Real-time detection of tomato pests and diseases is essential for precision agriculture, as it requires high accuracy, speed, and energy efficiency of edge-computing agricultural robots. This study proposes AgriLiteNet (Lightweight Networks for Agriculture), a lightweight neural network integrating MobileNetV3 for local feature extraction and a streamlined Swin Transformer for global modeling. AgriLiteNet is further enhanced by a lightweight channel–spatial mixed attention module and a feature pyramid network, enabling the detection of nine tomato pests and diseases, including small targets like spider mites, dense targets like bacterial spot, and large targets like late blight. It achieves a mean average precision at an intersection-over-union threshold of 0.5 of 0.98735, which is comparable to Suppression Mask R-CNN (0.98955) and Cas-VSwin Transformer (0.98874), and exceeds the performance of YOLOv5n (0.98249) and GMC-MobileV3 (0.98143). With 2.0 million parameters and 0.608 GFLOPs, AgriLiteNet delivers an inference speed of 35 frames per second and power consumption of 15 watts on NVIDIA Jetson Orin NX, surpassing Suppression Mask R-CNN (8 FPS, 22 W) and Cas-VSwin Transformer (12 FPS, 20 W). The model’s efficiency and compact design make it highly suitable for deployment in agricultural robots, supporting sustainable farming through precise pest and disease management.

This paper proposes a design and optimization scheme for an Internet of Things (IoT)-based cloud platform aimed at enhancing the communication efficiency and operational performance of autonomous agricultural machinery. The platform integrates the dual communication capabilities of Narrowband Internet of Things (NB-IoT) and 5G, where NB-IoT is utilized for low-power, reliable data transmission from environmental sensors, such as soil information and weather monitoring, while 5G supports high-bandwidth, low-latency tasks like task scheduling and path tracking to effectively address the diverse communication requirements of modern complex agricultural scenarios. The cloud platform improves operational efficiency and resource utilization through real-time task scheduling, dynamic optimization, and seamless coordination between devices. To accommodate the diverse operational demands of agricultural environments, the system incorporates a real-time data feedback mechanism leveraging sensor data for path tracking and adjustment, enhancing adaptability and stability. Furthermore, a multi-machine collaborative scheduling strategy combining Dijkstra’s algorithm and an improved Harris hawk optimization (IHHO) algorithm, along with a multi-objective optimized path tracking method, is introduced to further improve scheduling efficiency and resource utilization while improving path tracking accuracy and smoothness and reducing external interferences, including environmental fluctuations and sensor inaccuracies. Experimental results demonstrate that the IoT-based cloud platform excels in data transmission reliability, path tracking accuracy, and resource optimization, validating its feasibility in smart agriculture and providing an efficient and scalable solution for large-scale agricultural operations.

The intricate and dynamic nature of agricultural supply chains imposes stringent demands on optimization methodologies, necessitating multi-objective considerations, privacy safeguards, and decision transparency to address pivotal challenges in ensuring food security and sustainable development. This study introduces a Dynamic Heterogeneous Multi-Objective Graph Neural Network (DHMO-GNN) model, meticulously tailored for optimizing agricultural supply chains. It integrates five core modules: data preprocessing and heterogeneous graph construction, dynamic graph neural networks, multi-objective optimization, interpretability enhancement, and federated learning collaboration. The model adeptly captures temporal dynamics through sequential attention mechanisms and incremental updates, harmonizes cost, delivery time, and carbon emissions via multi-task learning and Pareto optimization, augments decision transparency with GNNExplainer and SHAP, and surmounts data silos by leveraging federated learning alongside differential privacy. Empirical evaluations on the Chengdu Hongguang Town Farmers’ Market dataset demonstrate that the centralized DHMO-GNN variant achieves a hypervolume indicator (HV) of 0.849, surpassing baseline models; the federated variant exhibits only a 2.6% decline under privacy constraints, underscoring its robustness. Ablation studies further corroborate the synergistic contributions of each module. This research furnishes an efficacious and trustworthy framework for the intelligent management of agricultural supply chains, holding profound implications for advancing digital transformation and green development.

