AI technology-based precise livestock breeding feeding management method and system

Abstract

The invention belongs to the technical field of livestock feeding management, and particularly relates to a livestock breeding feeding precise management method and system based on an AI technology. A fixed feeding mode based on a growth stage and a self-adaptive dynamic correction algorithm based on environment deviation and activity fluctuation are fused; a multi-dimensional data-driven feeding amount decision model is constructed, dependence of a single parameter is broken through, accurate feeding can be performed according to real-time monitoring data by adopting a self-adaptive feeding mode when the environment or activity amount fluctuates and deviates, the nutritional requirements of livestock are better met, and the feeding efficiency is improved. When data acquisition is abnormal or the environment and activity amount are stable and appropriate, a fixed feeding mode is adopted, and the fixed feeding mode can be periodically updated according to a self-adaptive feeding mode, so that the feeding accuracy and efficiency are improved, the feed waste is reduced, the stability of feed feeding is guaranteed, and the feeding efficiency is improved. And the feeding effect can be brought into an optimal state.

Description

Technical Field

[0001] The present invention belongs to the technical field of livestock feeding management, and specifically relates to a livestock breeding and feeding precision management method and system based on AI technology. Background Art

[0002] With the development of animal husbandry modernization, intelligent feeding systems have gradually become a key means to improve breeding efficiency and ensure livestock health. Accurately determining the feeding amount during intelligent feeding is the key to ensuring feeding results.

[0003] A variety of intelligent feeding solutions have been proposed in the prior art. For example, Chinese invention patent publication number CN115039711A proposes a calf feeding method and system. By obtaining the daily milk feed amount based on the age information of the target calf, the calf feeding equipment is controlled to quantitatively feed the target calf into the feeding box.

[0004] While the above-mentioned plan can effectively adapt to the nutritional needs of calves at different growth stages, its main drawback is that it relies solely on the calf's age as the sole basis for feeding, failing to consider the impact of dynamic factors such as ambient temperature and humidity, and livestock activity levels on food intake. This can easily lead to feed amounts deviating from actual needs. For example, when calves' appetites decrease in high temperatures, a fixed feed amount can result in feed waste. Therefore, in complex breeding environments, a feeding plan based solely on age may not meet actual needs, resulting in poor feeding results.

[0005] Another example is an intelligent livestock feeding system and feeding method proposed in Chinese invention patent publication number CN108012941A, which can actively monitor the breeding environment and livestock movement, dynamically judge and adjust the amount of feed required by livestock, and formulate corresponding feeding plans.

[0006] In theory, this approach can more comprehensively reflect the actual feeding needs of livestock, thereby improving feeding accuracy and effectiveness. However, a significant drawback is its heavy reliance on sensor-generated data. If there are malfunctions in the farming environment or livestock movement data collection, the system will be unable to obtain accurate information, resulting in unreliable feeding plans and even the possibility of feeding errors, which could affect the healthy growth of livestock.

[0007] It can be seen that although the existing intelligent feeding system has improved the accuracy and efficiency of feeding to a certain extent when determining the feeding amount, it is dependent on a single factor and cannot provide a reliable and stable feeding plan, which in turn affects the feeding effect to a certain extent. Summary of the Invention

[0008] The purpose of the present invention is to improve the deficiencies in the existing technology and provide a precise management method and system for livestock feeding based on AI technology. By integrating fixed feeding based on growth stage with adaptive feeding based on environment and activity level to construct a multi-dimensional data-driven feeding amount decision-making mechanism, it can provide a reliable and stable feeding plan, effectively solving the problems mentioned in the background technology.

[0009] The purpose of the present invention can be achieved through the following technical solutions: The first aspect of the present invention proposes a precise management method for livestock breeding and feeding based on AI technology, including: Step 1: Constructing a high-precision grid map based on the farm architectural drawings, and planning the feeding path in combination with obstacle identification.

[0010] Step 2: Control the feeding robot to move to the feeding area of ​​the pen according to the planned feeding path.

[0011] Step 3: Collect the environmental data and livestock activity data of the pen in real time through the environmental collection terminal and activity collection terminal set up in the pen.

[0012] Step 4: Identify abnormalities in the operating indicators of the collection end in real time and record the duration of the abnormality. At the same time, perform environmental deviation analysis on the collected environmental data, record the duration of environmental deviation, and perform activity fluctuation analysis on the collected livestock activity data.

[0013] Step 5: Trigger the switch from the adaptive feeding mode to the fixed feeding mode based on at least one of the following conditions:

[0014] a) The duration of the acquisition terminal abnormality exceeds the allowed duration;

[0015] b) The duration of environmental deviation is within the safe time period and the activity fluctuation is within the safe threshold;

[0016] Step 6: After the feeding robot completes the quantitative output of feed according to the selected feeding method, it will scatter the quantitatively output feed into the feeding area of ​​the pen.

[0017] The second aspect of the present invention proposes a precise livestock feeding management system based on AI technology, including: a path planning module, which constructs a high-precision grid map based on the farm architectural drawings, and plans the feeding path in combination with obstacle identification.

[0018] The mobile control module controls the feeding robot to move to the feeding area of ​​the pen according to the planned feeding path.

[0019] The data collection module is installed in the pen and includes an environment collection unit and a livestock activity detection unit.

[0020] Environmental collection unit: collects environmental data of the enclosure in real time through the environmental collection terminal.

[0021] Livestock activity detection unit: An activity collection terminal consisting of an infrared camera and a three-axis accelerometer. The infrared camera tracks the movement trajectory of livestock in the pen and calculates the distance moved per unit time. The three-axis accelerometer monitors the vibration frequency of the tags hanging on the necks of livestock in the pen.

