Pub Date : 2025-11-28DOI: 10.1016/j.biosystemseng.2025.104344
Wantong Zhang, Fa Song, Jiyu Sun
The population of animal pollinators has been steadily declining due to human activities, disrupting the balance of agricultural ecosystems. In response, UAV pollination technology has emerged as a promising option to mitigate the shortage of natural pollination services. This paper reviews the background of the development of UAV pollination technology, summarises and evaluates current UAV pollination technology and the intelligent technologies involved therein, highlighting the current state and development limitations. Finally, based on this analysis and the prevailing conditions of crop pollination and agricultural environments, the integration of intelligent technologies is proposed, particularly UAV-based systems, to improve the efficiency of pollination.
{"title":"A review of strategic enhancement of pollination with smart agriculture to counteract the decline of natural pollinators","authors":"Wantong Zhang, Fa Song, Jiyu Sun","doi":"10.1016/j.biosystemseng.2025.104344","DOIUrl":"10.1016/j.biosystemseng.2025.104344","url":null,"abstract":"<div><div>The population of animal pollinators has been steadily declining due to human activities, disrupting the balance of agricultural ecosystems. In response, UAV pollination technology has emerged as a promising option to mitigate the shortage of natural pollination services. This paper reviews the background of the development of UAV pollination technology, summarises and evaluates current UAV pollination technology and the intelligent technologies involved therein, highlighting the current state and development limitations. Finally, based on this analysis and the prevailing conditions of crop pollination and agricultural environments, the integration of intelligent technologies is proposed, particularly UAV-based systems, to improve the efficiency of pollination.</div></div>","PeriodicalId":9173,"journal":{"name":"Biosystems Engineering","volume":"261 ","pages":"Article 104344"},"PeriodicalIF":5.3,"publicationDate":"2025-11-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145615449","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-26DOI: 10.1016/j.biosystemseng.2025.104342
Xuan Zhao, Anbin Zhang, Fei Liu, Hongbin Bai, Yuxing Ren, Wenxue Dong, Shuhan Yang
Directional mechanised seeding of sunflower seeds plays a critical role in enhancing crop yield and minimizing seed loss. However, limited understanding of the rotational behaviour of irregularly shaped seeds, such as sunflower seeds, under airflow conditions has constrained the precision of current seeding equipment. In this study, a mathematical model describing seed rotation under airflow was developed using parametric equations that represent the geometry of sunflower seeds. A custom-built experimental setup, together with computational fluid dynamics (CFD) simulations in ANSYS Fluent, was employed to analyse the rotational dynamics in detail. Single-factor experiments revealed that suction hole diameter, negative pressure intensity, seed centre offset, equivalent diameter, and initial posture angle all significantly affected the torque and rotational direction of the seed. An empirical model was further established through orthogonal experiments to quantitatively describe the seed's rotational response under airflow. Validation experiments with sunflower seeds demonstrated the model's high predictive accuracy within the following parameter ranges: suction hole diameter (13–33 mm), negative pressure (1–5 kPa), seed centre offset (5–13 mm), equivalent diameter (6.19–9.28 mm), and initial posture angle (0–288°). These findings provide a theoretical basis for the design of pneumatic directional seeding systems and a technical reference for the targeted seeding of other irregularly shaped seeds, contributing to the advancement of precision agriculture.
