dense leaf occlusion, illustrating the visual complexity encountered during real world disease scanning; (d) TLDVLM based segmentation overlay of diseased leaf.
RowDetected LeavesSuccess (%)Mean Error (mm)Overspray (%)
Row 110493.318.15.8
Row 210891.719.46.5
Row 310092.018.66.0
Pooled31292.418.76.1
Table 3: Targeting performance of the MARTS platform across three greenhouse rows at the SILAL commercial facility, Abu Dhabi, UAE. Values are reported per row and pooled across all 312 detected diseased leaves.
ItemCost (USD)
12 V DC diaphragm pump78
Hollow-cone nozzle set (5 tips)64
2 L food-grade reservoir + cap22
Polyurethane tubing, fittings, relief valve41
MOSFET interface board56
Printed tool-flange adapter9
Table 4: Bill of materials for the modular spraying end-effector at single-unit retail pricing.
SystemCropTargeting (%)Vol. Red. (%)Cost (USD)Modular?
Modular robot + multispectral [ Oberti et al., 2016 ]Grapevine85–10065–85Not reportedPartial
Robotised patch sprayer [ Gonzalez-de-Soto et al., 2016 ]Row crops99.5 ∗56>> 3 000No
Smart tree-crop sprayer [ Partel et al., 2021 ]Citrus84 †28>> 2 500No
DL spraying robot + PWM [ Seol et al., 2022 ]Pear68.9 ‡—>> 4 000No
Smarter robotic sprayer [ Baltazar et al., 2021 ]Vineyard—40–60>> 3 500No
MARTS (ours)Tomato92.471412Yes
Table 5: Comparison of MARTS with state-of-the-art selective-spraying systems reported in the literature for greenhouse and protected cultivation applications. Targeting refers to the percentage of detected diseased or target regions that received a correctly placed spray pulse. Volume reduction is relative to each study’s own broadcast or conventional baseline. Cost refers to the spraying subsystem hardware only (excluding mobile base and manipulator).
Appendix figures & tables2 assets

Supplementary material from the paper’s appendix.

Appendix

SymbolDescriptionUnit
bbLeaf bounding box from VLMpixels
AABounding box areamm 2
(uc,vc)(u_{c},v_{c})Bounding box centre in pixels—
(fx,fy)(f_{x},f_{y})Camera focal lengthspixels
(cx,cy)(c_{x},c_{y})Camera principal pointpixels
zzMedian depth at leaf centroidm
Appendix
Table 6: Nomenclature: Geometric and control variables.
Abbrev.Full Form
BOMBill of Materials
CADComputer-Aided Design
DCDirect Current
DoFDegree of Freedom
FOVField of View
IMUInertial Measurement Unit
Appendix
Table 7: Nomenclature: Abbreviations.
More in this topic

Agricultural Robotics

Agricultural robotics develops autonomous machines that sense, navigate, and perform tasks such as monitoring, weeding, spraying, and harvesting crops.

See all 97 papers

Explore similar work

CardsList
Oct 17, 2026

Multi-Speed Field Evaluation of a Semi-Autonomous Precision Sprayer for Weed Control in Vegetable Crops

Machine vision-guided precision spraying is promising for sustainable vegetable production by enabling site-specific weed targeting, reducing herbicide application while protecting crop yield and quality. However, the effects of operating speed of sprayer systems on the integrated detection-tracking-actuation pipeline remain insufficiently characterized under field conditions. This study evaluated a semi-autonomous, tractor-based spraying system for targeted weed control in radish and quantified performance across three operating speeds. A five-season dataset comprising 10,599 images and 84,311 annotated instances across seven classes was curated for model development. A YOLOv11-medium model was integrated with ByteTrack tracking and a customized 12-nozzle sprayer on a semi-autonomous electric tractor. Field trials were conducted at 0.13, 0.27, and 0.45 m/s (i.e., 0.3, 0.6, and 1.0 mph) to characterize speed-dependent changes in weed detection, targeting, actuation, and crop damage. At the lowest speed of 0.13 m/s, the system achieved 94.6% mAP@50 and 93.2% recall for weed and crop detection. However, increasing platform speed substantially reduced system performance, with weed hit rate declining from 78.3% to 35.0% and crop contact rates increasing from 20.1% to 28.7%. Consequently, the system did not meet predefined commercial-viability thresholds. These results demonstrate that practical deployment requires improvements beyond weed detection alone, particularly coordinated improvements in motion-robust plant perception, spray-timing compensation, and crop-adjacent actuation precision. The identified performance bottlenecks provide specific engineering targets for advancing vision-guided precision spraying toward practical, high-speed vegetable weed management. The software programs developed in this study have been made publicly available to support future precision weeding research.

