Method and system for collaborative scheduling optimization of feeding equipment based on livestock breeding environment

Abstract

The application relates to the technical field of intelligent management of livestock breeding, and provides a feeding equipment collaborative scheduling optimization method and system based on a livestock breeding environment. The method sets each feeding equipment as an independent intelligent agent by collecting environmental data of livestock breeding, livestock growth data and feeding equipment state data, builds a collaborative scheduling model by relying on an improved multi-agent reinforcement learning algorithm, inputs dynamic scheduling basic data into a model state space, constructs a multi-objective reward function, generates a multi-device initial collaborative scheduling strategy, builds a digital twin model corresponding to the breeding environment and the feeding equipment, performs simulation pre-execution on the initial collaborative scheduling strategy, outputs an optimized scheduling strategy that passes verification, issues the optimized scheduling strategy to the entity feeding equipment for execution, continuously collects device feeding execution result data and breeding environment feedback data, continuously iterates and optimizes the collaborative scheduling model, and realizes dynamic self-adaptive updating of the feeding equipment scheduling strategy.

Description

Technical Field

[0001] This application relates to the field of intelligent management technology for livestock farming, and in particular to a method and system for collaborative scheduling optimization of feeding equipment based on the livestock farming environment. Background Technology

[0002] As livestock farming develops towards large-scale and intensive operations, the automated scheduling of feeding equipment has become a key factor in improving farming efficiency. In existing large-scale livestock farming scenarios, the scheduling of feeding equipment often relies on fixed time-series rules or the independent control of a single device. This fails to fully consider the dynamic changes in the farming environment, the growth status of livestock and poultry, and the collaborative needs of multiple devices. Consequently, problems such as path conflicts, insufficient feeding accuracy, serious feed waste, high energy consumption of equipment, and a lack of dynamic adaptive capabilities arise, making it impossible to meet the refined management requirements of modern smart livestock farming. Summary of the Invention

[0003] This application provides a method and system for collaborative scheduling optimization of feeding equipment based on the livestock farming environment. It aims to solve the problems in existing large-scale livestock farming scenarios, where the scheduling of feeding equipment mostly relies on fixed time sequence rules or independent control of a single device, without fully combining the dynamic changes of the farming environment, the growth status of livestock and poultry, and the collaborative needs of multiple devices. This can easily lead to path conflicts, insufficient feeding accuracy, serious feed waste, high energy consumption of equipment, and a lack of dynamic adaptive capabilities.

[0004] In a first aspect, embodiments of this application provide a method for collaborative scheduling optimization of feeding equipment based on a livestock farming environment, the method comprising: Collect environmental data, livestock growth data, and feeding equipment status data for livestock farming, integrate and process multi-dimensional data, eliminate data silos, and generate basic data for dynamic scheduling; Each feeding device is set as an independent intelligent agent. A collaborative scheduling model is built based on an improved multi-agent reinforcement learning algorithm. The basic data of dynamic scheduling is input into the state space of the model. A multi-objective reward function is constructed based on feeding accuracy, equipment energy consumption, feed utilization rate and livestock and poultry feed intake satisfaction to generate an initial collaborative scheduling strategy for multiple devices. A digital twin model corresponding to the breeding environment and feeding equipment is built. The initial collaborative scheduling strategy is simulated and pre-executed to identify and avoid operational risks such as equipment path conflicts, feed line blockages and overlapping feeding times. The optimized scheduling strategy that has passed the verification is output. The optimized scheduling strategy is then sent to the physical feeding equipment for execution. The equipment feeding execution results data and breeding environment feedback data are continuously collected. The feedback data is used as training samples to continuously iterate and optimize the collaborative scheduling model, so as to realize the dynamic adaptive update of the feeding equipment scheduling strategy.

[0005] In some embodiments, the collection of environmental data, livestock growth data, and feeding equipment status data for livestock farming includes: collecting temperature and humidity, ammonia concentration, and light intensity as environmental data through IoT sensors deployed in the pens; collecting livestock age, weight, historical feed intake, and health status as livestock growth data; and collecting feed line balance, equipment location, fault status, and energy consumption level as feeding equipment status data.

[0006] In some embodiments, the process of fusing multi-dimensional data to eliminate data silos and generate dynamic scheduling base data includes: performing outlier filtering and missing value filling on the collected multi-dimensional data to complete data cleaning; converting the cleaned data into a unified format to complete data standardization; aligning the standardized data with timestamps according to a unified time base; and integrating the aligned data to generate dynamic scheduling base data.

[0007] In some embodiments, setting each feeding device as an independent agent and building a collaborative scheduling model based on an improved multi-agent reinforcement learning algorithm includes: building a collaborative scheduling model using a centralized training distributed execution architecture; defining an observation space for each independent agent that includes environmental state, livestock state, and equipment state; defining an action space for each independent agent that includes feed amount, running speed, and movement path; and initializing the network parameters and training hyperparameters of the model.

[0008] In some embodiments, the step of inputting dynamic scheduling basic data into the model state space and constructing a multi-objective reward function based on feeding accuracy, equipment energy consumption, feed utilization rate, and livestock and poultry feed intake satisfaction to generate an initial collaborative scheduling strategy for multiple devices includes: inputting dynamic scheduling basic data into the observation space of each agent; calculating sub-reward values ​​for feeding accuracy, equipment energy consumption, feed utilization rate, and livestock and poultry feed intake satisfaction respectively; weighting and summing each sub-reward value to obtain a comprehensive reward value; and having each agent make decisions and interact based on the comprehensive reward value to generate an initial collaborative scheduling strategy for multiple devices.

[0009] In some embodiments, the construction of a digital twin model corresponding to the breeding environment and feeding equipment includes: collecting spatial layout parameters of the pen and geometric and physical parameters of each feeding equipment; constructing a three-dimensional virtual scene with a 1:1 scale to the actual breeding environment based on the collected parameters; mapping the real-time status of each feeding equipment to the corresponding virtual equipment in the three-dimensional virtual scene; and establishing a real-time data synchronization channel between the physical entity and the virtual model.

[0010] In some embodiments, the simulation pre-execution of the initial collaborative scheduling strategy to identify and avoid operational risks such as equipment path conflicts, material line congestion, and overlapping feeding times, and outputting a verified optimized scheduling strategy includes: inputting the initial collaborative scheduling strategy into the digital twin model; driving the virtual feeding equipment to simulate operation according to the initial collaborative scheduling strategy; detecting equipment path conflicts, material line congestion, and overlapping feeding times in real time during the simulation operation; adjusting the equipment operating parameters and execution sequence in the initial collaborative scheduling strategy according to the detected risks; re-inputting the adjusted strategy into the digital twin model for verification; repeating the risk detection and strategy adjustment steps until there are no operational risks during the simulation operation; and outputting a verified optimized scheduling strategy.

