Infrared physiotherapy cabin system and method based on animal and environment collaborative regulation

Abstract

The application provides an infrared physiotherapy cabin system and method based on animal and environment collaborative regulation, relates to the field of animal physiotherapy and intelligent monitoring and control, and comprises an animal state and cabin environment monitoring module, an infrared physiotherapy and cabin environment regulation module, an animal health assessment and curative effect assessment module, a personalized treatment and intelligent decision module, and a man-machine interaction and safety supervision module. Through multi-modal data synchronous alignment and data quality evaluation, a reliability result is formed, a health state representation is constructed under the reliability gating, and a risk level is output, infrared physiotherapy control parameters and cabin environment regulation control parameters are generated under the stage control boundary and treatment target constraint, and are executed, and online fine adjustment is performed on continuous adjustable parameters during physiotherapy, and rollback and interlock protection are performed when the risk is adjusted up or the reliability is decreased. After a cycle, curative effect is evaluated, data recording and tracing archiving are completed, and animal state and cabin environment collaborative closed-loop infrared physiotherapy regulation is realized.

Description

Technical Field

[0001] This invention relates to the field of animal physiotherapy and intelligent monitoring and control, specifically to an infrared physiotherapy chamber system and method based on the coordinated regulation of animals and the environment. Background Technology

[0002] In recent years, infrared therapy, as a non-pharmacological and non-invasive physical intervention, has been widely used in animal rehabilitation and care, aiding in the relief of pain and inflammation-related discomfort, postoperative or post-exercise recovery, and improving comfort. Compared to drug intervention, infrared therapy, under reasonable dosage and safety supervision, is relatively convenient to operate, repeatable, has a lower systemic drug burden, and is easier to control. Compared to some therapy methods that require complex contact or strong restraint, infrared irradiation can be implemented with reduced direct contact stimulation, thereby reducing animal resistance and stress risks to some extent. At the same time, with the increase in the number of companion animals and the growing demand for refined care, the requirements for safety, standardization, and assessability of animal therapy services for families and institutions are constantly increasing, making the integration of infrared therapy with condition monitoring, environmental protection, and risk management a practical necessity.

[0003] From an engineering perspective, placing animals in a relatively enclosed cabin for infrared intervention offers advantages such as more stable irradiation conditions, reduced external environmental disturbances, and easier unified safety monitoring. However, it also introduces issues like accumulated heat load within the cabin, temperature and humidity drift, insufficient ventilation, and fluctuations in animal condition due to changes in body position and stress behaviors. Significant individual differences among animals—breed, weight, age, and past health conditions—lead to varying tolerance ranges to heat stimulation and environmental changes. During therapy, animals may exhibit increased activity, restlessness, changes in breathing patterns, or altered body position, creating a dynamic coupling between infrared output and the cabin's temperature, humidity, airflow, and ventilation. Without effective monitoring and control, this can easily result in decreased comfort, unstable therapeutic effects, and even safety risks. Therefore, animal infrared therapy systems not only need to achieve controllable therapeutic output but also need to consider the stability of the cabin environment, timely detection of changes in animal condition, and comprehensive safety monitoring and traceability management.

[0004] Currently, similar implementation methods mainly include the following categories: First, open or semi-open infrared irradiation equipment, which irradiates animals locally or over the entire area using fixed power or a few power levels. This method has a simple structure but is greatly affected by the external environment, and changes in the animal's position can easily lead to unstable irradiation intensity distribution. Second, closed or semi-closed therapy chambers, which simultaneously irradiate animals by preset environmental parameters such as temperature and airflow. This can reduce external disturbances to a certain extent, but it relies heavily on experience-based settings and is difficult to cover individual differences and dynamic state changes in different animals. Third, therapy devices that introduce single physiological monitoring or simple video observation can provide a rough observation of the animal's condition, but often suffer from insufficient data reliability due to issues such as wearing stability, motion artifacts, occlusion, and asynchronous sampling, making it difficult to support refined process management. Fourth, comprehensive solutions based on manual inspection and intervention. Although these can handle abnormalities, they are highly dependent on the operator's experience and are difficult to form a standardized and reusable closed-loop process.

[0005] Meanwhile, existing systems generally suffer from independent data acquisition and processing at the data level: physiological, behavioral, and environmental data are often collected by different devices with inconsistent sampling rates, time bases, data formats, and output frequencies, making it difficult to complete correlation analysis, cross-validation, and consistency judgment on the same time scale. When sensors experience short-term detachment, obstruction, saturation, or communication anomalies, the system usually lacks online quality assessment and credibility expression mechanisms, which can easily lead to situations where judgments or controls are driven by low-quality data, thereby affecting safety and the stability of therapeutic efficacy.

[0006] Therefore, existing animal infrared therapy technologies still have the following shortcomings: they cannot consistently characterize and constrain the animal's physiological state, behavioral stress, and changes in the cabin environment during the therapy process, making it difficult to establish safety boundaries and process management that cater to individual differences; they cannot perform unified time alignment, online quality assessment, and reliability management of multi-source monitoring data, leading to delayed or misjudged anomaly identification; they cannot coordinate the infrared therapy output with cabin temperature and humidity, airflow, and ventilation regulation, easily resulting in comfort and safety risks caused by heat and humidity accumulation or sudden environmental changes; and they lack systematic support in alarm interlocking, manual intervention, data management, and full-process recording and verification, thus affecting the standardization, regulatory capacity, and traceability of the therapy process. Summary of the Invention

[0007] This invention provides an infrared therapy chamber system and method based on the coordinated regulation of animals and the environment. Addressing issues such as significant individual differences, dynamic fluctuations in animal condition, stress induced by heat, humidity, and airflow disturbances within the chamber, data obstruction, impact from detachment and movement, and the need for closed-loop support for risk identification and management during animal infrared therapy, this invention constructs a systematic architecture based on multimodal monitoring as input, data reliability gating, health assessment and risk prediction, personalized treatment parameter generation and online fine-tuning, and human-computer interaction and safety supervision. This achieves coordinated adaptation between infrared therapy and chamber environment regulation under safety constraints, forming a verifiable end-to-end data link.

[0008] An infrared physiotherapy chamber system based on the synergistic regulation of animals and the environment includes:

[0009] Among them, the animal status and cabin environment monitoring module is used to collect and output animal physiological data, animal activity and behavior data, and cabin environment data;

[0010] The infrared physiotherapy and cabin environment control module is used to receive effective control commands and execute infrared physiotherapy output and cabin environment control, while reading back the actuator working status and outputting it.

[0011] The animal health assessment and efficacy assessment module is used to synchronize and align the scattered time-series data output by the animal status and cabin environment monitoring module under a unified time reference to form a multimodal aligned time-series data stream. It also performs online data quality assessment and credibility management for each channel to output data quality and credibility results. Furthermore, under credibility gating, it outputs health status and risk prediction results as well as efficacy assessment results at the end of the cycle.