Md Khairul Islam, Boyang Deng, Yuzhen Lu +2

Department of Biosystems and Agricultural Engineering, Michigan State University, East Lansing, MI 48824, USA · Department of Horticulture, Michigan State University, East Lansing, MI 48824, USA

Precision AgricultureAgricultural RoboticsWeed Detection
AskPDF
Oct 6, 2026

A Swarm-Coordinated Multi-Robot System for Early Stress Detection in Agricultural Rows Using Multimodal Leaf Sensing

Early stress detection in crops is a necessity today to improve efficiency and reduce waste of time, money, and effort. However, most modern techniques, such as hyperspectral imaging and AI-based systems, are too costly and complex for medium and small-scale farmers to implement. This paper showcases CropSentry, a low-cost, ground-based multi-robot system that uses multimodal leaf sensing to continuously monitor crop health by tracking stress levels. The system comprises two autonomous bots that continuously detect leaf color and environmental data row by row. The observations are spatially mapped and sent over to the master bot, which uses color-coded row segments to generate a real-time web-based dashboard displaying crop health. After 63 observations were collected during the experiments, the results showed an overall crop health classification accuracy of 84.12%, with 82.60% for healthy plants, 88% for nutrient-deficient plants, and 80% for diseased plants. Also, 100% wireless communication success rate across 10 slave observations was achieved. Close-range leaf inspection across multiple bots can detect early stress in crops while remaining affordable, accessible, and scalable. It provides farmers with timely information to improve resource utilization and crop management.

Rishi Gupta, Astha Goyal, Vinay Vishwakarma

Delhi Public School Vasant Kunj, New Delhi, India · Assistant Professor, University of Delhi · Mentor, On My Own Technology Pvt. Ltd, Mumbai, India

Plant Disease ClassificationAgricultural Remote SensingAgricultural RoboticsPrecision Agriculture
AskPDF
Sep 9, 2026

Context-Aware Adaptive Pesticide Spraying for Agricultural Robots under Changing Weather and Terrain Using Vision-Language Models

Precision pesticide spraying is essential for optimizing application efficiency and ensuring uniform chemical distribution. Spraying performance is influenced by multiple factors, including environmental conditions such as temperature and wind speed, pesticide type, and the robot's capability to accurately perceive crops and target spray locations. Existing approaches predominantly emphasize crop detection and rely on predefined spraying parameters, whereas human operators dynamically adjust their spraying strategies by considering environmental conditions, region-specific crop characteristics, and the type of pesticide being applied. In this study, we propose a context-aware adaptive spraying framework based on Vision-Language Models (VLMs), which enables robots to leverage spatial reasoning and integrate information from multiple sources, including crop type, pesticide type, and weather data, to make adaptive and optimized spraying decisions. Subsequently, a trajectory-tracking controller based on Model Predictive Path Integral (MPPI) control is employed to ensure precise navigation and accurate spraying at crop locations. The comparative results demonstrate that the proposed method improves accuracy by at least 30% in detecting crop rows. In addition, the experimental evaluations conducted in two environments further demonstrate the robot's ability to flexibly adjust spraying volume and travel speed, while reducing pesticide drift.

Cong-Thanh Vu, Yen-Chen Liu

Department of Mechanical Engineering, National Cheng Kung University, Tainan, 70101, Taiwan

Agricultural RoboticsPrecision AgricultureVLMs for Robotics
AskPDF