Intelligent feeding devices constitute pivotal technological equipment driving the transformation of aquaculture toward intelligent and intensive operational paradigms. The fundamental mechanism centers on acquiring comprehensive environmental and aquaculture organism information through sophisticated multimodal perception systems, integrating advanced data modeling and decision-making algorithms for analytical processing, and implementing precision execution protocols to achieve highly efficient feeding operations. This study systematically reviews the development trajectory of intelligent feeding devices and establishes a robust system framework consisting of three interconnected architectural layers: the perception layer, decision-making layer, and execution layer. The perception layer focuses primarily on comprehensive monitoring of water quality parameters, meteorological conditions, and aquaculture organisms, while facilitating sophisticated multi-source information fusion processes. This layer incorporates advanced sensor networks including dissolved oxygen monitors, temperature sensors, pH meters, turbidity analyzers, underwater imaging systems, acoustic monitoring devices, and biometric measurement instruments. These integrated sensing technologies enable continuous real-time data acquisition regarding environmental fluctuations, fish behavioral patterns, growth performance indicators, and feeding response characteristics, providing essential foundational data for higher-level analytical processes. The decision-making layer operates as the cognitive nucleus of the system, utilizing sophisticated growth models and optimization algorithms to generate scientifically-based and individually-customized feeding strategies. It adopts advanced computational methodologies including machine learning algorithms, artificial neural networks, fuzzy logic controllers, genetic algorithms, and predictive modeling frameworks to analyze complex multi-dimensional datasets. The system processes information regarding species-specific nutritional requirements, growth kinetics, environmental conditions, feeding histories, and market demands to determine optimal feeding schedules, appropriate feed quantities, nutritional compositions, and distribution timing patterns. The execution layer implements physical operational capabilities through systematic improvements in device adaptability, sophisticated feed transportation mechanisms, and optimized distribution pattern designs, thereby enhancing overall system stability and broad applicability across diverse aquaculture environments. This layer encompasses precision mechanical engineering components, automated conveyor systems, programmable dispensing units, variable-speed distribution mechanisms, and intelligent control interfaces that ensure accurate feed delivery while maintaining consistent performance under varying operational conditions and environmental challenges. Current research reveals that existing technological implementations continue to confront significant challenges that constrain optimal performance and widespread adoption. These limitations include insufficient precision in multi-source data fusion processes, resulting in potential information loss and reduced analytical accuracy; limited generalization capabilities of existing computational models, restricting their adaptability across diverse aquaculture species, environmental conditions, and operational scales; and inadequate device reliability under harsh marine operational environments, leading to maintenance challenges, operational disruptions, and increased lifecycle costs. This study not only provides a comprehensive systematic framework for the design and optimization of intelligent feeding systems but also demonstrates substantial practical value in improving feed utilization efficiency, reducing operational costs, and promoting environmentally sustainable aquaculture practices. The proposed framework offers theoretical guidance for system integration, performance evaluation, and technological advancement, while addressing critical industry needs for enhanced productivity, cost-effectiveness, and environmental stewardship. Future research should prioritize strengthening interdisciplinary collaboration across marine biology, computer science, mechanical engineering, materials science, and environmental science. It should also emphasize rigorous validation of practical applications through extensive field testing and performance evaluation under real-world operational conditions. Research efforts must focus on: advancing algorithm robustness through sophisticated machine learning techniques; improving device performance through innovative engineering solutions; and developing standardized testing protocols for comprehensive system evaluation. These strategic initiatives will facilitate the large-scale commercial application of intelligent feeding technologies in aquaculture, accelerate the global aquaculture industry’s comprehensive intelligent transformation, and ultimately contribute to enhanced food security, environmental sustainability, and economic viability of modern aquaculture systems.