[0011] In some embodiments, the step of sending the optimized scheduling strategy to the physical feeding equipment for execution and continuously collecting the equipment feeding execution result data and the aquaculture environment feedback data includes: sending the corresponding instructions in the optimized scheduling strategy to the equipment according to the unique address of each physical feeding equipment; controlling the physical feeding equipment to execute the feeding operation according to the received instructions; collecting the actual feed amount, running time, and energy consumption level of each physical feeding equipment as the equipment feeding execution result data; and collecting the temperature, humidity, ammonia concentration, and light intensity in the pen as the aquaculture environment feedback data.

[0012] In some embodiments, the step of using feedback data as training samples to continuously iteratively optimize the collaborative scheduling model and realize the dynamic adaptive update of the feeding equipment scheduling strategy includes: organizing the collected equipment feeding execution result data and aquaculture environment feedback data into labeled training samples; inputting the labeled training samples into the collaborative scheduling model for incremental training; updating the network parameters of the collaborative scheduling model; and generating a new feeding equipment scheduling strategy based on the updated network parameters.

[0013] Secondly, this application provides a collaborative scheduling and optimization system for feeding equipment based on the livestock farming environment, the system comprising: The data generation unit is used to collect environmental data, livestock growth data, and feeding equipment status data of livestock farming, integrate and process multi-dimensional data, eliminate data silos, and generate basic data for dynamic scheduling. The strategy generation unit is used to set each feeding device as an independent intelligent agent, build a cooperative scheduling model based on the improved multi-agent reinforcement learning algorithm, input the dynamic scheduling basic data into the model state space, construct a multi-objective reward function based on feeding accuracy, equipment energy consumption, feed utilization rate and livestock and poultry feed intake satisfaction, and generate the initial cooperative scheduling strategy for multiple devices. The strategy update unit is used to build a digital twin model corresponding to the breeding environment and feeding equipment, simulate and pre-execute the initial collaborative scheduling strategy, identify and avoid operational risks such as equipment path conflicts, feed line blockages and overlapping feeding times, and output the optimized scheduling strategy that has passed the verification. The optimized scheduling strategy is then sent to the physical feeding equipment for execution, and the equipment feeding execution result data and breeding environment feedback data are continuously collected. The feedback data is used as training samples to continuously iterate and optimize the collaborative scheduling model, so as to realize the dynamic adaptive update of the feeding equipment scheduling strategy.

[0014] This application breaks down data silos between different dimensions through multi-dimensional data fusion processing, providing comprehensive and dynamic data support for scheduling decisions. It employs a multi-agent reinforcement learning algorithm to construct a collaborative scheduling model, designating each feeding device as an independent agent. Decisions are driven by a multi-objective reward function, achieving efficient collaboration among multiple devices while considering feeding accuracy, equipment energy consumption, feed utilization, and livestock feed satisfaction. Through a digital twin simulation pre-scheduling stage, potential equipment operation conflicts and process risks are identified and avoided in advance, significantly improving the reliability and stability of the scheduling scheme. Fourth, a closed-loop iterative optimization mechanism is established to continuously collect execution feedback data to update the model, enabling the scheduling strategy to adaptively optimize according to changes in the breeding scenario, the progression of livestock growth stages, and changes in equipment status, ensuring the long-term effectiveness and scenario adaptability of the scheduling scheme.

[0015] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic flowchart illustrating the steps of a collaborative scheduling optimization method for feeding equipment based on a livestock farming environment, provided in one embodiment of this application. Figure 2 This is a schematic diagram illustrating the principle of a collaborative scheduling optimization method for feeding equipment based on the livestock farming environment, provided in one embodiment of this application. Figure 3 This is a schematic block diagram of a collaborative scheduling optimization system for feeding equipment based on a livestock farming environment, provided in one embodiment of this application. Figure 4 This is a schematic block diagram of the structure of a computer device provided in an embodiment of this application.

[0018] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0021] It should be understood that, in order to clearly describe the technical solutions of the embodiments of the present invention, the terms "first" and "second" are used in the embodiments of the present invention to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and the terms "first" and "second" are not necessarily different.

[0022] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0023] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0024] As livestock farming develops towards large-scale and intensive operations, the automated scheduling of feeding equipment has become a key factor in improving farming efficiency. In existing large-scale livestock farming scenarios, the scheduling of feeding equipment often relies on fixed time-series rules or the independent control of a single device. This fails to fully consider the dynamic changes in the farming environment, the growth status of livestock and poultry, and the collaborative needs of multiple devices. Consequently, problems such as path conflicts, insufficient feeding accuracy, serious feed waste, high energy consumption of equipment, and a lack of dynamic adaptive capabilities arise, making it impossible to meet the refined management requirements of modern smart livestock farming.

[0025] Please refer to Figure 1 and Figure 2This application provides a method for collaborative scheduling optimization of feeding equipment based on a livestock farming environment, applied to computer equipment. The computer equipment can be deployed on a single server or a server cluster. It can also be deployed on handheld terminals, laptops, wearable devices, or robots, etc.

[0026] The provided method for collaborative scheduling optimization of feeding equipment based on livestock farming environment includes steps S101 to S103. Details are as follows: Step S101. Collect environmental data, livestock growth data, and feeding equipment status data for livestock farming, integrate the multi-dimensional data, eliminate data silos, and generate basic data for dynamic scheduling.

[0027] Specifically, this step provides the data foundation for the entire collaborative scheduling system, addressing the problem of insufficient decision-making basis caused by existing technologies that only collect operational data from a single device. This step is divided into two sub-stages: data acquisition and data fusion, which are executed sequentially.

[0028] The data acquisition sub-stage utilizes various sensing devices deployed at the aquaculture site to simultaneously collect dynamic data across three dimensions. The data acquisition process combines timed acquisition with event-triggered acquisition. Under normal conditions, data is collected at preset time intervals. When a sudden data change exceeds a preset threshold, the acquisition frequency is automatically increased to ensure the real-time nature and accuracy of the data. The collected raw data is transmitted to computer equipment via a wireless communication network for subsequent processing.

[0029] The data fusion sub-step standardizes the collected multi-dimensional raw data, eliminating format, time, and quality differences between different data sources to generate dynamic scheduling foundation data in a unified format. The data fusion process strictly follows the order of data cleaning, data standardization, timestamp alignment, and data integration to ensure that the final dynamic scheduling foundation data can be directly input into the subsequent collaborative scheduling model.

[0030] Step S102. Each feeding device is set as an independent intelligent agent. A collaborative scheduling model is built based on the improved multi-agent reinforcement learning algorithm. The basic data of dynamic scheduling is input into the state space of the model. A multi-objective reward function is constructed based on feeding accuracy, equipment energy consumption, feed utilization rate and livestock feeding satisfaction to generate an initial collaborative scheduling strategy for multiple devices.