[0012] The personalized treatment and intelligent decision-making module generates stage identifiers, stage control boundary tables, and treatment target sets based on the health status and risk prediction results output by the animal health assessment and efficacy assessment module, combined with the actuator working status read back by the infrared physiotherapy and cabin environment control module. Under the constraints of the stage control boundary table and treatment target sets, it generates infrared physiotherapy control parameters and cabin environment control parameters and sends them to the infrared physiotherapy and cabin environment control module for execution, and is used to fine-tune the control parameters online during the execution process.

[0013] The human-computer interaction and safety supervision module is used to provide basic animal information input, parameter verification or correction, manual intervention and control, alarm and interlock handling, data recording and traceability, and equipment operation and calibration management.

[0014] The animal status and cabin environment monitoring module, consisting of an animal physiological monitoring unit, an animal activity monitoring unit, and a cabin environment monitoring unit, is responsible for the acquisition and output of multi-source sensor data. The animal physiological monitoring unit collects and outputs vital signs data such as heart rate, blood oxygen saturation, respiratory rate, and body temperature. The animal activity monitoring unit collects information related to animal movement behavior and outputs position and movement parameters; a camera is used to acquire behavioral image streams and the visual input required for behavioral representation, while millimeter-wave radar is used for animal position and target tracking and stable acquisition of movement parameters. The cabin environment monitoring unit collects parameters related to cabin temperature, humidity, airflow, and ventilation; a temperature and humidity sensor provides temperature and humidity data, and an airflow sensor provides airflow and air volume data. The data output from each channel of the animal status and cabin environment monitoring module serves as the data source input for subsequent system evaluation.

[0015] The infrared physiotherapy and cabin environment control module, consisting of an infrared control unit and a cabin environment control unit, is responsible for executing and reading back the status of externally issued commands. The infrared control unit executes the output control commands of the infrared physiotherapy device and reads back the drive and equipment status to support monitoring and protection. The cabin environment control unit executes commands for temperature and humidity adjustment, as well as airflow and ventilation control within the cabin environment, and reads back the working status and related status of the environmental actuators, thus providing execution-side evidence for the system's safety links, process recording, and subsequent evaluation.

[0016] The animal health assessment and efficacy assessment module consists of a data quality assessment and credibility management unit, a health status assessment and risk prediction unit, and a cycle-end efficacy assessment unit. It is used to convert the multi-source observation of the animal status and cabin environment monitoring module and the execution readback of the infrared physiotherapy and cabin environment control module into credible status expressions, risk results, and cycle efficacy conclusions. The data quality assessment and credibility management unit is used to perform unified clock synchronization and time alignment on scattered time-series data from different sampling rates and channels, forming a multimodal aligned time-series data stream, and generating channel credibility and overall credibility for subsequent assessment, decision-making, and safety gating. The health status assessment and risk prediction unit is used to construct a unified health status vector based on credibility gating and individualized threshold configuration, outputting the contributions of thermal steady-state, cardiovascular load, oxygenation and ventilation, stress and comfort, and environmental adaptation sub-items, and forming a comprehensive health score. At the same time, based on the prediction model, it generates future health status vectors, future contribution values ​​of each sub-item, and future comprehensive health scores, and constructs scoring trend indicators to complete risk assessment and output risk level results of no risk, warning, high risk, and emergency shutdown. The prediction model incorporates constraints related to physiological rationality and consistency of the cabin thermal environment to improve the usability and stability of the prediction results. The cycle-end efficacy assessment unit is used to summarize and generalize process data, control execution records, and assessment results after a treatment cycle ends, forming quantitative efficacy conclusions and traceable cycle-level assessment records, providing a basis for subsequent protocol optimization and review.

[0017] The personalized treatment and intelligent decision-making module consists of a phase management and treatment goal generation unit, a personalized treatment plan generation unit, and an online strategy update and intelligent fine-tuning unit. It is used to complete the decision output of goal setting, plan output and online optimization within the safety boundary and to be executed in conjunction with the infrared physiotherapy and cabin environment control module. The phase management and treatment goal generation unit generates phase identifiers, phase control boundary tables, and treatment goal sets based on the health status and risk prediction data output by the animal health assessment and efficacy assessment modules, combined with actuator availability constraints. The personalized treatment plan generation unit generates infrared physiotherapy control parameters and cabin environment control parameters under the hard constraints of the phase control boundary table and treatment goal sets, and implements cross-channel collaborative control and linkage constraints on the two types of parameters to ensure that changes in infrared heat input are consistent with environmental temperature, humidity, airflow, and ventilation response. The online strategy update and intelligent fine-tuning unit performs small-scale online optimization of continuously adjustable parameters without exceeding the phase control boundaries and safety hard constraints, and implements gating control on the fine-tuning range based on overall credibility and key component credibility. When the risk is increased or the trend is unfavorable, it automatically switches to a conservative strategy or triggers a rollback to improve goal achievement and process stability.

[0018] The human-computer interaction and safety supervision module consists of a user interface interaction unit, a manual intervention and control unit, an alarm and notification unit, a data management and reporting unit, and an equipment operation and calibration unit. It provides operational access and safety assurance throughout the entire system operation process. The user interface interaction unit is used for basic animal information input, threshold and configuration loading, operational status presentation, and parameter verification and correction. The manual intervention and control unit provides channels for manual intervention, interlock authorization, or manual control in specific scenarios. The alarm and notification unit alerts and coordinates responses to changes in risk level, credibility anomalies, channel anomalies, and safety events. The data management and reporting unit links and archives multimodal data, evaluation data packages, control parameter outputs, online fine-tuning records, and interlock alarm events according to a unified time base and outputs traceable reports. The equipment operation and calibration unit performs self-checks and manages the effectiveness of door locks, emergency stops, power supply grounding, and sensor and actuator connections and calibrations, making critical safety conditions and availability verification a prerequisite for physical therapy execution.

[0019] The present invention has the following beneficial technical effects:

[0020] (1) This invention establishes a multimodal aligned time-series data stream, channel credibility sequence and overall credibility, and incorporates channel availability, signal quality and cross-channel consistency into the gating basis, so that health assessment, risk prediction and online fine-tuning no longer depend on the occasional fluctuations of a single channel, thereby reducing the risk of misjudgment and miscontrol caused by occlusion, detachment, motion interference and sampling anomalies, and improving the stability of closed-loop decision output.

[0021] (2) This invention uses a unified health status vector and a multi-item contribution system, combined with future status prediction and scoring trend indicators to output graded risk results, so that the system can form response and tightening strategies in advance when it is close to the risk boundary, thereby reducing the disturbance caused by sudden shutdowns and drastic rollbacks.

[0022] (3) This invention uses a three-layer structure of stage identifiers, stage control boundary tables and treatment target sets to solidify the control domain, change rate constraints and backoff conditions, so that infrared physiotherapy control parameters and cabin environment control parameters can be generated and executed synchronously within the same hard constraint system, thereby suppressing the accumulation of heat load and stress and discomfort induced by environmental changes, and improving the controllability and consistency of the treatment process.