[0022] The mode switching module includes an operation abnormality identification unit, an environment deviation analysis unit, an activity fluctuation analysis unit and a switching execution unit.

[0023] Operation abnormality identification unit: The real-time detection environment collection unit and livestock activity detection unit detect abnormalities in the operation indicators of the corresponding collection end, and record the duration of the abnormality when an abnormality is identified.

[0024] Environmental deviation analysis unit: performs environmental deviation analysis on the collected environmental data and records the duration of environmental deviation.

[0025] Activity fluctuation analysis unit: performs activity fluctuation analysis on the collected livestock activity data.

[0026] Switching execution unit: triggers the switch from the adaptive feeding mode to the fixed feeding mode according to at least one of the following conditions:

[0027] a) The duration of abnormal operation of the acquisition terminal exceeds the allowed duration;

[0028] b) The duration of environmental deviation is within the safe time period and the activity fluctuation is within the safe threshold.

[0029] Feeding control module: After the feeding robot completes the quantitative output of feed according to the selected feeding method, it will scatter the quantitatively output feed into the feeding area of ​​the pen.

[0030] Combining all the above technical solutions, the positive effects of the present invention are as follows: the present invention constructs a multi-dimensional data-driven feeding amount decision model by integrating a fixed feeding pattern based on the growth stage with an adaptive dynamic correction algorithm based on environmental deviation and activity fluctuation, breaking through the dependence on a single parameter. It not only improves the accuracy and efficiency of feeding while reducing feed waste, but also ensures the stability of feed feeding, which is conducive to achieving the best feeding effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The present invention is further described with reference to the accompanying drawings. However, the embodiments in the accompanying drawings do not constitute any limitation to the present invention. A person skilled in the art can obtain other drawings based on the following drawings without creative effort.

[0032] Figure 1This is a diagram of the implementation steps of a precise management method for animal husbandry and feeding based on AI technology in Example 1 of the present invention.

[0033] Figure 2 It is a structural schematic diagram of the feeding robot in the present invention.

[0034] Figure 3 This is a schematic diagram of the connection modules of an AI-based livestock breeding and feeding precision management system in Example 2 of the present invention.

[0035] Figure 4 This is a schematic diagram of the unit structure of the data acquisition module in the present invention.

[0036] Figure 5 Schematic diagram of the unit structure of the mode switching module in the present invention.

[0037] Description of the accompanying drawings: 1—mobile chassis, 2—frame, 3—storage bin, 4—feeding arm. DETAILED DESCRIPTION

[0038] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0039] Example 1

[0040] See also Figure 1 As shown, the present invention proposes a precise management method for animal husbandry and feeding based on AI technology, including: Step 1: constructing a high-precision grid map based on the farm architectural drawings, and planning the feeding path in combination with obstacle identification.

[0041] In the preferred implementation of the above solution, step 1 is specifically implemented as follows: S1. Construct a high-precision grid map based on the farm building drawings, and pre-mark known static obstacles on the map.

[0042] For example, static obstacles can be enclosure walls, feed pool edges, fixed equipment such as water pumps, distribution boxes, etc.

[0043] S2. Assign a unique identification code to each pen, associate it with the feeding cycle, and generate a feeding task priority list.

[0044] It should be noted that livestock feeding is usually arranged periodically at predetermined time intervals. For example, a day can be divided into multiple feeding periods according to the feeding time points as shown in Table 1 to ensure that the livestock receive the required feed at different time periods.

[0045] Table 1: Feeding period in a day

[0046] Feeding period Feeding time 5:00—9:00 9:00 9:00—13:00 13:00 13:00—17:00 17:00 17:00—21:00 21:00 21:00—1:00 1:00 1:00—5:00 5:00

[0047] As shown in Table 1, each feeding period is the time period between one feeding time point and the next, ensuring that livestock receive the feed they need to meet their physiological needs within a specific time period. The feeding time point refers to the specific time at the end of each feeding period, i.e., the actual feeding operation time for energy consumption during that time period.

[0048] However, the division of feeding periods may vary from pen to pen due to factors such as livestock species, growth stage, and health status, which results in different feeding priorities between pens within a day.

[0049] S3. Extract the target pen from the feeding task priority list, where the target pen is the pen to be fed at the current time, locate the position of the target pen, and simultaneously locate the current position of the robot using the positioning terminal built into the robot, and then mark the position on the constructed high-precision grid map.

[0050] S4. Perform initial feeding path planning based on static obstacles between the target enclosure and the robot's current location on the high-precision grid map.

[0051] It's important to note that building a high-precision grid map based on farm architectural drawings and pre-marking static obstacles provides the robot with precise spatial information. This helps avoid known fixed obstacles when planning the initial feeding path, reducing the risk of damage to the robot due to collisions.

[0052] S5. When the robot moves along the initial feeding path, the laser radar installed on the robot collects the point cloud data of the environment in which the robot moves in real time.

[0053] S6. Compare the point cloud data of the moving environment with the obstacle information in the map in real time to identify whether there are new obstacles, such as temporarily parked cleaning vehicles. If a new obstacle is identified, execute S7; otherwise, continue to move along the initial feeding path.

[0054] The above solution can enhance the system's adaptability to complex environments by dynamically identifying new obstacles during the robot's movement.

[0055] S7. Based on the contour extraction of the newly added obstacle, its volume is obtained, and based on this, it is judged whether the newly added obstacle occupies the path safety area, such as occupying 40% of the path width. If it occupies the path safety area, execute S8; otherwise, continue to move along the initial feeding path.

[0056] S8. Trigger local path planning with the current node as the center to form a new feeding path until the robot reaches the target pen.

