Pig house robot health monitoring and early warning system for pig raising platform

Abstract

The invention belongs to the technical field of intelligent breeding and robots, and particularly relates to a pig house robot health monitoring and early warning system for a pig raising platform. The system aims to solve the problem of early warning lag caused by discontinuity of existing monitoring data. The system comprises a static sentinel node subsystem which is used for continuously collecting data at high frequency, generating an abnormal score and sending a triage request; the mobile diagnosis and service unit subsystem is used for performing autonomous diagnosis after receiving the request, acquiring individual accurate data and generating a report; and the central control and prediction subsystem is used for fusing data, constructing a space-time atlas model, predicting epidemic disease transmission and generating prospective early warning. By means of the system, 7 * 24-hour uninterrupted monitoring of swinery is achieved, data faults are solved, common screening-fine diagnosis cooperation is formed, a space-time diagram neural network is used for providing prospective early warning, and an epidemic disease prevention and control gateway is remarkably advanced.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent farming and robotics technology, and specifically relates to a pig house robot health monitoring and early warning system for a pig farming platform. Background Art

[0002] The present invention belongs to the field of animal husbandry automation and intelligent monitoring technology, and specifically, relates to an animal health status monitoring and early warning system used in large-scale breeding environments. In modern animal husbandry, especially in the intensive and large-scale pig breeding industry, efficient and accurate disease prevention and control and individual health management are the core links to ensure industry stability and improve economic benefits. The traditional breeding model relies heavily on manual inspections to observe the health status of pigs. This method is not only labor-intensive and inefficient, but the observation results are easily affected by the experience level and subjective judgment of the breeders, making it difficult to achieve accurate and timely detection of early-stage, sporadic patients in large-scale groups.

[0003] To address these challenges, intelligent monitoring solutions based on automated equipment have been proposed and implemented. Specifically, these solutions typically utilize a piggery inspection robot equipped with multimodal sensors as a data collection platform. For example, by integrating an infrared thermal imager and a sound acquisition module into the robot, it can autonomously patrol the piggery according to a pre-set route and schedule (e.g., twice daily). During the inspection process, the infrared thermal imager non-contactly captures each pig's surface temperature distribution, providing an objective basis for screening for abnormal body temperatures. Simultaneously, the sound acquisition module records ambient sounds, particularly abnormal respiratory sounds such as coughing and wheezing. The collected multidimensional raw data is transmitted to a back-end data processing center, where deep learning models (such as the convolutional neural network-based ResNet model) intelligently analyze the thermal imaging data and audio signals to automatically identify individuals with temperatures exceeding a preset threshold (e.g., 39.5°C) or with abnormal respiratory rates. Once a pig with abnormal health is identified, the system generates an analysis report, combines it with its historical health records for comprehensive analysis, and ultimately sends an early warning message to the farm manager via the data platform. Compared to traditional manual inspections, this technical solution significantly improves the automation and objectivity of monitoring, to a certain extent solving the problem of delayed disease detection and representing a key direction for the development of intelligent farming.

[0004] However, with the continued expansion of livestock farming and increasingly stringent requirements for disease control, a deep-seated contradiction inherent in the aforementioned monitoring system, based on a scheduled inspection model, has gradually emerged. This stems from the fact that the core of this technical paradigm is based on a "discrete time point snapshot" data collection model, while the occurrence and spread of swine diseases is essentially a "continuous dynamic evolution." This fundamental mismatch between the temporal dimension of this monitoring model and the monitored objects creates an insurmountable bottleneck in the timeliness and foresight of the system's early warnings. Further analysis reveals that pigs experience an incubation period and subclinical phase from infection with a pathogen to the onset of overt clinical symptoms (such as a significant increase in body temperature or severe coughing). During the fixed interval of only twice-daily inspections, one or a few pigs may have already entered the early stages of infection, with subtle but indicative fluctuations in their body temperature or behavior. However, due to the discontinuous nature of inspections, these valuable early signals are highly likely to be missed. By the time the robot collects data at its next scheduled inspection point, the symptoms it detects may already be clear and delayed. By then, the virus may have already spread within the pen. The system's "warning" effectively becomes a "confirmation" of a fait accompli, missing the optimal window for precise intervention in the early stages of an epidemic. Furthermore, this discrete data collection method means that while the intelligent algorithms behind the system are powerful, they rely on sparse, isolated data points, making it impossible to construct a continuous health profile for individual pigs or even the entire herd. While the algorithm can determine whether a pig has a fever at a given moment, it struggles to dynamically model and predict rising temperature trends or the frequency of abnormal behaviors based on a continuous data stream, preventing the quantum leap from "diagnosis" to "prediction."

[0005] Therefore, how to design a method that can break through the limitations of the timed and discrete data collection mode and achieve continuous and high-frequency capture of the health status of pig herds in the breeding environment and accurate prediction of the dynamic evolution trend, so as to issue a truly forward-looking warning in the earliest stage of the occurrence of the epidemic, has become a key challenge and a technical problem that needs to be solved urgently by technical personnel in this field. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to overcome the discontinuous data collection defect of the pig house health monitoring system based on the timed and discrete inspection mode in the existing technology. This defect causes the system to be unable to capture the continuous dynamic evolution process of the pig health status, thereby causing serious delays in disease warning and missing the best time to intervene in the early stage of the disease spread.

[0007] In order to solve the technical problem of discontinuous monitoring of the health status of pig herds resulting in delayed warning in the above-mentioned background technology, the present invention provides a pig house robot health monitoring and early warning system for a pig farming platform.

[0008] The technical solution adopted by the present invention is: a pig house robot health monitoring and early warning system for pig farming platforms, characterized in that the system includes: A static sentinel node subsystem, consisting of at least one static sentinel node deployed in each independent pig pen. Each static sentinel node is used to continuously and frequently collect data on the environment and pig activities within its pen. Based on a preset end-side computing model, the collected raw data is analyzed in real time to generate a pen anomaly score that represents the overall abnormal state of the pen. When the pen anomaly score continuously exceeds a dynamically adjusted trigger threshold within a preset time window, the static sentinel node generates and sends a triage request containing the pen's unique identifier; A mobile diagnosis and service unit subsystem, which is a piggery inspection robot with autonomous navigation and refined operation capabilities. The mobile diagnosis and service unit subsystem is configured to execute two mutually switching working modes, namely baseline inspection mode and on-demand triage mode. In baseline inspection mode, the mobile diagnosis and service unit subsystem conducts inspections throughout the piggery according to a preset low-frequency scheduling strategy to obtain baseline health data of individual pigs and the environment. In on-demand triage mode, upon receiving a triage request sent by the static sentinel node subsystem, the mobile diagnosis and service unit subsystem immediately interrupts the current baseline inspection task, autonomously navigates to the pen specified by the triage request, and initiates a targeted, high-precision individual diagnostic protocol for the pigs in the pen to obtain accurate physiological and pathological data of the target individual pigs and generate a structured individual diagnostic report. And a central control and prediction subsystem, which conducts two-way data communication with the static sentinel node subsystem and the mobile diagnosis and service unit subsystem. The central control and prediction subsystem is used to receive and integrate the data stream of continuous pen abnormality scores sent by the static sentinel node subsystem and the discrete high-precision individual diagnosis report sent by the mobile diagnosis and service unit subsystem; the central control and prediction subsystem is further used to build and update a spatiotemporal map model of the health status of the pig herd in real time based on the received data, and use the model to predict the potential disease transmission path and future health status evolution trend in the pig herd, and generate forward-looking disease warning information.