[0031] Specifically, this step is the core decision-making stage of the entire collaborative scheduling system, addressing the problems of poor multi-device coordination and inability to simultaneously consider multiple optimization objectives caused by existing technologies that employ fixed rules or single-objective optimization. This step defines each feeding device as an independent intelligent agent and achieves distributed collaborative decision-making among multiple devices through a multi-agent reinforcement learning algorithm.

[0032] First, the basic architecture of the collaborative scheduling model is established, defining the observation space and action space for each agent, and initializing the model's network parameters and training hyperparameters. Then, the dynamic scheduling baseline data generated in step S101 is input into the model's state space as the basis for agent decision-making. Next, a multi-objective reward function is constructed, transforming four optimization objectives—feeding accuracy, equipment energy consumption, feed utilization, and livestock feed intake satisfaction—into quantifiable reward values ​​to guide the agent's decision-making direction. Finally, through interactive learning between agents, an initial multi-device collaborative scheduling strategy is generated.

[0033] Step S103. Build a digital twin model corresponding to the breeding environment and feeding equipment, simulate and pre-execute the initial collaborative scheduling strategy, identify and avoid operational risks such as equipment path conflicts, feed line blockages and overlapping feeding times, and output the verified optimized scheduling strategy; send the optimized scheduling strategy to the physical feeding equipment for execution, continuously collect the equipment feeding execution result data and breeding environment feedback data, and use the feedback data as training samples to continuously iterate and optimize the collaborative scheduling model to achieve dynamic adaptive updating of the feeding equipment scheduling strategy.

[0034] Specifically, this step is for ensuring the reliability and continuous optimization of the entire collaborative scheduling system, addressing the problems of existing technologies that directly issuing scheduling strategies can easily lead to operational risks and that scheduling strategies cannot be adaptively updated according to changes in the scenario. This step is divided into two sub-steps: digital twin simulation pre-verification and closed-loop iterative optimization, which are executed sequentially.

[0035] The digital twin simulation pre-verification sub-step first constructs a digital twin model that perfectly corresponds to the actual aquaculture environment and feeding equipment, achieving real-time data synchronization between the physical entity and the virtual model. Then, the initial collaborative scheduling strategy generated in step S102 is input into the digital twin model, driving the virtual feeding equipment to simulate operation according to the strategy. During the simulation, various potential operational risks are detected in real time, and the initial scheduling strategy is adjusted based on the detection results. The simulation verification and strategy adjustment process is repeated until there are no operational risks during the simulation, at which point the verified optimized scheduling strategy is output.

[0036] The closed-loop iterative optimization sub-step first distributes the validated optimized scheduling strategy to the physical feeding equipment for execution, controlling the equipment to complete the actual feeding operation. During equipment execution, feeding execution result data and feedback data from the aquaculture environment are continuously collected. The collected feedback data is organized into labeled training samples and input into the collaborative scheduling model for incremental training, updating the model's network parameters. Based on the updated model parameters, a new scheduling strategy is generated, achieving dynamic adaptive updating of the scheduling strategy.

[0037] In some embodiments, the collection of environmental data, livestock growth data, and feeding equipment status data for livestock farming includes: collecting temperature and humidity, ammonia concentration, and light intensity as environmental data through IoT sensors deployed in the pens; collecting livestock age, weight, historical feed intake, and health status as livestock growth data; and collecting feed line balance, equipment location, fault status, and energy consumption level as feeding equipment status data.

[0038] Environmental data such as temperature, humidity, ammonia concentration, and light intensity are collected using IoT sensors deployed within the enclosures. Temperature and humidity sensors are positioned 1.5 meters above the ground within the enclosures, with at least one sensor in each enclosure, collecting data once per minute. Ammonia sensors are deployed in poorly ventilated areas within the enclosures, with at least one sensor in each enclosure, collecting data once every 5 minutes. Light intensity sensors are deployed in unobstructed locations within the enclosures, with at least one sensor in each enclosure, collecting data once every 10 minutes.

[0039] Livestock and poultry age, weight, historical feed intake, and health status are collected as livestock and poultry growth data. Age data is entered into the livestock management system and updated daily. Weight data is collected using weighing equipment deployed at the pen exits, with weight data collected at least weekly for each batch of livestock and poultry. Historical feed intake data is generated from the feeding records of the feeding equipment and updated daily. Health status data is entered by farm staff or collected using intelligent visual recognition equipment and updated in real time.

[0040] The feeding equipment's status data includes feed line level, equipment location, fault status, and energy consumption. Feed line level data is collected every 30 minutes via a level sensor deployed within the feed hopper. Equipment location data is collected once per second via a positioning module deployed on the feeding equipment. Equipment fault status data is collected in real-time via the feeding equipment's built-in diagnostic module, and fault information is uploaded accordingly. Equipment energy consumption data is collected hourly via an energy metering module deployed on the equipment's power supply line.

[0041] In some embodiments, the process of fusing multi-dimensional data to eliminate data silos and generate dynamic scheduling base data includes: performing outlier filtering and missing value filling on the collected multi-dimensional data to complete data cleaning; converting the cleaned data into a unified format to complete data standardization; aligning the standardized data with timestamps according to a unified time base; and integrating the aligned data to generate dynamic scheduling base data.

[0042] This embodiment provides a detailed description of the specific implementation method of data fusion processing in step S101.

[0043] Data cleaning was performed on the collected multi-dimensional data, including outlier filtering and missing value imputation. Outlier filtering employed the 3σ principle, calculating the mean and standard deviation of each data sequence. Data exceeding three times the standard deviation of the mean were identified as outliers and removed. Missing value imputation used linear interpolation, calculating the imputation value for the missing position based on the values ​​of the two valid data points before and after the missing value. For sequences with more than three consecutive missing data points, the historical average value of the sequence was used for imputation.

[0044] The cleaned data is converted into a unified format to complete data standardization. Data standardization uses the min-max standardization method, converting data of different dimensions into values ​​between 0 and 1. For indicators where larger values ​​are better, a forward standardization formula is used; for indicators where smaller values ​​are better, a backward standardization formula is used. Standardized data is retained to four decimal places to ensure data accuracy.

[0045] The standardized data is timestamped according to a unified time base. Coordinated Universal Time (UTC) is used as the unified time base, and the timestamps of all data are converted to UTC format. Based on the different data collection frequencies, the data is aligned at 1-minute intervals. For data collected more than 1 minute apart, the average value within that time interval is used as the aligned data; for data collected less than 1 minute apart, forward padding is used to fill in any missing data within that time interval.

[0046] The aligned data is then integrated to generate basic data for dynamic scheduling. Environmental data, livestock growth data, and feeding equipment status data at the same time point are combined into a single complete data record, arranged in timestamp order. Each data record contains a unique timestamp and data fields for all dimensions. The generated basic data for dynamic scheduling is stored in a time-series database for subsequent use by the collaborative scheduling model.