[0057] The above solution can dynamically adjust the path during movement, ensuring that the robot always moves along the optimal path. Even if it encounters unexpected obstacles, it can ensure the smooth completion of the task. In addition, when an obstacle is sensed, local path planning is used. Compared with global planning, it can reduce computing overhead and ensure the reliability of the robot under resource-constrained conditions.

[0058] Step 2: Control the feeding robot to move to the feeding area of ​​the pen according to the planned feeding path.

[0059] In the improved implementation of the above solution, the present invention uses a feeding robot in the feeding process, please refer to Figure 2 As shown, the feeding robot includes: a mobile chassis 1, which is installed under a frame 2 and is used to achieve autonomous movement according to path planning instructions.

[0060] It's important to understand that the mobile chassis 1 is the mobile foundation of the entire robotic system, responsible for autonomous movement according to path planning instructions. It can flexibly navigate the farm and ensure that feed is accurately delivered to the designated location.

[0061] As an example, a mobile chassis structure can be composed of drive wheels, obstacle avoidance sensors, and a suspension system, where the drive wheels usually adopt a multi-wheel design with independent motors and reducers to achieve precise speed control and steering.

[0062] Obstacle avoidance sensors are laser radars or ultrasonic sensors installed around the mobile chassis 1 to detect obstacles ahead and adjust the path in real time.

[0063] The suspension system adopts an independent suspension system to ensure that the robot can travel smoothly on uneven ground.

[0064] The frame 2 is fixed on the mobile chassis 1 and serves as a supporting base.

[0065] It is important to understand that the frame 2 serves as the supporting base of the entire robot, fixing and connecting all other components.

[0066] The storage bin 3 is fixed on the frame 2 and is used to store feed. It has a built-in feed metering and control device. The feed weight in the bin is monitored by the feed metering device and compared with the target feeding amount determined by the currently selected feeding mode. According to the comparison result, the control device is used to output the feed quantitatively through a closed-loop feedback mechanism.

[0067] It should be understood that the storage bin 3 serves as a feed storage unit for achieving accurate quantitative output of feed.

[0068] As an example, the storage silo structure may consist of a silo body, a weighing sensor, a screw feeder, a flow control valve, and a discharge area.

[0069] The discharge area is used to temporarily store feed that is about to be discharged.

[0070] The screw feeder is located at the bottom of the silo and is driven by a motor to adjust the output speed of the feed. It is responsible for transporting the feed from the silo to the flow control valve.

[0071] The flow control valve is installed at the outlet of the screw feeder and connected to the output area bin. It can adjust the valve opening to control the feed entering the discharge area.

[0072] The weighing sensor is installed at the bottom of the discharge area to monitor the feed weight in the discharge area in real time and feed the data back to the control system. When the detected feed amount reaches the target feeding amount, the system will stop feed output.

[0073] It should be noted that the feed metering device mentioned above includes a weighing sensor and a discharge area, which stores and monitors the output feed volume in real time, and feeds back the data to the control system to determine whether feed needs to be output and how much feed to output.

[0074] The aforementioned control devices include a screw feeder and a flow control valve, which work together to precisely regulate the feed output through a closed-loop feedback mechanism.

[0075] The feeding arm 4 is hinged to the side of the frame 2 and is used to transport the quantitatively output feed from the storage bin 3 to the feeding area of ​​the pen, specifically to transport the feed stored in the discharge area of ​​the storage bin 3 to the feeding area of ​​the pen.

[0076] Step 3: Collect the environmental data and livestock activity data of the pen in real time through the environmental collection terminal and activity collection terminal set up in the pen.

[0077] Specifically, the environment collection end is a multi-source environment sensor, and the activity quantity collection end includes an infrared camera and a three-axis accelerometer.

[0078] It should be added that the present invention is applicable to farms with pens as basic breeding units. Each pen houses a predetermined number of livestock, such as cattle, and is equipped with corresponding environment collection terminals and activity collection terminals to achieve all-round monitoring of the livestock's growth environment, daily activities, and growth and development.

[0079] More specifically, the multi-source environmental sensor is composed of a temperature sensor, a humidity sensor, and a light sensor, and the environmental data obtained are temperature, humidity, and light intensity, respectively.

[0080] It should be noted that the reason for choosing temperature, humidity, and light intensity as environmental data for the pen is that these have a crucial impact on livestock feeding. Specifically, temperature has a significant impact on the metabolic rate and appetite of livestock. Excessively high temperatures can cause heat stress in livestock, reduce appetite, and may even lead to dehydration and heatstroke; while excessively low temperatures increase livestock's energy consumption to maintain body temperature, thereby reducing the energy used for eating.

[0081] Humidity and temperature work together to influence air quality and livestock comfort within pens. High humidity increases the risk of heat stress, especially in hot conditions, leading to a decrease in appetite. Low humidity can dry out the respiratory tract, affecting overall health and indirectly affecting feeding.

[0082] Light intensity significantly affects the behavioral rhythms and physiological states of livestock. Adequate light helps promote the synthesis of vitamin D, strengthen the immune system, and regulate circadian rhythms, which in turn affects livestock's feeding behavior. Unsuitable light conditions can disrupt livestock's circadian clocks, affecting their normal feeding times and amounts.

[0083] The livestock activity data mentioned above include movement distance per unit time and vibration frequency, wherein the infrared camera tracks the movement trajectory of the livestock in the pen and calculates the movement distance per unit time, and the three-axis accelerometer monitors the vibration frequency of the tag hanging on the neck of the livestock in the pen.

[0084] It's important to understand that infrared cameras use image processing and computer vision technology to track livestock movements and calculate distance traveled per unit time. The specific implementation process is as follows: the infrared camera continuously captures a video stream or continuous image frames within the pen. These image frames contain information about the livestock's location at different points in time.

[0085] Image processing algorithms such as background subtraction and contour detection are used to identify the location of livestock from each frame of the image.

