Development of an Autonomous Weed-Weeding Robot Utilizing Edge-AI and Multispectral Computer Vision for Row-Crop Farming
Abstract
Weeds represent one of the most economically damaging biotic stresses in row-crop production, yet current chemical control strategies face mounting regulatory pressure and resistance evolution. This paper presents the design, integration, and preliminary field evaluation of an autonomous ground robot for in-season mechanical weed control in row-crop systems, combining Edge-AI inference with a five-band multispectral vision pipeline. The system architecture comprises a four-wheel differential-drive platform, a MicaSense-class multispectral imaging module, an NVIDIA Jetson Orin edge-computing unit, and a servo-actuated inter-row/intra-row tine mechanism. A YOLOv8-nano model, trained on a field-collected and synthetically augmented dataset of 14,200 annotated multispectral frames covering four principal row-crop weed species (Chenopodium album, Amaranthus retroflexus, Galium aparine, Convolvulus arvensis), is deployed for real-time per-frame weed localisation at 18 frames per second on the embedded platform. Vegetation index fusion combining NDVI and NDRE channels provides a secondary spectral discriminator that reduces false-positive detections on soil and crop residue by 34% relative to RGB-only inference. RTK-GNSS-guided row-following navigation achieves lateral deviation of ≤1.8 cm (1σ) at 0.6 m/s travel speed. Field trials on maize and soybean demonstrate a weed removal efficacy of 82–89% with a crop damage rate below 1.2%, at a treatment throughput of 0.45 ha/h.
References
[1] E. C. Oerke, "Crop losses to pests," J. Agric. Sci., vol. 144, no. 1, pp. 31–43, Feb. 2006.
[2] V. Balaska, Z. Adamidou, Z. Vryzas, and A. Gasteratos, "Sustainable crop protection via robotics and artificial intelligence solutions," Machines, vol. 11, no. 8, pp. 774, Aug. 2023.
[3] A. Mouazen, I. Alexandridis, N. Ioannidis, and M. Ruiz-Garcia, "Sensing and perception in robotic weeding: Innovations and limitations for digital agriculture," Sensors, vol. 24, no. 17, pp. 5686, Sep. 2024.
[4] W. Zhang, Z. Miao, N. Li, C. He, and T. Sun, "Review of current robotic approaches for precision weed management," Curr. Robot. Rep., vol. 3, no. 3, pp. 139–151, Sep. 2022.
[5] F. Visentin, S. Cremasco, M. Sozzi, L. Signorini, M. Signorini, F. Marinello, and R. Muradore, "A mixed-autonomous robotic platform for intra-row and inter-row weed removal for precision agriculture," Comput. Electron. Agric., vol. 214, pp. 108270, Nov. 2023.
[6] L. Quan, W. Jiang, H. Li, H. Li, Q. Wang, and L. Chen, "Intelligent intra-row robotic weeding system combining deep learning technology with a targeted weeding mode," Biosyst. Eng., vol. 216, pp. 13–31, Apr. 2022.
[7] N. Rai, Y. Zhang, B. G. Ram, L. Schumacher, R. K. Yellavajjala, S. Bajwa, and X. Sun, "Applications of deep learning in precision weed management: A review," Comput. Electron. Agric., vol. 206, pp. 107698, Mar. 2023.
[8] G. Jocher, A. Chaurasia, and J. Qiu, "Ultralytics YOLOv8," GitHub, 2023. [Online]. Available: https://github.com/ultralytics/ultralytics
[9] F. Dang, D. Chen, Y. Lu, and Z. Li, "YOLOWeeds: A novel benchmark of YOLO object detectors for multi-class weed detection in cotton production systems," Comput. Electron. Agric., vol. 205, pp. 107655, Feb. 2023.
[10] X. Fan, X. Chai, J. Zhou, and T. Sun, "Deep learning based weed detection and target spraying robot system at seedling stage of cotton field," Comput. Electron. Agric., vol. 214, pp. 108317, Nov. 2023.
[11] C. Mwitta and G. C. Rains, "Evaluation of inference performance of deep learning models for real-time weed detection in an embedded computer," Sensors, vol. 24, no. 2, pp. 514, Jan. 2024.
[12] V. Balaska, Z. Adamidou, Z. Vryzas, and A. Gasteratos, "Sustainable crop protection via robotics and artificial intelligence solutions," Machines, vol. 11, no. 8, pp. 774, Aug. 2023.
[13] N. Genze, T. Bhattarai, R. Ajekwe, M. Grieb, and D. G. Grimm, "WeedsGalore: A multispectral and multitemporal UAV-based dataset for crop and weed segmentation in agricultural maize fields," arXiv:2502.13103, 2025.
[14] N. Genze, R. Ajekwe, Z. Güreli, F. Haselbeck, M. Grieb, and D. G. Grimm, "Deep learning-based early weed segmentation using motion blurred UAV images of sorghum fields," Comput. Electron. Agric., vol. 202, pp. 107388, Nov. 2022.
[15] G. Coleman, W. Salter, and M. Walsh, "OpenWeedLocator (OWL): An open-source, low-cost device for fallow weed detection," Sci. Rep., vol. 12, no. 1, pp. 170, Jan. 2022.
[16] G. Iannaccone, C. Sbrana, I. Morelli, and S. Strangio, "Power electronics based on wide-bandgap semiconductors: Opportunities and challenges," IEEE Access, vol. 9, pp. 139446–139456, 2021.
[17] T. Ayoub Shaikh, T. Rasool, and F. Rasheed Lone, "Towards leveraging the role of machine learning and artificial intelligence in precision agriculture and smart farming," Comput. Electron. Agric., vol. 198, pp. 107119, Jul. 2022.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 BAUDER Press
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
Authors submitting a manuscript retain copyright of their work upon acceptance for publication in Aintelia Science Notes Journal (ASNJ). By submitting, authors agree that:
- Copyright Retention: Copyright of the published article remains with the author(s). BAUDER Press is granted a non-exclusive license to publish, reproduce, distribute, and archive the work.
- Licensing: The article is published under CC BY-NC 4.0, allowing third parties to share and adapt the work for non-commercial purposes, provided appropriate credit is given to the author(s) and the journal.
- Author Rights: Authors are free to deposit, distribute, and reuse their article in any medium (repositories, theses, conferences, teaching materials) without restriction, provided the journal is cited as the original publication venue.