[0047] In some embodiments, setting each feeding device as an independent agent and building a collaborative scheduling model based on an improved multi-agent reinforcement learning algorithm includes: building a collaborative scheduling model using a centralized training distributed execution architecture; defining an observation space for each independent agent that includes environmental state, livestock state, and equipment state; defining an action space for each independent agent that includes feed amount, running speed, and movement path; and initializing the network parameters and training hyperparameters of the model.

[0048] This embodiment provides a detailed explanation of the specific implementation method for building the collaborative scheduling model in step S102.

[0049] A centralized training and distributed execution architecture is adopted to build a collaborative scheduling model. In the centralized training phase, the experience data of all agents is uploaded to the central training module, which trains the global value network based on the experience data of all agents. In the distributed execution phase, each agent independently makes decisions and executes corresponding actions based on its local observation data and the trained global value network.

[0050] An observation space is defined for each independent agent, encompassing environmental status, livestock status, and equipment status. Environmental status includes data in three dimensions: temperature and humidity, ammonia concentration, and light intensity. Livestock status includes data in four dimensions: age, average weight, historical average feed intake, and health status. Equipment status includes data in four dimensions: feed level, current location, operating status, and current energy consumption. The total dimensions of the observation space are 11.

[0051] Define an action space for each independent agent, including feed quantity, operating speed, and movement path. Feed quantity is a continuous action, ranging from 0 to the maximum theoretical feed intake of livestock in the pen. Operating speed is a continuous action, ranging from the minimum to the maximum operating speed of the equipment. Movement path is a discrete action, including all preset feasible path options.

[0052] Initialize the model's network parameters and training hyperparameters. The model uses a deep neural network as the value network and policy network. The network structure includes an input layer, three hidden layers, and an output layer. The number of neurons in the input layer matches the dimension of the observation space, each hidden layer contains 128 neurons, and the number of neurons in the output layer matches the dimension of the action space. The training hyperparameters are set as follows: learning rate 0.001, discount factor 0.99, experience replay buffer size 100,000, batch size 64, and training iterations 10,000.

[0053] In some embodiments, the step of inputting dynamic scheduling basic data into the model state space and constructing a multi-objective reward function based on feeding accuracy, equipment energy consumption, feed utilization rate, and livestock and poultry feed intake satisfaction to generate an initial collaborative scheduling strategy for multiple devices includes: inputting dynamic scheduling basic data into the observation space of each agent; calculating sub-reward values ​​for feeding accuracy, equipment energy consumption, feed utilization rate, and livestock and poultry feed intake satisfaction respectively; weighting and summing each sub-reward value to obtain a comprehensive reward value; and having each agent make decisions and interact based on the comprehensive reward value to generate an initial collaborative scheduling strategy for multiple devices.

[0054] This embodiment provides a detailed explanation of the specific implementation of the initial scheduling strategy generation in step S102.

[0055] The dynamic scheduling basic data is input into the observation space of each agent. The latest dynamic scheduling basic data is read from the time-series database in timestamp order, and the environmental state, livestock state, and equipment state fields from the data are input into the observation space of the corresponding agent.

[0056] The following sub-reward values ​​are calculated separately: feeding accuracy sub-reward value, equipment energy consumption sub-reward value, feed utilization rate sub-reward value, and livestock feed intake satisfaction sub-reward value. The feeding accuracy sub-reward value is calculated based on the difference between the actual feed input and the theoretical feed intake of livestock; the smaller the difference, the higher the reward value. The equipment energy consumption sub-reward value is calculated based on the ratio of the actual energy consumption of the equipment to the baseline energy consumption; the smaller the ratio, the higher the reward value. The feed utilization rate sub-reward value is calculated based on the ratio of the remaining feed to the total feed input; the smaller the ratio, the higher the reward value. The livestock feed intake satisfaction sub-reward value is calculated based on the feeding duration and feeding behavior activity of livestock; the closer the feeding duration is to the preset optimal duration and the more active the feeding behavior, the higher the reward value.

[0057] The weighted sum of the four sub-reward values ​​yields the overall reward value. The weighting coefficients for the four sub-reward values ​​are set according to the actual needs of the farming scenario, with the default settings being 0.3 for feeding accuracy, 0.2 for equipment energy consumption, 0.3 for feed utilization, and 0.2 for livestock feed intake satisfaction. The sum of the weighting coefficients is 1.

[0058] Each agent interacts and makes decisions based on the comprehensive reward value to generate an initial collaborative scheduling strategy for multiple devices. Each agent, based on current observation data and the comprehensive reward value, outputs an action probability distribution through the policy network and selects the appropriate action according to the probability distribution. Agents share their action information through a communication module to avoid resource contention and operational conflicts. The combination of actions from all agents forms the initial collaborative scheduling strategy for multiple devices.

[0059] In some embodiments, the construction of a digital twin model corresponding to the breeding environment and feeding equipment includes: collecting spatial layout parameters of the pen and geometric and physical parameters of each feeding equipment; constructing a three-dimensional virtual scene with a 1:1 scale to the actual breeding environment based on the collected parameters; mapping the real-time status of each feeding equipment to the corresponding virtual equipment in the three-dimensional virtual scene; and establishing a real-time data synchronization channel between the physical entity and the virtual model.

[0060] This embodiment provides a detailed explanation of the specific implementation method for building the digital twin model in step S103.

[0061] The spatial layout parameters of the feeding pens and the geometric and physical parameters of each feeding device were collected. The spatial layout parameters of the feeding pens included length, width, height, pen divisions, and aisle locations. The geometric parameters of the feeding devices included their external dimensions and the relative positions of their components. The physical parameters of the feeding devices included their maximum operating speed, maximum feed rate, energy consumption characteristics, and probability of failure. All parameters were obtained through on-site measurements and by referring to the equipment manuals.

[0062] A 1:1 scale 3D virtual scene was constructed based on the collected parameters, mirroring the actual aquaculture environment. 3D modeling software was used to create 3D models of the pens and feeding equipment, which were then placed within the virtual scene according to the actual spatial layout parameters. The virtual scene contains virtual counterparts of all the physical pens and feeding equipment, with the virtual equipment's geometry and physical characteristics perfectly matching the physical equipment.

[0063] The real-time status of each feeding device is mapped to the corresponding virtual device in the 3D virtual scene. A one-to-one correspondence is established between physical and virtual devices. Real-time operational status data of the physical devices, including device location, operating speed, feed quantity, and fault status, is acquired through a data communication interface. The acquired status data is then updated in real-time to the attributes of the corresponding virtual devices, ensuring that the operational status of the virtual devices remains synchronized with that of the physical devices.

[0064] Establish a real-time data synchronization channel between the physical entity and the virtual model. A message queue communication protocol is used to achieve data transmission between the physical entity and the virtual model. The data synchronization channel supports bidirectional data transmission, allowing both the uploading of status data from the physical device to the virtual model and the sending of control commands from the virtual model to the physical device. Data synchronization latency is controlled within 100 milliseconds to ensure real-time synchronization.