[0009] As a preferred embodiment of the present invention, the static sentinel node is an integrated, low-power electronic device whose hardware components include a central microcontroller unit, a wireless communication transceiver module, a power supply module, and a sensor array. The sensor array specifically includes: a three-axis microelectromechanical system (MEMS) accelerometer with a sampling frequency set to no less than 4096Hz and a measurement range of ±16g, which is used to continuously capture weak vibration signals of the pen structure caused by pigs coughing, panting, and abnormal limb movements; a temperature and humidity composite sensor with a temperature measurement accuracy of ±0.2°C and a humidity measurement accuracy of ±2%RH, which is used to monitor microenvironmental changes within the pen; and an 8x8 pixel array infrared thermopile sensor with a 60-degree field of view and a temperature measurement range of 20°C to 100°C, which is used to obtain a low-resolution thermal map of the overall surface temperature distribution of the pigs in the pen. The static sentinel node is enclosed in an IP68-rated, corrosion-resistant, and gnaw-proof housing made of injection-molded polycarbonate and covered with a 304 stainless steel mesh.

[0010] Furthermore, the end-side computing model embedded within the static sentinel node is specifically a pen anomaly score calculation module. The pen anomaly score calculation module operates as follows: It performs a fast Fourier transform (FFT) on the raw vibration data collected by the MEMS accelerometer within continuous 1-second time windows, converting the time-domain signal into a frequency-domain signal. Within the frequency domain, the module integrates the energy within the preset characteristic frequency bands associated with pig coughing (50Hz to 200Hz) and panting (400Hz to 800Hz) to obtain a vibration anomaly index. It calculates the mean and standard deviation of the temperature output of the 64 pixels from the infrared thermopile sensor and compares this with a dynamic baseline temperature issued by the central control and prediction subsystem to obtain a thermal map anomaly index. The vibration anomaly index and the thermal map anomaly index are weighted and summed, with the weight coefficients dynamically configured by the central control and prediction subsystem based on historical data and the pig herd's daily activity rhythm. The summation results in the current pen anomaly score. The pen abnormality score is a dimensionless normalized value ranging from 0 to 1.

[0011] As another embodiment of the present invention, wireless communication between the static sentinel node subsystem and the central control and prediction subsystem utilizes the LoRaWAN protocol. The static sentinel node subsystem serves as the terminal device, while a LoRaWAN gateway is deployed on the central control and prediction subsystem side. When the moving average of a static sentinel node's pen anomaly score remains above the dynamically adjusted trigger threshold set by the central control and prediction subsystem for that pen for 300 consecutive seconds, the static sentinel node is triggered and an uplink data packet is constructed. The uplink data packet is a structured data frame whose payload contains a fixed 16-bit pen unique identifier, a 32-bit timestamp, an 8-bit trigger cause code (indicating that the anomaly is primarily due to vibration or thermal imaging), and a 16-bit floating-point number representing the current anomaly score. This data packet constitutes a triage request and is sent to the central control and prediction subsystem via the LoRaWAN network.

[0012] Specifically, the hardware platform of the mobile diagnosis and service unit subsystem includes a four-wheel differential drive chassis, an inertial measurement unit (IMU) and a battery management system. Its core mission payloads include: a 3D laser radar (LiDAR) for autonomous navigation and three-dimensional environment perception, with a scanning frequency of 15Hz and a ranging accuracy of ±2cm; a high-resolution infrared thermal imager, whose detector is an uncooled vanadium oxide (VOx) microbolometer with a resolution of 640x512 pixels, a thermal sensitivity (NETD) of less than 40mK, and a temperature measurement accuracy of ±2cm; The accuracy is ±0.3°C after calibration; a planar circular microphone array consisting of 8 MEMS microphones with a spacing of 10 cm, mounted on a two-dimensional pan-tilt head, is used to locate the sound source; a long-range high-frequency RFID reader that complies with the ISO11784 / 11785 standard and operates at a frequency of 134.2kHz, is equipped with a directional antenna; and a service end effector mounted on the end of a six-axis collaborative robotic arm, which integrates a wireless charging coil and a high-pressure cleaning nozzle for energy replenishment and lens cleaning of static sentinel nodes.

[0013] Furthermore, the individual diagnosis protocol executed by the mobile diagnosis and service unit subsystem is a deterministic procedure consisting of multiple steps. After receiving the dispatch instruction forwarded by the central control and prediction subsystem containing the target pen identifier, the protocol is activated: the first step is path planning and arrival. The unit uses 3D lidar data and pre-stored high-precision maps to plan and execute an optimal collision-free path to reach the entrance of the target pen; the second step is individual segmentation and locking. After the unit enters the pen, it uses 3D lidar to scan the space inside the pen, and combines its RGB-D depth camera to perform three-dimensional point cloud instance segmentation on all pigs in the pen. At the same time, the microphone array is started to locate the sound source of the abnormal acoustic signal (such as coughing) indicated in the triage request. By fusing the sound source localization vector and the three-dimensional point cloud segmentation result, the individual pig that makes the abnormal sound is locked as the primary diagnostic target; the third step is non-contact vital signs precise measurement. The robotic arm adjusts its posture and accurately positions the high-resolution infrared thermal imager. The unit focuses on the core area of ​​the head and torso of the targeted individual, collects a continuous 5-second video stream, and uses an RFID reader to read the unique identification code in the ear tag of the individual; the fourth step is data analysis and report generation. The unit's onboard computing unit (an embedded computer with an integrated graphics processing unit) processes the collected infrared video stream, extracts multiple quantitative indicators such as the maximum body surface temperature, eye corner temperature, and uneven distribution of body surface temperature, and combines the read individual identification code to generate a structured individual diagnosis report containing multi-dimensional information such as individual ID, diagnosis time, maximum body surface temperature, thermal map snapshot, and acoustic characteristics; the fifth step is data reporting. The unit sends this report to the central control and prediction subsystem via a 5G or Wi-Fi6 wireless network, and then continues to execute a simplified diagnostic process for other individuals in the pen or decides on the next action based on new instructions.