{"title":"Modelling and analysis of airflow-induced rotational behaviour of sunflower seeds for directional sowing","authors":"Xuan Zhao, Anbin Zhang, Fei Liu, Hongbin Bai, Yuxing Ren, Wenxue Dong, Shuhan Yang","doi":"10.1016/j.biosystemseng.2025.104342","DOIUrl":"10.1016/j.biosystemseng.2025.104342","url":null,"abstract":"<div><div>Directional mechanised seeding of sunflower seeds plays a critical role in enhancing crop yield and minimizing seed loss. However, limited understanding of the rotational behaviour of irregularly shaped seeds, such as sunflower seeds, under airflow conditions has constrained the precision of current seeding equipment. In this study, a mathematical model describing seed rotation under airflow was developed using parametric equations that represent the geometry of sunflower seeds. A custom-built experimental setup, together with computational fluid dynamics (CFD) simulations in ANSYS Fluent, was employed to analyse the rotational dynamics in detail. Single-factor experiments revealed that suction hole diameter, negative pressure intensity, seed centre offset, equivalent diameter, and initial posture angle all significantly affected the torque and rotational direction of the seed. An empirical model was further established through orthogonal experiments to quantitatively describe the seed's rotational response under airflow. Validation experiments with sunflower seeds demonstrated the model's high predictive accuracy within the following parameter ranges: suction hole diameter (13–33 mm), negative pressure (1–5 kPa), seed centre offset (5–13 mm), equivalent diameter (6.19–9.28 mm), and initial posture angle (0–288°). These findings provide a theoretical basis for the design of pneumatic directional seeding systems and a technical reference for the targeted seeding of other irregularly shaped seeds, contributing to the advancement of precision agriculture.</div></div>","PeriodicalId":9173,"journal":{"name":"Biosystems Engineering","volume":"261 ","pages":"Article 104342"},"PeriodicalIF":5.3,"publicationDate":"2025-11-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145615451","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-21DOI: 10.1016/j.biosystemseng.2025.104341
Rongrong Li , Hongwen Li , Jin He , Yingbo Wang , Caiyun Lu , Zhengyang Wu , Shan Jiang , Zongfu Yang
Based on the principle of dilute-phase pneumatic conveying, this study proposes an innovative design for a pneumatic straw suction device for conservation tillage to reduce pressure drop and improve conveying efficiency. A coupled simulation method combining the Discrete Element Method (DEM) with Computational Fluid Dynamics (CFD) was employed, and its accuracy was validated through bench tests. The distribution and variation of airflow, straw particle motion, and the interaction of these parameters were investigated to achieve low pressure drop and high efficiency. By analysing the sources of pressure drop, it was found that the primary factors affecting airflow and particle conveying were the inner diameter of the air duct or the upper surface of the suction chamber, the bending diameter ratio of the elbow, and the fan rotational speed. The response surface optimisation revealed that the best low-loss, high-efficiency performance was achieved when the diameter of the air duct, the bending diameter ratio, and the fan rotational speed were set to 200 mm, 1.54, and 2900 rpm, respectively. Under these conditions, the pressure drop down, pressure drop up, and the percentage of straw mass were 12.58 Pa, 17.12 Pa, and 3.15 %, respectively. This study provides new insights into the interaction between straw particles and airflow in pneumatic conveying systems.
{"title":"Airflow variations and particle conveying characteristics in pneumatic straw suction device based on CFD-DEM","authors":"Rongrong Li , Hongwen Li , Jin He , Yingbo Wang , Caiyun Lu , Zhengyang Wu , Shan Jiang , Zongfu Yang","doi":"10.1016/j.biosystemseng.2025.104341","DOIUrl":"10.1016/j.biosystemseng.2025.104341","url":null,"abstract":"<div><div>Based on the principle of dilute-phase pneumatic conveying, this study proposes an innovative design for a pneumatic straw suction device for conservation tillage to reduce pressure drop and improve conveying efficiency. A coupled simulation method combining the Discrete Element Method (DEM) with Computational Fluid Dynamics (CFD) was employed, and its accuracy was validated through bench tests. The distribution and variation of airflow, straw particle motion, and the interaction of these parameters were investigated to achieve low pressure drop and high efficiency. By analysing the sources of pressure drop, it was found that the primary factors affecting airflow and particle conveying were the inner diameter of the air duct or the upper surface of the suction chamber, the bending diameter ratio of the elbow, and the fan rotational speed. The response surface optimisation revealed that the best low-loss, high-efficiency performance was achieved when the diameter of the air duct, the bending diameter ratio, and the fan rotational speed were set to 200 mm, 1.54, and 2900 rpm, respectively. Under these conditions, the pressure drop down, pressure drop up, and the percentage of straw mass were 12.58 Pa, 17.12 Pa, and 3.15 %, respectively. This study provides new insights into the interaction between straw particles and airflow in pneumatic conveying systems.</div></div>","PeriodicalId":9173,"journal":{"name":"Biosystems Engineering","volume":"261 ","pages":"Article 104341"},"PeriodicalIF":5.3,"publicationDate":"2025-11-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145570001","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-21DOI: 10.1016/j.biosystemseng.2025.104329