[0023] (4) This invention optimizes continuously adjustable parameters in small steps within hard constraints through online strategy updates and intelligent fine-tuning, and uses credibility and risk level as gating and freezing conditions. When the data is reliable and the safety margin is sufficient, the system improves the achievement of the target. When the risk is increased, the trend is unfavorable or the credibility is reduced, it automatically switches to a conservative strategy and triggers a rollback, thereby forming an adaptive optimization capability under the premise of safety priority. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the embodiments will be briefly described below.

[0025] Figure 1 This is a system framework diagram of the present invention;

[0026] Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and beneficial effects of the embodiments of the present invention clearer, the system operation process is described below in conjunction with the embodiments of the present invention. It should be understood that the described embodiments are only some embodiments of the present invention, and not all embodiments. Other implementation methods obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention. Unless otherwise defined, the technical terms used in this specification should be understood as meanings commonly understood by those skilled in the art.

[0028] like Figure 1 As shown, the infrared therapy chamber system based on the coordinated regulation of animals and the environment of the present invention generally includes five modules: Module 1 is used to collect and output multi-source data on animal physiology, animal behavior and position, and chamber environment status; Module 2 is used to receive effective control commands issued from the upper level and execute infrared therapy and chamber environment regulation, while reading back the actuator status for monitoring and recording; Module 3 is used to perform multimodal temporal alignment, quality assessment and credibility management on multi-source data, and complete health status assessment, risk prediction and cycle end efficacy assessment under credibility gating; Module 4 is used to generate stage identifiers, stage control boundaries and treatment goals based on the status and risk results output by Module 3, and further generate infrared and environmental control parameter sequences, while performing online strategy updates and intelligent fine-tuning within hard constraints; Module 5 is used to provide support such as user interface interaction, manual intervention and control, alarm notification, data management reports and equipment operation and calibration, so that the system has the ability to operate, control and trace.

[0029] Module 1 is the animal status and cabin environment monitoring module, consisting of an animal physiological monitoring unit, an animal activity monitoring unit, and a cabin environment monitoring unit. It is responsible for the acquisition and output of multi-source sensor data. The animal physiological monitoring unit collects and outputs vital signs data such as animal heart rate, blood oxygen saturation, respiratory rate, and body temperature. The animal activity monitoring unit collects information related to animal movement behavior and outputs position and movement parameters. A camera is used to acquire behavioral image streams and the visual input required for behavioral representation, while millimeter-wave radar is used for animal position and target tracking and stable acquisition of movement parameters. The cabin environment monitoring unit collects parameters related to cabin temperature, humidity, airflow, and ventilation. A temperature and humidity sensor provides temperature and humidity data, and an airflow sensor provides airflow and air volume data. The data output from each channel of Module 1 serves as the data source input for subsequent system evaluation.

[0030] Module 2 is the infrared physiotherapy and cabin environment control module, consisting of an infrared control unit and a cabin environment control unit. It is responsible for executing and reading back the status of externally issued commands. The infrared control unit executes the output control commands of the infrared physiotherapy device and reads back the drive and equipment status to support monitoring and protection. The cabin environment control unit executes the temperature, humidity, airflow, and ventilation control commands of the cabin environment and reads back the working status and related status of the environmental actuators, thereby providing execution-side basis for the system's safety link, process recording, and subsequent evaluation.

[0031] Module 3 is the animal health assessment and efficacy assessment module, which consists of a data quality assessment and credibility management unit, a health status assessment and risk prediction unit, and a cycle-end efficacy assessment unit. It is used to transform the multi-source observations of Module 1 and the execution readback of Module 2 into credible status expressions, risk results, and cycle efficacy conclusions. The data quality assessment and credibility management unit is used to perform unified clock synchronization and time alignment on scattered time-series data from different sampling rates and channels, forming a multimodal aligned time-series data stream, and generating channel credibility and overall credibility for subsequent assessment, decision-making, and safety gating. The health status assessment and risk prediction unit is used to construct a unified health status vector based on credibility gating and individualized threshold configuration, outputting the contributions of thermal steady-state, cardiovascular load, oxygenation and ventilation, stress and comfort, and environmental adaptation sub-items, and forming a comprehensive health score. At the same time, based on the prediction model, it generates future health status vectors, future contribution values ​​of each sub-item, and future comprehensive health scores, and constructs scoring trend indicators to complete risk assessment and output risk level results of no risk, warning, high risk, and emergency shutdown. The prediction model incorporates constraints related to physiological rationality and consistency of the cabin thermal environment to improve the usability and stability of the prediction results. The cycle-end efficacy assessment unit is used to summarize and generalize process data, control execution records, and assessment results after a treatment cycle ends, forming quantitative efficacy conclusions and traceable cycle-level assessment records, providing a basis for subsequent protocol optimization and review.

[0032] Module 4 is the personalized treatment and intelligent decision-making module, consisting of a phase management and treatment goal generation unit, a personalized treatment plan generation unit, and an online strategy update and intelligent fine-tuning unit. It is used to complete the decision output process—goal setting, plan output, and online optimization—within the safety boundaries and is executed in conjunction with Module 2. The phase management and treatment goal generation unit generates phase identifiers, phase control boundary tables, and treatment goal sets based on the health status and risk prediction data output from Module 3, combined with actuator availability constraints. The personalized treatment plan generation unit generates infrared physiotherapy control parameters and cabin environment control parameters under the hard constraints of the phase control boundary table and treatment goal sets, and implements cross-channel collaborative control and linkage constraints on these two types of parameters to ensure that changes in infrared heat input are consistent with environmental temperature, humidity, airflow, and ventilation response. The online strategy update and intelligent fine-tuning unit performs small-scale online optimization of continuously adjustable parameters without exceeding the phase control boundaries and safety hard constraints, and gates the fine-tuning range based on overall credibility and key component credibility. When the risk increases or the trend becomes unfavorable, it automatically switches to a conservative strategy or triggers a rollback to improve goal achievement and process stability.

[0033] Module 5 is the Human-Machine Interaction and Safety Supervision Module, consisting of a user interface interaction unit, a manual intervention and control unit, an alarm and notification unit, a data management and reporting unit, and an equipment operation and calibration unit. It provides operational access and safety assurance throughout the entire system operation process. The user interface interaction unit is used for basic animal information input, threshold and configuration loading, operational status presentation, and parameter verification and correction. The manual intervention and control unit provides channels for manual intervention, interlock authorization, or manual control in specific scenarios. The alarm and notification unit alerts and coordinates responses to changes in risk level, credibility anomalies, channel anomalies, and safety events. The data management and reporting unit links and archives multimodal data, evaluation data packages, control parameter outputs, online fine-tuning records, and interlock alarm events according to a unified time base and outputs traceable reports. The equipment operation and calibration unit performs self-checks and manages the effectiveness of door locks, emergency stops, power supply grounding, and sensor and actuator connections and calibrations, making critical safety conditions and availability verification a prerequisite for physical therapy execution.

[0034] like Figure 2 As shown in this embodiment, the operation process of an infrared therapy chamber system based on the coordinated regulation of animals and the environment during an animal therapy cycle may include the following steps:

[0035] S101: Cabin pre-inspection and system initialization.