[0065] In some embodiments, the simulation pre-execution of the initial collaborative scheduling strategy to identify and avoid operational risks such as equipment path conflicts, material line congestion, and overlapping feeding times, and outputting a verified optimized scheduling strategy includes: inputting the initial collaborative scheduling strategy into the digital twin model; driving the virtual feeding equipment to simulate operation according to the initial collaborative scheduling strategy; detecting equipment path conflicts, material line congestion, and overlapping feeding times in real time during the simulation operation; adjusting the equipment operating parameters and execution sequence in the initial collaborative scheduling strategy according to the detected risks; re-inputting the adjusted strategy into the digital twin model for verification; repeating the risk detection and strategy adjustment steps until there are no operational risks during the simulation operation; and outputting a verified optimized scheduling strategy.

[0066] This embodiment provides a detailed explanation of the specific implementation method of the simulation pre-execution and optimization scheduling strategy output in step S103.

[0067] The initial cooperative scheduling strategy is input into the digital twin model. The action instructions of each agent in the initial cooperative scheduling strategy are parsed into control instructions that can be executed by the virtual devices, and the control instructions are sent to the corresponding virtual devices in chronological order.

[0068] The virtual feeding equipment is driven to operate in a simulated manner according to the initial collaborative scheduling strategy. Based on received control commands, the virtual equipment simulates and executes corresponding actions in a 3D virtual scene, including movement, feeding, and starting / stopping. During the simulation, the operating status of the virtual equipment and changes in environmental parameters are calculated in real time.

[0069] The system detects and manages the risks of equipment path conflicts, material line blockages, and overlapping feeding times during real-time simulation. Equipment path conflict detection calculates the movement trajectories of different virtual devices to determine if there are intersecting trajectories and overlapping times. Material line blockage detection simulates the material flow process to determine if material accumulation exceeds a preset threshold. Overlapping feeding times are detected by analyzing the feeding time periods of different devices within the same pen to determine if there are overlapping time periods.

[0070] Based on detected risks, adjust the equipment operating parameters and execution sequence in the initial collaborative scheduling strategy. For equipment path conflicts, adjust the operating speed or movement path of the conflicting equipment to stagger their movement trajectories in time. For material line blockages, adjust the feeding speed or feeding sequence of relevant equipment to avoid concentrated material conveying. For overlapping feeding times, adjust the feeding start time of overlapping equipment to ensure that their feeding time periods do not overlap.

[0071] The adjusted strategy is then re-entered into the digital twin model for verification. The risk detection and strategy adjustment steps are repeated until no operational risks are detected during the simulation. The strategy is considered validated when no operational risks are detected in three consecutive simulations.

[0072] Output the validated optimized scheduling strategy. The optimized scheduling strategy contains the operating parameters and execution sequence of all feeding equipment, stored in a structured data format, and can be directly sent to the physical equipment for execution.

[0073] In some embodiments, the step of sending the optimized scheduling strategy to the physical feeding equipment for execution and continuously collecting the equipment feeding execution result data and the aquaculture environment feedback data includes: sending the corresponding instructions in the optimized scheduling strategy to the equipment according to the unique address of each physical feeding equipment; controlling the physical feeding equipment to execute the feeding operation according to the received instructions; collecting the actual feed amount, running time, and energy consumption level of each physical feeding equipment as the equipment feeding execution result data; and collecting the temperature, humidity, ammonia concentration, and light intensity in the pen as the aquaculture environment feedback data.

[0074] This embodiment provides a detailed explanation of the specific implementation method of optimizing the scheduling strategy issuance and feedback data collection in step S103.

[0075] The corresponding instructions in the optimized scheduling strategy are sent to each feeding device based on its unique address. Each feeding device is assigned a unique device address, and each control instruction in the optimized scheduling strategy includes that address. The computer then transmits the control instructions to the corresponding device via a wireless communication network.

[0076] The physical feeding equipment is controlled to execute feeding operations according to received instructions. Upon receiving a control instruction, the physical feeding equipment parses the instruction content and performs corresponding actions according to the operating parameters and execution sequence specified in the instruction, including moving to a designated location, feeding according to a specified amount of feed, and running at a specified speed. During the execution process, the equipment provides real-time feedback on its execution status to the computer.

[0077] The actual feed input, runtime, and energy consumption levels of each feeding device are collected as the feeding execution result data. Actual feed input data is collected through the device's built-in weighing module, runtime data is generated from the device's operation log, and energy consumption data is collected through the electricity metering module. All data is uploaded to the computer at preset time intervals.

[0078] Temperature, humidity, ammonia concentration, and light intensity within the enclosures are collected as environmental feedback data for the aquaculture environment. Environmental data is collected using IoT sensors deployed within the enclosures, with the collection frequency consistent with that in step S101. The collected environmental data is uploaded to a computer in real time, serving as the basis for model iterative optimization.

[0079] In some embodiments, the step of using feedback data as training samples to continuously iteratively optimize the collaborative scheduling model and realize the dynamic adaptive update of the feeding equipment scheduling strategy includes: organizing the collected equipment feeding execution result data and aquaculture environment feedback data into labeled training samples; inputting the labeled training samples into the collaborative scheduling model for incremental training; updating the network parameters of the collaborative scheduling model; and generating a new feeding equipment scheduling strategy based on the updated network parameters.

[0080] This embodiment provides a detailed explanation of the specific implementation method of the iterative optimization of the collaborative scheduling model in step S103.

[0081] The collected data on equipment feeding execution results and aquaculture environment feedback were compiled into labeled training samples. Each training sample consists of two parts: input data and labeled data. The input data is the basic dynamic scheduling data before the scheduling strategy is implemented, and the labeled data is the actual effect data after the scheduling strategy is implemented, including actual feeding accuracy, actual equipment energy consumption, actual feed utilization rate, and actual livestock feed intake satisfaction.

[0082] Labeled training samples are input into the collaborative scheduling model for incremental training. An incremental learning method is employed, where the model is trained using new training samples based on the existing model parameters. During training, the parameters of the first two layers of the model are frozen, and only the parameters of the last three layers are updated, preserving the general knowledge already learned by the model while adapting to new scenario data.

[0083] Update the network parameters of the collaborative scheduling model. When the loss function value of incremental training converges to below a preset threshold, stop training and save the updated model parameters. The updated model parameters automatically replace the original model parameters for use in subsequent scheduling strategy generation.

[0084] A new feeding device scheduling strategy is generated based on the updated network parameters. At the start of the next scheduling cycle, the cooperative scheduling model generates a new initial cooperative scheduling strategy for multiple devices based on the latest dynamic scheduling base data and the updated network parameters, and then proceeds to the next round of simulation verification and execution.