Microexpressions have extremely high due value in national security, public safety, medical, and other fields. However, microexpressions have characteristics that are obviously different from macroexpressions, such as short duration and weak changes, which greatly increase the difficulty of microexpression recognition work. In this paper, we propose a microexpression recognition method based on multimodal fusion through a comparative study of traditional microexpression recognition algorithms such as LBP algorithm and CNN and LSTM deep learning algorithms. The method couples the separate microexpression image information with the corresponding body temperature information to establish a multimodal fusion microexpression database. This paper firstly introduces how to build a multimodal fusion microexpression database in a laboratory environment, secondly compares the recognition accuracy of LBP, LSTM, and CNN + LSTM networks for microexpressions, and finally selects the superior CNN + LSTM network in the comparison results for model training and testing on the test set under separate microexpression database and multimodal fusion database. The experimental results show that a microexpression recognition method based on multimodal fusion designed in this paper is more accurate than unimodal recognition in multimodal recognition after feature fusion, and its recognition rate reaches 75.1%, which proves that the method is feasible and effective in improving microexpression recognition rate and has good practical value.

Precise curvilinear path tracking remains a persistent challenge for autonomous agricultural machines, where conventional Model Predictive Control (MPC) suffers from poor adaptability to varying curvatures and high computational overhead in unstructured farmland environments. This paper proposes a soft-constrained MPC framework enhanced by a two-layer fuzzy architecture and Recursive Least Squares filtering to address these limitations simultaneously. The first fuzzy layer dynamically adjusts the MPC prediction horizon in response to real-time path curvature, enabling proactive steering on complex curved trajectories. The second fuzzy layer tunes the state weighting matrix online based on lateral and heading deviations, improving transient tracking accuracy without increasing computational cost. Recursive Least Squares filtering is further integrated to suppress sensor noise and compensate for tire slip dynamics inherent to farmland operation. The proposed framework is validated using MATLAB simulations on both constant-curvature semicircular paths and variable-curvature S-curve trajectories at operational speeds of 2.0 and 2.5 m/s, followed by outdoor field trials on a scaled autonomous robot platform. Simulation results demonstrate average tracking error reductions of 52.7–55.9% on constant-curvature paths and 10.8–18.2% on variable-curvature paths compared to fixed-parameter soft-constrained MPC. Field experiments confirm practical viability, achieving an RMS lateral error of 0.131 m over a 50 m curved route on natural terrain. These results demonstrate that the hierarchical decomposition of adaptation objectives yields substantial accuracy gains while preserving real-time feasibility on resource-constrained embedded platforms.

Soil moisture prediction is essential for precision irrigation in water-limited agricultural systems. This study presents a deep learning-driven irrigation framework for winter wheat, integrating a novel Informer–UNet model with a Comprehensive Irrigation Index for adaptive water management. The Informer–UNet combines ProbSparse self-attention mechanisms with UNet’s multi-scale feature fusion, enabling simultaneous prediction of soil moisture at 27 monitoring points across three depths, 10, 30, and 50 cm, while quantifying prediction uncertainty through Monte Carlo Dropout. A Comprehensive Irrigation Index incorporating moisture deviation, spatial variance, and confidence interval width was developed, with weights optimized via genetic algorithm. Field experiments were conducted in Chengdu, China, over two winter wheat growing seasons. The Informer–UNet achieved superior prediction accuracy, R² greater than 0.98, RMSE less than 0.65, compared to LSTM, Transformer, and standard Informer models, with the fastest convergence and lowest validation loss. The proposed DeepIndexIrr strategy maintained soil moisture within the target range, 55% to 75%, for over 81% of the irrigation period, reducing water consumption by 38.2% compared to fixed-threshold control and 19.2% compared to expert manual scheduling. These results demonstrate that integrating spatially distributed deep learning predictions with uncertainty-informed decision rules offers a promising approach for sustainable precision irrigation.