[0036] After the system is powered on, a pre-check and initialization of the cabin are performed before entering the working state. The equipment maintenance and calibration unit in module 5 performs self-checks on the cabin door locks, emergency stop switch, power supply, and grounding status, and verifies the connection status, response status, and calibration validity of all sensors and actuators. If the verification passes, calibration parameters are loaded, and a zero-point reset is performed if necessary. The cabin is kept clean and unobstructed, with no foreign objects obstructing the camera's field of view, the millimeter-wave radar beam coverage area, or the ventilation ducts; any obstruction is considered a condition prohibiting startup.

[0037] S102: Establishment of pre-entry environmental reference parameters.

[0038] After the S101 self-test passes, the cabin environment monitoring unit of the system startup module 1 enters the sampling stage. Based on the confirmed calibration parameters, it collects cabin temperature, humidity and airflow data, and performs sliding average processing within a preset time window to form "pre-entry environment reference data". This reference data serves as the benchmark and basis for subsequent environmental closed-loop control and anomaly identification.

[0039] S103: Animal entry and basic information entry.

[0040] After the cabin pre-inspection is completed, guide the animal into the cabin. Before the animal enters the cabin, enter the animal's identity and basic information, including breed, weight, age, medical history and contraindications, in the user interface interaction unit of module 5, and record it as metadata for this infrared therapy, which is used to load the threshold table, select model configuration and determine control boundaries.

[0041] S104: Sensor Wearing and Multimodal Monitoring Establishment.

[0042] The animal physiological monitoring unit in Module 1 monitors vital signs according to the process of "wearing guidance - data acquisition establishment - access confirmation - continuous acquisition". The "access confirmation" is determined by the data quality assessment and reliability management unit in Module 3. The PPG sensor is installed on the adjustable neck wearer, and the optical window is stably fitted to the skin or sparse hair area of ​​the neck. After completing the acquisition of PPG segments of preset duration, the animal physiological monitoring unit in Module 1 submits the segment data to the data quality assessment and reliability management unit in Module 3 for wear access assessment. The unit outputs a channel availability conclusion based on quality indicators such as waveform amplitude, baseline stability, period consistency, and optical saturation / underexposure status. When the access threshold is not reached, the user interface interaction unit in Module 5 prompts the operator to adjust the wearing position and tightness and re-acquire and assess. After access is passed, the animal physiological monitoring unit in Module 1 enters the continuous acquisition stage and calculates and outputs real-time results of heart rate and blood oxygen saturation, and marks the channel as "access valid" for inclusion in the subsequent infrared physiotherapy process. Respiratory monitoring uses a chest strap sensor fixed to the periphery of the thorax. After the animal physiological monitoring unit of module 1 completes the respiratory signal acquisition and baseline initialization, it submits the signal segment to the data quality assessment and reliability management unit of module 3 for admission assessment. The unit determines the availability based on the periodic characteristic stability and frequency domain energy concentration. Under the condition that the admission is passed and the animal is relatively still, the animal physiological monitoring unit of module 1 completes the initial calibration and gives a stable respiratory rate result. Body temperature monitoring employs a laser body temperature sensor for non-contact measurement of the target area on the flank. The temperature sensor is mounted on an adjustable bracket inside the cabin, with its measurement field of view aligned with the target area. The animal activity monitoring unit in module 1 aligns the temperature measurement area using millimeter-wave radar and visual positioning, and submits the body temperature segment and alignment status to the data quality assessment and reliability management unit in module 3 for admission evaluation. This unit outputs a conclusion on the alignment validity of the body temperature channel. During the physiotherapy process, when changes in body position cause the temperature measurement point to deviate, the animal activity monitoring unit in module 1 updates the temperature measurement area in real time and automatically repositions it, continuously outputting the real-time body temperature value. Simultaneously, the data quality assessment and reliability management unit in module 3 updates the availability marker of the body temperature data segment based on the alignment confidence level to ensure the consistency of the temperature measurement area and the continuity of the body temperature sequence.

[0043] S105: Generate a multimodal aligned timing data stream.

[0044] Animal physiological data, animal activity and behavior data, and cabin environment data are collected and output as distributed time-series data by the animal physiological monitoring unit, animal activity monitoring unit, and cabin environment monitoring unit of Module 1, respectively. The system uses the master clock as a unified time reference. Each monitoring unit generates a timestamp for each channel's data during sampling and outputs it to the buffer queue in chronological order. After receiving the distributed time-series data, the data quality assessment and reliability management unit of Module 3 performs unified clock synchronization and time alignment processing, resampling, interpolating, and sliding window registration for data at different sampling rates to form a "multimodal aligned time-series data stream." This multimodal aligned time-series data stream serves as a unified time reference for sharing and interaction among functional units within the system. Each functional unit carries the same timestamp and writes it into the record when performing data sharing and closed-loop collaborative calculations, thereby ensuring the consistency of multi-directional real-time data interaction and closed-loop collaboration.

[0045] S106: Door safety confirmation and start-up interlock.

[0046] The system uses real-time image streams from the animal activity monitoring unit in Module 1 and millimeter-wave radar echo information to determine if the animal is within the effective monitoring area and maintains a preset safe distance from the door's movement range, the cabin walls, and the infrared therapy transmitter. When this condition is met, the manual intervention and manual control unit in Module 5 outputs an interlock permission to close the door, which is then used by the main controller to drive the door control mechanism to close the door. After the door closes, the system verifies the door lock closure and locking status and sets door lock closure as a prerequisite for starting infrared therapy. If either status is not met, the system prohibits entering the infrared therapy execution state, and the user interface interaction unit in Module 5 displays an error message and suggested solutions.

[0047] S107: Online data quality assessment and credibility management.

[0048] The data quality assessment and reliability management unit of module 3 receives the dispersed time-series data output by module 1 throughout the infrared therapy process, synchronously aligns it to form a multimodal aligned time-series data stream, and performs online quality assessment and reliability updates for each channel. The multimodal aligned time-series data stream includes heart rate, blood oxygen saturation, respiratory rate, body temperature, visual behavioral characteristics, millimeter-wave radar positioning / motion parameters, and environmental parameters such as cabin temperature, humidity, and airflow. Simultaneously, the unit loads corresponding threshold tables and weight configurations based on the animal basic information entered by the user interface interaction unit of module 5, serving as the parameter basis for channel quality judgment and anomaly identification.

[0049] Module 3's data quality assessment and reliability management unit extracts quality features from each channel using a sliding assessment window and calculates channel reliability scores (normalized to 0-1). Channel reliability calculation employs a hierarchical discrimination and fusion approach. The first layer detects packet loss, timestamp continuity, sampling rate deviation, and latency based on data integrity and temporal characteristics, and combines this with sensor state variables / state bits (including dropout, occlusion, saturation, and communication anomalies) to form basic availability criteria. The second layer, under the premise of satisfying basic availability, calculates signal quality indicators for different channels and maps them to quality sub-scores. These quality indicators include: PPG signal signal-to-noise ratio, baseline jitter, period consistency, and motion artifacts; respiratory signal main frequency energy concentration and period stability; body temperature signal temperature measurement alignment confidence and abnormal jump rate; visual signal clarity and occlusion ratio; millimeter-wave radar signal echo signal-to-noise ratio and target tracking continuity; and environmental channel short-term stability and abnormal change rate. The third layer applies range and rate of change constraints to key physiological and environmental quantities based on the loaded configuration, and introduces cross-channel consistency constraints to reduce the weight of suspicious windows. Then, it performs weighted fusion according to the quality index weight of each channel to obtain the window-level channel credibility, and performs time smoothing update on the channel credibility score to suppress frequent switching caused by single window fluctuations, forming a stable channel credibility sequence and channel availability label.

