Editorial: New trends in distributed and autonomous intelligent systems for crop production

Abstract

Our research topic has inspired studies on new results on lightweight computer vision (CV) models and edge deployment strategies for crop production. With the proliferation of unmanned aerial vehicles (UAVs), autonomous harvesters, and Internet-of-Things (IoT) devices in agriculture, the demand for efficient, accurate, and deployable perception models at the edge has become increasingly urgent. Our research topic has published four outstanding papers dealing with the design of lightweight deep learning models for different crop production tasks, including sugarcane node detection, garlic damage detection, olive fruit detection, and clustered litchi segmentation. These studies have considered the most advanced models like YOLOv8, YOLOv11n, and U-Net, and provided refined designs to adapt the models for resource-constrained edge devices and complex field environments. Specifically, all authors have conducted extensive experiments to validate the improved performance. For example, the model EdgeSugarcane proposed in the paper "EdgeSugarcane: A Lightweight High-Precision Method for Real-Time Sugarcane Node Detection in Edge Computing Environments" achieves a precision of 93.5%, a recall of 80.0%, and a mAP of 87.0% on the test set, with the inference time reduced to only 1.9 ms after TensorRT-based FP16 quantization, which is 3.3 times faster than before optimization. Meanwhile, the Garlic-YOLO-DD model proposed in "Garlic-YOLO-DD: A Lightweight Object Detection Algorithm for Garlic Damage Detection" reduces the number of parameters to 57.96% of YOLOv11n, decreases computational load by 20.63%, increases inference speed by 15.97%, and improves mAP@50 by 27.64 percentage points. The YOLO-TinyFuse model in "A Lightweight YOLO-TinyFuse Model for Small Target Detection of Olive Fruits" achieves an mAP50 of 92.3% and a recall of 84.5%, outperforming YOLOv8n by 2.6% and 3.2%, respectively, while reducing parameters by 6.76%. Furthermore, the DSD-UNet model in "A Lightweight DSD-UNet-Based Method for Clustered Litchi Segmentation" achieves an mIoU of 91.40% and an F1-score of 95.42%, with inference speed increased by 138.9% and parameter count reduced by 61.5% compared with the original U-Net. Generally, these studies extend the design of lightweight CV models and edge deployment strategies for more robust and efficient autonomous systems in crop production.Besides, our special issue also collected impressive studies on intelligent decisionmaking and knowledge-driven systems for crop production. The first study, entitled "Research on Plant Knowledge Graph Reasoning Based on Dual-Channel Attention and Topological Perception," proposes the KRGAI-PLANT model, which addresses the challenge of long-range dependency reasoning in plant knowledge graphs by integrating a global attention mechanism with local topological perception. This work provides an effective reasoning tool for transforming multi-source plant data into actionable knowledge, supporting association prediction and decision-making in precision agriculture, ecological monitoring, and intelligent plant protection. The second study, entitled "Challenges and Strategies for Harnessing Large Language Models in Plant Protection," provides a comprehensive perspective on the opportunities and challenges of applying LLMs in plant protection. The authors discuss key application areas such as pest and disease monitoring, control decision support, plant quarantine risk assessment, and pesticide development, and propose strategies including domain-specific fine-tuning, knowledge-graph grounding, GraphRAG, multi-agent collaboration, model compression, and hierarchical edge-cloud deployment. These studies have set fundamental supports for building distributed and autonomous intelligent systems in crop production, ranging from structured knowledge reasoning to natural language-based decision support.Generally, our research topic has provided a forum sharing pioneering ideas and investigations on distributed and autonomous intelligent systems for crop production, and collected a batch of AI models and frameworks specifically tailored for intelligent crop production. During the whole process, the research topic receives 2,165 downloads, 7,708 article views and 12K topic views. We believe that the research topic and these accepted papers will shed light on the subsequent studies on cutting-edge techniques for distributed and autonomous intelligent systems in crop production.