[0085] Please see Figure 3 As shown, Figure 3 This is a schematic diagram of the structure of a collaborative scheduling and optimization system 200 for feeding equipment based on a livestock farming environment, provided in this application embodiment. This collaborative scheduling and optimization system 200 for feeding equipment based on a livestock farming environment is used to execute the steps of the collaborative scheduling and optimization method for feeding equipment based on a livestock farming environment shown in the above embodiments. This collaborative scheduling and optimization system 200 for feeding equipment based on a livestock farming environment can be a single server or a server cluster, or it can be a terminal, such as a handheld terminal, a laptop computer, a wearable device, or a robot.

[0086] like Figure 3 As shown, the collaborative scheduling optimization system 200 for feeding equipment based on the livestock farming environment includes: The data generation unit 201 is used to collect environmental data, livestock and poultry growth data and feeding equipment status data of livestock farming, integrate and process multi-dimensional data, eliminate data silos, and generate basic data for dynamic scheduling. The strategy generation unit 202 is used to set each feeding device as an independent intelligent agent, build a cooperative scheduling model based on the improved multi-agent reinforcement learning algorithm, input the dynamic scheduling basic data into the model state space, construct a multi-objective reward function based on feeding accuracy, equipment energy consumption, feed utilization rate and livestock feeding satisfaction, and generate the initial cooperative scheduling strategy for multiple devices. The strategy update unit 203 is used to build a digital twin model corresponding to the breeding environment and feeding equipment, simulate and pre-execute the initial collaborative scheduling strategy, identify and avoid operational risks such as equipment path conflicts, feed line blockages and overlapping feeding times, and output the optimized scheduling strategy that has passed the verification. The optimized scheduling strategy is then sent to the physical feeding equipment for execution, and the equipment feeding execution result data and breeding environment feedback data are continuously collected. The feedback data is used as training samples to continuously iterate and optimize the collaborative scheduling model, so as to realize the dynamic adaptive update of the feeding equipment scheduling strategy.

[0087] In some embodiments, the collection of environmental data, livestock growth data, and feeding equipment status data for livestock farming includes: collecting temperature and humidity, ammonia concentration, and light intensity as environmental data through IoT sensors deployed in the pens; collecting livestock age, weight, historical feed intake, and health status as livestock growth data; and collecting feed line balance, equipment location, fault status, and energy consumption level as feeding equipment status data.

[0088] In some embodiments, the process of fusing multi-dimensional data to eliminate data silos and generate dynamic scheduling base data includes: performing outlier filtering and missing value filling on the collected multi-dimensional data to complete data cleaning; converting the cleaned data into a unified format to complete data standardization; aligning the standardized data with timestamps according to a unified time base; and integrating the aligned data to generate dynamic scheduling base data.

[0089] In some embodiments, setting each feeding device as an independent agent and building a collaborative scheduling model based on an improved multi-agent reinforcement learning algorithm includes: building a collaborative scheduling model using a centralized training distributed execution architecture; defining an observation space for each independent agent that includes environmental state, livestock state, and equipment state; defining an action space for each independent agent that includes feed amount, running speed, and movement path; and initializing the network parameters and training hyperparameters of the model.

[0090] In some embodiments, the step of inputting dynamic scheduling basic data into the model state space and constructing a multi-objective reward function based on feeding accuracy, equipment energy consumption, feed utilization rate, and livestock and poultry feed intake satisfaction to generate an initial collaborative scheduling strategy for multiple devices includes: inputting dynamic scheduling basic data into the observation space of each agent; calculating sub-reward values ​​for feeding accuracy, equipment energy consumption, feed utilization rate, and livestock and poultry feed intake satisfaction respectively; weighting and summing each sub-reward value to obtain a comprehensive reward value; and having each agent make decisions and interact based on the comprehensive reward value to generate an initial collaborative scheduling strategy for multiple devices.

[0091] In some embodiments, the construction of a digital twin model corresponding to the breeding environment and feeding equipment includes: collecting spatial layout parameters of the pen and geometric and physical parameters of each feeding equipment; constructing a three-dimensional virtual scene with a 1:1 scale to the actual breeding environment based on the collected parameters; mapping the real-time status of each feeding equipment to the corresponding virtual equipment in the three-dimensional virtual scene; and establishing a real-time data synchronization channel between the physical entity and the virtual model.

[0092] In some embodiments, the simulation pre-execution of the initial collaborative scheduling strategy to identify and avoid operational risks such as equipment path conflicts, material line congestion, and overlapping feeding times, and outputting a verified optimized scheduling strategy includes: inputting the initial collaborative scheduling strategy into the digital twin model; driving the virtual feeding equipment to simulate operation according to the initial collaborative scheduling strategy; detecting equipment path conflicts, material line congestion, and overlapping feeding times in real time during the simulation operation; adjusting the equipment operating parameters and execution sequence in the initial collaborative scheduling strategy according to the detected risks; re-inputting the adjusted strategy into the digital twin model for verification; repeating the risk detection and strategy adjustment steps until there are no operational risks during the simulation operation; and outputting a verified optimized scheduling strategy.

[0093] In some embodiments, the step of sending the optimized scheduling strategy to the physical feeding equipment for execution and continuously collecting the equipment feeding execution result data and the aquaculture environment feedback data includes: sending the corresponding instructions in the optimized scheduling strategy to the equipment according to the unique address of each physical feeding equipment; controlling the physical feeding equipment to execute the feeding operation according to the received instructions; collecting the actual feed amount, running time, and energy consumption level of each physical feeding equipment as the equipment feeding execution result data; and collecting the temperature, humidity, ammonia concentration, and light intensity in the pen as the aquaculture environment feedback data.

[0094] In some embodiments, the step of using feedback data as training samples to continuously iteratively optimize the collaborative scheduling model and realize the dynamic adaptive update of the feeding equipment scheduling strategy includes: organizing the collected equipment feeding execution result data and aquaculture environment feedback data into labeled training samples; inputting the labeled training samples into the collaborative scheduling model for incremental training; updating the network parameters of the collaborative scheduling model; and generating a new feeding equipment scheduling strategy based on the updated network parameters.

[0095] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the above-described collaborative scheduling optimization system for feeding equipment based on the livestock farming environment and its various modules can be referred to the corresponding contents in the various embodiments of the above-described collaborative scheduling optimization method for feeding equipment based on the livestock farming environment, and will not be repeated here.

[0096] The aforementioned method for collaborative scheduling optimization of feeding equipment based on the livestock farming environment can be implemented as a computer program, which can be used in various ways, such as... Figure 3 It runs on the system shown.

[0097] Please see Figure 4 , Figure 4 This is a schematic block diagram of the structure of a computer device provided in an embodiment of this application. The computer device includes a processor, a memory, and a network interface connected via a device bus, wherein the memory may include a storage medium and internal memory.

