Livestock feeding control method, system and equipment based on multi-source information decision

Abstract

The invention provides a livestock feeding control method, system and equipment based on multi-source information decision. The method comprises the following steps: acquiring multi-source information corresponding to target livestock; the multi-source information comprises historical physiological information, current physiological information, current behavior information and current environment information; according to the historical physiological information, predicting a first feeding scheme of the target livestock in the current feeding period, and according to the current physiological information and the current environment information, predicting a second feeding scheme of the target livestock in the current feeding period; acquiring a real-time health state of the target livestock according to the current physiological information and the current behavior information; adjusting the second feeding scheme according to the real-time health state and the first feeding scheme to obtain a target feeding scheme, and controlling a feeding pump associated with the target livestock to execute feeding operation according to the target feeding scheme. According to the invention, the feeding scheme is comprehensively decided through the multi-modal information, so that the individuation, precision and intelligence of livestock feeding are improved.

Description

Livestock feeding control methods, systems, and equipment based on multi-source information decision-making Technical Field

[0001] This invention relates to the field of livestock management technology, and in particular to a livestock feeding control method, system and equipment based on multi-source information decision-making. Background Technology

[0002] As the livestock industry transforms towards large-scale and intelligent operations, higher demands are placed on the precision of nutrition supply and the timeliness of health management in the raising of young livestock (such as calves) in order to improve feed conversion rates and reduce the risk of disease.

[0003] Currently, the feeding control of young livestock mainly adopts a feeding method based on human experience. The specific implementation plan is usually as follows: the feeder determines the feeding amount based on the age or current weight of the young livestock by referring to a fixed feeding standard table; or, the feeder simply observes the growth status of the young livestock through manual inspections and conducts extensive feeding management and disease screening based on experience.

[0004] However, this method is prone to nutritional imbalance or feed waste, and it is difficult to track individual differences in young animals. It is also impossible to make targeted interventions and adjustments to the feeding programs for different young animals, resulting in poor accuracy in feeding control, which in turn affects the growth performance and survival rate of young animals. Summary of the Invention

[0005] This invention provides a livestock feeding control method, system, and device based on multi-source information decision-making, which solves the shortcomings of existing technologies that rely on human experience for livestock feeding control, resulting in poor accuracy of feeding control, and achieves improved precision and intelligence in livestock feeding.

[0006] This invention provides a livestock feeding control method based on multi-source information decision-making, comprising: acquiring multi-source information corresponding to a target livestock; the multi-source information including historical physiological information, current physiological information, current behavioral information, and current environmental information; the historical physiological information being physiological information collected in each historical feeding cycle, and the current physiological information, current behavioral information, and current environmental information being physiological information, behavioral information, and environmental information collected in the current feeding cycle, respectively; predicting a first feeding plan for the target livestock in the current feeding cycle based on the historical physiological information, and predicting a second feeding plan for the target livestock in the current feeding cycle based on the current physiological information and current environmental information; acquiring the real-time health status of the target livestock based on the current physiological information and current behavioral information; adjusting the second feeding plan based on the real-time health status and the first feeding plan to obtain a target feeding plan; and controlling a feeding pump associated with the target livestock to perform feeding operations according to the target feeding plan.

[0007] According to the present invention, a livestock feeding control method based on multi-source information decision-making includes controlling a feeding pump associated with the target livestock to perform feeding operations according to the target feeding scheme. The method comprises: calculating the actual flow rate error between the real-time output flow rate of the feeding pump and the target flow rate corresponding to the target feeding scheme, and the actual level error between the real-time liquid level of the feeding container associated with the target livestock and the target liquid level corresponding to the target feeding scheme; inputting the actual flow rate error to a first PID controller to obtain a flow rate control component; inputting the actual liquid level error to a second PID controller to obtain a liquid level control component; weighting and synthesizing the flow rate control component and the liquid level control component according to the target value of the target weight coefficient, and generating a target feeding signal based on the weighted synthesis result; determining the feeding speed of the feeding pump according to the correlation between the target feeding signal and the motor speed of the feeding pump; and controlling the feeding pump to perform feeding operations according to the feeding speed.

[0008] According to the present invention, a livestock feeding control method based on multi-source information decision-making includes the following steps for determining the target value of the target weight coefficient: adjusting the flow rate control parameters of the first PID controller; simulating the feeding operation of the feeding pump according to the adjusted flow rate control parameters until the response index value of the simulated output flow rate of the feeding pump meets the rapidity threshold; and determining the optimal flow rate control parameter from the last adjusted flow rate control parameters; adjusting the liquid level control parameters of the second PID controller; simulating the feeding operation of the feeding pump according to the adjusted liquid level control parameters until the change index of the simulated liquid level of the feeding container meets the stability threshold; and determining the optimal liquid level control parameter from the last adjusted liquid level control parameters; generating test control commands corresponding to different preset values ​​of the target weight coefficient based on the optimal flow rate control parameters and the optimal liquid level control parameters; simulating the feeding operation of the feeding pump according to each test control command; obtaining the comprehensive control error generated by the simulated operation of the feeding pump under each test control command; and determining the preset value corresponding to the minimum comprehensive control error as the target value of the target weight coefficient.

[0009] According to the present invention, a livestock feeding control method based on multi-source information decision-making, wherein generating a target feeding signal based on a weighted synthesis result includes: acquiring real-time temperature data of the feeding material in the feeding container and the real-time sucking frequency of the target livestock; determining a flow rate feedforward compensation amount corresponding to the real-time temperature data based on the correlation between the real-time temperature data and the flow rate; superimposing the flow rate feedforward compensation amount with the weighted synthesis result to generate a feedback control signal; generating a fluctuation signal synchronized with the real-time sucking frequency, and superimposing the fluctuation signal with the feedback control signal to obtain the target feeding signal.

[0010] According to the present invention, a livestock feeding control method based on multi-source information decision-making, wherein predicting a first feeding plan for a target livestock in the current feeding cycle based on the historical physiological information includes: acquiring historical growth data of the target livestock in each of the historical feeding cycles from the historical physiological information; fitting a maximum weight parameter, a growth rate parameter, and a regulation coefficient suitable for the target livestock based on each of the historical growth data; constructing a growth model based on the maximum weight parameter, the growth rate parameter, and the regulation coefficient, and predicting a growth curve of the target livestock based on the growth model; calculating predicted growth data of the target livestock in the current feeding cycle based on the growth curve; and inputting the predicted growth data into a fitted regression model to obtain the first feeding plan; wherein the fitted regression model is constructed based on fitting each of the historical growth data and the historical feeding data of the target livestock in each of the historical feeding cycles.

[0011] According to the present invention, a livestock feeding control method based on multi-source information decision-making includes predicting a second feeding plan for the target livestock in the current feeding cycle based on the current physiological information and the current environmental information. The method comprises: acquiring multiple sets of training samples; each set of training samples includes historical physiological information of each sample livestock, historical environmental information of each sample livestock, and historical feeding data of each sample livestock; performing multiple linear regression fitting on the multiple sets of training samples using the least squares method to obtain a proportionality coefficient matrix corresponding to the demand model, and constructing the demand model based on the proportionality coefficient matrix; extracting data from the current physiological information and the current environmental information to obtain a first data set; the first data set includes breed data, age data, weight data, body temperature data, sex data, body length data, body size data, temperature data, and humidity data; and inputting the first data set into the demand model to obtain the second feeding plan.

[0012] According to the present invention, a livestock feeding control method based on multi-source information decision-making, the step of obtaining the real-time health status of the target livestock based on the current physiological information and the current behavioral information includes: extracting data from the current physiological information and the current behavioral information to obtain a second data set; the second data set includes body length data, body size data, posture data, respiratory rate data, lung murmur data, carbon dioxide concentration data, and excrement morphology data; inputting the second data set into a health decision-making model to obtain the real-time health status output by the health decision-making model through multi-node classification decision-making on the second data set; wherein, the health decision-making model is trained based on multiple sets of historical physiological data and historical behavioral data with health status labels corresponding to each sample livestock.

[0013] According to the present invention, a livestock feeding control method based on multi-source information decision-making, wherein adjusting a second feeding plan according to the real-time health status and a first feeding plan to obtain a target feeding plan includes: when the real-time health status is healthy, modifying the recommended feeding amount in the second feeding plan based on the recommended feeding amount in the first feeding plan and the feed intake rate of the target livestock in the previous feeding cycle to obtain the target feeding plan; when the real-time health status is abnormal, obtaining a drug addition instruction and a feeding amount adjustment coefficient corresponding to the real-time health status according to the abnormality type of the real-time health status, calculating the attenuation of the recommended feeding amount in the second feeding plan according to the feeding amount adjustment coefficient and the feed intake rate, and generating the target feeding plan according to the calculation result and the drug addition instruction.