[0014] A core technical feature of this invention is the spatiotemporal graph model of pig health status within the central control and prediction subsystem, specifically a spatiotemporal graph neural network (ST-GNN). The model's construction and operation mechanism is as follows: During the model construction phase, the system abstracts the entire pig farm into a dynamic graph structure G = (V, E, X). V is the set of nodes in the graph, with each node v_i∈V uniquely corresponding to a pig wearing an RFID ear tag. E is the set of edges in the graph, with an edge e_ij∈E indicating a potential disease transmission link between nodes v_i and v_j. The initial weight of this link is set based on the physical pen adjacency of the pigs (e.g., shared stalls, shared aisles) and the fixed air circulation simulation path. X is the node feature matrix, and Xi(t) is the feature vector of node v_i at time t. This feature vector contains historical health data for the pig, such as age, weight, vaccination history, and other static information. During the real-time model update phase, the data fusion module of the central control and prediction subsystem performs the following operations: It continuously receives a data stream of pen anomaly scores from all static sentinel nodes. When the anomaly score S_k(t) for pen k increases, the system increases the weight of the edges e_ij connecting all nodes v_i∈V_k within pen k. At the same time, S_k(t) is added to the feature vectors X_i(t) of all nodes within that pen, acting as a time-varying pen-level environmental feature. Upon receiving an individual diagnostic report for pig v_i generated by the mobile diagnostic and service unit subsystem, the system directly updates the node's latest feature vector X_i(t) with the precise physiological and pathological data (such as precise body temperature T_i and cough frequency C_i) from the report, using this data as the node's true state label at that moment. Finally, during the model prediction phase, the ST-GNN model performs a forward propagation at a fixed time step (e.g., every 10 minutes). The model propagates node information spatially through graph convolution operations, aggregating the health status of neighboring nodes. Simultaneously, a gated recurrent unit (GRU) or long short-term memory (LSTM) network models the feature sequence of each node in the temporal dimension to capture the evolving trends in its health status. The model output is a prediction matrix P(t+Δt), where P_i(t+Δt) is a vector containing a series of predictive indicators, such as the probability that pig v_i's body temperature will exceed the clinical threshold within the next Δt period (for example, the next 12 hours) and the probability of severe coughing.

[0015] Furthermore, the central control and prediction subsystem includes a dynamic threshold and scheduling decision module. This module calculates and distributes a private, time-varying trigger threshold for each static sentinel node. Dynamic threshold calculations are based on the statistical distribution of the pen's historical anomaly scores, the overall activity rhythm of the pig herd (e.g., determining whether the current period is feeding, sleeping, or peak activity based on time), and global environmental parameters (e.g., overall pen temperature). By dynamically adjusting the threshold, the system can appropriately raise the threshold during active periods to reduce false alarms and lower the threshold during quiet periods to increase sensitivity. Furthermore, when multiple triage requests arrive simultaneously, the scheduling decision module prioritizes tasks within the mobile diagnosis and service unit subsystem based on the requested anomaly scores, the value of the pigs within the pen (e.g., gilts or finishing pigs), and the ST-GNN model's predicted disease spread risk level for the pen, generating an optimal scheduling sequence.

[0016] The beneficial effects of the present invention are: First, by deploying a static sentinel node subsystem, the present invention achieves 7x24 hours of uninterrupted, high-frequency monitoring of pig health-related indicators, fundamentally solving the data gap problem caused by the traditional discrete inspection mode, and ensuring that any weak, early group abnormal signals can be captured in time.

[0017] Second, this invention establishes a two-tiered diagnostic model, combining "general screening and intensive diagnosis," with static sentinel nodes working in tandem with mobile robots. The static nodes are responsible for low-cost, wide-coverage abnormality detection, while the mobile robots, acting as high-precision diagnostic resources, are dispatched on-demand and precisely. This significantly improves the utilization of scarce diagnostic resources and achieves an optimal balance between overall system performance and cost.

[0018] Third, by introducing a predictive model based on a spatiotemporal graph neural network, this invention transforms monitoring data from isolated "snapshots" into a continuous, interconnected "spatiotemporal data stream." The system no longer merely identifies and alerts to symptoms that have already occurred; instead, by learning the dynamics of disease transmission within pig populations, it can quantitatively predict and proactively warn of future health risks. This provides unprecedented decision-making support for livestock managers to implement ultra-early intervention measures and disrupt disease transmission chains, significantly shifting the focus of disease prevention and control. Fourth, the system architecture of this invention is highly scalable and adaptable. A dynamic threshold adjustment mechanism enables the system to adapt to the activity patterns of pigs at different growth stages, reducing interference from environmental changes. The modular design facilitates future integration of new sensors (such as gas sensors to monitor ammonia concentrations) or more advanced predictive algorithms, ensuring the system's long-term technological advancement. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1It is a block diagram of the overall architecture of the system of the present invention.

[0020] Figure 2 It is a structural diagram of a single static sentinel node in the static sentinel node subsystem of the present invention.

[0021] Figure 3 It is a structural diagram of the mobile diagnosis and service unit subsystem in the present invention.

[0022] Figure 4 It is a schematic diagram of the workflow of the system in the present invention.

[0023] Figure 5 It is a schematic diagram of the spatiotemporal atlas model of the pig herd health status in the present invention.

[0024] The accompanying drawings are numbered as follows: 10. Static sentinel node subsystem; 11. Central microcontroller unit; 12. Wireless communication transceiver module; 13. Power supply module; 14. Sensor array; 15. Corrosion-resistant and anti-bite shell; 20. Mobile diagnosis and service unit subsystem; 21. Four-wheel differential drive chassis; 22. 3D laser radar; 23. High-resolution infrared thermal imager; 24. Planar circular microphone array; 25. Long-range high-frequency RFID reader / writer; 26. Six-axis collaborative robotic arm; 30. Central control and prediction subsystem. DETAILED DESCRIPTION

[0025] To make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0026] See also Figure 1 The figure shows the overall architecture of a piggery robotic health monitoring and early warning system for pig farming platforms, as provided by the present invention. This system aims to overcome the data discontinuity issues associated with traditional discrete inspection models. Through an innovative multi-level collaborative mechanism, it enables continuous monitoring, accurate diagnosis, and forward-looking predictions of pig health status. The system is logically divided into three core subsystems that collaborate with each other and enable bidirectional data flow: a static sentinel node subsystem 10, a mobile diagnosis and service unit subsystem 20, and a central control and prediction subsystem 30.

[0027] Specifically, the static sentinel node subsystem 10 is the cornerstone of the entire monitoring network. Its physical entity is composed of a large number of static sentinel nodes strategically deployed in each independent pig pen. Each pen is equipped with at least one such node, and its core task is to conduct uninterrupted, high-sampling-frequency data collection on the microenvironment and pig activity status within its jurisdiction. These nodes are not simple data forwarding units, but intelligent sensor terminals with end-side computing capabilities. Based on the internal solidified computing model, they perform real-time preliminary analysis and feature extraction on the massive amount of raw data collected. Their output is not the original data stream, but a highly concentrated quantitative indicator that characterizes the overall health risk of the pen, namely the pen abnormality score. When this score continues to be higher than a dynamically set trigger threshold within a preset time window, the static sentinel node will actively generate and send a triage request to the upper layer of the system. The triage request is a structured data packet, and its payload must contain the unique identifier of the pen, thereby accurately indicating the geographical location where the abnormal event occurred.