[0098] The storage medium can store operating devices and computer programs. The computer program includes program instructions that, when executed, cause the processor to perform any method for coordinating and optimizing the scheduling of feeding equipment based on the livestock farming environment.

[0099] The processor provides computing and control capabilities, supporting the operation of the entire computer device.

[0100] Internal memory provides an environment for the execution of computer programs in non-volatile storage media. When the computer program is executed by the processor, it enables the processor to execute any method for the coordinated scheduling optimization of feeding equipment based on the livestock farming environment.

[0101] This network interface is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the terminal to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0102] It should be understood that the processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, a general-purpose processor can be a microprocessor or any conventional processor.

[0103] In one embodiment, the processor is configured to run a computer program stored in memory to perform the following steps: Collect environmental data, livestock growth data, and feeding equipment status data for livestock farming, integrate and process multi-dimensional data, eliminate data silos, and generate basic data for dynamic scheduling; Each feeding device is set as an independent intelligent agent. A collaborative scheduling model is built based on an improved multi-agent reinforcement learning algorithm. The basic data of dynamic scheduling is input into the state space of the model. A multi-objective reward function is constructed based on feeding accuracy, equipment energy consumption, feed utilization rate and livestock and poultry feed intake satisfaction to generate an initial collaborative scheduling strategy for multiple devices. A digital twin model corresponding to the breeding environment and feeding equipment is built. The initial collaborative scheduling strategy is simulated and pre-executed to identify and avoid operational risks such as equipment path conflicts, feed line blockages and overlapping feeding times. The optimized scheduling strategy that has passed the verification is output. The optimized scheduling strategy is then sent to the physical feeding equipment for execution. The equipment feeding execution results data and breeding environment feedback data are continuously collected. The feedback data is used as training samples to continuously iterate and optimize the collaborative scheduling model, so as to realize the dynamic adaptive update of the feeding equipment scheduling strategy.

[0104] In some embodiments, the collection of environmental data, livestock growth data, and feeding equipment status data for livestock farming includes: collecting temperature and humidity, ammonia concentration, and light intensity as environmental data through IoT sensors deployed in the pens; collecting livestock age, weight, historical feed intake, and health status as livestock growth data; and collecting feed line balance, equipment location, fault status, and energy consumption level as feeding equipment status data.

[0105] In some embodiments, the process of fusing multi-dimensional data to eliminate data silos and generate dynamic scheduling base data includes: performing outlier filtering and missing value filling on the collected multi-dimensional data to complete data cleaning; converting the cleaned data into a unified format to complete data standardization; aligning the standardized data with timestamps according to a unified time base; and integrating the aligned data to generate dynamic scheduling base data.

[0106] In some embodiments, setting each feeding device as an independent agent and building a collaborative scheduling model based on an improved multi-agent reinforcement learning algorithm includes: building a collaborative scheduling model using a centralized training distributed execution architecture; defining an observation space for each independent agent that includes environmental state, livestock state, and equipment state; defining an action space for each independent agent that includes feed amount, running speed, and movement path; and initializing the network parameters and training hyperparameters of the model.

[0107] In some embodiments, the step of inputting dynamic scheduling basic data into the model state space and constructing a multi-objective reward function based on feeding accuracy, equipment energy consumption, feed utilization rate, and livestock and poultry feed intake satisfaction to generate an initial collaborative scheduling strategy for multiple devices includes: inputting dynamic scheduling basic data into the observation space of each agent; calculating sub-reward values ​​for feeding accuracy, equipment energy consumption, feed utilization rate, and livestock and poultry feed intake satisfaction respectively; weighting and summing each sub-reward value to obtain a comprehensive reward value; and having each agent make decisions and interact based on the comprehensive reward value to generate an initial collaborative scheduling strategy for multiple devices.

[0108] In some embodiments, the construction of a digital twin model corresponding to the breeding environment and feeding equipment includes: collecting spatial layout parameters of the pen and geometric and physical parameters of each feeding equipment; constructing a three-dimensional virtual scene with a 1:1 scale to the actual breeding environment based on the collected parameters; mapping the real-time status of each feeding equipment to the corresponding virtual equipment in the three-dimensional virtual scene; and establishing a real-time data synchronization channel between the physical entity and the virtual model.

[0109] In some embodiments, the simulation pre-execution of the initial collaborative scheduling strategy to identify and avoid operational risks such as equipment path conflicts, material line congestion, and overlapping feeding times, and outputting a verified optimized scheduling strategy includes: inputting the initial collaborative scheduling strategy into the digital twin model; driving the virtual feeding equipment to simulate operation according to the initial collaborative scheduling strategy; detecting equipment path conflicts, material line congestion, and overlapping feeding times in real time during the simulation operation; adjusting the equipment operating parameters and execution sequence in the initial collaborative scheduling strategy according to the detected risks; re-inputting the adjusted strategy into the digital twin model for verification; repeating the risk detection and strategy adjustment steps until there are no operational risks during the simulation operation; and outputting a verified optimized scheduling strategy.

[0110] In some embodiments, the step of sending the optimized scheduling strategy to the physical feeding equipment for execution and continuously collecting the equipment feeding execution result data and the aquaculture environment feedback data includes: sending the corresponding instructions in the optimized scheduling strategy to the equipment according to the unique address of each physical feeding equipment; controlling the physical feeding equipment to execute the feeding operation according to the received instructions; collecting the actual feed amount, running time, and energy consumption level of each physical feeding equipment as the equipment feeding execution result data; and collecting the temperature, humidity, ammonia concentration, and light intensity in the pen as the aquaculture environment feedback data.

[0111] In some embodiments, the step of using feedback data as training samples to continuously iteratively optimize the collaborative scheduling model and realize the dynamic adaptive update of the feeding equipment scheduling strategy includes: organizing the collected equipment feeding execution result data and aquaculture environment feedback data into labeled training samples; inputting the labeled training samples into the collaborative scheduling model for incremental training; updating the network parameters of the collaborative scheduling model; and generating a new feeding equipment scheduling strategy based on the updated network parameters.

[0112] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to implement the steps of the collaborative scheduling optimization method for feeding equipment based on the livestock farming environment provided in any embodiment of this application.

[0113] The computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiments, such as the hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, SmartMedia Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the computer device.

[0114] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for collaborative scheduling optimization of feeding equipment based on livestock farming environment, characterized in that, include: Collect environmental data, livestock growth data, and feeding equipment status data for livestock farming, integrate and process multi-dimensional data, eliminate data silos, and generate basic data for dynamic scheduling; Each feeding device is set as an independent intelligent agent. A collaborative scheduling model is built based on an improved multi-agent reinforcement learning algorithm. The basic data of dynamic scheduling is input into the state space of the model. A multi-objective reward function is constructed based on feeding accuracy, equipment energy consumption, feed utilization rate and livestock and poultry feed intake satisfaction to generate an initial collaborative scheduling strategy for multiple devices. Build a digital twin model of the breeding environment and feeding equipment, simulate and pre-execute the initial collaborative scheduling strategy, identify and avoid operational risks such as equipment path conflict, feed line blockage and feeding time overlap, and output the verified optimized scheduling strategy. The optimized scheduling strategy is sent to the physical feeding equipment for execution. The data on the feeding execution results of the equipment and the feedback data of the breeding environment are continuously collected. The feedback data is used as training samples to continuously iterate and optimize the collaborative scheduling model, so as to realize the dynamic adaptive update of the feeding equipment scheduling strategy.