[0014] This invention also provides a livestock feeding control system based on multi-source information decision-making, comprising: a data acquisition unit for acquiring multi-source information corresponding to a target livestock; the multi-source information includes historical physiological information, current physiological information, current behavioral information, and current environmental information; the historical physiological information is physiological information collected in each historical feeding cycle, and the current physiological information, current behavioral information, and current environmental information are physiological information, behavioral information, and environmental information collected in the current feeding cycle, respectively; a prediction unit for predicting a first feeding plan for the target livestock in the current feeding cycle based on the historical physiological information, and predicting a second feeding plan for the target livestock in the current feeding cycle based on the current physiological information and the current environmental information; a decision-making unit for acquiring the real-time health status of the target livestock based on the current physiological information and the current behavioral information; and a control unit for adjusting the second feeding plan based on the real-time health status and the first feeding plan to obtain a target feeding plan, and controlling a feeding pump associated with the target livestock to perform feeding operations according to the target feeding plan.

[0015] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the livestock feeding control method based on multi-source information decision-making as described above.

[0016] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the livestock feeding control method based on multi-source information decision-making as described above.

[0017] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the livestock feeding control method based on multi-source information decision-making as described above.

[0018] The livestock feeding control method, system, and equipment based on multi-source information decision-making provided by this invention achieve comprehensive real-time perception of individual information, environmental information, and health status of young livestock by integrating multi-source sensors. By combining historical growth trends and current nutritional needs for dual prediction and incorporating real-time health status for dynamic adjustment, this method effectively solves the problems of nutritional imbalance caused by traditional feeding and high mortality rates of young livestock due to delayed disease detection. It can dynamically customize feeding programs for different individual differences and environmental changes, effectively improving the precision and intelligence of livestock feeding, helping to increase daily weight gain, promote rumen development, significantly reduce disease risk, and save on breeding management costs and improve management efficiency. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 is one of the flowcharts of the livestock feeding control method based on multi-source information decision-making provided by the present invention.

[0021] Figure 2 is a second schematic flowchart of the livestock feeding control method based on multi-source information decision-making provided by the present invention.

[0022] Figure 3 is one of the structural schematic diagrams of the livestock feeding control system based on multi-source information decision-making provided by the present invention.

[0023] Figure 4 is a schematic diagram of the functional mapping relationship of each module in the livestock feeding control system provided by the present invention.

[0024] Figure 5 is one of the flowcharts of the feeding control steps based on the dual-path parallel PID control architecture provided by the present invention.

[0025] Figure 6 is a second schematic flowchart of the feeding control steps based on the dual-path parallel PID control architecture provided by the present invention.

[0026] Figure 7 is a flowchart illustrating the method for determining the first feeding program provided by the present invention.

[0027] Figure 8 is a flowchart illustrating the second feeding program determination method provided by the present invention.

[0028] Figure 9 is a flowchart of the health status decision-making provided by the present invention.

[0029] Figure 10 is a second schematic diagram of the livestock feeding control system based on multi-source information decision-making provided by the present invention.

[0030] Figure 11 is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0032] In recent years, the livestock industry has gradually shifted from a traditional, small-scale, free-range, and primarily manual farming model to an intelligent and precise integrated farming model. This has led to increasingly higher demands on intensive young animal husbandry in terms of nutrition, immunization, and management efficiency. Simultaneously, the industry's requirements for the comprehensiveness and precision of young animal feeding techniques are also gradually increasing. Currently, most livestock farming relies on traditional young animal feeding techniques, using extensive management based on human experience or single parameters (such as age and weight). This traditional management approach has many drawbacks: firstly, it is difficult to track individual differences in young animals, leading to low feed utilization, nutritional imbalances, and limited growth performance; secondly, the lack of effective real-time monitoring methods results in delayed identification of early disease signs such as diarrhea and lung murmurs, leading to insufficient health risk management. Consequently, the accuracy of feeding control is poor, affecting the growth performance and survival rate of young animals.

[0033] To achieve precise feeding and timely early disease management, comprehensive decision-making through multi-source information fusion is essential. Therefore, there is an urgent need to develop a comprehensive decision-making method that integrates young animal health monitoring, growth environment perception, and machine learning models to realize precise and intelligent feeding of young animals.

[0034] Figure 1 is one of the flowcharts of the livestock feeding control method based on multi-source information decision-making provided by the present invention. Figure 2 is another flowchart of the livestock feeding control method based on multi-source information decision-making provided by the present invention.

[0035] To address the key technical challenges of low feed utilization, poor nutritional suitability, and insufficient health risk management in existing technologies, as shown in Figure 1, this application provides a livestock feeding control method based on multi-source information decision-making. This method primarily focuses on refined, intelligent, and differentiated feeding control for young livestock. The core objective of this method is to achieve intelligent and precise feeding technology in intensive young livestock farming. By saving breeding costs, increasing daily weight gain in young livestock, promoting rumen development, and reducing disease risk, it comprehensively controls feeding to improve breeding efficiency.

[0036] As shown in Figure 2, specifically, this application differs from the traditional extensive management method that relies on a single parameter or static model, and constructs a closed-loop process that integrates multi-source information collection of young livestock, intelligent analysis and decision-making of young livestock data, and precise feeding execution of young livestock. The technical concept is as follows: First, multi-sensor real-time collection of cross-modal data includes physiological information (such as age, weight, and body temperature), behavioral information (whether the animal frequently lies down, has diarrhea, or has lung murmurs), and environmental information (such as temperature and humidity). Then, a growth curve prediction algorithm is used to analyze the multi-source data to predict a first feeding plan that meets the animal's growth needs. A regression prediction model is then used to analyze the multi-source data to predict a second feeding plan that meets the animal's personalized needs at the current instant. A decision tree model is used to intelligently analyze the multi-source data to obtain the animal's health status. Based on the health status and the first feeding plan, the feed ratio and amount in the second feeding plan are dynamically adjusted. Finally, data from flow and level sensors during the feeding process are used to perform dual-path error synthesis proportional-integral-derivative (PID) control on the feeding pump, accurately outputting the predicted feed amount, thereby achieving personalized, precise, and intelligent feeding.

[0037] This method is applied to a livestock feeding control system based on multi-source information decision-making (hereinafter referred to as the feeding control system or system). The system integrates young animal health monitoring, growth environment perception and execution mechanisms, which constitute the hardware foundation of the method in this application.

[0038] Figure 3 is one of the structural schematic diagrams of the livestock feeding control system based on multi-source information decision-making provided by the present invention. As shown in Figure 3, the system mainly consists of three parts: a multi-source information acquisition module for young livestock, a smart data analysis and decision-making module for young livestock (integrated in the control cabinet), and a precision feeding execution module for young livestock. The hardware of each part and its specific functions are as follows: Figure 4 is a schematic diagram of the functional mapping relationship of each module in the livestock feeding control system provided by the present invention. As shown in Figures 3 and 4, the multi-source information acquisition module for young livestock is responsible for sensing the physiological, behavioral, and environmental information of the target livestock (such as calves) in real time, providing multi-source data support for subsequent decision-making. Specifically, it includes, but is not limited to, a radio frequency identification (RFID) reader, a weighing sensor, an infrared depth (RedGreen Blue-Depth, RGB-D) binocular depth camera, a temperature sensor, a sound sensor, a carbon dioxide sensor, and a temperature and humidity sensor.

