EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

Airflow-assisted robotic strawberry harvesting system integrating deep learning detection and swarm-based optimization

Qi Gao, Yuhao Cui, Zixuan Xiao, Ruibo Liu, Huang Junjie, Zishen Liu, Ziliang Li, Shoutian Dong, Lifeng Guo, Miao Zhang, Jinggong Wei

Results in Engineering · 2026-12 · 32 · 113807

DOI: 10.1016/j.rineng.2026.113807 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Results in Engineering, 2026-12. Verfasst von Qi Gao und 10 weiteren Autor:innen. Die Arbeit nennt 52 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0). Gefördert von: Philippine Council for Industry Energy and Emerging Technology Research and Development.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Strawberry harvesting remains a labor-intensive task due to the fruit’s fragile structure, dense planting patterns, and complex greenhouse environments. To improve harvesting efficiency while reducing mechanical damage, this study proposes an integrated perception–decision–execution framework for a semi-contact strawberry harvesting robot. First, an improved lightweight detection model, YOLOv12n-DK, is developed by incorporating a Dynamic Large-Kernel Attention (DLKA) module into the baseline YOLOv12n architecture. The DLKA module is introduced to improve contextual feature representation and potentially enhance detection robustness under occlusion, dense fruit distribution, and varying illumination conditions, with limited additional computational complexity. Second, particle swarm optimization (PSO) is employed to refine the Cartesian offsets of predefined harvesting-trajectory control points by minimizing a weighted objective that considers path length, smoothness, reference tracking, and kinematic feasibility. Visual target localization and CFD-selected airflow conditions are integrated into the robotic workflow but are not treated as decision variables in the PSO optimization. Third, an airflow-assisted semi-contact end-effector is designed to stabilize strawberries during peduncle cutting, thereby reducing mechanical compression and surface damage during harvesting. Experimental results demonstrate that the proposed YOLOv12n-DK model achieves a precision of 93.0%, recall of 94.4%, and mAP50 of 97.6%, outperforming several lightweight detection models. In greenhouse experiments without canopy occlusion, the robotic system achieved an average harvesting success rate of 91.8% with an average harvesting time of 10.4 s per fruit. The proposed framework demonstrates the feasibility of integrating lightweight vision models, swarm intelligence optimization, and semi-contact robotic manipulation for intelligent soft-fruit harvesting in precision agriculture.

Förderung

  • Philippine Council for Industry Energy and Emerging Technology Research and Development

Zitieren

APA

Gao, Q., Cui, Y., Xiao, Z., Liu, R., Junjie, H., Liu, Z., Li, Z., Dong, S., Guo, L., Zhang, M., & Wei, J. (2026). Airflow-assisted robotic strawberry harvesting system integrating deep learning detection and swarm-based optimization. Results in Engineering, 32, 113807. https://doi.org/10.1016/j.rineng.2026.113807

Vancouver

Gao Q, Cui Y, Xiao Z, Liu R, Junjie H, Liu Z, et al. Airflow-assisted robotic strawberry harvesting system integrating deep learning detection and swarm-based optimization. Results in Engineering. 2026;32:113807. doi:10.1016/j.rineng.2026.113807

BibTeX
@article{Gao2026,
  author = {Gao, Qi and Cui, Yuhao and Xiao, Zixuan and Liu, Ruibo and Junjie, Huang and Liu, Zishen and Li, Ziliang and Dong, Shoutian and Guo, Lifeng and Zhang, Miao and Wei, Jinggong},
  title = {Airflow-assisted robotic strawberry harvesting system integrating deep learning detection and swarm-based optimization},
  journal = {Results in Engineering},
  year = {2026},
  volume = {32},
  pages = {113807},
  doi = {10.1016/j.rineng.2026.113807}
}
Kostenlos starten →