[0050] The data quality assessment and credibility management unit in Module 3 identifies anomaly types based on channel credibility sequences and anomaly duration, and classifies data problems into four categories: transient artifacts, short-term missing data, persistent offsets, and systematic anomalies. At the same time, it updates and improves the multimodal aligned time-series data stream. For transient artifacts, the unit identifies those with short duration and isolated spikes or short-term jumps, performs robust filtering and outlier suppression on their data segments, and replaces them with neighborhood estimates. The replaced samples are then written into the traceability index. For short-term missing data, the unit determines it to be a short-term recoverable missing data based on timestamp continuity and the missing span. If the missing span does not exceed a preset upper limit and cross-channel consistency constraints are met, interpolation or prediction based on historical windows is performed on the missing interval, and the completed samples are written into the traceability index. For persistent offsets, the unit identifies baseline drift, long-term deviation, or slow trend accumulation. Drift is determined by trend slope, cumulative deviation, and comparison with the entry baseline, and online offset compensation is performed on correctable items. When drift cannot be corrected by the algorithm or is related to wearing / alignment, the unit generates a "re-wearing / calibration / obstruction removal" event. The alarm and notification unit of module 5, in conjunction with the user interface interaction unit, outputs a handling prompt and records the event timestamp and affected channel. In the event of a systemic anomaly, when multiple critical channels are simultaneously in a state of low confidence, critical physiological channels fail and persist for more than a preset duration, or a short-term loss evolves into an unrecoverable loss, the unit triggers a systemic anomaly handling process and generates a stop interlock trigger record, which is used to perform protective actions on the security link and retain traceability.

[0051] Module 3's data quality assessment and credibility management unit performs time-smooth updates on the channel credibility scores calculated within the sliding window for each channel, and weights and summarizes the channel credibility scores according to channel weights to obtain the overall credibility. The channel credibility and overall credibility are uniformly packaged into a "data quality and credibility data package" for external output; at the same time, the channel anomaly type and duration, the triggering conditions and methods for replacement / supplementation / compensation, the corresponding time range, and the handling events and interlock trigger information are all written to the log and associated with the traceability index.

[0052] S108: Health status assessment and risk prediction.

[0053] After obtaining the multimodal aligned time-series data stream and data quality and credibility data package output by the data quality assessment and credibility management unit of module 3, the health status assessment and risk prediction unit of module 3 performs a dual transformation of "individualized baseline normalization + safety threshold distance mapping" on each element involved in the assessment: using the pre-entry environmental reference data and animal health baseline as references, the current observation value is offset corrected and scaled to obtain the deviation feature that characterizes the degree of deviation relative to the baseline; the current observation value is mapped to the normalized threshold distance feature relative to the upper / lower safety limit, and a segmented penalty mapping is used for the part that exceeds the safety boundary to enhance the sensitivity to edge risk states, so that the health assessment can respond in advance to states that are close to the threshold but have not yet exceeded the limit.

[0054] The health status assessment and risk prediction unit in module 3 constructs a unified health status vector based on the deviation feature and threshold distance feature within a sliding assessment window. It then calculates window statistics for each dimension of this health status vector to form a window-level feature set. These window statistics include short-term rate of change, trend slope, volatility / stability measures, and mutation measures. According to a preset sub-item mapping rule, the window-level feature set is divided into feature subsets corresponding to each sub-item, and sub-item contribution calculations are performed on each feature subset. The sub-item mapping rules are solidified in the form of "sub-item - feature subset - weight / function", specifically: the thermal steady-state sub-item corresponds to the first feature subset composed of body temperature deviation, body temperature change rate, ambient temperature and airflow characteristics, and physiotherapy control quantity change characteristics; the cardiovascular load sub-item corresponds to the second feature subset composed of heart rate deviation, heart rate change rate, and activity intensity characteristics; the oxygenation and ventilation sub-item corresponds to the third feature subset composed of blood oxygen saturation deviation, respiratory rate deviation, respiratory fluctuation, and activity change characteristics; the stress and comfort sub-item corresponds to the fourth feature subset composed of activity agitation level, postural stability, and persistence of abnormal behavior characteristics; and the environmental adaptation sub-item corresponds to the fifth feature subset composed of temperature and humidity deviation, airflow mutation characteristics, and environmental control quantity change characteristics. The feature subsets of each sub-item are calculated according to a preset nonlinear mapping function to obtain the contribution values ​​of the thermal steady-state sub-item, cardiovascular load sub-item, oxygenation and ventilation sub-item, stress and comfort sub-item, and environmental adaptation sub-item, respectively. Based on the credibility scores of each channel, corresponding sub-item credibility is generated for each sub-item, and the sub-item credibility is used for gating and fusion adjustment in the comprehensive health scoring stage. Specifically, when the sub-item credibility corresponding to a sub-item is in the high credibility range, the sub-item contribution value participates in the weighted fusion of the comprehensive health score according to its basic fusion weight; when the sub-item credibility is in the usable range and decreases, the effective weight of the sub-item in the weighted fusion is continuously reduced to achieve flexible weight reduction before the comprehensive health score is calculated; when there is a sub-item credibility below the usable threshold, the sub-item contribution value is removed from the comprehensive health score link and retained only in reference form. At the same time, a cautious bias is introduced into the comprehensive health score, and the uncertainty expression of the comprehensive health score is output in the form of "score value + interval boundary". The interval boundary is jointly determined by the overall credibility, sub-item credibility, and window-level feature volatility.