[0039] The system includes RFID readers installed near feeding points to read electronic ear tags worn by livestock, obtaining physiological information such as breed, date of birth, age, and sex. Weighing sensors are installed at the bottom of the livestock's standing area, such as under the feeding pen floor, to collect real-time weight data. This allows the system to calculate the daily weight gain of young animals based on continuously collected real-time weight data. By comparing the daily weight gain with a standard growth curve, problems such as slow growth and developmental delays can be detected promptly, providing a direct basis for adjusting nutritional intake. An RGB-D binocular depth camera, mounted above the feeding pen, is used to acquire images and depth information of the target livestock. Its functions include monitoring behavioral information (such as frequent lying down and standing posture), acquiring body size data (body length, body size), and assisting in health monitoring (identifying the color and shape of excrement to determine if diarrhea is present). Body size data helps determine slow growth, and image recognition of fecal characteristics accurately locates diarrhea symptoms. Combined with historical data, it helps analyze the cause of diarrhea (such as indigestion or viral infection), thereby guiding adjustments to the feeding formula (such as adding antidiarrheal medication or electrolytes). A temperature sensor is used to collect the body temperature information of the target livestock. Real-time temperature monitoring directly identifies whether young animals have elevated body temperature (fever), a fundamental physiological indicator for determining infection or stress response. Carbon dioxide sensors monitor the concentration of exhaled carbon dioxide in young livestock, serving as an auxiliary basis for assessing respiratory metabolism and health status. Based on periodic changes in concentration, they indirectly monitor respiratory rate and simultaneously monitor the air quality of the respiratory environment. Sound sensors collect sound signals (such as coughs and breathing sounds) emitted by the target livestock. Audio analysis determines respiratory rate and the presence of lung murmurs, aiding in early pneumonia warnings. The combination of data from both carbon dioxide and sound sensors provides a comprehensive assessment of the respiratory system status of young livestock, effectively identifying respiratory diseases. Temperature and humidity sensors monitor the temperature and humidity of the rearing environment in real time. This monitored data is used not only for feed quantity prediction (heat and cold stress) but also, combined with sound sensor data, for a comprehensive assessment of respiratory environmental quality. For example, high humidity and high temperature environments easily lead to bacterial growth; the system can then provide environmental improvement suggestions to prevent disease.

[0040] The personalized precision feeding execution module for young livestock is responsible for executing specific feeding operations based on the decision results and providing feedback on the status data during the execution process. This module includes, but is not limited to, feeding pumps, milk tanks and medicine containers, feeding buckets (also known as feeding containers), flow sensors, and liquid level sensors.

[0041] The feeding pump, as the core actuator, is controlled by the control cabinet and is responsible for transporting feed materials (such as milk) from the storage tank to the feeding end. The milk tank and medicine container store regular liquid feed (such as milk) and therapeutic / health medications, respectively. The feeding bucket is the feeding container located in front of the target livestock. A flow sensor, installed on the output pipe of the feeding pump, is used to measure the actual output flow rate and velocity of the feeding pump in real time, providing feedback data for PID flow rate control. A level sensor, installed inside or beside the feeding bucket, is used to measure the real-time liquid level in the feeding bucket. Its functions include: monitoring whether the target liquid level has been reached, providing feedback data for PID flow rate control, calculating the feed intake (feed intake rate) of young livestock, and monitoring the remaining feed. By analyzing prolonged feeding time or increased remaining feed, the system can determine that the young livestock have a decreased appetite, thus providing an early warning of declining health.

[0042] The intelligent data analysis and decision-making module (or control cabinet) for young livestock integrates a processor, memory, and related circuits. Its main functions include data aggregation (i.e., receiving multi-source information collected by all the aforementioned sensors), comprehensive decision-making (i.e., generating a target feeding plan), and closed-loop control (i.e., using a dual-path error synthesis PID algorithm based on feedback data from flow and level sensors to precisely adjust the motor speed of the feeding pump, ensuring that the actual feeding amount matches the target feeding amount).

[0043] Based on the above system architecture, this application realizes a complete intelligent feeding process from perception to decision-making to execution through multi-sensor collaboration and closed-loop control. The following will elaborate on the method flow of this application based on this system. As shown in Figure 1, the method includes: Step 110, acquiring multi-source information corresponding to the target livestock; the multi-source information includes historical physiological information, current physiological information, current behavioral information, and current environmental information; the historical physiological information is physiological information collected in each historical feeding cycle, and the current physiological information, current behavioral information, and current environmental information are respectively physiological information, behavioral information, and environmental information collected in the current feeding cycle.

[0044] In this embodiment, the target livestock refers to young livestock individuals that require meticulous feeding management, also known as young animals, such as calves and lambs. For ease of description, the following description uses calves as an example, but it is not limited to this.

[0045] Multi-source information refers to cross-modal data collected using heterogeneous sensors, aiming to comprehensively perceive the individual state of livestock and the state of the environment. This multi-source information includes at least historical physiological information, current physiological information, current behavioral information, and current environmental information collected for the target livestock.

[0046] The historical physiological information refers to the physiological data collected and stored in the database during each historical feeding cycle, used to reflect the long-term growth trend of the target livestock. This information is obtained by: identifying the electronic ear tag worn by the target livestock using an RFID reader to obtain its identification (ID), and retrieving its historical physiological information from the background database based on this ID. This physiological information includes, but is not limited to, breed data, age data, weight data, body temperature data, sex data, body length data, and body size data.

[0047] Current physiological information, current behavioral information, and current environmental information refer to data collected in real time during the current feeding cycle (i.e., this feeding cycle), which are used to reflect the instantaneous state and external conditions of the target livestock.

[0048] For obtaining current physiological information, the target livestock's current breed, age, and sex data can be obtained by identifying it with an RFID reader; the target livestock's current weight data can be measured in real time by a weighing sensor; the target livestock's current body temperature data can be collected by a temperature sensor; and the target livestock's body length and body size data can be calculated from the image data collected by an RGB-D binocular depth camera.

[0049] Current behavioral information includes, but is not limited to, posture data, excrement characteristics data, respiratory rate data, lung murmur data, and carbon dioxide concentration data. To acquire current behavioral information, the target livestock can be identified by using an RGB-D binocular depth camera to obtain its posture data (such as whether it lies down for a long time or is unsteady on its feet) and excrement characteristics data (such as color and shape). On-site audio is collected by a sound sensor and analyzed to obtain respiratory rate data and lung murmur data. Carbon dioxide concentration data of the target livestock's exhaled carbon dioxide is obtained by a carbon dioxide sensor.

[0050] Current environmental information includes, but is not limited to, temperature and humidity data. To obtain current environmental information, temperature and humidity data of the current feeding environment can be collected by temperature and humidity sensors.

[0051] Step 120: Based on the historical physiological information, predict the first feeding plan for the target livestock in the current feeding cycle; based on the current physiological information and the current environmental information, predict the second feeding plan for the target livestock in the current feeding cycle.

[0052] Optionally, after obtaining multi-source information corresponding to the target livestock, historical physiological information can be used for fitting and prediction to calculate the predicted weight that the target livestock should reach under the current feeding cycle, based on its genetic potential and historical developmental trajectory, as well as the theoretical nutrient supply required to maintain this growth rate. This theoretical nutrient supply constitutes the first feeding plan.

[0053] In addition, by using current physiological and environmental information, a multivariate regression analysis model is used to predict and output the personalized demand that matches the current instantaneous state, which constitutes the second feeding plan.

[0054] Step 130: Obtain the real-time health status of the target livestock based on the current physiological information and the current behavioral information.

[0055] Optionally, the real-time health status here serves as the basis for deciding whether to proceed with normal feeding or disease intervention. This embodiment achieves early identification of disease symptoms by integrating current physiological and behavioral information.

[0056] Specifically, multiple judgments are made based on the current physiological and behavioral information obtained in step 110 to determine whether the target livestock is currently in a healthy state or in an abnormal state (i.e., a disease state). The health status identification here can be obtained through multiple judgments using an artificial intelligence model (such as a neural network model or a decision number model) constructed based on historical physiological information, historical behavioral information, and corresponding historical actual health status; alternatively, it can be obtained by judging the current physiological and behavioral information against different health status thresholds one by one. This embodiment does not specifically limit this approach. The real-time health status here includes, but is not limited to, healthy and abnormal states. Abnormal states can be further subdivided into developmental abnormalities, digestive abnormalities, behavioral abnormalities, respiratory abnormalities, and other state categories.

[0057] Step 140: Based on the real-time health status and the first feeding plan, adjust the second feeding plan to obtain a target feeding plan. Based on the target feeding plan, control the feeding pump associated with the target livestock to perform feeding operations.

[0058] Optionally, after obtaining the real-time health status, the first feeding plan, and the second feeding plan, the second feeding plan can be adjusted using the real-time health status and the first feeding plan to ultimately generate the target feeding plan. In this process, different adjustment logics can be applied to the second feeding plan based on different real-time health statuses. For example, when the animal is in a healthy state, it is determined that the target animal can digest and absorb nutrients normally. The second feeding plan, reflecting the current environmental needs, is then used as the primary approach, and modified with reference to the first feeding plan, which reflects long-term growth trends (e.g., using a weighted average or limiting range method). This generates a target feeding plan that includes the recommended feed amount, thereby ensuring that the feed amount adapts to environmental changes without deviating from the growth trajectory. When an abnormal state is present (such as developmental abnormality, digestive abnormality, behavioral abnormality, or respiratory abnormality), the corresponding treatment strategy and feeding amount adjustment coefficient are queried according to the type of abnormality. The recommended amount in the second feeding plan is adjusted based on the feeding amount adjustment coefficient. The feeding formula in the second feeding plan is also adjusted in combination with the drug addition information in the treatment strategy to obtain the target feeding plan. This embodiment does not specifically limit this.