[0086] Once the position of the livestock in each frame is determined, a continuous motion trajectory is generated.

[0087] According to the generated motion trajectory, the displacement distance between all frames is accumulated to obtain the moving distance per unit time.

[0088] It's important to note that livestock activity levels significantly influence their feeding behavior. When livestock activity levels increase, their energy expenditure increases, which in turn increases hunger and prompts them to eat more to replenish their energy. Conversely, when livestock activity levels decrease, their energy expenditure decreases, and their food needs also decrease, as they don't need to eat to replenish the large amounts of energy they consume.

[0089] It should be noted that the reason for using the movement distance per unit time and vibration frequency as activity indicators when detecting livestock activity is that it can reflect livestock activity data from both macro and micro levels, which is more comprehensive. The movement distance per unit time reflects the activity intensity of livestock from a macro level. A higher movement distance usually indicates that the livestock is more active, and the vibration frequency captures the subtle movement information of the livestock.

[0090] It should be emphasized that if there are multiple livestock in a pen, the activity level of each livestock needs to be tested separately.

[0091] Step 4: Identify abnormalities in the operating indicators of the collection end in real time and record the duration of the abnormality. At the same time, analyze the environmental deviation of the collected environmental data and record the duration of the environmental deviation. Analyze the activity fluctuation of the collected livestock activity data.

[0092] In one feasible way of implementing the above steps, the abnormal identification of the operating indicators of the collection end refers to the following operation: obtain the current operating indicators from the environment collection end and the activity collection end in real time, and compare them with the normal operating indicator range. If an operating indicator of a certain collection end exceeds the normal operating indicator range, the operation of the collection end is identified as abnormal.

[0093] It's important to note that since the data acquisition terminal typically consists of electronic components, its operating indicators include electrical operating indicators such as voltage and current, as well as performance indicators such as response time and sensor accuracy. The data acquisition terminal may experience abnormal operation due to unstable power supply or device failure. The normal operating indicator range can be obtained from the data acquisition terminal's user manual or technical specifications.

[0094] It should be noted that abnormal operation of the acquisition terminal not only affects its own functions, but may also directly lead to distortion of the environmental data it collects. For example, in the event of voltage fluctuations or sensor failure, the acquisition terminal may output incorrect temperature and humidity values, thereby triggering a false environmental deviation judgment. Therefore, before performing environmental deviation analysis, it is necessary to ensure that the acquisition terminal is in normal operation. Once an operational abnormality is detected in the acquisition terminal, the environmental data collected during this period will be considered unreliable. To ensure the continuity of environmental monitoring, valid data from adjacent time periods can be used to interpolate the environmental parameter values ​​at the time of the current operational abnormality, and then the interpolated environmental data can be used to perform environmental deviation analysis.

[0095] Another feasible method of the above steps is to perform environmental deviation analysis on the collected environmental data and record the environmental deviation duration as follows: the real-time collected environmental data is compared with the preset suitable environment to calculate the percentage absolute deviation of each environmental parameter.

[0096] The percentage absolute deviation formula mentioned above is , where Indicates the environmental data collected in real time. Indicates the preset suitable environment.

[0097] It should be emphasized that since livestock have strict range requirements for their eating environment, which cannot be too large or too small, absolute deviation is used in environmental deviation analysis to calculate the degree of deviation of the real-time collected environmental data from the preset suitable environment.

[0098] It is important to know that the preset suitable environment refers to the environmental conditions that are pre-set according to the type and growth stage of livestock and are conducive to the healthy eating of livestock.

[0099] Based on historical feeding records, a correlation model between food intake and each environmental parameter was constructed, and the correlation coefficient between food intake and each environmental parameter was extracted through the correlation model as the weight factor of the environmental parameter.

[0100] In the manner in which the above scheme can be implemented, the association model between food intake and each environmental parameter is constructed as follows: a weight sensor is set below the feeding area (such as a feeding pool) of each pen to measure the weight of feed before and after feeding in real time during the feeding period, and the food intake during the feeding period is obtained by calculating the weight difference.

[0101] The amount of food consumed during each feeding period is combined with environmental data and livestock activity data collected in the pen to form a feeding record.

[0102] From all feeding records, select records with similar livestock activity levels as backup feeding records.

[0103] As a preferred embodiment, an activity deviation threshold can be set to compare the livestock activity in each record, and feeding records with activity deviations within the set threshold can be screened out as records with similar livestock activity.

[0104] The environmental data and food intake in the backup feeding records were used to apply relevant algorithms to construct a correlation model between food intake and each environmental parameter.

[0105] It should be noted that when establishing the correlation model between food intake and environmental parameters, feeding records from animals with similar activity levels were selected as data support. This is because establishing a correlation model based on data with similar activity levels can avoid interference introduced by differences in livestock activity levels, thereby improving the accuracy and reliability of the model.

[0106] It should be further added that the feeding records mentioned above all come from the same pen, and the association model between food intake and environmental parameters established based on these records is also specially customized for the pen, that is, each pen can establish an association model between food intake and environmental parameters.

[0107] Various machine learning models can be used to construct correlation models, such as random forests, support vector machines, and neural networks. Through training, these models can be used to extract correlation coefficients between food intake and each environmental parameter. These correlation coefficients reflect the degree of association between each environmental parameter and food intake, representing the degree of influence of each environmental parameter on food intake.

[0108] In a specific example implementation, the correlation coefficient is obtained based on the constructed association model as follows: the environmental data in the backup feeding record is used as the independent variable, the food intake is used as the dependent variable, and the Pearson correlation coefficient formula is used to calculate the correlation coefficient between each environmental parameter and the food intake.