Jia-Hao He , S.K. Mickelson , J.L. Hatfield , Mehari Z. Tekeste
Digitised soil tilth is a numerical index that quantifies soil's physical state after tillage operations using digital imaging or sensor technology. Limited research exists on extracting features from tilled soil to achieve soil tilth digitisation for Artificial Intelligence (AI)-driven smart tillage decision support for soil management and crop productivity. Visual soil images, infrared soil images, and soil temperature readings were collected on tilled soil field behind two tillage systems, mouldboard plough (MP) (sample size 432), and a disk ripper (DP) (sample size 432). Three feature extraction methods were employed to derive soil attributes. These extracted features were then applied to six machine learning models to assess their effectiveness in AI applications, aiming for AI-driven smart tillage decision system. The study also proposed implementing feature fusion, which combines features into 20-dimensional vectors from visual and infrared images, and soil temperature readings. This fusion approach creates a distinct separation between the two tillage systems in the feature space, enhancing AI applications by improving classification accuracy from 79 % to 99 %, which translates to a reliable decision-making system with 95 % reduction in misclassification errors compared to using random forest model with individual sensor features. The results indicate the initial success of the proposed fusion approach and extraction methods in AI applications, showing promise for further use in AI-driven smart tillage decision system. Besides the model-based classification method development, computation capability and selection of sensor availability were also assessed for accelerated implementation of the methodology to field digital tillage applications.
{"title":"Tillage-induced soil feature extraction and multi-sensors fusion for tillage system classification","authors":"Jia-Hao He , S.K. Mickelson , J.L. Hatfield , Mehari Z. Tekeste","doi":"10.1016/j.biosystemseng.2025.104329","DOIUrl":"10.1016/j.biosystemseng.2025.104329","url":null,"abstract":"<div><div>Digitised soil tilth is a numerical index that quantifies soil's physical state after tillage operations using digital imaging or sensor technology. Limited research exists on extracting features from tilled soil to achieve soil tilth digitisation for Artificial Intelligence (AI)-driven smart tillage decision support for soil management and crop productivity. Visual soil images, infrared soil images, and soil temperature readings were collected on tilled soil field behind two tillage systems, mouldboard plough (MP) (sample size 432), and a disk ripper (DP) (sample size 432). Three feature extraction methods were employed to derive soil attributes. These extracted features were then applied to six machine learning models to assess their effectiveness in AI applications, aiming for AI-driven smart tillage decision system. The study also proposed implementing feature fusion, which combines features into 20-dimensional vectors from visual and infrared images, and soil temperature readings. This fusion approach creates a distinct separation between the two tillage systems in the feature space, enhancing AI applications by improving classification accuracy from 79 % to 99 %, which translates to a reliable decision-making system with 95 % reduction in misclassification errors compared to using random forest model with individual sensor features. The results indicate the initial success of the proposed fusion approach and extraction methods in AI applications, showing promise for further use in AI-driven smart tillage decision system. Besides the model-based classification method development, computation capability and selection of sensor availability were also assessed for accelerated implementation of the methodology to field digital tillage applications.</div></div>","PeriodicalId":9173,"journal":{"name":"Biosystems Engineering","volume":"261 ","pages":"Article 104329"},"PeriodicalIF":5.3,"publicationDate":"2025-11-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145570002","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-19DOI: 10.1016/j.biosystemseng.2025.104340
Arian van Westreenen , Ningyi Zhang , Leo F.M. Marcelis , Elias Kaiser
The intensity of sunlight incident on leaves often fluctuates, affecting physiological processes and plant growth at various temporal and spatial scales. However, sunlight is often diffuse due to clouds and aerosols, and the extent to which the characteristics of fluctuating light intensity change under diffuse compared to direct light is not well quantified. Making use of a glass that converts ca. 45 % of incoming light into diffuse light in a commercial tomato greenhouse, light intensity above the crop was recorded at high frequency (10 Hz) for 4.5 months under both clear and diffuse glass. Dips in light intensity below an upper, calculated baseline intensity were marked as shadeflecks, and their daily number, duration, frequency, amplitude and light integral were recorded. Diffuse glass reduced the number of shadeflecks (44 day−1 vs. 112 day−1 under direct glass), and increased their average length (460 s vs. 250 s per shadefleck). Under both glass types, most shadeflecks were very short (<1 s), and were fewer and weaker in winter than in spring and summer. Short shadeflecks (0.1–0.4 s length) occurred 60–110 % more often under direct than under diffuse glass. It was concluded that glass which makes approximately half of all incoming light diffuse reduces the number of shadeflecks, and tends to increase their length as well as reduce their amplitude. However, despite these effects, fluctuations in light intensity are still surprisingly many and short under diffuse glass.