[0059] After obtaining the target feeding plan, the target feeding plan can be converted into a hardware control signal so as to send a motor drive signal to the feeding pump according to the target flow rate and target liquid level determined in the target feeding plan.

[0060] It should be noted that in this control, the feeding pump can be driven directly to perform the feeding operation based on the target flow rate and target liquid level. In addition, in order to ensure that the amount of feed delivered to the feeding hopper is accurately consistent with the target feeding plan, the real-time flow rate fed back by the flow sensor and the real-time liquid level fed back by the liquid level sensor can also be received during the control process. A closed-loop control algorithm is used to dynamically adjust the speed of the feeding pump to ensure that the amount of feed delivered to the feeding hopper is accurately consistent with the target feeding plan.

[0061] The method provided in this embodiment integrates multi-source sensors to achieve comprehensive real-time perception of individual information, environmental information, and health status of young livestock. By combining historical growth trends and current nutritional needs for dual prediction and incorporating real-time health status for dynamic adjustment, this method effectively solves the problems of nutritional imbalance caused by traditional feeding and high mortality rates of young livestock due to delayed disease detection. It can dynamically customize feeding programs for different individual differences and environmental changes, effectively improving the precision and intelligence of livestock feeding, helping to increase daily weight gain, promote rumen development, significantly reduce disease risk, and save on breeding management costs and improve management efficiency.

[0062] In some embodiments, step 140 specifically includes: step 141, calculating the actual flow rate error between the real-time output flow rate of the feeding pump and the target flow rate corresponding to the target feeding scheme, and the actual liquid level error between the real-time liquid level of the feeding container associated with the target livestock and the target liquid level corresponding to the target feeding scheme.

[0063] Step 142: Input the actual flow rate error to the first PID controller to obtain the flow rate control component; input the actual liquid level error to the second PID controller to obtain the liquid level control component; Step 143: Based on the target value of the target weight coefficient, perform weighted synthesis on the flow rate control component and the liquid level control component, and generate a target feeding signal based on the weighted synthesis result; Step 144: Based on the correlation between the target feeding signal and the motor speed of the feeding pump, determine the feeding speed of the feeding pump, and control the feeding pump to perform feeding operations based on the feeding speed.

[0064] It should be noted that during livestock feeding, the actual feeding amount often differs significantly from the target feeding amount, resulting in material waste or slow livestock growth and development. Dynamic adjustments cannot be made, failing to meet the needs of precise livestock feeding. Considering these issues, this embodiment designs a dynamic feedback feeding control mechanism for young livestock based on a PID control algorithm for a feeding pump and real-time monitoring by multiple sensors. Specifically, the actual feeding amount of young livestock is monitored in real-time by level and flow sensors. Combined with the PID control algorithm, real-time feedback and dynamic adjustment of the feeding pump's feeding speed are achieved. This allows for the synergy of multiple sensors and the PID control algorithm for the young livestock feeding pump, accurately outputting the target feeding amount, reducing milk waste, and saving feeding costs.

[0065] Figure 5 is one of the flowcharts of the feeding control steps based on the dual-path parallel PID control architecture provided by the present invention; Figure 6 is another flowchart of the feeding control steps based on the dual-path parallel PID control architecture provided by the present invention.

[0066] As shown in Figures 5 and 6, during the feeding process, the fluid output of the feeding pump and the liquid level of the feeding container can be monitored in real time by using a flow sensor and a liquid level sensor, respectively, so as to obtain the real-time output flow rate of the feeding pump and the real-time liquid level of the feeding container.

[0067] Then, the real-time output flow rate of the feeding pump was calculated. Target flow rate corresponding to the target feeding program The difference between them yields the actual flow velocity error, i.e. This error reflects whether the current delivery rate of the feed pump is up to standard.

[0068] In addition, the real-time liquid level of the feeding container (such as a feeding bucket) is calculated. Target liquid level corresponding to the target feeding program The difference between them yields the actual liquid level error. ,Right now This error reflects the progress of the total feeding amount.

[0069] Subsequently, to simultaneously ensure the stability of the flow rate and the accuracy of the total volume, a dual-path parallel PID control architecture can be used to calculate the flow rate control component and the liquid level control component. The first PID controller (also known as the flow rate PID controller) is primarily responsible for the system's rapid response. It receives the actual flow rate error and outputs the flow rate control component through proportional (P), integral (I), and derivative (D) calculations. Its function is to suppress flow rate fluctuations and ensure the stability of feed output.

[0070] The second PID controller (also known as the liquid level PID controller) is mainly responsible for the steady-state accuracy of the system. It receives the actual liquid level error and, through PID calculations, outputs the liquid level control component. Its function is to eliminate accumulated errors and ensure that the liquid level height when feeding is finally stopped is accurately in line with the target value.

[0071] Furthermore, after obtaining the flow rate control component and the liquid level control component, in order to solve the overshoot or response lag problems that may be caused by a single control dimension, a target value of the target weight coefficient is introduced. The two control components are linearly weighted and fused based on this coefficient to calculate the synthesized control signal. The specific calculation formula is as follows: Among them, the target value of the target weight coefficient It is a value between 0 and 1, used to balance the weight of flow rate control and level control in the overall decision-making. The target value of the target weight coefficient here can be set manually based on experience, or it can be obtained through testing, such as by combining single-variable tests that shield flow rate error and test only level error, single-variable tests that shield level error and test only flow rate error, and bivariate tests that combine flow rate error and level error, etc. This embodiment does not specifically limit it in this way.

[0072] Furthermore, based on the weighted synthesis result This can be used to generate a target feeding signal for driving the actuator. In the process of generating the target feeding signal, the weighted synthesis result can be used directly as the target feeding signal; or other parameters that affect feeding performance, such as the temperature of the feed material and / or the real-time sucking frequency of the target livestock, can be superimposed on the weighted synthesis result. This embodiment does not specifically limit this.

[0073] Furthermore, since feed pumps are typically driven by motors, their delivery speed has a specific functional relationship with the motor speed (such as a linear relationship). The system pre-stores this correlation; therefore, after acquiring the target feed signal, the corresponding motor speed can be calculated in reverse based on the target feed signal, thereby adjusting the feed pump's feeding speed.

[0074] Through this real-time closed-loop regulation, when the flow sensor detects a decrease in flow rate, the first PID controller will quickly increase the flow rate control component to increase the pump speed; when the liquid level approaches the target value, the second PID controller will gradually decrease the liquid level control component to reduce the pump speed to prevent overflow or overfeeding; when feeding is finished, the motor speed is controlled to 0.

[0075] In addition, at the end of each feeding phase, the amount of feed consumed and the amount of feed remaining in the target livestock can be monitored in real time by a liquid level sensor to obtain the feeding efficiency of the target livestock, which facilitates the dynamic adjustment of the amount of feed for the next feeding of the target livestock.

[0076] The method provided in this embodiment effectively solves the limitations of single-variable control in dynamic feeding scenarios through a dual-variable fusion PID control strategy. It effectively avoids overfeeding or underfeeding, reduces resource waste, and improves feeding efficiency. It ensures rapid response in the feeding process and achieves precise control of the total amount of feed, minimizing feed waste and enhancing the intelligence level of the feeding system.

[0077] Based on the above embodiments, this embodiment provides a detailed explanation of the specific process for determining the target value of the target weight coefficient involved in step 143. The selection of the target weight coefficient is directly related to the dynamic performance and steady-state accuracy of the PID control system. In order to obtain the optimal control effect, this embodiment adopts an offline or online parameter determination strategy of single-variable calibration and bivariate optimization.

[0078] In some embodiments, the step of determining the target value of the target weight coefficient includes: step 143-1, adjusting the flow rate control parameters of the first PID controller, simulating the control of the feeding pump to perform feeding operations according to the adjusted flow rate control parameters, until the response index value of the simulated output flow rate of the feeding pump meets the speed threshold, and determining the optimal flow rate control parameters from the last adjusted flow rate control parameters.

[0079] This step is the single-variable calibration process for flow rate. At the start of calibration, the system temporarily sets the target weight coefficient to 0 (or a value close to 0). At this time, the control system is mainly dominated by the first PID controller (flow rate control loop) to shield or minimize the influence of liquid level error.