[0109] Because the Pearson correlation coefficient ranges from -1 to 1, its absolute value reflects the strength of the linear association between variables. Specifically, the closer the absolute value is to 1, the stronger the correlation between the environmental parameter and food intake; the closer the absolute value is to 0, the weaker the linear correlation between the two. When using the correlation coefficient to quantify the impact of environmental parameters on food intake, it is important to focus on the strength of the association rather than the direction of the correlation. Therefore, the absolute value of the correlation coefficient should be used as the weight factor for the environmental parameter to reflect the weight of the parameter's impact on food intake.

[0110] The environmental deviation is compared with the set warning value, where the warning value is initially set by the system. When the environmental deviation at a certain moment reaches the warning value, the environmental deviation duration is recorded.

[0111] In another feasible manner of the above steps, the collected livestock activity data is subjected to activity fluctuation analysis as follows: based on the current feeding period, activity data of the same feeding period is extracted from historical feeding records and historical mean values ​​are calculated.

[0112] The activity fluctuation was calculated by comparing the unit time movement distance and vibration frequency collected during the current feeding period with the corresponding historical averages.

[0113] For example, the activity fluctuation calculation formula is: , where Indicates activity fluctuations. 、 Respectively represent the moving distance per unit time and the vibration frequency, 、 They respectively represent the historical mean of the distance moved per unit time and the historical mean of the vibration frequency during the same feeding period.

[0114] It should be explained that when analyzing the fluctuation of livestock activity, the detected livestock activity is compared with the livestock activity during the same historical feeding period rather than with the appropriate activity. This is because there are differences in the physiological state, health status and behavioral patterns among different livestock individuals, and the appropriate activity level is difficult to determine uniformly. Suitable environmental parameters are relatively easy to determine through experiments and literature, and have high repeatability and consistency. By comparing the current livestock activity with the activity during the same historical feeding period, it is possible to identify whether there are abnormal fluctuations in the current activity. This method uses historical data as a benchmark, which can better reflect the normal range of changes in actual operations and avoid uncertainties caused by individual differences and external factors.

[0115] It is further explained that the above activity fluctuation formula 、 They respectively reflect the deviation of the moving distance per unit time and the vibration frequency relative to the historical mean. The overall activity fluctuation is obtained by combining these two deviations using the square root of the sum of squares.

[0116] Step 5: Trigger the switch from the adaptive feeding mode to the fixed feeding mode based on at least one of the following conditions:

[0117] a) The duration of the abnormality at the collection end exceeds the allowed duration;

[0118] b) The duration of environmental deviation is within the safe time period and the activity fluctuation is within the safe threshold;

[0119] It should be understood that the implementation process of feeding mode switching is: when feeding is needed, if the collection end in the current feeding period continues to operate abnormally or the environment deviation is within the safe time and the activity fluctuation is within the safe threshold when the collection end is operating normally, a fixed feeding mode can be used to improve feeding efficiency. On the contrary, if the collection end operates normally but the environment deviation is large or the activity fluctuation is large, in this case the fixed feeding mode may not meet the actual feeding needs. At this time, the adaptive feeding mode can be used to accurately feed according to real-time monitoring data to better meet the nutritional needs of livestock.

[0120] From this, it can be seen that the fixed feeding mode is a fallback strategy when the collection end has an operational abnormality or operates normally with a stable and appropriate environment and activity level. It can ensure the continuity and stability of feeding, simplify the operating process, and provide emergency response capabilities in emergencies.

[0121] It should be noted that the above-mentioned allowed time, safe time and safety threshold are all system initial settings.

[0122] Among them, the fixed feeding mode: the benchmark feeding amount is obtained based on the preset livestock growth stage and feeding amount association table. These benchmark feeding amounts are determined based on scientific feeding standards and historical data. The association table is periodically updated according to the feeding record data of the adaptive feeding mode to ensure its accuracy and effectiveness.

[0123] Specifically, the fixed feeding mode is implemented as follows: the livestock growth cycle is divided into multiple growth stages, each growth stage is associated with a reference feeding amount, and a livestock growth stage and feeding amount association table is formed.

[0124] Illustratively, the growth stages mentioned above may be the cub stage, the growth stage, the fattening stage, and the like.

[0125] It should be noted that the benchmark feeding amount mentioned above refers to the amount of feed that a single livestock eats in a single feeding period.

[0126] The three-dimensional morphological data of livestock is monitored in real time through monitoring terminals installed in the pen. The three-dimensional morphological data includes but is not limited to parameters such as body shape, body length, body height, and body width. The current growth stage of the livestock is determined by the pre-established mapping relationship between the three-dimensional morphology of the livestock and the growth stage.

[0127] As a specific implementation of the above solution, the mapping relationship between the three-dimensional morphology of livestock and growth stages can be set according to the standards of the animal husbandry industry. Taking beef cattle as an example, the "NY / T 815-2004 Beef Cattle Feeding Standard" can be referred to. This standard clearly defines the growth stage division method based on morphological parameters such as body length and body height. For example, if the body length is And tall , it is determined to be the growth period.

[0128] On this basis, the feeding amount correlation table between livestock growth stages and benchmark feeding amounts can also be constructed based on the feeding recommendation data in this standard.

[0129] The current baseline feeding amount is obtained by combining the current growth stage of the livestock with the number of livestock in the pen and the livestock growth stage and feeding amount association table.

[0130] Adaptive feeding mode: Dynamic feeding amount is calculated by combining environmental deviation status, activity fluctuation and historical feeding records.

[0131] Specifically, the adaptive feeding mode is implemented as follows: while analyzing the degree of environmental deviation, the deviation environment parameters and their corresponding direction of environmental deviation are marked to form an environmental deviation information set, wherein the environmental deviation information set includes the degree of environmental deviation, the duration of environmental deviation, the deviation environment parameters and the direction of environmental deviation.