[0028] The Mobile Diagnosis and Service Unit (MDSU) subsystem 20 physically manifests as a piggery inspection robot with highly autonomous navigation and precise operational capabilities. The design concept of the MDSU subsystem 20 is to centralize and mobilize high-value, high-precision diagnostic resources, and dispatch them on demand based on the intelligence of the static sentinel node subsystem 10. Therefore, the MDSU subsystem 20 is configured to dynamically switch between two distinct operating modes: baseline inspection mode and on-demand triage mode. In baseline inspection mode, the robot follows a pre-set, low-frequency global scheduling strategy, performing routine inspections within the common passageways of the entire piggery. Its primary purpose is to collect data for constructing and calibrating a global health baseline, such as updating a high-precision three-dimensional map, recording ambient background noise levels in different areas, and sampling the normal activity patterns of the pigs when not in an abnormal state. In on-demand triage mode, upon receiving a triage request from the static sentinel node subsystem 10, the MDSU subsystem 20 immediately suspends its current baseline task and prioritizes the triage request as the highest priority. Using its autonomous navigation capabilities, it quickly and precisely navigates to the designated pen indicated by the triage request. Upon arrival, it transitions from a broad scan to a highly targeted, structured, and highly accurate individual diagnostic protocol. This protocol is designed to collect in-depth physiological and pathological data on specific pigs within the pen, ultimately generating a structured, multi-dimensional individual diagnostic report, which is then uploaded.

[0029] The central control and prediction subsystem 30 is the "brain" of the entire system. It maintains continuous two-way data exchange with the aforementioned static sentinel node subsystem 10 and mobile diagnosis and service unit subsystem 20 via a wireless communication network. The core function of the static sentinel node subsystem 10 lies in data fusion and in-depth analysis. On the one hand, it continuously receives and aggregates the continuous data stream of pen abnormality scores from all static sentinel nodes; on the other hand, it receives discrete but highly valuable individual diagnostic reports sent by the mobile diagnosis and service unit subsystem 20 after completing its mission. The key innovation of the central control and prediction subsystem 30 is that it does not process this data in isolation, but instead utilizes these heterogeneous data sources to construct and maintain a real-time spatiotemporal map model of the pig herd's health status. The spatiotemporal graph model of pig herd health status abstracts the health status of the entire pig herd into a dynamically evolving network. By analyzing the changes in node status and the mutual influence between nodes in the network, the system is able to identify potential disease transmission paths and make probabilistic predictions on the evolution trend of the pig herd's health status within a specific time window in the future. Ultimately, it generates forward-looking disease warning information, thereby providing decision-making support for managers.

[0030] As a preferred embodiment of the present invention, please refer to Figure 2 , which shows the internal structure of the static sentinel node subsystem 10 in more detail. The static sentinel node subsystem 10 is designed as a highly integrated electronic device optimized for low-power operation. Its hardware core is a central microcontroller unit 11, such as a microcontroller based on the ARM Cortex-M4 core, which is responsible for performing all data processing, logic control and communication tasks within the node. Connected to the central microcontroller unit 11 is a wireless communication transceiver module 12, which is used to realize the data link with the upper system. At the same time, a power supply module 13 provides energy for the entire node. The power supply module 13 can be designed to be powered by a long-life lithium thionyl chloride battery, or integrated with an energy harvesting circuit to cooperate with the wireless charging function of the mobile diagnostic unit.

[0031] The core functionality of the static sentinel node subsystem 10 lies in its integrated sensor array 14. Carefully selected and configured, this sensor array 14 captures multimodal signals most relevant to swine respiratory illnesses and stress in a non-contact, non-invasive manner. Specifically, the sensor array 14 includes a three-axis microelectromechanical system (MEMS) accelerometer with strictly defined technical specifications. The sampling frequency is set to no less than 4096Hz to ensure full waveform capture of weak vibration signals from structures such as the pen fence, floor, or trough, caused by high-frequency events like coughing or wheezing. Its measurement range is set to ±16g to withstand large-amplitude vibrations, such as pig impacts, without saturation. Furthermore, the array includes a combined temperature and humidity sensor with a temperature measurement accuracy better than ±0.2°C and a relative humidity measurement accuracy better than ±2%RH. This allows for precise monitoring of temperature and humidity changes within the pen microenvironment, which are closely related to swine comfort and the occurrence of certain diseases. Furthermore, the sensor array 14 also incorporates an 8x8 pixel infrared thermopile sensor. This sensor provides low-resolution thermal imaging with a 60-degree field of view, covering key areas in most standard pens. Its temperature measurement range, from 20°C to 100°C, is sufficient to monitor both normal and abnormal body temperatures in pigs experiencing fever. While the sensor does not pursue high resolution, its purpose is to capture overall surface temperature trends within the pen, serving as an early indicator of group fever.

[0032] To ensure long-term stable operation in the harsh environment of a piggery, characterized by high humidity, high corrosion, and the risk of animal gnawing, the entire static sentinel node is enclosed in a rugged, corrosion- and gnaw-resistant housing 15 with an IP68 protection rating. The main structure of this housing 15 is precision-injected polycarbonate (PC) for excellent waterproof sealing and impact resistance. An additional layer of woven 304 stainless steel mesh is applied to the exterior, particularly in areas potentially accessible to pigs. This mesh effectively resists gnawing and abrasion, protecting the sensor window and housing from physical damage.

[0033] Furthermore, the end-side computing model embedded within the central microcontroller unit 11 of the static sentinel node subsystem 10 is specifically implemented as a pen anomaly score calculation module. The introduction of this pen anomaly score calculation module aims to convert raw, high-bandwidth sensor data into low-bandwidth, high-information-density feature indicators directly at the data source, significantly reducing wireless communication bandwidth requirements and overall node power consumption. The internal execution flow of the pen anomaly score calculation module is a deterministic algorithm sequence. The module performs a fast Fourier transform (FFT) on 4096 raw vibration data points collected by the MEMS accelerometer within each consecutive 1-second time window, converting the complex time-domain vibration signal into a more easily analyzable frequency domain signal. In the frequency domain spectrum, the module focuses on two pre-defined characteristic frequency bands that are highly correlated with common respiratory symptoms in pigs: one from 50 Hz to 200 Hz, whose energy changes primarily correspond to coughing events; the other from 400 Hz to 800 Hz, whose energy changes are associated with wheezing symptoms. The module numerically integrates the energy within these two frequency bands and normalizes the result to produce a quantitative vibration anomaly index. The module processes temperature readings from 64 pixels of the 8x8 infrared thermopile sensor, calculating the arithmetic mean and standard deviation of these 64 data points. The mean reflects the average surface temperature within the field of view, while the standard deviation indicates the dispersion or unevenness of the temperature distribution. The module compares these two calculated values ​​with a dynamic baseline temperature for the pen, regularly issued by the central control and prediction subsystem 30 (for example, the expected average temperature of the pen's healthy pigs at the current time), to produce a thermal map anomaly index. In this crucial step, the module performs a weighted summation of the aforementioned vibration anomaly index and the thermal map anomaly index. The weighting coefficients are not fixed but are dynamically calculated and remotely configured by the central control and prediction subsystem 30 based on historical data for the pen, the age of the pigs, and the current daily activity rhythm (such as eating, sleeping, and peak activity). For example, when pigs are sleeping, the weight of the vibration index is increased to increase sensitivity to nighttime coughing. The final summation result is normalized to form a dimensionless floating-point number between 0 and 1, which is the current pen abnormality score.