[0080] In this state, the system controls the flow rate parameters of the first PID controller, including the proportional coefficient. Integral coefficient and differential coefficients The process involves iterative adjustments. After each parameter adjustment, the feeding pump is simulated to perform feeding operations through a simulation environment or a physical no-load test, and the step response curve of the simulated output flow rate of the feeding pump is monitored. Response index values, such as rise time and overshoot, are calculated. If the response index values ​​do not meet the preset speed threshold (e.g., rise time less than 0.5 seconds and overshoot less than 5%), the parameters are adjusted further. The parameters after the last adjustment are determined as the optimal flow rate control parameters when the conditions are met. This process ensures that the flow rate control loop has sufficient fast tracking capability.

[0081] Step 143-2: Adjust the liquid level control parameters of the second PID controller. Based on the adjusted liquid level control parameters, simulate the feeding pump to perform feeding operations until the simulated liquid level change index of the feeding container meets the stability threshold. The optimal liquid level control parameters are then determined from the last adjusted liquid level control parameters.

[0082] Specifically, this step is the single-variable calibration process for liquid level. The system temporarily sets the target weight coefficient to 1 (or a value close to 1). At this time, the control system is mainly dominated by the second PID controller (liquid level control loop) to shield or minimize the influence of flow rate error.

[0083] In this state, the system's level control parameters for the second PID controller include the proportional coefficient. Integral coefficient and differential coefficients The process involves iterative adjustments. The feeding pump is simulated to perform the feeding operation, with a focus on monitoring the changes in the simulated liquid level within the feeding container. Indicators of this change, such as steady-state error and the number of oscillations, are calculated. If these indicators do not meet a preset stability threshold (e.g., a steady-state error of less than 10 ml with no significant oscillations), the parameters are continuously adjusted. Once the conditions are met, the parameters from the last adjustment are determined as the optimal liquid level control parameters. This process ensures that the liquid level control loop possesses sufficient accuracy and stability.

[0084] Step 143-3: Based on the optimal flow rate control parameters and the optimal liquid level control parameters, generate test control commands corresponding to different preset values ​​of the target weight coefficient; simulate and control the feeding pump to perform feeding operations according to each test control command; obtain the comprehensive control error generated by the simulated operation of the feeding pump under each test control command; and determine the preset value corresponding to the minimum comprehensive control error as the target value of the target weight coefficient.

[0085] Optionally, after determining the optimal intrinsic parameters of the two controllers, the system enters the bivariate fusion optimization stage. At this point, the optimal flow rate control parameter and the optimal liquid level control parameter are locked unchanged. The system sets a series of preset values ​​(e.g., 0.1, 0.2, ..., 0.9) within the range of the target weight coefficient (e.g., 0 to 1). For each preset value, a corresponding test control command is generated to prepare for a full-process simulated feeding test.

[0086] During the test, the system sequentially applies the weighting coefficients corresponding to each preset value to simulate the joint control of simultaneously input flow rate error and liquid level error.

[0087] During each simulation run, the system records the control performance throughout the entire process and calculates the overall control error. This overall control error can be a weighted integral index that comprehensively considers the dynamic error caused by flow velocity fluctuations and the static error caused by the final liquid level deviation.

[0088] Then, the comprehensive control error under all preset values ​​is compared, and the preset value corresponding to the test with the smallest value is selected as the target value of the final target weight coefficient.

[0089] The method provided in this embodiment scientifically determines the key coupling parameters in the dual PID control architecture by first calibrating independently and then optimizing by fusion. This enables the system to perfectly balance the contradictory requirements of fast flow rate response and accurate liquid level control in actual operation, further improving the robustness and accuracy of feeding control.

[0090] In some embodiments, step 143 specifically includes: acquiring real-time temperature data of the feeding material in the feeding container and the real-time sucking frequency of the target livestock; determining a flow rate feedforward compensation amount corresponding to the real-time temperature data based on the correlation between the real-time temperature data and the flow rate; superimposing the flow rate feedforward compensation amount with the weighted synthesis result to generate a feedback control signal; generating a fluctuation signal synchronized with the real-time sucking frequency, and superimposing the fluctuation signal with the feedback control signal to obtain the target feeding signal.

[0091] Optionally, before generating the final control signal, the system first acquires real-time temperature data of the feeding material (such as milk) using a temperature and humidity sensor or a dedicated liquid temperature probe. Since the viscosity of fluids changes significantly with temperature (thickening at low temperatures and thinning at high temperatures), this directly affects pumping efficiency.

[0092] Meanwhile, the system analyzes sucking sounds through a sound sensor or monitors the instantaneous rate of liquid level drop through a liquid level sensor to calculate the real-time sucking frequency of the target livestock.

[0093] Because the system pre-stores the correlation between the temperature and flow rate of the feed material, after acquiring real-time temperature data, this correlation can be queried to determine the additional compensation value required to maintain the target flow rate at that temperature, i.e., the flow rate feedforward compensation. For example, when the temperature is low, the fluid viscosity is high, requiring an increase in compensation to improve the pump speed.

[0094] The calculated flow rate feedforward compensation is directly superimposed on the weighted synthesis result of the dual PID controller output, thus forming a feedforward control path. This path can offset temperature interference in advance before the PID feedback regulation takes effect, significantly improving the system's anti-interference capability and generating a temperature-corrected feedback control signal.

[0095] Furthermore, in order to improve the feeding comfort and digestive efficiency of young animals, a wave signal, such as a pulse wave signal or a sine wave signal, can be generated based on the real-time sucking frequency.

[0096] Finally, the fluctuation signal is superimposed on the feedback control signal to form the final target feeding signal.

[0097] In this way, the signal driving the feeding pump is no longer a constant DC level, but a pulsating signal with biomimetic rhythm. This makes the fluid output by the feeding pump exhibit a pulsating flow that matches the sucking rhythm of the young animals, ensuring both the accuracy of the total amount and improving biocompatibility, which helps to promote the sucking reflex and digestive juice secretion of the young animals.

[0098] The method provided in this embodiment solves the control error caused by changes in fluid physical properties through temperature feedforward compensation, and improves the bio-friendlyness of the feeding process through biomimetic pulse superposition, demonstrating a deep integration of precise control and bio-adaptation.

[0099] This embodiment, based on the aforementioned embodiments, provides a detailed explanation of the specific implementation process of step 120, which involves predicting the first feeding plan for the target livestock in the current feeding cycle based on the historical physiological information. This process aims to tap into the growth potential of young livestock and formulate a benchmark feeding plan that conforms to their natural developmental patterns.

[0100] Figure 7 is a flowchart illustrating the method for determining the first feeding program provided by the present invention; as shown in Figure 7, in some embodiments, step 120 specifically includes: step 121, obtaining the historical growth data of the target livestock in each of the historical feeding cycles from the historical physiological information.

[0101] The historical growth data here refers to the sequence of weight changes of the target livestock recorded continuously by weighing sensors over a period of time (e.g., within 15-30 days after birth). Specifically, it can be extracted from a database using electronic ear tags on the target livestock.

[0102] Step 122: Based on the historical growth data, fit the maximum weight parameter, growth rate parameter and adjustment coefficient that are suitable for the target livestock.

[0103] In this embodiment, the Gompertz growth model can be used to describe the growth pattern of young animals. The mathematical expression of this model is: ;in, For the first Weight data for each feeding cycle; Maximum weight parameter For adjustment coefficients, This is a growth rate parameter.

[0104] Specifically, historical growth data can be used to fit the Gompertz growth model using nonlinear least squares and / or the Levenberg-Marquardt algorithm to calculate the maximum body weight, growth rate, and adjustment coefficients suitable for the specific target livestock.

[0105] Step 123: Construct a growth model based on the maximum body weight parameter, the growth rate parameter, and the adjustment coefficient, and predict the growth curve of the target livestock based on the growth model.

[0106] Optionally, since growth curve prediction is crucial in precision feeding systems for young livestock, it effectively helps predict future weight changes and adjusts feeding strategies accordingly, thereby improving feeding efficiency. Growth curves primarily describe the pattern of weight gain in young livestock over a certain period. Combined with analysis of historical feeding data, they can predict feed requirements for the next 3-7 days, enabling precise adjustments. Therefore, after obtaining the maximum suitable weight parameters, growth rate parameters, and adjustment coefficients for the target livestock, a Gompertz growth model can be constructed using these parameters, and a growth curve can be plotted or generated accordingly. This curve reflects the future weight gain trend of the young livestock under ideal feeding conditions.

[0107] Step 124: Calculate the predicted growth data of the target livestock in the current feeding cycle based on the growth curve.

[0108] By substituting the current feeding cycle's corresponding time point (such as the current age) into the constructed growth model, the theoretical weight of the target livestock at that time point is calculated, which is the predicted growth data. This algorithm can usually be used to predict weight changes in the next 3-7 days.

