Routing decisions made by unmanned aerial vehicle (UAV) swarms are affected by complex and dynamically changing topologies. A centralized routing algorithm imposes the entire computational burden on one module, and the high data dimensionality renders computation burdensome. In this letter, we develop a multi-agent reinforcement learning-based routing algorithm for a UAV swarm. The UAVs are trained in a data-driven manner to make distributed routing decisions. Factors that include channel quality, UAV movement, UAV overhead, and the extent of neighbor variation are incorporated into link quality assessment. Long short-term memory is used to improve the Actor and Critic networks, and more information on temporal continuity is added to facilitate adaptation to the dynamically changing environment. Simulations show that the proposed routing algorithm reduces data transmission delay and enhances the transmission rate compared with traditional routing algorithms.

Routing protocols do not respond quickly to environmental changes due to the high mobility of nodes in the Flying Ad Hoc Network (FANET), to obtain reliable transmission links. This paper proposes an adaptive link maintenance method based on deep reinforcement learning (DRL-MLsA), which can dynamically adjust the time interval of broadcasting Hello packets. This method can cope with the highly dynamic network environment, and adapt to both active routing and table-driven routing protocols. The method considers the channel model of the signal and investigates the impact of UAV communication range on link maintenance. We can get an agent by perceiving the degree of changes in the number of neighbors in a dynamic environment. The optimal broadcast cycle was obtained to maximize the energy of the node to send and receive task data. We substituted the single-output network model with a competitive network to overcome the reward overestimation problem, which also improves the convergence speed of the algorithm. Simulation results showed that DRL-MLsA can reduce the communication overhead for link maintenance, while at the same time increase the throughput of the network and decrease the packet loss of transmission.

Recent advancements in protocol design and optimization for Flying Ad-hoc Networks have shown significant progress. However, the existing centralized optimization control algorithms pose conflicts with the distributed nature of these systems. To address this challenge, this paper proposes a novel distributed routing optimization algorithm based on multi-agent reinforcement learning, integrated with an adaptive multi-mode routing framework incorporating multi-protocol cooperation mechanisms. The algorithm adopts the Decentralized Execution Multi-Agent Deep Deterministic Policy Gradient (DE-MADDPG) algorithm framework, enabling individual unmanned aerial vehicles to directly adjust the protocols and protocol parameters of the current node based solely on local network information. This optimization enhances network structure and overall performance without considering the transmission delay of control signals. The ns3-gym simulation platform is employed for performance evaluation, comparing the proposed algorithm with Deep Q-Network (DQN), Multi-Agent Deep Deterministic Policy Gradient (MADDPG), and other algorithms. The experimental results demonstrate that compared to the routing algorithm optimized based on DQN, the proposed algorithm exhibits the best overall performance, with a 26.31% reduction in energy consumption and a 19.69% decrease in end-to-end delay.

This study presents an improved weed detection model, ADL-YOLOv8, designed to enhance detection accuracy for small targets while achieving model lightweighting. It addresses the challenge of attaining both high accuracy and low memory usage in current intelligent weeding equipment. By overcoming this issue, the research not only reduces the hardware costs of automated impurity removal equipment but also enhances software recognition accuracy, contributing to reduced pesticide use and the promotion of sustainable agriculture. The ADL-YOLOv8 model incorporates a lighter AKConv network for better processing of specific features, an ultra-lightweight DySample upsampling module to improve accuracy and efficiency, and the LSKA-Attention mechanism for enhanced detection, particularly of small targets. On the same dataset, ADL-YOLOv8 demonstrated a 2.2% increase in precision, a 2.45% rise in recall, a 3.07% boost in mAP@0.5, and a 1.9% enhancement in mAP@0.95. The model’s size was cut by 15.77%, and its computational complexity was reduced by 10.98%. These findings indicate that ADL-YOLOv8 not only exceeds the original YOLOv8n model but also surpasses the newer YOLOv9t and YOLOv10n in overall performance. The improved algorithm model makes the hardware cost required for embedded terminals lower.

Mulberry leaves, when infected by pathogens, can suffer significant yield loss or even death if early disease detection and timely spraying are not performed. To enhance the detection performance of mulberry leaf diseases in natural environments and to precisely locate early small lesions, we propose a high-precision, high-efficiency disease detection algorithm named YOLOv8-RFMD. Based on improvements to You Only Look Once version 8 (YOLOv8), we first proposed the Multi-Dimension Feature Attention (MDFA) module, which integrates important features at the pixel-level, spatial, and channel dimensions. Building on this, we designed the RFMD Module, which consists of the Conv-BatchNomalization-SiLU (CBS) module, Receptive-Field Coordinated Attention (RFCA) Conv, and MDFA, replacing the Bottleneck in the model’s Residual block. We then employed the ADown down-sampling structure to reduce the model size and computational complexity. Finally, to improve the detection precision of small lesion features, we replaced the Complete Intersection over Union (CIOU) loss function with the Normalized Wasserstein Distance (NWD) loss function. Results show that the YOLOv8-RFMD model achieved a mAP50 of 94.3% and a mAP50:95 of 67.8% on experimental data, representing increases of 2.9% and 4.3%, respectively, compared to the original model. The model size was reduced by 0.53 MB to just 5.45 MB, and the GFLOPs were reduced by 0.3 to only 7.8. YOLOv8-RFMD has displayed great potential for application in real-world mulberry leaf disease detection systems and automatic spraying operations.

During the rice harvesting process, severe occlusion and adhesion exist among multiple targets, such as rice, straw, and leaves, making it difficult to accurately distinguish between rice grains and impurities. To address the current challenges, a lightweight semantic segmentation algorithm for impurities based on an improved SegFormer network is proposed. To make full use of the extracted features, the decoder was redesigned. First, the Feature Pyramid Network (FPN) was introduced to optimize the structure, selectively fusing the high-level semantic features and low-level texture features generated by the encoder. Secondly, a Part Large Kernel Attention (Part-LKA) module was designed and introduced after feature fusion to help the model focus on key regions, simplifying the model and accelerating computation. Finally, to compensate for the lack of spatial interaction capabilities, Bottleneck Recursive Gated Convolution (B-${\mathrm{g}}^n$Conv) was introduced to achieve effective segmentation of rice grains and impurities. Compared with the original model, the improved model’s pixel accuracy (PA) and F1 score increased by 1.6% and 3.1%, respectively. This provides a valuable algorithmic reference for designing a real-time impurity rate monitoring system for rice combine harvesters.