[0055] The health status assessment and risk prediction unit in Module 3 uses a Physical Information Neural Network (PINN) to predict multimodal alignment time-series data. The input consists of the current multimodal alignment time-series data and the control quantity time-series data from the infrared control unit and the cabin environment control unit in Module 2. The output is a future multimodal alignment time-series sequence isomorphic to the input. During model training and online adaptation, soft physical and physiological constraints are introduced to ensure the prediction results are physiologically reasonable and consistent with the cabin thermal environment. These constraints include: the coupling relationship between body temperature, thermal environment, and infrared power; the consistency constraint between respiratory rate and blood oxygen saturation; and the coupling relationship between heart rate and activity intensity. Based on the future multimodal alignment time-series sequence output by PINN, a future health status vector and related statistics are generated using the same algorithm. The scoring weights are flexibly adjusted using the original channel confidence and overall confidence, and the future contribution values ​​of each component and the future comprehensive health score are calculated accordingly. Based on the current contribution values ​​of each component, the current comprehensive health score, the future contribution values ​​of each component, and the future comprehensive health score, a scoring trend index is constructed, and a risk assessment is completed. Scoring trend indicators include the difference between the current and future comprehensive health scores, the rate of change or trend slope of the scores, the threshold distance between the future score interval and each risk boundary, the time position when the lower bound of the future score interval reaches the risk boundary, and the duration or length of the crossover after the lower bound of the future score interval crosses the risk boundary. Risks are categorized based on these scoring trend indicators into four levels: None, Warning, High Risk, and Emergency Shutdown. None indicates that the current overall health score is within a safe range and the scoring trend indicator does not show a continuous unfavorable trend approaching the risk boundary; the future overall health score range does not reach the warning boundary. Warning indicates that the future overall health score shows a continuous downward trend, and the magnitude, slope, or threshold distance meets the warning threshold; or the lower bound of the future score range approaches the high-risk boundary but has not yet crossed it. High risk indicates that the future overall health score range has reached the high-risk range, or the lower bound of the future score range has crossed the high-risk boundary and meets the required duration; or there is a continuous unfavorable contribution value for future sub-items that is consistent with the scoring trend indicator. Emergency Shutdown indicates that the lower bound of the future score range crosses the emergency boundary within the prediction window, and the overall credibility and key sub-item credibility meet the set thresholds; or the current observation has triggered the emergency safety boundary and triggered a stop interlock. When overall credibility is insufficient, sensitivity to adverse trends is increased while the strength of definitive statements is reduced. The alarm and notification unit of module 5, in conjunction with the user interface interaction unit, outputs a "suspected risk" and provides handling suggestions. The contribution values ​​of each sub-item, the overall health score, future contribution values ​​of each sub-item, the future overall health score, the score trend indicators, and the risk level are uniformly packaged into a "health status and risk prediction data package" for external output.

[0056] S109: Phase Management and Treatment Goal Generation.

[0057] Module 4's phase management and treatment goal generation unit uses health status and risk prediction data packets as input for phase determination. It also reads the actuator operating status from the infrared control unit and cabin environment control unit of Module 2, and issues corresponding phase identifiers and phase control boundary tables accordingly. If any critical actuator is unavailable or in a protected locked state, Module 4's phase management and treatment goal generation unit limits the phase plan to a safe, stable, executable range. It also uses Module 5's alarm and notification unit, in conjunction with the user interface interaction unit, to remind the operator of the fault location and handling steps. The system prohibits entry into the treatment execution phase and optimization / enhancement phase until the actuator status is restored and the availability check is passed again. Animal infrared therapy is divided into a safe, stable phase, a treatment execution phase, an optimization / enhancement phase, and an abnormal / termination phase. When the risk level is high risk or an emergency shutdown occurs, the stage is directly marked as an abnormal / termination stage, and a termination stage control boundary table is issued, immediately imposing hard constraints on control variables; its infrared output power is set to zero, infrared mode switching is locked (switching is prohibited), environmental settings are rolled back to the safe setting combination, and ventilation and airflow are increased to above the guaranteed lower limit and maintained with stable airflow (airflow change rate is limited). When the risk level is a warning, or the overall credibility / key component credibility decreases, leading to increased assessment uncertainty, or when a steady-state baseline needs to be re-established during the initial startup and after stage switching, the system marks the stage as a safe steady-state stage and issues a low-disturbance stage control boundary table. The boundary table explicitly provides: upper and lower limits of infrared output power (low power range) and upper and lower limits of duty cycle (low duty cycle range), specifying the use of either "continuous low power" or "intermittent low duty cycle template"; upper limits of power increase rate of change, power decrease rate of change, duty cycle rate of change, and minimum interval for mode switching; upper limits of ambient temperature setting change rate, ambient humidity setting change rate, and airflow change rate; and a fixed activation of the ventilation guarantee lower limit. If the overall credibility or the credibility of any key component approaches the lower limit of availability, a uniform tightening coefficient is activated within the boundary table. The upper limits of relevant parameters are tightened proportionally according to the uniform tightening coefficient, and the "prohibit entry into the treatment execution stage" flag is set to true until the credibility recovers to the preset credibility threshold. When the risk level is "none," the comprehensive health score is within the target working range, the scoring trend does not show a continuous approach to the risk boundary, and the credibility of key components is not lower than the available threshold, the stage is marked as the treatment execution stage, and an execution control boundary table is issued. Closed-loop tracking and adjustment are performed within the infrared output and environmental regulation control domain specified in the boundary table to achieve the treatment goal. If the risk level is raised to the warning level or the scoring trend turns unfavorable during execution, the stage is rolled back to the safe steady-state stage according to the rollback rules.When the risk level is currently zero and remains stable, the overall health score is within a high safety margin range, the score trend meets the enhancement threshold (stable trend, volatility below the upper limit) for several consecutive assessment windows, and the overall confidence and the confidence of each key component are within a high confidence range, the stage is designated as the optimization and enhancement stage, and an enhancement control boundary table is issued. This boundary table makes slight enhancements to the infrared output or environmental settings without exceeding the infrared power safety limit and the environmental safety limit, while also having stricter perturbation constraints, namely, a smaller upper limit for the power change rate, a smaller upper limit for the duty cycle change rate, a longer minimum interval for mode switching, and smaller upper limits for the temperature setting change rate, humidity setting change rate, and airflow change rate. When any exit trigger signal occurs, the optimization and enhancement stage is immediately exited, and the system reverts to the treatment execution stage.

[0058] Module 4's stage management and treatment goal generation unit continues to generate the treatment goal set for this cycle. The treatment goal set outputs state goals item by item in the form of engineering parameters. Each item includes a target range or threshold, allowable deviation, upper limit of allowable rate of change, recovery time window, applicable stage range, triggering conditions, and rollback conditions. The target load is fixed as body temperature, heart rate, blood oxygen saturation, respiratory rate, cabin ambient temperature, cabin ambient humidity, and airflow and ventilation intensity. Simultaneously, behavioral indicators output from Module 1's animal activity monitoring unit are incorporated into the goal system as "state constraints and triggers," including activity intensity, agitation or stress intensity, postural stability, and duration of abnormal behavior. Goal generation employs explicit mapping rules: the contribution value of each item determines priority and tightening direction; the comprehensive health score and trend determine the tightening magnitude; and the risk level determines the target upper limit boundary and rollback strategy. When the contribution of thermal steady-state components is unfavorable or the trend is unfavorable, the treatment target set tightens the upper limit of the allowable rate of change in body temperature and extends the body temperature recovery time window, while giving coordinated targets for ambient temperature and airflow to suppress the accumulation of heat load; when the contribution of cardiovascular load components is unfavorable or the trend is unfavorable, the treatment target set tightens the upper limit of allowable deviation in heart rate and the upper limit of heart rate change rate, and sets the trigger condition of "infrared output can only change smoothly"; when the contribution of oxygenation and ventilation components is unfavorable or the trend is unfavorable, the treatment target set increases the priority of the lower limit of blood oxygen and respiratory stability targets, clarifies the lower limit of ventilation intensity and airflow stability targets, and gives the upper limit of allowable airflow fluctuation and the airflow recovery time window; when the contribution of stress and comfort components is unfavorable, the treatment target set gives thresholds and duration thresholds for activity intensity, stress intensity, postural instability and duration of abnormal behavior, and writes them as target triggers. Once the limits are exceeded, the target implementation boundary table tightening is triggered; when the contribution of environmental adaptation components is unfavorable, the treatment target set tightens the allowable deviation of temperature and humidity and extends the environmental recovery time window, and increases the priority of mutation suppression. If multiple sub-items are unfavorable at the same time, the treatment target set generates a joint conservative target, that is, the body temperature, heart rate, oxygenation, ventilation and environmental targets are uniformly rolled back to the center of the safe range and a longer recovery time window, and the triggered sub-item combination, the corresponding target parameter results and rollback conditions are recorded in the log to ensure traceability.