[0109] Step 125: Input the predicted growth data into the fitting regression model to obtain the first feeding scheme; wherein, the fitting regression model is constructed based on the historical growth data and the historical feeding data of the target livestock in each historical feeding cycle.

[0110] In addition, it is necessary to obtain fitting parameters by fitting the historical growth data of the target livestock in each historical feeding cycle collected by the weighing sensor and the historical feeding data of the target livestock in each historical feeding cycle collected by the liquid level sensor, so as to build a fitting regression model based on the fitting parameters and establish the mapping relationship between body weight and feed intake.

[0111] For example, the established fitted regression model is ;in, For the first Weight data for each feeding cycle, For the first Feeding data for each feeding cycle and These are the fitting parameters.

[0112] The acquired predicted growth data (predicted weight) By inputting the fitted regression model, the corresponding first feeding plan can be calculated, which includes the predicted feed amount based on growth potential. Feeding program.

[0113] This embodiment utilizes the Gompertz model combined with historical data regression to achieve a scientific prediction of the future growth needs of young animals, ensuring that the feeding program can support their optimal growth rate.

[0114] Based on the above embodiments, this embodiment provides a detailed description of the specific implementation process of step 120, which involves predicting the second feeding plan for the target livestock in the current feeding cycle based on the current physiological information and the current environmental information. This process aims to develop a refined, on-demand feeding plan in response to environmental changes and current physical differences, thereby achieving precise on-demand feeding.

[0115] Figure 8 is a flowchart illustrating the method for determining the second feeding plan provided by the present invention. As shown in Figure 8, in some embodiments, step 120 includes: acquiring multiple sets of training samples; each set of training samples includes historical physiological information of each sample animal, historical environmental information of each sample animal, and historical feeding data of each sample animal; using the least squares method to perform multiple linear regression fitting on the multiple sets of training samples to obtain the proportional coefficient matrix corresponding to the demand model, and constructing the demand model based on the proportional coefficient matrix; extracting data from the current physiological information and the current environmental information to obtain a first data set; the first data set includes breed data, age data, weight data, body temperature data, sex data, body length data, body size data, as well as temperature data and humidity data; inputting the first data set into the demand model to obtain the second feeding plan.

[0116] Optionally, livestock nutritional requirements are influenced by multiple factors, such as breed data, age data, weight data, sex data, environmental temperature and humidity data, body length data, and body size data. To provide precise nutritional support, it is necessary to establish personalized feed amount prediction models based on these factors, thereby accurately calculating the feed amount for each livestock at different growth stages.

[0117] Specifically, firstly, historical data of N sample livestock are collected using multi-source sensors, including but not limited to feeding data (such as milk consumption) for each sample livestock during each historical feeding cycle. Variety data Age data Weight data W, gender data Body temperature data Ambient temperature data Ambient humidity data (RH), body length data Body size data Historical data is collected to form multiple training samples. The collected training samples are then used to construct a matrix, resulting in a historical data matrix X (each row corresponds to the observation data of one livestock sample, and each column corresponds to an independent variable, i.e., historical physiological data and historical environmental data) and a response vector 𝑦 (each element corresponds to the historical actual feeding data of each livestock sample). The constructed historical data matrix X and response vector 𝑦 are then solved using the least squares method to obtain the parameter combination that minimizes the sum of squared errors, thus yielding the scaling factor matrix. This process involves determining the proportional coefficient for each influencing factor. Finally, a milk feeding amount-linear regression prediction model (also known as a demand model) is constructed using the proportional coefficient matrix. This model inputs breed data, age data, weight data, body temperature data, sex data, body length data, body size data, as well as temperature and humidity data obtained from the target livestock during the current feeding cycle into the demand model to determine the appropriate feeding amount for the target livestock during the current feeding cycle. Perform precise calculations.

[0118] The specific calculation formula for the demand model is as follows: ;in, For the predicted feeding amount, , , , , , , , , represents the fitting coefficient.

[0119] The method provided in this embodiment, by constructing a multiple linear regression model, fully considers the comprehensive impact of factors such as environmental temperature and humidity, body size differences, and breed age on nutritional requirements. Compared to traditional extensive feeding based solely on age, this approach can output personalized and precise feeding amounts that match the current instantaneous state, effectively improving the scientific nature and adaptability of nutritional supply.

[0120] Based on the above embodiments, this embodiment provides a detailed explanation of the specific implementation process of obtaining the real-time health status of the target livestock in step 130 based on the current physiological information and the current behavioral information. Addressing the problem that current livestock health monitoring relies on human experience and suffers from high mortality rates due to delayed disease detection (such as diarrhea and lung infections), this embodiment proposes an automated health decision-making mechanism based on a decision tree model using multi-source perception and classification and regression trees (CART).

[0121] Figure 9 is a flowchart of the health status decision-making provided by the present invention. As shown in Figure 9, in some embodiments, step 130 specifically includes: step 131, extracting data from the current physiological information and the current behavioral information to obtain a second data set.

[0122] Step 132: Input the second data set into the health decision model to obtain the real-time health status output by the health decision model on the second data set through multi-node classification decision.

[0123] It should be noted that current monitoring and early warning of calf health status mainly relies on manual experience, often leading to calf deaths due to delayed treatment of diseases such as diarrhea and lung infections. This embodiment uses an RGB-D depth camera and carbon dioxide and sound sensors to monitor the health status of calves in real time and makes decisions based on a decision tree model. This model can monitor multiple health indicators of calves, including body length, body size, fecal color and shape, behavior, respiratory rate, and lung murmurs. Based on this data, it can provide early disease warnings and dynamically adjust the feeding plan according to the warning results. In other words, it can associate specific health conditions with specific plans. When abnormal health conditions are detected (such as substandard body size, abnormal fecal color or shape, abnormal behavior, abnormal exhaled carbon dioxide concentration, or lung murmurs), the model will make targeted decisions. According to the decision plan, the system will automatically adjust the feeding plan based on a combination of the feeding amount predicted by the growth curve and the feeding amount predicted by the regression prediction model.

[0124] Specifically, during the data acquisition phase, the system first determines whether each sensor device is online. If the device is online, it synchronously collects various data of the target livestock through multi-source heterogeneous sensors, and extracts features highly correlated with health from them to form a second dataset.

[0125] The second dataset specifically includes feature data in the following dimensions: First, morphological data, including but not limited to body length data and body size data; this data is obtained by capturing images and measuring them using an RGB-D depth camera, and is used to assess whether the physical development of young animals meets their age standards.

[0126] Second, behavioral and appearance data, including but not limited to excrement characteristics data and posture data; the excrement characteristics data are obtained by image processing of fecal images captured by an RGB-D depth camera, extracting the fecal color (e.g., whether it is green, whether it has white foam) and shape (e.g., whether it is loose). Posture data is behavioral posture monitored by an RGB-D depth camera, such as whether the person is weak, has difficulty standing (difficulty in standing), or frequently lies down.

[0127] Third, physiological and respiratory characteristic data, including but not limited to carbon dioxide concentration data, respiratory rate data, and lung murmur data; among them, carbon dioxide concentration data is collected by a carbon dioxide sensor and reflects the respiratory metabolic level. Respiratory rate data and lung murmur data are obtained by collecting respiratory audio through a sound sensor and extracting respiratory frequency (such as whether it is rapid) and lung sound characteristics (such as whether there are moist rales or other murmurs) through spectrum analysis.

[0128] Furthermore, to achieve accurate health profiling, a health decision-making model was introduced, constructed using the CART algorithm. This model contains multiple levels of decision nodes, simulating the diagnostic logic of a veterinarian, and making layer-by-layer judgments on the input second data set through multiple decision branches.

[0129] One branch is used to monitor livestock development; specifically, it determines whether the body length and size do not conform to the age-standard measurements. If so, it is determined to be a developmental abnormality (such as growth retardation or malnutrition).

[0130] Branch 2 is used for monitoring the digestive system of livestock; specifically, it determines whether the characteristics of excrement are abnormal (e.g., loose stools, green color, or white foam). If the results are positive, it is determined to be an abnormal digestive condition (e.g., diarrhea, constipation, or other digestive system problems).

[0131] Branch 3 is used to monitor the limb and mental state of livestock; specifically, it determines whether posture data is abnormal (such as weakness, frequent lying down). If so, it is determined to be a behavioral abnormality (such as abdominal pain, joint problems, etc.).