[0034] As another embodiment of the present invention, the wireless data link between the static sentinel node subsystem 10 and the central control and prediction subsystem 30 preferably utilizes the LoRaWAN (Long Range Wide Area Network) protocol. This choice is based on a combination of considerations: the complex piggery environment, wide coverage, and extremely stringent node power consumption requirements. In this architecture, each static sentinel node serves as an end device in the LoRaWAN network, while the central control and prediction subsystem 30 deploys one or more LoRaWAN gateways, responsible for receiving data sent by all nodes within the wide area. The mechanism for triggering triage requests is designed to be both robust and sensitive. Specifically, a static sentinel node is considered triggered only when the stall anomaly score calculated by its internal algorithm remains above a specific trigger threshold over a continuous 300-second (5-minute) window. This trigger threshold is not a global static value, but rather a private dynamic threshold calculated and issued independently by the central control and prediction subsystem 30 for each stall, enabling adaptive adaptation to the characteristics of different pig herds. Once the trigger conditions are met, the static sentinel node immediately constructs a LoRaWAN uplink data packet. This packet is a highly compact structured data frame, with a carefully designed payload that carries the most critical information in the smallest possible number of bytes. Its fixed format includes: a 16-bit unique identifier for precise location; a 32-bit Unix timestamp recording the time of the event; an 8-bit trigger cause code indicating whether the increase in the anomaly score is primarily due to vibration or heatmap metrics, providing preliminary clues for subsequent detailed diagnosis; and a 16-bit compressed floating-point value representing the current anomaly score. This uplink data packet is sent across the LoRaWAN network, passing through the gateway and ultimately arriving at the central control and prediction subsystem 30. It itself constitutes the aforementioned triage request.

[0035] Specifically, see Figure 3 , which shows the hardware platform composition of the mobile diagnosis and service unit subsystem 20. In addition to conventional mobile robot platform components, such as a four-wheel differential drive chassis 21 for flexible movement and on-the-spot turning, an inertial measurement unit (IMU) for posture estimation, and a battery management system (BMS) responsible for energy distribution and monitoring, its core value lies in the set of highly specialized mission payloads it carries. This payload includes: a 3D laser radar 22 for high-precision autonomous navigation and three-dimensional environmental perception, such as a 16-line laser radar with a scanning frequency set to 15Hz, which can build a point cloud map of the surrounding environment in real time during the robot's movement. Its ranging accuracy reaches ±2cm, which is crucial for collision-free path planning in narrow pig house passages.

[0036] The mission payload also includes a high-resolution infrared thermal imager 23. Its detector uses an uncooled vanadium oxide (VOx) microbolometer with a physical resolution of 640x512 pixels, providing detailed thermal images. Its thermal sensitivity (NETD) is less than 40mK, meaning it can discern extremely small temperature differences, which is crucial for detecting early signs of fever. After rigorous blackbody calibration, its absolute temperature accuracy within the target temperature range reaches ±0.3°C, sufficient to meet veterinary diagnostic requirements.

[0037] To accurately locate abnormal sounds, the mobile diagnosis and service unit subsystem 20 is equipped with a planar circular microphone array 24. This array consists of eight high-signal-to-noise ratio MEMS microphones, evenly spaced (for example, 10 cm) along the circumference of a plane. The array is mounted on a two-dimensional pan-tilt platform capable of both tilt and horizontal rotation. By processing the minute time difference (TDOA) between the signals from the same sound source received by the eight microphones, the system can calculate the precise azimuth and elevation angles of the sound source.

[0038] To uniquely identify each individual pig, the mobile diagnostic and service unit subsystem 20 also integrates a long-range, high-frequency RFID reader / writer 25 that complies with the ISO11784 / 11785 standards. Operating at a frequency of 134.2 kHz and coupled with a directional antenna with high gain and good directivity, the long-range, high-frequency RFID reader / writer 25 can reliably read the unique identification code in the RFID electronic ear tag implanted in the pig's ear from a distance (e.g., 0.5 to 1 meter).

[0039] Furthermore, to enable more complex interactive and service functions, some of the aforementioned diagnostic sensors are mounted on the end of a six-axis collaborative robotic arm 26. This six-axis collaborative robotic arm 26 offers a high degree of flexibility, enabling precise movement and alignment of sensors to target locations. A wireless charging coil and a high-pressure cleaning nozzle are further integrated into its end effector. The wireless charging coil provides contactless energy replenishment for low-battery static sentinel nodes during the robot's inspection process, while the high-pressure cleaning nozzle uses cleaning fluid to clean sensor windows at static sentinel nodes or the robot's own sensors that are obscured by dirt, ensuring the system's long-term autonomous operation.

[0040] Furthermore, the individual diagnosis protocol executed by the mobile diagnosis and service unit subsystem 20 after entering the on-demand triage mode is a strict deterministic procedure consisting of multiple logical steps. Figure 4As shown in the workflow diagram, the individual diagnosis protocol is officially activated after the central control and prediction subsystem 30 forwards a dispatch instruction containing the target pen identifier. The first step is path planning and arrival. The robot uses the real-time point cloud data collected by its onboard 3D lidar 22 to match its location with the internally stored high-precision map of the pig pen. It then invokes path planning algorithms such as A or D to plan a collision-free, time-optimized path from its current position to the target pen entrance and autonomously executes the optimal path. The second step is individual segmentation and locking. Once the robot enters the target pen, it first uses the 3D lidar 22 to scan the interior of the pen one or more times. Simultaneously, it uses its onboard RGB-D depth camera to obtain 3D point cloud data and color images of all pigs within the pen. Using point cloud clustering and instance segmentation algorithms, the system is able to segment each pig as an independent entity in three-dimensional space. Simultaneously, it activates the planar circular microphone array 24 to continuously monitor and analyze ambient sound. Once an acoustic signal matching the abnormality type indicated in the triage request (such as a typical cough) is captured, the sound source localization algorithm immediately calculates the direction of the sound's origin. By spatially fusing and aligning this sound source localization vector with the three-dimensional point cloud segmentation results, the system can pinpoint the specific pig emitting the abnormal sound with a high probability, identifying it as the primary diagnostic target. The third step involves precise, non-contact measurement of vital signs. Once the target individual is located, the six-axis collaborative robotic arm 26 immediately adjusts its posture and precisely focuses its high-resolution infrared thermal imager 23 on the core areas of the individual's head and torso, specifically the canthus, which more accurately reflects core body temperature. This captures an infrared video stream for at least five seconds. While capturing thermal images, the robotic arm also adjusts its posture to align its long-range, high-frequency RFID reader / writer 25 with the individual, reading the unique identification code embedded in the ear tag and ultimately confirming the individual's identity. The fourth step involves data analysis and report generation. The robot's onboard computing unit (typically an embedded computer with a graphics processing unit (GPU), such as the NVIDIA Jetson series) processes the captured high-resolution infrared video stream in real time. Algorithms automatically extract multiple quantitative biometrics, such as the maximum body surface temperature during the 5-second observation period, the average temperature in the canthus region, and the unevenness (standard deviation) of the body surface temperature distribution in the torso. These metrics are integrated with the individual's ID number, diagnosis time, and acoustic feature analysis results (such as cough frequency and spectral characteristics) to generate a standardized, machine-readable, structured individual diagnosis report. The fifth step is data reporting. After the diagnosis and report are generated, the robot uploads this complete individual diagnosis report in real time to the central control and prediction subsystem via its onboard high-speed wireless communication module (such as 5G or Wi-Fi 6).30After completing the report, the robot will decide whether to continue to perform a simplified diagnostic process on other individuals in the pen (for example, only conduct a temperature check) or end the triage task and return to the baseline inspection mode based on the new instructions issued by the central system.