{"title":"Rapid irradiance fluctuations in a greenhouse: Effects of diffuse glass on shadeflecks","authors":"Arian van Westreenen , Ningyi Zhang , Leo F.M. Marcelis , Elias Kaiser","doi":"10.1016/j.biosystemseng.2025.104340","DOIUrl":"10.1016/j.biosystemseng.2025.104340","url":null,"abstract":"<div><div>The intensity of sunlight incident on leaves often fluctuates, affecting physiological processes and plant growth at various temporal and spatial scales. However, sunlight is often diffuse due to clouds and aerosols, and the extent to which the characteristics of fluctuating light intensity change under diffuse compared to direct light is not well quantified. Making use of a glass that converts ca. 45 % of incoming light into diffuse light in a commercial tomato greenhouse, light intensity above the crop was recorded at high frequency (10 Hz) for 4.5 months under both clear and diffuse glass. Dips in light intensity below an upper, calculated baseline intensity were marked as shadeflecks, and their daily number, duration, frequency, amplitude and light integral were recorded. Diffuse glass reduced the number of shadeflecks (44 day<sup>−1</sup> vs. 112 day<sup>−1</sup> under direct glass), and increased their average length (460 s vs. 250 s per shadefleck). Under both glass types, most shadeflecks were very short (<1 s), and were fewer and weaker in winter than in spring and summer. Short shadeflecks (0.1–0.4 s length) occurred 60–110 % more often under direct than under diffuse glass. It was concluded that glass which makes approximately half of all incoming light diffuse reduces the number of shadeflecks, and tends to increase their length as well as reduce their amplitude. However, despite these effects, fluctuations in light intensity are still surprisingly many and short under diffuse glass.</div></div>","PeriodicalId":9173,"journal":{"name":"Biosystems Engineering","volume":"261 ","pages":"Article 104340"},"PeriodicalIF":5.3,"publicationDate":"2025-11-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145570005","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-19DOI: 10.1016/j.biosystemseng.2025.104339
Jun Wei, Zhan Zhao, En Lu, Sisi Liu, Xinyu Hu, Qianqian Zhou, Cheng Xu
Due to variations in the soil environment and path curvature, it remains challenging for agricultural machinery to accurately track complex trajectories in field operations. In this paper, an adaptive backstepping tracking control approach for differential drive agricultural vehicles is presented. The extended Kalman filter (EKF) is used to estimate the longitudinal slip rate, and a dynamic model of the vehicle under longitudinal slipping conditions is established. Using the integral of time and square tracking error (ITSE) as the performance evaluation index, the dynamic characteristics of trajectory tracking under different backstepping control law coefficients are analysed. Then, with the desired velocity, trajectory curvature, lateral error, longitudinal error, and heading angle error as inputs, an optimisation method for the control law coefficients based on the adaptive-network-based fuzzy inference system (ANFIS) is proposed. Finally, the trajectory tracking simulations and practical experiments were performed employing a differential-drive vehicle. The results indicated that the accuracy and stability of trajectory tracking can be significantly improved by incorporating the proposed slip compensation and adaptive adjustment of the control law coefficients, particularly when the desired trajectory curvature was discontinuous or changed sharply.