[0132] Branch 4 is used for monitoring the respiratory system of livestock; specifically, it determines whether data such as carbon dioxide concentration, respiratory rate, or lung murmurs are abnormal (e.g., abnormal breathing, presence of lung murmurs). If the results are found to be abnormal, the condition is identified as an abnormal respiratory state (e.g., bronchopneumonia, foreign body pneumonia, viral infection).

[0133] Furthermore, after obtaining the health decision model, the second dataset can be input into the health decision model so that the health decision model can output the classification result of the leaf node, i.e., the real-time health status, according to the traversal path of the input data in the decision tree.

[0134] The health decision-making model is trained based on multiple sets of historical physiological and behavioral data with health status labels for each sample of livestock. In other words, through training with a large amount of historical case data, the model has a high accuracy rate in disease identification.

[0135] Furthermore, based on the different abnormality types identified in the decision, corresponding medication addition instructions and feeding amount adjustment coefficients can be generated. These instructions and coefficients allow for adjustments to the feeding plan to arrive at the final target feeding plan. For example, for developmental abnormalities, a medication addition instruction can be generated to "increase protein and energy density in milk replacer." For digestive abnormalities, an instruction can be generated to "adjust milk replacer formula, increase antibiotics, and reduce protein." For behavioral abnormalities, an instruction can be generated to "adjust milk replacer formula, increase calcium and other vitamins." For respiratory abnormalities, an instruction can be generated to "isolate for treatment and administer broad-spectrum antibiotics such as penicillin, streptomycin, and cefotaxime."

[0136] If the equipment is offline or the model cannot make a judgment, it will switch to manual monitoring mode to manually troubleshoot and diagnose the problem, in order to generate a corresponding feeding intervention additive formula to adjust the feeding program and obtain the final target feeding program.

[0137] The method provided in this embodiment, through multi-dimensional perception using RGB-D cameras, sound and gas sensors, combined with logical reasoning of CART decision trees, enables comprehensive real-time monitoring of young animals from body size development to digestive and respiratory systems. This allows for the timely detection of early signs such as substandard body size, diarrhea, and pneumonia, and the generation of targeted intervention suggestions. This provides a scientific and timely basis for dynamically adjusting feeding programs, significantly reducing the risk of death in young animals due to delayed medical treatment.

[0138] This embodiment, based on the aforementioned embodiments, provides a detailed explanation of the specific logic in step 140 of adjusting the second feeding plan according to the real-time health status and the first feeding plan to obtain the target feeding plan.

[0139] In some embodiments, step 140 specifically includes: step 141, when the real-time health status is healthy, adjusting the recommended feeding amount in the second feeding plan based on the recommended feeding amount in the first feeding plan and the feed intake rate of the target livestock in the previous feeding cycle to obtain the target feeding plan.

[0140] Optionally, when the target livestock is determined to be in a healthy state in the current feeding cycle, a robust growth strategy is adopted to determine the target feeding plan. Specifically, a second feeding plan reflecting current needs is used as a benchmark, and a first feeding plan reflecting long-term trends is introduced for weighted adjustment. At the same time, feed intake rate is considered to calculate the final target feeding plan, thereby ensuring that the feed amount is both adapted to the environment and in line with the growth plan, and does not exceed the livestock's eating desire.

[0141] Step 142: When the real-time health status is abnormal, obtain the drug addition instruction and feeding amount adjustment coefficient corresponding to the real-time health status according to the abnormality type of the real-time health status. According to the feeding amount adjustment coefficient and the feed intake rate, perform a reduction calculation on the recommended feeding amount in the second feeding plan, and generate the target feeding plan according to the calculation result and the drug addition instruction.

[0142] When an abnormality is detected, the feeding program is adjusted to a disease intervention strategy to ultimately generate a target feeding program that includes instructions for adding medication.

[0143] Specifically, based on the type of abnormality, the system queries the preset treatment strategy library to obtain the corresponding drug addition instructions (such as adding electrolytes) and feeding amount adjustment coefficient (e.g., 0.8).

[0144] Secondly, using the feeding amount adjustment coefficient and feed intake rate, the recommended amount in the second feeding program is reduced to obtain the safe feeding amount.

[0145] Finally, the safe feeding amount is combined with the medication addition instructions to generate a target feeding program that includes a specific formula and dosage.

[0146] The method provided in this embodiment realizes a closed-loop logic of monitoring-early warning-intervention, which can automatically adjust the diet structure of young animals in the early stage of illness, assist in treatment, and effectively reduce the risk of disease deterioration caused by improper feeding.

[0147] The livestock feeding control system based on multi-source information decision-making provided by the present invention will be described below. The livestock feeding control system based on multi-source information decision-making described below can be referred to in correspondence with the livestock feeding control method based on multi-source information decision-making described above.

[0148] Figure 10 is a second schematic diagram of the livestock feeding control system based on multi-source information decision-making provided by the present invention; as shown in Figure 10, the system includes: a data acquisition unit 1010 for acquiring multi-source information corresponding to the target livestock; the multi-source information includes historical physiological information, current physiological information, current behavioral information, and current environmental information; the historical physiological information is physiological information collected in each historical feeding cycle, and the current physiological information, the current behavioral information, and the current environmental information are physiological information, behavioral information, and environmental information collected in the current feeding cycle, respectively; a prediction unit 1020 is used to predict based on the historical... The system uses physiological information to predict a first feeding plan for the target livestock in the current feeding cycle, and predicts a second feeding plan for the target livestock in the current feeding cycle based on the current physiological information and the current environmental information; the decision unit 1030 is used to obtain the real-time health status of the target livestock based on the current physiological information and the current behavioral information; the control unit 1040 is used to adjust the second feeding plan based on the real-time health status and the first feeding plan to obtain a target feeding plan, and control the feeding pump associated with the target livestock to perform feeding operations according to the target feeding plan.

[0149] The system provided in this embodiment integrates multi-source sensors to achieve comprehensive real-time perception of individual information, environmental information, and health status of young livestock. By combining historical growth trends and current nutritional needs for dual prediction, and incorporating real-time health status for dynamic adjustment, this method effectively solves the problems of nutritional imbalance caused by traditional feeding and high mortality rates of young livestock due to delayed disease detection. It can dynamically customize feeding programs for different individual differences and environmental changes, effectively improving the precision and intelligence of livestock feeding, helping to increase daily weight gain, promote rumen development, significantly reduce disease risk, and save on breeding management costs and improve management efficiency.

[0150] The system provided by this invention is used to execute the above-described method embodiments. For specific processes and details, please refer to the above embodiments, which will not be repeated here.

[0151] Figure 11 illustrates a schematic diagram of the physical structure of an electronic device. As shown in Figure 11, the electronic device may include: a processor 1110, a communications interface 1120, a memory 1130, and a communication bus 1140, wherein the processor 1110, the communications interface 1120, and the memory 1130 communicate with each other through the communication bus 1140. The processor 1110 can call logic instructions in the memory 1130 to execute a livestock feeding control method based on multi-source information decision-making. The method includes: acquiring multi-source information corresponding to a target livestock; the multi-source information includes historical physiological information, current physiological information, current behavioral information, and current environmental information; the historical physiological information is physiological information collected in each historical feeding cycle, and the current physiological information, current behavioral information, and current environmental information are physiological information, behavioral information, and environmental information collected in the current feeding cycle, respectively; predicting a first feeding plan for the target livestock in the current feeding cycle based on the historical physiological information; predicting a second feeding plan for the target livestock in the current feeding cycle based on the current physiological information and the current environmental information; acquiring the real-time health status of the target livestock based on the current physiological information and the current behavioral information; adjusting the second feeding plan based on the real-time health status and the first feeding plan to obtain a target feeding plan; and controlling a feeding pump associated with the target livestock to perform feeding operations based on the target feeding plan.

[0152] Furthermore, the logical instructions in the aforementioned memory 1130 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0153] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the livestock feeding control method based on multi-source information decision-making provided by the above methods. The method includes: acquiring multi-source information corresponding to the target livestock; the multi-source information includes historical physiological information, current physiological information, current behavioral information, and current environmental information; the historical physiological information is physiological information collected in each historical feeding cycle, and the current physiological information, the current behavioral information, and the current environmental information are respectively... The system collects physiological, behavioral, and environmental information during the current feeding cycle; based on the historical physiological information, it predicts a first feeding plan for the target livestock during the current feeding cycle; based on the current physiological and environmental information, it predicts a second feeding plan for the target livestock during the current feeding cycle; based on the current physiological and behavioral information, it obtains the real-time health status of the target livestock; based on the real-time health status and the first feeding plan, it adjusts the second feeding plan to obtain a target feeding plan; and based on the target feeding plan, it controls the feeding pump associated with the target livestock to perform feeding operations.