2. The collaborative scheduling method for livestock feeding equipment based on multi-agent reinforcement learning according to claim 1, characterized in that, The collected environmental data, livestock growth data, and feeding equipment status data for livestock farming include: The system collects environmental data such as temperature, humidity, ammonia concentration, and light intensity using IoT sensors deployed in the pens; and collects livestock and poultry age, weight, historical feed intake, and health status as livestock and poultry growth data. The remaining feed line level, equipment location, fault status, and energy consumption level of the feeding equipment are collected as status data for the feeding equipment.

3. The collaborative scheduling method for livestock feeding equipment based on multi-agent reinforcement learning according to claim 1, characterized in that, The process of fusing multi-dimensional data, eliminating data silos, and generating dynamic scheduling foundational data includes: Data cleaning is performed on the collected multi-dimensional data by filtering outliers and filling in missing values; the cleaned data is then converted into a unified format to complete data standardization; the standardized data is then timestamped according to a unified time base; and the aligned data is then integrated to generate dynamic scheduling base data.

4. The collaborative scheduling method for livestock feeding equipment based on multi-agent reinforcement learning according to claim 1, characterized in that, The step of setting each feeding device as an independent intelligent agent and building a cooperative scheduling model based on an improved multi-agent reinforcement learning algorithm includes: A centralized training and distributed execution architecture is used to build a collaborative scheduling model; an observation space including environmental state, livestock and poultry state and equipment state is defined for each independent agent; an action space including feed amount, running speed and movement path is defined for each independent agent; and the network parameters and training hyperparameters of the model are initialized.

5. The collaborative scheduling method for livestock feeding equipment based on multi-agent reinforcement learning according to claim 4, characterized in that, The process involves inputting dynamic scheduling basic data into the model state space, constructing a multi-objective reward function based on feeding accuracy, equipment energy consumption, feed utilization rate, and livestock feed intake satisfaction, and generating an initial collaborative scheduling strategy for multiple devices, including: The dynamic scheduling base data is input into the observation space of each agent; Calculate the reward values ​​for feeding accuracy, equipment energy consumption, feed utilization, and livestock and poultry feed intake satisfaction, respectively. The weighted sum of each sub-reward value is used to obtain the comprehensive reward value; each agent makes decisions and interacts based on the comprehensive reward value to generate an initial collaborative scheduling strategy for multiple devices.

6. The collaborative scheduling method for livestock feeding equipment based on multi-agent reinforcement learning according to claim 1, characterized in that, The construction of the digital twin model corresponding to the breeding environment and feeding equipment includes: Collect spatial layout parameters of the pen and geometric and physical parameters of each feeding device; A 3D virtual scene with a 1:1 scale to the actual aquaculture environment is constructed based on the collected parameters. The real-time status of each feeding device is mapped to the corresponding virtual device in the three-dimensional virtual scene; Establish a real-time data synchronization channel between physical entities and virtual models.

7. The collaborative scheduling method for livestock feeding equipment based on multi-agent reinforcement learning according to claim 6, characterized in that, The simulation pre-execution of the initial collaborative scheduling strategy identifies and avoids operational risks such as equipment path conflicts, material line congestion, and overlapping material feeding times, and outputs a verified optimized scheduling strategy, including: Input the initial cooperative scheduling strategy into the digital twin model; Drive the virtual feeding device to simulate operation according to the initial cooperative scheduling strategy; Real-time detection of risks such as equipment path conflicts, material line blockages, and overlapping feeding times during simulated operation; Adjust the device operating parameters and execution sequence in the initial collaborative scheduling strategy based on the detected risks; The adjusted strategy is re-input into the digital twin model for verification; the risk detection and strategy adjustment steps are repeated until there is no operational risk during the simulation; the verified optimized scheduling strategy is output.

8. The collaborative scheduling method for livestock feeding equipment based on multi-agent reinforcement learning according to claim 1, characterized in that, The process involves distributing the optimized scheduling strategy to the physical feeding equipment for execution, and continuously collecting data on the equipment's feeding execution results and feedback data from the aquaculture environment, including: The corresponding instructions in the optimized scheduling strategy will be sent to the devices according to the unique address of each entity's feeding device. Control the physical feeding equipment to perform feeding operations according to the received instructions; The actual feed input, runtime, and energy consumption level of each feeding device are collected as data on the feeding execution results. Temperature, humidity, ammonia concentration, and light intensity inside the enclosures are collected as feedback data for the aquaculture environment.

9. The collaborative scheduling method for livestock feeding equipment based on multi-agent reinforcement learning according to claim 1, characterized in that, The step of using feedback data as training samples to continuously iteratively optimize the collaborative scheduling model and achieve dynamic adaptive updates of the feeding equipment scheduling strategy includes: The collected data on equipment feeding results and aquaculture environment feedback were compiled into labeled training samples; The labeled training samples are input into the collaborative scheduling model for incremental training. Update the network parameters of the cooperative scheduling model; A new feeding equipment scheduling strategy is generated based on the updated network parameters.

10. A collaborative scheduling optimization system for feeding equipment based on livestock farming environment, used to implement the method as described in any one of claims 1-9, characterized in that, include: The data generation unit is used to collect environmental data, livestock growth data, and feeding equipment status data of livestock farming, integrate and process multi-dimensional data, eliminate data silos, and generate basic data for dynamic scheduling. The strategy generation unit is used to set each feeding device as an independent intelligent agent, build a cooperative scheduling model based on the improved multi-agent reinforcement learning algorithm, input the dynamic scheduling basic data into the model state space, construct a multi-objective reward function based on feeding accuracy, equipment energy consumption, feed utilization rate and livestock and poultry feed intake satisfaction, and generate the initial cooperative scheduling strategy for multiple devices. The strategy update unit is used to build a digital twin model of the breeding environment and feeding equipment, simulate and pre-execute the initial collaborative scheduling strategy, identify and avoid operational risks such as equipment path conflicts, feed line blockage and feeding time overlap, and output the verified optimized scheduling strategy. The optimized scheduling strategy is sent to the physical feeding equipment for execution. The data on the feeding execution results of the equipment and the feedback data of the breeding environment are continuously collected. The feedback data is used as training samples to continuously iterate and optimize the collaborative scheduling model, so as to realize the dynamic adaptive update of the feeding equipment scheduling strategy.

Images