{"title":"Adaptive backstepping tracking control for differential drive vehicles under longitudinal slipping conditions","authors":"Jun Wei, Zhan Zhao, En Lu, Sisi Liu, Xinyu Hu, Qianqian Zhou, Cheng Xu","doi":"10.1016/j.biosystemseng.2025.104339","DOIUrl":"10.1016/j.biosystemseng.2025.104339","url":null,"abstract":"<div><div>Due to variations in the soil environment and path curvature, it remains challenging for agricultural machinery to accurately track complex trajectories in field operations. In this paper, an adaptive backstepping tracking control approach for differential drive agricultural vehicles is presented. The extended Kalman filter (EKF) is used to estimate the longitudinal slip rate, and a dynamic model of the vehicle under longitudinal slipping conditions is established. Using the integral of time and square tracking error (ITSE) as the performance evaluation index, the dynamic characteristics of trajectory tracking under different backstepping control law coefficients are analysed. Then, with the desired velocity, trajectory curvature, lateral error, longitudinal error, and heading angle error as inputs, an optimisation method for the control law coefficients based on the adaptive-network-based fuzzy inference system (ANFIS) is proposed. Finally, the trajectory tracking simulations and practical experiments were performed employing a differential-drive vehicle. The results indicated that the accuracy and stability of trajectory tracking can be significantly improved by incorporating the proposed slip compensation and adaptive adjustment of the control law coefficients, particularly when the desired trajectory curvature was discontinuous or changed sharply.</div></div>","PeriodicalId":9173,"journal":{"name":"Biosystems Engineering","volume":"261 ","pages":"Article 104339"},"PeriodicalIF":5.3,"publicationDate":"2025-11-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145570004","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-18DOI: 10.1016/j.biosystemseng.2025.104328
Lang Qiao , Jiahao Fan , Jose G. Franco , Alison J. Duff , Emily J. Diaz-Vallejo , Tong Yu , Zhou Zhang
Alfalfa is an important high-quality livestock feed around the world, and timely and accurate yield prediction is crucial for precision harvest management. Hyperspectral remote sensing (RS) is an efficient method for non-destructive alfalfa yield prediction. However, the high cost and the relatively low spatial resolution remain the main obstacles to its widespread adoption. Therefore, this study aims to propose a hyperspectral reconstruction method based on UAV RGB imagery to reduce the data acquisition cost and improve the performance of alfalfa yield prediction. Firstly, three features selection methods including competitive adaptive reweighted sampling (CARS), variable importance in subsets selection algorithm (VISSA), and recursive feature elimination (RFE) are evaluated for their potential in selecting important bands from hyperspectral data for alfalfa yield. Secondly, the combination of the CARS and Multi-stage Spectral-wise Transformer (MST++) is used to reconstruct the important hyperspectral band images from RGB images for alfalfa yield. Finally, the reconstructed hyperspectral features and RGB spatial features are integrated to enhance the model accuracy. The experiments conducted in the Prairie du Sac farm in 2021 and 2022 showed that the hyperspectral features reconstructed using the proposed method exhibited strong consistency with the original features and achieved similar accuracy in predicting alfalfa yield (R2 = 0.717, RMSE = 476 kg ha−1, MAE = 376 kg ha−1). Also, combining the reconstructed hyperspectral features with the RGB spatial features could further improve the performance of yield prediction (R2 = 0.745, RMSE = 452 kg ha−1, MAE = 348 kg ha−1). Furthermore, the generalisation of the proposed method was validated using an independent alfalfa dataset from Arlington farm in 2023.