[0154] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the livestock feeding control method based on multi-source information decision-making provided by the above methods. The method includes: acquiring multi-source information corresponding to a target livestock; the multi-source information includes historical physiological information, current physiological information, current behavioral information, and current environmental information; the historical physiological information is physiological information collected in each historical feeding cycle, and the current physiological information, current behavioral information, and current environmental information are physiological information, behavioral information, and environmental information collected in the current feeding cycle, respectively; predicting a first feeding plan for the target livestock in the current feeding cycle based on the historical physiological information; predicting a second feeding plan for the target livestock in the current feeding cycle based on the current physiological information and the current environmental information; acquiring the real-time health status of the target livestock based on the current physiological information and the current behavioral information; adjusting the second feeding plan based on the real-time health status and the first feeding plan to obtain a target feeding plan; and controlling a feeding pump associated with the target livestock to perform feeding operations according to the target feeding plan.

[0155] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0156] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0157] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A livestock feeding control method based on multi-source information decision-making, characterized in that, include: Obtain multi-source information corresponding to the target livestock; the multi-source information includes historical physiological information, current physiological information, current behavioral information, and current environmental information; The historical physiological information refers to the physiological information collected in each historical feeding cycle, and the current physiological information, the current behavioral information, and the current environmental information refer to the physiological information, behavioral information, and environmental information collected in the current feeding cycle, respectively. Based on the historical physiological information, a first feeding plan for the target livestock in the current feeding cycle is predicted; based on the current physiological information and the current environmental information, a second feeding plan for the target livestock in the current feeding cycle is predicted. Based on the current physiological information and the current behavioral information, the real-time health status of the target livestock is obtained; Based on the real-time health status and the first feeding plan, the second feeding plan is adjusted to obtain a target feeding plan. Based on the target feeding plan, the feeding pump associated with the target livestock is controlled to perform feeding operations.

2. The livestock feeding control method based on multi-source information decision-making according to claim 1, characterized in that, The step of controlling the feeding pump associated with the target livestock to perform feeding operations according to the target feeding plan includes: calculating the actual flow rate error between the real-time output flow rate of the feeding pump and the target flow rate corresponding to the target feeding plan, and the actual liquid level error between the real-time liquid level of the feeding container associated with the target livestock and the target liquid level corresponding to the target feeding plan; inputting the actual flow rate error to a first PID controller to obtain a flow rate control component; inputting the actual liquid level error to a second PID controller to obtain a liquid level control component; performing weighted synthesis of the flow rate control component and the liquid level control component according to the target value of the target weight coefficient, and generating a target feeding signal according to the weighted synthesis result; determining the feeding speed of the feeding pump according to the correlation between the target feeding signal and the motor speed of the feeding pump, and controlling the feeding pump to perform feeding operations according to the feeding speed.

3. The livestock feeding control method based on multi-source information decision-making according to claim 2, characterized in that, The steps for determining the target value of the target weight coefficient include: adjusting the flow rate control parameters of the first PID controller; simulating the feeding operation of the feeding pump based on the adjusted flow rate control parameters until the response index value of the simulated output flow rate of the feeding pump meets the rapidity threshold; and determining the optimal flow rate control parameters from the last adjusted flow rate control parameters; adjusting the liquid level control parameters of the second PID controller; simulating the feeding operation of the feeding pump based on the adjusted liquid level control parameters until the change index of the simulated liquid level of the feeding container meets the stability threshold; and determining the optimal liquid level control parameters from the last adjusted liquid level control parameters; generating test control commands corresponding to different preset values ​​of the target weight coefficient based on the optimal flow rate control parameters and the optimal liquid level control parameters; simulating the feeding operation of the feeding pump based on each test control command; obtaining the comprehensive control error generated by the simulated operation of the feeding pump under each test control command; and determining the preset value corresponding to the minimum comprehensive control error as the target value of the target weight coefficient.

4. The livestock feeding control method based on multi-source information decision-making according to claim 2, characterized in that, The step of generating a target feeding signal based on the weighted synthesis result includes: acquiring real-time temperature data of the feeding material in the feeding container and the real-time sucking frequency of the target livestock; determining a flow rate feedforward compensation amount corresponding to the real-time temperature data based on the correlation between the real-time temperature data and the flow rate; superimposing the flow rate feedforward compensation amount with the weighted synthesis result to generate a feedback control signal; generating a fluctuation signal synchronized with the real-time sucking frequency, and superimposing the fluctuation signal with the feedback control signal to obtain the target feeding signal.

5. The livestock feeding control method based on multi-source information decision-making according to any one of claims 1-4, characterized in that, The step of predicting a first feeding plan for the target livestock in the current feeding cycle based on the historical physiological information includes: obtaining historical growth data of the target livestock in each of the historical feeding cycles from the historical physiological information; fitting a maximum weight parameter, growth rate parameter, and adjustment coefficient suitable for the target livestock based on each of the historical growth data; constructing a growth model based on the maximum weight parameter, the growth rate parameter, and the adjustment coefficient, and predicting a growth curve for the target livestock based on the growth model; calculating predicted growth data for the target livestock in the current feeding cycle based on the growth curve; and inputting the predicted growth data into a fitted regression model to obtain the first feeding plan; wherein the fitted regression model is constructed based on fitting each of the historical growth data and the historical feeding data of the target livestock in each of the historical feeding cycles.

6. The livestock feeding control method based on multi-source information decision-making according to any one of claims 1-4, characterized in that, The step of predicting a second feeding plan for the target livestock in the current feeding cycle based on the current physiological information and the current environmental information includes: acquiring multiple sets of training samples; each set of training samples includes historical physiological information of each sample livestock, historical environmental information of each sample livestock, and historical feeding data of each sample livestock; using the least squares method to perform multiple linear regression fitting on the multiple sets of training samples to obtain a proportional coefficient matrix corresponding to the demand model, and constructing the demand model based on the proportional coefficient matrix; extracting data from the current physiological information and the current environmental information to obtain a first data set; the first data set includes breed data, age data, weight data, body temperature data, sex data, body length data, body size data, temperature data, and humidity data; and inputting the first data set into the demand model to obtain the second feeding plan.

7. The livestock feeding control method based on multi-source information decision-making according to any one of claims 1-4, characterized in that, The step of obtaining the real-time health status of the target livestock based on the current physiological information and the current behavioral information includes: extracting data from the current physiological information and the current behavioral information to obtain a second data set; the second data set includes body length data, body size data, posture data, respiratory rate data, lung murmur data, carbon dioxide concentration data, and excrement morphology data; inputting the second data set into a health decision model to obtain the real-time health status output by the health decision model through multi-node classification decision on the second data set; wherein, the health decision model is trained based on multiple sets of historical physiological data and historical behavioral data with health status labels corresponding to each sample livestock.

8. The livestock feeding control method based on multi-source information decision-making according to any one of claims 1-4, characterized in that, The step of adjusting the second feeding plan based on the real-time health status and the first feeding plan to obtain a target feeding plan includes: when the real-time health status is healthy, modifying the recommended feeding amount in the second feeding plan based on the recommended feeding amount in the first feeding plan and the feed intake rate of the target livestock in the previous feeding cycle to obtain the target feeding plan; when the real-time health status is abnormal, obtaining the drug addition instruction and feeding amount adjustment coefficient corresponding to the real-time health status according to the abnormality type of the real-time health status, calculating the attenuation of the recommended feeding amount in the second feeding plan according to the feeding amount adjustment coefficient and the feed intake rate, and generating the target feeding plan according to the calculation result and the drug addition instruction.

9. A livestock feeding control system based on multi-source information decision-making, characterized in that, include: The data acquisition unit is used to acquire multi-source information corresponding to the target livestock; The multi-source information includes historical physiological information, current physiological information, current behavioral information, and current environmental information; The historical physiological information refers to the physiological information collected in each historical feeding cycle, and the current physiological information, the current behavioral information, and the current environmental information refer to the physiological information, behavioral information, and environmental information collected in the current feeding cycle, respectively. The prediction unit is configured to predict a first feeding plan for the target livestock in the current feeding cycle based on the historical physiological information, and to predict a second feeding plan for the target livestock in the current feeding cycle based on the current physiological information and the current environmental information. The decision unit is used to obtain the real-time health status of the target livestock based on the current physiological information and the current behavioral information; the control unit is used to adjust the second feeding plan based on the real-time health status and the first feeding plan to obtain the target feeding plan, and control the feeding pump associated with the target livestock to perform feeding operations according to the target feeding plan.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the livestock feeding control method based on multi-source information decision-making as described in any one of claims 1 to 8.

Images