[0132] It needs to be explained that the direction of environmental deviation means whether the deviation environmental parameters are higher or lower than the appropriate environmental parameters.

[0133] While analyzing the activity volume fluctuation, the fluctuating activity volume indicator and its corresponding fluctuation direction are marked to form an activity volume fluctuation information set, wherein the activity volume fluctuation information set includes the activity volume fluctuation, the fluctuating activity volume indicator and the fluctuation direction.

[0134] It needs to be explained that the direction of fluctuation means whether the fluctuation activity indicator is higher or lower than the historical average.

[0135] Based on the environmental deviation information set and the activity fluctuation information set, historical feeding records under the same conditions are selected from the historical feeding records as reference feeding records.

[0136] It should be noted that when selecting reference feeding records from historical feeding records, it is best to select historical feeding records from the same feeding period to ensure the accuracy and reliability of the reference data. For example, assuming the current feeding period is 9:00-13:00, the environmental deviation information set and activity fluctuation information set composed of the current feeding period are compared with the environmental data and activity data of historical feeding records from the same feeding period. The historical feeding records whose environmental data and activity data both meet the environmental deviation information set and activity fluctuation information set are selected as reference feeding records.

[0137] It should be noted that the historical feeding records in which the environmental data and activity data are screened as reference feeding records actually refer to the historical feeding records with completely consistent environmental deviations and activity fluctuations from the records of the same historical feeding period, which are matched based on the environmental deviation information set and activity fluctuation information set collected during the current feeding period as matching conditions.

[0138] It's important to emphasize that when using the adaptive feeding mode, the system uses the environmental deviation information set and activity level fluctuation information set for the current feeding period to filter out reference feeding records with the same conditions from historical feeding records. The environmental and activity levels corresponding to these filtered feeding records are highly consistent with those under the current adaptive feeding mode. Furthermore, the system can use the livestock's food intake from these filtered feeding records as a reference for the current feeding amount, significantly improving the accuracy and effectiveness of feeding, ensuring that the feeding amount is more closely aligned with the livestock's actual needs, and enhancing the system's adaptability and flexibility.

[0139] The coefficient of variation of the livestock feed intake in all reference feeding records is calculated, where the coefficient of variation is the ratio of the mean to the standard deviation, which reflects and is compared with the configured critical value. If the coefficient of variation does not reach the critical value, it indicates that the feed intake in the reference feeding records is relatively stable. In this case, the mean of the livestock feed intake in all reference feeding records is taken as the dynamic feeding amount. Otherwise, it indicates that there is a large variability in the feed intake in the reference feeding records. In this case, the median of the livestock feed intake in all reference feeding records is taken as the dynamic feeding amount.

[0140] The association table is periodically updated according to the feeding record data of the adaptive feeding mode as follows: a feeding record is formed after feeding is implemented according to the dynamic feeding amount, and the actual food intake is extracted therefrom.

[0141] Calculate the deviation between the actual food intake and the dynamic feeding amount, and compare it with the set allowable deviation. By comparison, if the deviation meets the allowable deviation, the dynamic feeding amount is considered to be more accurate, and the dynamic feeding amount is used as the corrected benchmark feeding amount. If the deviation does not meet the allowable deviation, the dynamic feeding amount is considered to be not accurate enough, and the actual food intake is used as the corrected benchmark feeding amount.

[0142] In the example implementation of the above solution, the allowable deviation can be determined based on a statistical analysis of the distribution of deviations between the amount fed and the actual amount consumed in a large number of historical feeding records. By analyzing the central trend of deviations in the historical data, a representative deviation range can be identified, thereby setting a reasonable allowable deviation threshold.

[0143] Specifically, the deviation between the actual food intake and the given feeding amount in each historical feeding record can be clustered and analyzed. By classifying historical feeding records with similar deviation characteristics into the same category, the data distribution pattern under different deviation modes can be identified. Then, among all clustering results, the deviation corresponding to the main cluster with the largest number of samples is selected as the set allowable deviation.

[0144] The environmental deviation information set and the activity level fluctuation information set are used as constraints for revising the baseline feeding amount to ensure that the revised baseline feeding amount can better adapt to the current environment and activity status.

[0145] The revised baseline feeding amount and its constraint conditions are stored in the associated table as preliminary update data for the baseline feeding amount.

[0146] It should be noted that the correction and update of the above-mentioned benchmark feeding amount does not use the prepared updated benchmark feeding amount in all cases. It is mainly applicable in the following situations: the adaptive feeding mode is started in the current feeding period, and the system calculates the deviation between the actual food intake and the dynamic feeding amount, and judges that it meets the allowable deviation. At this time, the dynamic feeding amount is used as the corrected benchmark feeding amount. When the collection terminal operation is abnormal and the fixed feeding mode needs to be started, the system will compare the environmental information and activity level before the collection terminal operation is abnormal with the environmental deviation information set and the activity level fluctuation information set. If this information is consistent with the environmental deviation information set and the activity level fluctuation information set, the corrected benchmark feeding amount will be used as the benchmark feeding amount in the fixed feeding mode. In this way, the dynamic update of the benchmark feeding amount in the fixed feeding mode is realized, and the accuracy and effectiveness of feed delivery can be ensured even when adaptive feeding is not possible.

[0147] If the reason for starting the fixed feeding mode is that the collection terminal is operating normally and the environment has not deviated or the activity level has not fluctuated, then in this case there is no need to use the forecast updated benchmark feeding amount. The benchmark feeding amount in the association table can be used directly for feeding. This can simplify the operating process and ensure feeding efficiency.

[0148] It's important to note that the present invention doesn't factor livestock health into its feeding decisions. While health does affect livestock's feeding behavior, it's not considered in this invention. Specifically, factors influencing livestock's feeding are limited to the environment and activity level, not their health.