苜蓿是世界范围内重要的优质家畜饲料，及时准确的产量预测对精准收获管理至关重要。高光谱遥感（RS）是一种有效的无损预测紫花苜蓿产量的方法。然而，高成本和相对较低的空间分辨率仍然是其广泛采用的主要障碍。因此，本研究旨在提出一种基于无人机RGB图像的高光谱重建方法，以降低数据采集成本，提高紫花苜蓿产量预测的性能。首先，对竞争自适应重加权采样（CARS）、可变重要度子集选择算法（VISSA）和递归特征消除（RFE）三种特征选择方法在高光谱数据中选择重要波段的潜力进行了评价。其次，将CARS与多级光谱变换（Multi-stage Spectral-wise Transformer, mst++）相结合，从RGB图像中重构出苜蓿产量的重要高光谱波段图像；最后，将重建的高光谱特征与RGB空间特征相结合，提高模型精度。在2021年和2022年在Prairie du Sac农场进行的实验表明，使用该方法重建的高光谱特征与原始特征具有较强的一致性，并且在预测苜蓿产量方面具有相似的精度（R2 = 0.717, RMSE = 476 kg ha - 1, MAE = 376 kg ha - 1）。此外，将重建的高光谱特征与RGB空间特征相结合可以进一步提高产量预测的性能（R2 = 0.745, RMSE = 452 kg ha - 1, MAE = 348 kg ha - 1）。此外，使用2023年阿灵顿农场的独立苜蓿数据集验证了所提出方法的泛化。
{"title":"Hyperspectral reconstruction based on low-cost UAV RGB imagery for alfalfa yield prediction","authors":"Lang Qiao , Jiahao Fan , Jose G. Franco , Alison J. Duff , Emily J. Diaz-Vallejo , Tong Yu , Zhou Zhang","doi":"10.1016/j.biosystemseng.2025.104328","DOIUrl":"10.1016/j.biosystemseng.2025.104328","url":null,"abstract":"<div><div>Alfalfa is an important high-quality livestock feed around the world, and timely and accurate yield prediction is crucial for precision harvest management. Hyperspectral remote sensing (RS) is an efficient method for non-destructive alfalfa yield prediction. However, the high cost and the relatively low spatial resolution remain the main obstacles to its widespread adoption. Therefore, this study aims to propose a hyperspectral reconstruction method based on UAV RGB imagery to reduce the data acquisition cost and improve the performance of alfalfa yield prediction. Firstly, three features selection methods including competitive adaptive reweighted sampling (CARS), variable importance in subsets selection algorithm (VISSA), and recursive feature elimination (RFE) are evaluated for their potential in selecting important bands from hyperspectral data for alfalfa yield. Secondly, the combination of the CARS and Multi-stage Spectral-wise Transformer (MST++) is used to reconstruct the important hyperspectral band images from RGB images for alfalfa yield. Finally, the reconstructed hyperspectral features and RGB spatial features are integrated to enhance the model accuracy. The experiments conducted in the Prairie du Sac farm in 2021 and 2022 showed that the hyperspectral features reconstructed using the proposed method exhibited strong consistency with the original features and achieved similar accuracy in predicting alfalfa yield (<em>R</em><sup><em>2</em></sup> = 0.717, RMSE = 476 kg ha<sup>−1</sup>, MAE = 376 kg ha<sup>−1</sup>). Also, combining the reconstructed hyperspectral features with the RGB spatial features could further improve the performance of yield prediction (<em>R</em><sup><em>2</em></sup> = 0.745, RMSE = 452 kg ha<sup>−1</sup>, MAE = 348 kg ha<sup>−1</sup>). Furthermore, the generalisation of the proposed method was validated using an independent alfalfa dataset from Arlington farm in 2023.</div></div>","PeriodicalId":9173,"journal":{"name":"Biosystems Engineering","volume":"261 ","pages":"Article 104328"},"PeriodicalIF":5.3,"publicationDate":"2025-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145570003","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-15DOI: 10.1016/j.biosystemseng.2025.104322
Yanan Zhang , Dongqing Wang , Yuefeng Du , Changkai Wen , Linze Wang , Zhikang Wu
Demand for high-power power-shift tractors (exceeding 100 hp) continues to grow annually in modern agriculture. The accuracy of wet clutch pressure control significantly affects shift quality and overall tractor performance. However, direct measurement of piston displacement in a wet clutch is not feasible, making it difficult to apply many contemporary control strategies for effective clutch pressure management. To address this sensorless control challenge, a control framework based on physical and digital twins is proposed. A physical twin of the clutch is constructed, and a virtual clutch developed using data from the physical twin. From these twins, a mechanistic model was established, and a novel clutch pressure controller designed. A complete clutch twinning system was built, and experimental tests conducted to validate the proposed method. Compared to proportional-integral-derivative (PID) control algorithm, the approach reduced jerk and slipping friction work during tractor start-up by 46.6 % and 1.1 % respectively, and exhibited robust performance across different temperature conditions. This approach offers a promising reference for sensorless clutch pressure control in high-power tractors.