Agricultural yield forecast can often rely mainly on the single model approach at present. The single model is trained on all available features for prediction. However, the approach cannot capture the complex nonlinear relationships between meteorological factors and yield. Traditional decomposition can also be limited to the long-term trend fitting (e.g., moving average, high-pass filtering). In this study, an integrated model was developed to forecast the peanut yield in the southwestern Guangdong Province, China. Meteorological data (temperature, precipitation, sunshine duration, wind speed, and relative humidity) and yield records were collected from the 16 test regions from 2000 to 2023. Polynomial regression and stacking ensemble learning were integrated using three systematic procedures: 1) Long-term trend modeling with polynomial regression was used to quantify the impact of technological advancements and agricultural management on the yield; 2) Dimensionality reduction via principal component analysis (PCA) was employed to extract the 12 principal components with a cumulative variance contribution of 90% from normalized meteorological data; 3) The base learners were set as the stacked generalization framework with the K-nearest neighbors (KNN), random forest (RF), and gradient boosting regressors (GBR). While the Lasso regression was set as the meta-learner. The cross-validation was optimized for the meteorological yield analysis. The model showed excellent performance, with a mean absolute percentage error reduction of 0.22~0.68 percentage points compared to the combination of polynomial regression and a single machine learning method. Specifically, the polynomial regression with the stacked model was achieved in the lowest MAPE (2.09%), MAE (57.10 kg/hm²), and RMSE (78.55 kg/hm²), compared with the rest models, such as the KNN (MAPE: 2.70%), RF (MAPE: 2.31%), and GBR (MAPE: 2.77%). In addition, its R² value is as high as 0.96, indicated that the combined model can explain 96% of the variance in the actual production data, demonstrating its high accuracy of prediction, demonstrating its high accuracy of prediction. According to the pre-August meteorological inputs (two months pre-harvest), early forecast testing also maintained a high accuracy (R²=0.94), with a MAPE of 2.91% and MAE of 71.88 kg/hm². The high effectiveness of the improved was provided for the early yield forecast. A series of trials were carried out to validate the improved model for 2020－2023. The robustness of the model was further confirmed, with an average MAPE of 4.62% in the different regions. However, regional variations were also observed in the accuracy of the forecast. The MAPE ranged from 0.25% in Yunfu to 8.15% in Zhanjiang. There was also the strong influence of regional heterogeneity and non-meteorological factors, such as soil properties and farming practices. Different decomposition of the trend was also compared, including the moving average, exponential smoothing, high-pass filtering, and polynomial regression. Polynomial regression outperformed the rest. Among them, the long-term yield trends driven by technological advancements were accurately captured to smooth out the effects of extreme weather events. A more stable and accurate separation of trend and meteorological yield was obtained for the precise forecast. In conclusion, the polynomial regression was integrated for the trend analysis. A stacked ensemble model was suitable for the meteorological yield forecast. A robust and accurate framework was offered to forecast the peanut yield. Early forecasts were also provided for the regional variations during agricultural management and market strategy. Future research can further enhance to incorporate the additional variables, such as the soil properties and satellite data. The region-specific models can also be expected to consider the local agricultural practices and environmental conditions.

Path planning for field agricultural robots must satisfy several criteria: establishing feeding routes, maintaining gentle slopes, approaching multiple livestock observation points, ensuring timely environmental monitoring, and achieving high efficiency. The complex terrain of outdoor farming areas poses a challenge. Traditional A* algorithms, which generate only the shortest path, fail to meet these requirements and often produce paths that lack smoothness. Therefore, identifying the most suitable path, rather than merely the shortest one, is essential. This study introduced a path-planning algorithm tailored to field-based livestock farming environments, building upon the traditional A* algorithm. It constructed a digital elevation model, integrated an artificial potential field for evaluating multiple target points, calculated terrain slope, optimized the search neighborhood based on robot traversability, and employed Bézier curve segmentation for path optimization. This method segmented the path into multiple curves by evaluating the slopes of the lines connecting adjacent nodes, ensuring a smoother and more efficient route. The experimental results demonstrate its superiority to traditional A*, ensuring paths near multiple target points, significantly reducing the search space, and resulting in over 69.4% faster search speeds. Bézier curve segmentation delivers smoother paths conforming to robot trajectories.