[0149] Step 6: After the feeding robot completes the quantitative output of feed according to the selected feeding method, it will scatter the quantitatively output feed into the feeding area of ​​the pen.

[0150] Example 2

[0151] See also Figure 3 As shown, the present invention proposes a precise livestock feeding management system based on AI technology, including: a path planning module, which constructs a high-precision grid map based on the farm architectural drawings, and plans the feeding path in combination with obstacle identification.

[0152] The mobile control module controls the feeding robot to move to the feeding area of ​​the pen according to the planned feeding path.

[0153] See also Figure 4 As shown, the data acquisition module is arranged in the pen and includes an environment acquisition unit and a livestock activity detection unit.

[0154] Environmental collection unit: collects environmental data of the enclosure in real time through the environmental collection terminal.

[0155] Livestock activity detection unit: An activity collection terminal consisting of an infrared camera and a three-axis accelerometer. The infrared camera tracks the movement trajectory of livestock in the pen and calculates the distance moved per unit time. The three-axis accelerometer monitors the vibration frequency of the tags hanging on the necks of livestock in the pen.

[0156] See also Figure 5 As shown, the mode switching module includes an operation abnormality identification unit, an environment deviation analysis unit, an activity fluctuation analysis unit and a switching execution unit.

[0157] Operation abnormality identification unit: The real-time detection environment collection unit and livestock activity detection unit detect abnormalities in the operation indicators of the corresponding collection end, and record the duration of the abnormality when an abnormality is identified.

[0158] Environmental deviation analysis unit: performs environmental deviation analysis on the collected environmental data and records the duration of environmental deviation.

[0159] Activity fluctuation analysis unit: performs activity fluctuation analysis on the collected livestock activity data.

[0160] Switching execution unit: triggers the switch from the adaptive feeding mode to the fixed feeding mode according to at least one of the following conditions:

[0161] a) The duration of abnormal operation of the acquisition terminal exceeds the allowed duration;

[0162] b) The duration of environmental deviation is within the safe time period and the activity fluctuation is within the safe threshold;

[0163] Feeding control module: After the feeding robot completes the quantitative output of feed according to the selected feeding method, it will scatter the quantitatively output feed into the feeding area of ​​the pen.

[0164] The parameters involved in the above formula are all dimensionless and calculated numerically. The formula is a formula obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0165] The above embodiments may be implemented in whole or in part through software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0166] Those skilled in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0167] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0168] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0169] Finally, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A precise management method for animal husbandry and feeding based on AI technology, characterized in that: include: Step 1: Build a high-precision grid map based on the farm architectural drawings and plan the feeding path by combining obstacle identification; Step 2: Control the feeding robot to move to the feeding area of ​​the pen according to the planned feeding path; Step 3: Collect the environmental data and livestock activity data of the pen in real time through the environmental collection terminal and activity collection terminal set up in the pen; Step 4: Identify abnormalities in the operating indicators of the collection end in real time and record the duration of the abnormality. At the same time, analyze the environmental deviation of the collected environmental data and record the duration of the environmental deviation. Analyze the activity fluctuation of the collected livestock activity data. Step 5: Trigger the switch from the adaptive feeding mode to the fixed feeding mode based on at least one of the following conditions: a) The duration of the acquisition terminal abnormality exceeds the allowed duration; b) The duration of environmental deviation is within the safe time period and the activity fluctuation is within the safe threshold; Step 6: After the feeding robot completes the quantitative output of feed according to the selected feeding method, it will scatter the quantitatively output feed into the feeding area of ​​the pen.

2. The AI-based precise management method for animal husbandry and feeding according to claim 1, characterized in that: The step 1 is specifically implemented as follows: S1. Build a high-precision grid map based on the farm architectural drawings and pre-mark known static obstacles on the map; S2. Assign a unique identification code to each pen and generate a priority list of feeding tasks associated with the feeding cycle; S3 extracts the target pen from the feeding task priority list and locates the target pen. At the same time, the robot's current position is located by the positioning terminal built into the robot, and then the position is marked on the constructed high-precision grid map; S4. Initial feeding path planning based on static obstacles between the target enclosure and the robot's current position on the high-precision grid map; S5. The robot moves along the initial feeding path by collecting real-time point cloud data of the environment in which the robot is moving through the laser radar installed on the robot; S6. Compare the point cloud data of the moving environment with the obstacle information in the map in real time to identify whether there are new obstacles. If a new obstacle is identified, execute S7; otherwise, continue moving according to the initial feeding path; S7. Based on the contour extraction of the newly added obstacle, its volume is obtained, and based on this, it is judged whether the newly added obstacle occupies the path safety area. If it occupies the path safety area, then S8 is executed. Otherwise, the initial feeding path is continued; S8. Trigger local path planning with the current node as the center to form a new feeding path until the robot reaches the target pen.

3. The AI-based precise management method for animal husbandry and feeding according to claim 1, characterized in that: For abnormal identification of the operating indicators of the acquisition end, please refer to the following operations: The current operating indicators are obtained from the environment collection end and the activity collection end in real time, and compared with the normal operating indicator range. If an operating indicator of a collection end exceeds the normal operating indicator range, the collection end is identified as operating abnormally.

4. The method for precise management of animal husbandry and feeding based on AI technology according to claim 1, characterized in that: The process of analyzing the environmental deviation of the collected environmental data and recording the duration of the environmental deviation is as follows: Compare the real-time collected environmental data with the preset suitable environment to calculate the percentage absolute deviation of each environmental parameter; Based on historical feeding records, a correlation model between food intake and each environmental parameter was constructed, and the correlation coefficient between food intake and each environmental parameter was extracted through the correlation model as the weight factor of the environmental parameter; The environmental deviation is calculated by taking the weighted average of the percentage absolute deviation of each environmental parameter and the corresponding weight factor, and compared with the set warning value. When the environmental deviation at a certain moment reaches the warning value, the duration of the environmental deviation is recorded.