[0041] As a core technical feature of the present invention, the spatiotemporal graph model of the pig herd health status deployed in the central control and prediction subsystem 30 is implemented by using a spatiotemporal graph neural network (ST-GNN). Figure 5 The concept of the spatiotemporal model of pig herd health status is schematically illustrated in the figure. The construction and operation mechanism of the spatiotemporal model of pig herd health status reflects the advancement of pig herd health management from an individual, isolated perspective to a group, interconnected, and systematic perspective.

[0042] During the model initialization phase, the system abstracts all pigs on the farm into a dynamic graph structure G = (V, E, X). In this structure, V represents the set of nodes in the graph, where each node v_i∈V uniquely corresponds to a pig wearing an RFID ear tag. E represents the set of edges in the graph, where the presence of an edge e_ij∈E indicates a potential disease transmission link between node v_i (pig i) and node v_j (pig j). The initial weight of this link is not set arbitrarily but is quantified based on the physical proximity of the pigs (e.g., sharing the same stall, upstream and downstream of the same airflow path, sharing the same sewage duct, etc.) and the aerosol diffusion paths within the piggery simulated using computational fluid dynamics (CFD). X represents the node feature matrix, where Xi(t) is an eigenvector describing the state of node v_i at time t. This eigenvector not only contains static or slowly changing information about the individual, such as age, breed, weight, and historical vaccination records, but also reserves space for dynamic health indicators.

[0043] The data fusion module of the central control and prediction subsystem 30 plays a key role in the real-time model update phase. The data fusion module performs the following operations: It continuously receives a data stream of pen anomaly scores from all static sentinel nodes within the pig farm. When a significant and sustained increase in the anomaly score S_k(t) for a particular pen k is detected, the system performs two updates: First, the weights of the edges e_ij connecting all pig nodes (v_i, v_j∈V_k) within pen k in the dynamic graph model are enhanced, reflecting the sharp increase in the risk of internal transmission within the pen due to group anomalies; second, the anomaly score S_k(t) is added as a time-varying, shared environmental feature to the feature vector X_i(t) of all pig nodes within the pen. At the same time, when the system receives a high-precision individual diagnosis report for a specific pig v_i generated by the mobile diagnosis and service unit subsystem 20, it directly and with high priority updates the precise physiological and pathological data contained in the report (for example, precise body temperature T_i = 40.8°C, coughing frequency C_i = 3 times / minute, etc.) to the latest feature vector X_i(t) of that node. More importantly, this precise diagnosis report is considered the "ground truth" of the node at that moment and will be used for subsequent model training or calibration.

[0044] Finally, during the model prediction phase, this constructed and updated ST-GNN model performs a full forward propagation calculation at a fixed time step (for example, every 10 minutes). In each calculation, the model first propagates and aggregates node information in the spatial dimension through graph convolution operations such as graph convolutional networks (GCNs) or graph attention networks (GATs). This means that each node's feature vector incorporates the health status information of all its neighboring nodes (i.e., pigs associated with transmission risk). Next, the model models each node's own feature vector sequence in the temporal dimension using a gated recurrent unit (GRU) or long short-term memory network (LSTM) unit to capture the inherent trends and patterns in the evolution of its health status over time. Through this joint spatiotemporal modeling, the final output of the model is a prediction matrix P(t+Δt), where each element P_i(t+Δt) of the matrix is ​​a prediction vector, which contains a series of forward-looking quantitative indicators, such as the probability of a specific health event occurring in the pig v_i within a preset time window Δt in the future (for example, the next 12 hours or 24 hours), such as the probability of body temperature exceeding the clinical fever threshold (such as 40.5°C), the probability of severe coughing symptoms, and the comprehensive probability of being judged as a high-risk infection individual.

[0045] Furthermore, to make the entire system more intelligent and efficient, the central control and prediction subsystem 30 also integrates a dynamic threshold and scheduling decision module. The scheduling decision module has two core functions. The first is to calculate and issue a private, time-varying trigger threshold for each static sentinel node. The calculation of the dynamic threshold is a multi-factor decision process based on: the statistical distribution of the pen's historical anomaly scores (mean, variance, percentiles, etc.) over the past few weeks or months; the overall activity rhythm of the pigs inferred from information provided by the farm production management system. For example, the system knows that 8:00 a.m. to 9:00 a.m. is peak feeding time, when pig activity and noise levels are naturally high. Therefore, the trigger threshold is appropriately raised during this period to reduce false alarms. During the deep sleep period from midnight to 4:00 a.m., the threshold is significantly lowered to maximize sensitivity to detecting subtle anomalies; and global environmental parameters obtained from the pig house environmental control system, such as the average temperature and ventilation rate of the entire pig house, which also affect the pigs' baseline behavior. Through this adaptive dynamic threshold adjustment, the system can effectively separate true abnormal signals from normal physiological and environmental fluctuations.

[0046] The second core function of the scheduling decision module is to intelligently prioritize and prioritize triage requests from multiple pens when the system simultaneously receives them. This prioritized sequence is then generated, instructing the mobile diagnosis and service unit subsystem 20 to execute tasks according to this optimal sequence. This prioritization decision is also not based on a single metric, but rather on a comprehensive evaluation of multiple factors: first, the magnitude and duration of the anomaly score carried by the request; second, the economic value of the pigs within the pen, as determined by the backend database (for example, a pen of gilts awaiting calving clearly has a higher diagnostic priority than a pen of fattening pigs about to be slaughtered); and, most crucially, the ST-GNN model's predicted risk level for future disease spread for that pen. A pen with a medium anomaly score, but predicted by the model to be a source of high risk, may receive a higher diagnostic priority than a pen with a high anomaly score but predicted to be an isolated incident.

[0047] Example 1 This embodiment describes the specific application of the present invention in a modern pig farm comprising 20 separate farrowing pens, each housing one sow and ten suckling piglets.

[0048] The system is deployed as follows: A static sentinel node is installed on the wall of each farrowing pen approximately 1.5 meters above the trough. The entire farrowing area is covered by a mobile diagnostic and service unit subsystem. The central control and prediction subsystem is deployed in the farm's central control room and communicates with the 20 static sentinel nodes via a LoRaWAN network and with the mobile diagnostic unit via a Wi-Fi 6 network.

[0049] At 2:30 a.m. one morning, the pigs were generally at rest. Based on historical rhythm data, the central control and prediction subsystem had lowered the trigger threshold of static sentinel nodes in all farrowing pens from 0.65 during the day to 0.40.

[0050] The MEMS accelerometer inside the static sentinel node in pen No. 12 began detecting weak but rhythmic, atypical vibrations emanating from the pen's metal fencing. The client-side computing module performed FFT analysis on the atypical vibration signal and detected a persistent peak in the 50Hz-200Hz frequency band, with energy levels exceeding 3 standard deviations above the normal background noise. Based on this, the module calculated a vibration anomaly index of 0.75. Simultaneously, data from its 8x8 infrared thermopile sensor showed that the temperature of five of the 64 pixels was consistently slightly higher than that of surrounding pixels, causing the thermal map anomaly index to rise from a typical 0.1 to 0.35. After weighted summation (weighting nighttime vibration at 0.7 and the thermal map at 0.3), the instantaneous pen anomaly score for this node reached 0.63. Over the following 300 seconds, the sliding average of this score stabilized at 0.55, consistently exceeding the dynamically adjusted trigger threshold of 0.40.