{"title":"Sensorless wet clutch pressure control method for high-power tractors using physical and digital twins","authors":"Yanan Zhang , Dongqing Wang , Yuefeng Du , Changkai Wen , Linze Wang , Zhikang Wu","doi":"10.1016/j.biosystemseng.2025.104322","DOIUrl":"10.1016/j.biosystemseng.2025.104322","url":null,"abstract":"<div><div>Demand for high-power power-shift tractors (exceeding 100 hp) continues to grow annually in modern agriculture. The accuracy of wet clutch pressure control significantly affects shift quality and overall tractor performance. However, direct measurement of piston displacement in a wet clutch is not feasible, making it difficult to apply many contemporary control strategies for effective clutch pressure management. To address this sensorless control challenge, a control framework based on physical and digital twins is proposed. A physical twin of the clutch is constructed, and a virtual clutch developed using data from the physical twin. From these twins, a mechanistic model was established, and a novel clutch pressure controller designed. A complete clutch twinning system was built, and experimental tests conducted to validate the proposed method. Compared to proportional-integral-derivative (PID) control algorithm, the approach reduced jerk and slipping friction work during tractor start-up by 46.6 % and 1.1 % respectively, and exhibited robust performance across different temperature conditions. This approach offers a promising reference for sensorless clutch pressure control in high-power tractors.</div></div>","PeriodicalId":9173,"journal":{"name":"Biosystems Engineering","volume":"261 ","pages":"Article 104322"},"PeriodicalIF":5.3,"publicationDate":"2025-11-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145518619","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-15DOI: 10.1016/j.biosystemseng.2025.104326
Uddhav Bhattarai , Rajkishan Arikapudi , Steven A. Fennimore , Frank N. Martin , Stavros G. Vougioukas
Manual fruit harvesting is common in agriculture, but the amount of time pickers spend on non-productive activities can make it very inefficient. Accurately identifying picking vs. non-picking activity is crucial for estimating picker efficiency and optimising labour management and harvest processes. In this study, a practical system was developed to calculate the efficiency of pickers in commercial strawberry harvesting. Instrumented picking carts (iCarritos) were developed to record the harvested fruit weight, geolocation, and iCarrito movement in real time. The iCarritos were deployed during the commercial strawberry harvest season in Santa Maria, CA. The collected data was then used to train a CNN-LSTM-based deep neural network to classify a picker’s activity into “Pick” and “NoPick” classes. Experimental evaluations showed that the CNN-LSTM model showed promising activity recognition performance with an F1 score of 0.97. The recognition results were then used to compute picker efficiency and the time required to fill a tray. Analysis of the season-long harvest data showed that the average picker efficiency was 75.09% with an estimation accuracy of 97.23%. Furthermore, the average tray fill time was 6.85 min with an estimation accuracy of 96.78%. When integrated into commercial harvesting, the proposed technology can aid growers in monitoring automated worker activity and optimising harvests to reduce non-productive time and enhance overall harvest efficiency.