[0059] Module 4's stage management and treatment goal generation unit adopts a process of "expert rules - manual correction - data-driven parameter refinement" to output stage identifiers, stage control boundary tables, and treatment goal sets. In the stage where training samples are insufficient, the health status and risk prediction data packets output by Module 3's health status assessment and risk prediction unit are first mapped to stage and parameter results based on expert rules, along with the actuator working states returned by Module 2's infrared control unit and cabin environment control unit. Then, the operator performs item-by-item correction on the stage division, stage control boundary table, and treatment goal set through the user interface interaction unit in Module 5. Module 5's data management and reporting unit records the parameter versions before and after correction, forming structured samples linked by a unified timestamp. As samples accumulate, the data-driven model learns the mapping relationship between "health status and risk prediction data and actuator working states to manually corrected stage and goal outputs"; the model output is validated by passing the hard constraints of the stage control boundary table. During the initial stage of insufficient sample coverage, the data-driven model participates with low fusion weights and only performs limited parameter refinement on the rule outputs. As sample coverage increases, its fusion weights gradually increase according to a preset strategy to improve consistency with manual correction, and the final output is solidified into the configuration and logs in the form of engineering parameters.

[0060] S110: Personalized treatment plan generation.

[0061] The personalized treatment plan generation unit in Module 4 uses the stage control boundary table and the treatment target set as constraint inputs to generate a sequence of infrared physiotherapy control parameters. The control parameter sequence of the infrared physiotherapy device is determined in the order of "band - power - mode - irradiation method" and verified by boundary table constraints item by item: First, the band or band combination selection result is output, and the band switching rules are given (including the allowed switching stage range, minimum switching time interval, switching trigger conditions and prohibition conditions); then, the power setting trajectory is output, which is represented by a discrete time point sequence, including the initial power, target power, upper limit of power rise slope, upper limit of power fall slope, upper power limit and lower power limit, and the climbing strategy is specified (single-segment climbing, segmented climbing or step climbing, number of segments and duration of each segment); then, the irradiation mode parameters are output, clarifying the selection result of continuous, pulse or intermittent mode, and giving the duty cycle, period, single irradiation duration and interval time. The mode parameters are selected from the stage template and constrained by the rate of change; finally, the irradiation method parameters are output, clarifying the selection result of point-targeted or coverage irradiation, and giving the target area position, target area, scanning path or coverage area parameters.

[0062] Module 4's personalized treatment plan generation unit synchronously generates a sequence of cabin environment control parameters and implements a "coordinated, intensity-coupled, and disturbance-mutually-suppressive" linkage constraint with the infrared control parameter sequence. Near points of change in infrared band (wavelength) switching, output intensity (power / duty cycle) adjustment, irradiation mode switching, or irradiation method (targeted / coverage) switching, and near target area scanning parameters, the environmental side prioritizes smooth compensation for airflow and ventilation, gradually adjusting according to the lower limit of ventilation assurance and the upper limit of airflow change rate to suppress local heat load accumulation in the target area and avoid stress induced by sudden airflow changes. When the thermal steady-state components or trends indicate a high heat load, the environmental side prioritizes a combination of small-step temperature setting reduction and enhanced airflow stability, while limiting the rate of temperature setting change and the rate of airflow change, ensuring that "cooling and heat dissipation" are completed without introducing sudden changes. When oxygenation and ventilation are unfavorable and increased ventilation is required, the environmental control unit increases the lower limit of ventilation assurance while limiting airflow jumps. Simultaneously, it implements power limiting or intermittent extension with the infrared control unit, ensuring that "oxygenation" and "load reduction" occur synergistically and do not cancel each other out. The environmental control parameter sequence is determined in the order of "temperature-humidity-ventilation and airflow-purification," and each parameter is verified through boundary table constraints: outputting the ambient temperature setting trajectory and the upper limit of the temperature setting change rate; outputting the ambient humidity setting trajectory, humidity allowable deviation, and the upper limit of the humidity setting change rate; outputting the ventilation intensity setting value or setting trajectory, the upper limit of the change rate, and the minimum ventilation assurance lower limit; outputting the filter level and limiting its switching frequency to maintain stable air quality changes. After the infrared therapy and cabin environment control parameter sequence completes the stage control boundary table verification, it is submitted to the user interface interaction unit of module 5 for manual correction. After the correction generates an effective version, it is sent to the infrared control unit and cabin environment control unit of module 2, and the data management and reporting unit of module 5 records the parameters before and after correction and their effective versions, and archives them.

[0063] The personalized treatment plan generation unit in Module 4 follows the process of "expert rules - manual correction - data-driven parameter refinement". Based on sample accumulation, the data-driven model learns the mapping relationship between "stage control boundary table and treatment target set to the control parameter sequence of the manually corrected effective version". The model output refines the parameters of the set trajectory, duty cycle, ramp curve parameters and time window without breaking the hard constraints of the stage control boundary table, and participates in the generation of the effective version according to the fusion weight. The fusion weight gradually increases with the increase of sample coverage according to the preset strategy, so that the generation mechanism of control parameter sequence gradually transitions from expert rule-based to data-driven.

[0064] S111: The actuator executes the effective control parameters.

[0065] The infrared control unit of module 2 generates the effective version of the control parameter sequence according to the personalized treatment plan generation unit of module 4. It performs band (wavelength) selection and switching, output intensity (power / duty cycle) modulation, and irradiation mode control on the infrared physiotherapy device, and reads back the status quantities such as drive voltage / current and light source temperature rise for closed-loop correction and overheat protection. The band switching component executes the switching rules of the effective version and meets the minimum switching interval. The beam pointing is based on the fusion positioning of vision and millimeter-wave radar to achieve target area tracking and scanning for fixed-point targeting or coverage irradiation. The cabin environment control unit of module 2 performs temperature control, humidity control, and ventilation control. Temperature control is implemented by the cabin air conditioning unit and the air duct mixing valve, and is strictly constrained by the upper limit of the set temperature change rate. Humidity control is implemented by the humidifier and dehumidifier, and is strictly constrained by the upper limit of the set humidity change rate and the condensation risk threshold. Ventilation is implemented by the ventilation fan and electric air valve, and is strictly constrained by the minimum ventilation guarantee lower limit, the upper limit of the air volume change rate, and the maximum wind speed stimulation threshold. The configured filter purifier operates according to the purification level and its switching frequency, and reads back the level and air volume status for stability correction.