[0051] At 2:35 AM, the static sentinel node in pen 12 automatically constructed and sent a triage request. The packet content was: (pen ID: 0x000C, timestamp: 1678822500, trigger reason: 0x01 (primarily vibration), anomaly score: 0.55).

[0052] The central control and prediction subsystem receives a triage request. If there are no other requests, the system immediately issues a dispatch instruction to the mobile diagnosis and service unit subsystem on standby at the charging station.

[0053] After receiving the command, the mobile diagnostic unit autonomously planned a route and arrived at the entrance of pen 12 within two minutes. Once inside, it initiated the individual segmentation and lock-on procedure. The microphone array quickly located the continuous coughing sounds of a piglet in the corner of the pen, and the sound source localization vector pointed to the target. Simultaneously, a 3D lidar 22 and a depth camera scanned the pen, segmenting the piglet from its sow and littermates into a separate point cloud instance. After fusing the sound source and point cloud data, the target was uniquely locked.

[0054] The robotic arm aimed a high-resolution infrared thermal imager 23 at the piglet's head, capturing a five-second video stream. The onboard computer analyzed the piglet's maximum body temperature, which was 40.8°C, and the temperature at the corner of its eye was 40.6°C. Simultaneously, the RFID reader successfully read the piglet's ear tag ID as "3201-12-07."

[0055] At 2:41 a.m., the mobile diagnostic unit generated and submitted an individual diagnosis report. Upon receiving the report, the central control and prediction subsystem 30 immediately updated the node status representing piglet "3201-12-07" in the spatiotemporal graph model, marking it as confirmed to have a fever and cough. The ST-GNN model performed a forward propagation prediction, indicating that the probability of the other nine piglets in the same litter who had close contact with this piglet developing fever symptoms within the next 12 hours increased to 75%. The risk of infection for the piglets in adjacent pen 11 was 20%.

[0056] The system immediately sent a high-priority alert to the farm manager's mobile device: "Warning: Piglet ID 3201-12-07 in farrowing room 12 has been confirmed to have a high fever (40.8°C) and cough. The model predicts an extremely high risk of infection among littermates. Immediate isolation and observation, along with intervention for the entire pen, is recommended." Comparative Example 1 This comparative example employed a traditional monitoring solution that relied solely on regular inspections by mobile robots. This solution was deployed in the same pig farm environment as Example 1, but without any static sentinel nodes. The mobile robots were set to conduct routine inspections of all 20 farrowing pens every four hours.

[0057] Similarly, at 2:30 a.m. one day, piglet "3201-12-07" in pen No. 12 began to show the same early symptoms as in Example 1.

[0058] However, since there were no static sentinel points for continuous monitoring, the system was unaware of this. The robot's last inspection was completed at midnight, and the next scheduled inspection was at 4:00 am.

[0059] During this 90-minute "monitoring blind spot", the sick piglet's condition worsened, and it transmitted the pathogen to the other three piglets in the same litter through direct contact and aerosols.

[0060] At 4:00 AM, the mobile robot arrived at Pen 12 as planned for a routine inspection. During the inspection, its onboard infrared thermal imager scanned the entire pen and discovered multiple hotspots. The robot then initiated a detailed diagnostic program, measuring each piglet individually and confirming that four, including the first piglet, had temperatures exceeding 40.5°C.

[0061] At 4:05 a.m., the robot completed its diagnosis and reported the results, and the system sent an alert to management.

[0062] Data comparative analysis In order to more intuitively demonstrate the beneficial effects of the technical solution of the present invention, the key performance indicators of Example 1 and Comparative Example 1 are quantitatively compared, and the specific data are shown in the following table.

[0063]

[0064] Through the detailed description and data comparison of the above embodiments and comparative examples, it can be seen that the pig house robot health monitoring and early warning system for pig farming platforms provided by the present invention, through the collaborative working mechanism of static sentinel points and mobile robots, significantly shortens the delay time from the occurrence of epidemics to the discovery and early warning, greatly moves the intervention window forward, effectively controls the spread of epidemics within the group, and greatly improves the utilization efficiency of high-value diagnostic resources. Its technological advancement and practical value are obvious.

Claims

1. A pig house robot health monitoring and early warning system for pig farming platforms, characterized by: The system includes: a static sentinel node subsystem (10), which is composed of at least one static sentinel node deployed in each independent pig pen, and is used to continuously collect pen environment and pig activity data at a high frequency, generate pen abnormality scores based on real-time analysis of the pen based on the end-side computing model, and send a triage request containing a pen identifier when the pen abnormality score continues to exceed a dynamically adjusted trigger threshold; The mobile diagnosis and service unit subsystem (20) is a pig house inspection robot with autonomous navigation capabilities, which performs a baseline inspection mode and an on-demand triage mode: in the baseline inspection mode, it conducts a full-area inspection according to a low-frequency scheduling strategy to obtain baseline health data; in the on-demand triage mode, it responds to triage requests and autonomously navigates to the target pen, executes individual diagnosis protocols to obtain accurate physiological and pathological data of the target pigs, and generates individual diagnosis reports; The central control and prediction subsystem (30) conducts two-way data communication with the static sentinel node subsystem (10) and the mobile diagnosis and service unit subsystem (20), receives and integrates the continuous pen abnormality scores of the static sentinel node and the individual diagnosis reports of the mobile diagnosis and service unit subsystem (20), constructs a spatiotemporal map model of the pig herd health status, predicts the disease transmission path and health evolution trend, and generates early warning information.

2. The system according to claim 1, wherein: The static sentinel node is an integrated electronic device, and its hardware structure includes a central microcontroller unit (11), a wireless communication transceiver module (12), a power supply module (13) and a group of sensor arrays (14). The central microcontroller unit (11) is electrically connected to the wireless communication transceiver module (12), the power supply module (13) supplies power to all hardware units, the central microcontroller unit (11) is a central control chip, and a group of sensor arrays (14) is electrically connected to the central microcontroller unit (11). The static sentinel node is encapsulated in an anti-corrosion and anti-bite shell (15) with a protection grade of IP68, which is made of polycarbonate injection molding and covered with 304 stainless steel metal mesh; The sensor array (14) specifically includes: a three-axis micro-electromechanical system accelerometer, whose sampling frequency is set to be no less than 4096 Hz and whose measurement range is ±16g, and is used to continuously capture the weak vibration signals of the pen structure caused by the pigs' coughing, panting and abnormal limb movements, and the weak vibration signals of the pen structure caused by the pigs' coughing, panting and abnormal limb movements are the original vibration data; A temperature and humidity composite sensor with a temperature measurement accuracy of ±0.2°C and a humidity measurement accuracy of ±2%RH is used to monitor microenvironmental changes in the pen; and an 8x8 pixel array infrared thermopile sensor with a field of view of 60 degrees and a temperature measurement range of 20°C to 100°C is used to obtain a low-resolution thermal map of the overall surface temperature distribution of the pigs in the pen.