5. The method for precise management of animal husbandry and feeding based on AI technology according to claim 4, characterized in that: The process of building a correlation model between food intake and each environmental parameter based on historical feeding records is as follows: A weight sensor is installed below the feeding area of ​​each pen to measure the weight of feed before and after feeding in real time. The food intake during the feeding period is obtained by calculating the weight difference. The amount of food consumed during each feeding period is combined with environmental data and activity data collected in the pen to form a feeding record; Select records with similar activity levels from all feeding records as backup feeding records; The environmental data and food intake in the backup feeding records were used to apply relevant algorithms to construct a correlation model between food intake and each environmental parameter.

6. The AI-based precise management method for animal husbandry and feeding according to claim 1, characterized in that: The activity fluctuation analysis of the collected livestock activity data is performed as follows: Based on the current feeding period, activity data of the same feeding period is extracted from the historical feeding records and the historical mean is calculated; The activity fluctuation was calculated by comparing the unit time movement distance and vibration frequency collected during the current feeding period with the corresponding historical averages.

7. The method for precise management of animal husbandry and feeding based on AI technology according to claim 1, characterized in that: The fixed feeding mode is to obtain a reference feeding amount based on a preset livestock growth stage and feeding amount association table, wherein the association table is periodically updated according to the feeding record data of the adaptive feeding mode, wherein the fixed feeding mode is specifically implemented as follows: Divide the livestock growth cycle into multiple growth stages, associate each growth stage with a baseline feeding amount, and form a correlation table between livestock growth stages and feeding amounts; The three-dimensional morphological data of livestock is monitored in real time through the monitoring terminal in the pen, and the current growth stage of the livestock is determined by using the preset mapping relationship between the three-dimensional morphology of livestock and growth stage; The current baseline feeding amount is obtained from the association table based on the number of livestock in the pen and the current growth stage of the livestock.

8. The method for precise animal husbandry and feeding management based on AI technology according to claim 1, characterized in that: The adaptive feeding mode is to calculate the dynamic feeding amount by combining the environmental deviation state, activity fluctuation and historical feeding records. The adaptive feeding mode is specifically implemented as follows: While analyzing the environmental deviation, the deviated environmental parameters and their corresponding environmental deviation directions are marked to form an environmental deviation information set; While analyzing the activity fluctuation, the fluctuating activity indicators and their corresponding fluctuation directions are marked to form an activity fluctuation information set; Based on the environmental deviation information set and the activity fluctuation information set, historical feeding records under the same conditions are selected from the historical feeding records as reference feeding records; The coefficient of variation of the livestock feed intake in all reference feeding records is calculated and compared with the configured critical value. If the coefficient of variation does not reach the critical value, the mean of the livestock feed intake in all reference feeding records is taken as the dynamic feeding amount. Otherwise, the median of the livestock feed intake in all reference feeding records is taken as the dynamic feeding amount.

9. The method for precise management of animal husbandry and feeding based on AI technology according to claim 7, characterized in that: The association table is periodically updated according to the feeding record data of the adaptive feeding mode as follows: After feeding according to the dynamic feeding amount, a feeding record is formed and the actual food intake is extracted from it; Calculate the deviation between the actual food intake and the dynamic feeding amount, and compare it with the set allowable deviation. If the deviation meets the allowable deviation, the dynamic feeding amount will be used as the corrected reference feeding amount. If the deviation does not meet the allowable deviation, the actual food intake will be used as the corrected reference feeding amount. The environmental deviation information set and the activity fluctuation information set are used as constraints for correcting the baseline feeding amount; The revised baseline feeding amount and its constraint conditions are stored in the associated table as preliminary update data for the baseline feeding amount.

10. A precise livestock breeding and feeding management system based on AI technology, characterized in that: include: The path planning module builds a high-precision grid map based on the farm architectural drawings and plans the feeding path in combination with obstacle identification; The mobile control module controls the feeding robot to move to the feeding area of ​​the pen according to the planned feeding path; The data collection module is installed in the pen and includes an environment collection unit and a livestock activity detection unit; Environmental collection unit: collects environmental data of the enclosure in real time through the environmental collection terminal; Livestock activity detection unit: This includes an activity acquisition terminal consisting of an infrared camera and a three-axis accelerometer. The infrared camera tracks the movement of livestock in the pen and calculates the distance traveled per unit time. The three-axis accelerometer monitors the vibration frequency of the tags hanging on the necks of livestock in the pen. A mode switching module, comprising an operation anomaly identification unit, an environment deviation analysis unit, an activity fluctuation analysis unit, and a switching execution unit; Operation abnormality identification unit: The real-time detection environment collection unit and livestock activity detection unit detect abnormalities in the operation indicators of the corresponding collection end, and record the duration of the abnormality when an abnormality is identified; Environmental deviation analysis unit: performs environmental deviation analysis on the collected environmental data and records the duration of environmental deviation; Activity fluctuation analysis unit: analyzes the activity fluctuation of the collected livestock activity data; Switching execution unit: triggers the switch from the adaptive feeding mode to the fixed feeding mode according to at least one of the following conditions: a) The duration of abnormal operation of the acquisition terminal exceeds the allowed duration; b) The duration of environmental deviation is within the safe time period and the activity fluctuation is within the safe threshold; Feeding control module: After the feeding robot completes the quantitative output of feed according to the selected feeding method, it will scatter the quantitatively output feed into the feeding area of ​​the pen.

Images