{"title":"Data-driven worker activity recognition and efficiency estimation in manual fruit harvesting","authors":"Uddhav Bhattarai , Rajkishan Arikapudi , Steven A. Fennimore , Frank N. Martin , Stavros G. Vougioukas","doi":"10.1016/j.biosystemseng.2025.104326","DOIUrl":"10.1016/j.biosystemseng.2025.104326","url":null,"abstract":"<div><div>Manual fruit harvesting is common in agriculture, but the amount of time pickers spend on non-productive activities can make it very inefficient. Accurately identifying picking vs. non-picking activity is crucial for estimating picker efficiency and optimising labour management and harvest processes. In this study, a practical system was developed to calculate the efficiency of pickers in commercial strawberry harvesting. Instrumented picking carts (iCarritos) were developed to record the harvested fruit weight, geolocation, and iCarrito movement in real time. The iCarritos were deployed during the commercial strawberry harvest season in Santa Maria, CA. The collected data was then used to train a CNN-LSTM-based deep neural network to classify a picker’s activity into “Pick” and “NoPick” classes. Experimental evaluations showed that the CNN-LSTM model showed promising activity recognition performance with an F1 score of 0.97. The recognition results were then used to compute picker efficiency and the time required to fill a tray. Analysis of the season-long harvest data showed that the average picker efficiency was 75.09% with an estimation accuracy of 97.23%. Furthermore, the average tray fill time was 6.85 min with an estimation accuracy of 96.78%. When integrated into commercial harvesting, the proposed technology can aid growers in monitoring automated worker activity and optimising harvests to reduce non-productive time and enhance overall harvest efficiency.</div></div>","PeriodicalId":9173,"journal":{"name":"Biosystems Engineering","volume":"261 ","pages":"Article 104326"},"PeriodicalIF":5.3,"publicationDate":"2025-11-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145518620","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The rugged topography of hilly and mountainous regions presents significant challenges for conventional chassis systems, limiting agricultural mechanization and productivity. Here, a novel three-degree-of-freedom (3-DOF) agricultural chassis with a passive adaptive suspension is proposed that integrates adaptive all-wheel attachment and vibration damping to maintain excellent traction and smooth movement on uneven terrain. Based on the Lagrange method, a suspension vibration model incorporating both adaptive and vertical damping was developed to analyse the system's response to ground excitation. Subsequently, a chassis dynamics model accounting for coupled pitch-roll vibrations was established, and its effectiveness was verified through bump road experiments. Furthermore, a genetic algorithm was employed for multi-objective optimisation of the suspension system, yielding optimal parameters. Experimental validation confirmed the significant vibration reduction performance of the novel passive adaptive suspension, with reductions of 28.5 % in vertical acceleration, 14.2 % in pitch angular velocity, and 17.3 % in roll angular velocity. The developed dynamic model served as a valuable theoretical reference for vibration control and performance analysis. The proposed chassis demonstrated potential for diverse agricultural operations in hilly and mountainous terrain, including seeding, spraying, harvesting, and transportation.
{"title":"Vibration characteristics of a terrain-adaptive agricultural chassis for hilly and mountainous terrain","authors":"Xiaoliang Zhang, Yujie Huang, Peixiang Wang, Longjin Liang, Yiheng Cheng, Pingyi Liu, Liang Sun","doi":"10.1016/j.biosystemseng.2025.104338","DOIUrl":"10.1016/j.biosystemseng.2025.104338","url":null,"abstract":"<div><div>The rugged topography of hilly and mountainous regions presents significant challenges for conventional chassis systems, limiting agricultural mechanization and productivity. Here, a novel three-degree-of-freedom (3-DOF) agricultural chassis with a passive adaptive suspension is proposed that integrates adaptive all-wheel attachment and vibration damping to maintain excellent traction and smooth movement on uneven terrain. Based on the Lagrange method, a suspension vibration model incorporating both adaptive and vertical damping was developed to analyse the system's response to ground excitation. Subsequently, a chassis dynamics model accounting for coupled pitch-roll vibrations was established, and its effectiveness was verified through bump road experiments. Furthermore, a genetic algorithm was employed for multi-objective optimisation of the suspension system, yielding optimal parameters. Experimental validation confirmed the significant vibration reduction performance of the novel passive adaptive suspension, with reductions of 28.5 % in vertical acceleration, 14.2 % in pitch angular velocity, and 17.3 % in roll angular velocity. The developed dynamic model served as a valuable theoretical reference for vibration control and performance analysis. The proposed chassis demonstrated potential for diverse agricultural operations in hilly and mountainous terrain, including seeding, spraying, harvesting, and transportation.</div></div>","PeriodicalId":9173,"journal":{"name":"Biosystems Engineering","volume":"261 ","pages":"Article 104338"},"PeriodicalIF":5.3,"publicationDate":"2025-11-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145518617","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号