3. The system according to claim 2, characterized in that The client-side computing model solidified within the static sentinel node is specifically a barn anomaly score calculation module. Its execution process is as follows: Perform fast Fourier transform on the raw vibration data collected by the triaxial microelectromechanical system accelerometer in continuous 1-second time windows to convert the time domain signal into the frequency domain signal; In the frequency domain, integrating the energy of the first characteristic frequency band and the second characteristic frequency band to obtain a vibration abnormality index, wherein the frequency band range of the first characteristic frequency band is 50 Hz to 200 Hz, and the frequency band range of the second characteristic frequency band is 400 Hz to 800 Hz; The mean and standard deviation of the temperature of 64 pixels output by the infrared thermopile sensor of the 8x8 pixel array are calculated and compared with a dynamic baseline temperature issued by the central control and prediction subsystem (30) to obtain the thermal map anomaly index; The vibration abnormality index and the thermal map abnormality index are weighted and summed. The weight coefficient is dynamically configured by the central control and prediction subsystem (30) based on historical data and the daily activity rhythm of the pig herd. The sum result is the current pen abnormality score.

4. The system according to claim 1, wherein: The wireless communication between the static sentinel node subsystem (10) and the central control and prediction subsystem (30) adopts the LoRaWAN protocol, the static sentinel node serves as the terminal device, and a LoRaWAN gateway is deployed on the central control and prediction subsystem (30); The conditions for generating a triage request are: the moving average of the pen abnormality score of a static sentinel node is continuously higher than the trigger threshold set by the central control and prediction subsystem (30) for the pen for 300 consecutive seconds; The triage request is a structured uplink data packet, the data payload of which fixedly contains: a 16-bit column unique identifier, a 32-bit timestamp, an 8-bit trigger reason code, and a 16-bit floating point number representing the current anomaly score value.

5. The system according to claim 1, wherein: The hardware platform of the mobile diagnosis and service unit subsystem (20) includes a four-wheel differential drive chassis (21), an inertial measurement unit and a battery management system. The four-wheel differential drive chassis (21) is a transmission part of the piggery inspection robot and is used to provide power to the piggery inspection robot. The inertial measurement unit is used to compensate for the inertia of the piggery inspection robot during inspection. Core mission payloads include: A 3D laser radar (22) for autonomous navigation and three-dimensional environment perception, with a scanning frequency of 15 Hz and a ranging accuracy of ±2 cm; A high-resolution infrared thermal imager (23) whose detector is an uncooled vanadium oxide microbolometer with a resolution of 640x512 pixels, a thermal sensitivity of less than 40mK, and a temperature measurement accuracy of ±0.3°C after calibration; a planar circular microphone array (24) consisting of 8 MEMS microphones with a spacing of 10cm, mounted on a two-dimensional pan-tilt platform, for localization of sound sources; A long-range high-frequency RFID reader (25) compliant with ISO11784 / 11785 standards and operating at a frequency of 134.2 kHz, coupled with a directional antenna; and a six-axis collaborative robotic arm (26).

6. The system according to claim 5, characterized in that An individual diagnostic protocol consists of multiple steps, including: Step 1: Path planning and arrival: The mobile diagnosis and service unit subsystem (20) uses the scanning data of the 3D laser radar (22) and the pre-stored high-precision map to plan and execute a collision-free optimal path to reach the target fence entrance; The second step is individual segmentation and locking: After the mobile diagnosis and service unit subsystem (20) enters the pen, it uses the 3D laser radar (22) to scan the space inside the pen, and combines it with the RGB-D depth camera to perform three-dimensional point cloud instance segmentation on all pigs in the pen; at the same time, it starts the planar circular microphone array (24) to locate the sound source of the abnormal acoustic signal indicated in the triage request; by fusing the direction vector obtained by the sound source positioning with the three-dimensional point cloud instance segmentation result, the individual pig that makes the abnormal sound is locked as the primary diagnostic target; The third step is non-contact and precise measurement of vital signs: the six-axis collaborative robotic arm (26) adjusts its posture and accurately aligns the high-resolution infrared thermal imager (23) with the core area of ​​the head and trunk of the locked pig individual to capture a continuous 5-second video stream, and uses a long-range high-frequency RFID reader (25) to read the unique identification code in its ear tag; Step 4: Data analysis and report generation: The onboard computing unit of the mobile diagnosis and service unit subsystem (20) processes the collected infrared video stream, extracts the maximum body surface temperature, eye corner temperature, and quantitative indicators of uneven distribution of body surface temperature, and combines it with the read individual identity code to generate a structured individual diagnosis report; The fifth step is data reporting. After the diagnosis and report are generated, the robot uploads this complete individual diagnosis report to the central control and prediction subsystem in real time through its high-speed wireless communication module (30).

7. The system according to claim 5, characterized in that The end of the six-axis collaborative robotic arm (26) is equipped with a service end effector, which is integrated with a wireless charging coil for contactless energy replenishment of static sentinel nodes and a high-pressure cleaning nozzle for cleaning the static sentinel nodes or the mobile diagnosis and service unit subsystem (20) itself.

8. The system according to claim 1, wherein: The spatiotemporal graph model of pig herd health status is specifically a spatiotemporal graph neural network, which is constructed as follows: Abstract the entire pig farm into a dynamic graph structure G=(V,E,X); Where V is the node set of the graph, and each node v_i∈V uniquely corresponds to a pig wearing an RFID ear tag; E is the edge set of the graph. An edge e_ij∈E indicates a potential disease transmission association between nodes v_i and v_j. The initial weight of the association is set based on the physical adjacency of the pigs and the fixed air circulation simulation path. X is the feature matrix of the node, Xi(t) is the feature vector of node vi at time t, and the feature vector contains the historical health data of the pig.

9. The system according to claim 8, characterized in that The real-time update mechanism of the spatiotemporal map model of pig herd health status includes: Continuously receive the data stream of pen anomaly scores from all static sentinel nodes. When the anomaly score S_k(t) of pen k increases, increase the weight of the edge e_ij connecting all nodes v_i∈V_k in pen k. At the same time, treat S_k(t) as a time-varying pen-level environmental feature and append it to the feature vector Xi_i(t) of all nodes in the pen. Furthermore, upon receiving an individual diagnosis report for pig v_i generated by the mobile diagnosis and service unit subsystem (20), the precise physiological and pathological data in the individual diagnosis report are directly updated to the latest feature vector X_i(t) of this node, and used as the true state label of this node at this moment.

10. The system according to claim 1, wherein: The central control and prediction subsystem (30) further includes a dynamic threshold and scheduling decision module, which is used to: Calculate and issue a private, time-varying trigger threshold for each static sentinel node. The trigger threshold is calculated based on the historical statistical distribution of abnormal scores for the pen, the overall activity rhythm of the pig herd, and global environmental parameters. Furthermore, when multiple triage requests arrive at the same time, the scheduling decision module is responsible for prioritizing the tasks of the mobile diagnosis and service unit subsystem (20) based on the abnormality score of the request, the value of the pigs in the pen, and the disease spread risk level of the pen predicted by the spatiotemporal graph model, and generating the optimal scheduling sequence.

Images