[0066] S112: Online strategy updates and intelligent fine-tuning.

[0067] Module 4's online strategy update and intelligent fine-tuning unit, within the current treatment cycle, uses the stage control boundary table and treatment target set as hard constraints. Through reinforcement learning, it performs small-scale online fine-tuning of the effective control parameters executed by Module 2's infrared control unit and cabin environment control unit, optimizing the execution control parameters without changing the treatment target or exceeding the action boundary. The state input of the reinforcement learning model is a health status and risk prediction data package, and the action output is the increment of the effective control parameters. This increment is preferably limited to a continuously adjustable subset of parameters (including infrared power, duty cycle, intermittent time, ambient temperature and humidity settings, airflow, and ventilation intensity), and does not switch discrete strategy items such as band, mode, and irradiation method when the stage control boundary table does not allow it. The reward function focuses on treatment target error and stability, and imposes penalty terms on the magnitude and rate of change of control quantities, behavioral disturbances, and risk trends. The unit reads the latest status in real time and calculates the target deviation and trend; it performs gating based on the overall credibility and the credibility of key components. When the credibility decreases, it automatically enters a conservative mode and tightens the fine-tuning amplitude; then, the reinforcement learning strategy outputs action increments, which are pruned by the stage control boundary table to meet the upper and lower limits and the upper limit of the rate of change. After behavior gating verification, executable fine-tuning instructions are generated and sent to the infrared control unit and cabin environment control unit of module 2, and the execution results are read back and written to the log. If the risk level is raised to the warning level or the scoring trend turns unfavorable, gain-type fine-tuning is immediately stopped and switched to de-scrambling fine-tuning. If necessary, stage rollback is triggered. Reinforcement learning adopts a combination of offline pre-training and online small-step updates. In the offline stage, the initial strategy is trained using historical structured samples; in the online stage, adaptive updates are performed with a small learning rate only when the risk level is currently zero and the credibility continuously meets the threshold and the control margin is sufficient. Once the warning / high-risk level is entered or the credibility falls below the threshold, online updates are frozen and the strategy is rolled back to the offline version. Each fine-tuning and strategy update records a unified timestamp, stage identifier, state representation parameters, and action increment, which are then linked and archived by the data management and reporting unit in module 5.

[0068] S113: Efficacy assessment at the end of the cycle.

[0069] The efficacy assessment unit at the end of the cycle in Module 3 uses the physiological baseline upon entry and the environmental reference data before entry as a baseline comparison. It performs cycle-level statistical induction on key assessment elements in the health status and risk prediction data package, calculating the difference between the cycle start window and the cycle end window, the mean drift over the entire cycle, quantile changes, and stability measures. These statistical results are then mapped to a quantitative representation of changes in the contribution values ​​of individual components and the net gain of the overall health score. Furthermore, this unit performs trend assessment, calculating the trend slope, rate of change, volatility, and abrupt change of the overall health score within the cycle, making a discriminative judgment on the efficacy pattern. The efficacy patterns include short-term improvement but unfavorable trend, no short-term improvement but improving trend, short-term improvement and stable improving trend, slow improvement but stable trend, unstable improvement and unfavorable trend, and unstable improvement but improving trend. Furthermore, the unit performs steady-state recovery and target achievement assessment, statistically analyzing the achievement rate of key state quantities, the number and duration of exceedances, and the recovery time window satisfaction rate based on the execution process records. It also performs periodic-level summarization of the smoothness, rate-of-change saturation ratio, and trimming trigger ratio of infrared and environmental control quantities to evaluate the convergence quality and disturbance control level of the closed-loop tracking of the treatment target. Furthermore, the unit performs risk burden quantification, periodically summarizing the risk level sequence, warning / high-risk / emergency shutdown trigger records, interlocking, and rollback events to form risk proportion, longest continuous duration, and trigger precursor characterization indicators. These trigger precursor characterization indicators include the degree of threshold approach and the early warning time for lower bound exceedances.

[0070] The cycle-end efficacy assessment unit in Module 3 registers the online fine-tuning process generated by the online strategy update and intelligent fine-tuning unit in Module 4 with the health status and risk prediction data package under the same time benchmark, forming a traceable action-state response sequence. Based on this, and combining the calculated baseline control results, trend assessment results, steady-state recovery and goal achievement statistics, and risk burden characterization indicators, a sample validity discrimination criterion is constructed. A conservative admission strategy is adopted for sample screening. When the overall credibility and key component credibility continuously meet the preset thresholds, and the baseline control and goal achievement evaluation meet the preset improvement discrimination thresholds, while the trend assessment does not show an unfavorable trend of continuously approaching the risk boundary, and the steady-state recovery indicators meet the preset convergence thresholds and the risk burden does not increase (allowing it to remain flat or decrease), the corresponding fine-tuning segments and cycle entries are marked as positive valid samples. When there is saturation of pruning or frequent triggering of gating that restricts executable actions, or when there is insufficient steady-state recovery, unstable goal achievement, or increased risk burden, the samples are marked as restricted samples according to the rules. The resulting positive valid samples are written into the sample library by the data management and reporting unit of module 5, and used by the online policy update and intelligent fine-tuning unit of module 4 for offline playback and policy update.

[0071] S114: Data review at the end of treatment.

[0072] Module 5 comprises a user interface interaction unit, a manual intervention and control unit, an alarm and notification unit, a data management and reporting unit, and an equipment operation and calibration unit, and runs throughout the entire operation of the infrared physiotherapy system. After the treatment, the operator can initiate a periodic review request through the user interface interaction unit of Module 5; the data management and reporting unit of Module 5 performs correlation retrieval and review presentation of all data and operation process for this period based on a unified timestamp; the unit supports report presentation to meet the needs of efficacy review and subsequent period parameter configuration optimization.

Claims

1. An infrared physiotherapy chamber system based on the synergistic regulation of animals and the environment, characterized in that, include: The animal status and cabin environment monitoring module is used to collect and output animal physiological data, animal activity and behavior data, and cabin environment data. The infrared physiotherapy and cabin environment control module is used to receive effective control commands and execute infrared physiotherapy output and cabin environment control, while reading back the actuator working status and outputting it. The animal health assessment and efficacy assessment module is used to synchronize and align the time-series data output by the animal status and cabin environment monitoring module to form a multimodal aligned time-series data stream, and to perform online data quality assessment and credibility management for each channel. Under credibility gating, it outputs health status and risk prediction results as well as efficacy assessment results at the end of the cycle. The personalized treatment and intelligent decision-making module generates stage identifiers, stage control boundary tables, and treatment target sets based on health status and risk prediction results and combined with the actuator working status. Under the constraints of the stage control boundary table and treatment target sets, it generates infrared physiotherapy control parameters and cabin environment regulation control parameters and sends them to the infrared physiotherapy and cabin environment regulation modules for execution. It is also used to fine-tune the control parameters online during the execution process. The human-computer interaction and safety supervision module is used for information input, parameter verification or correction, manual intervention and control, alarm and interlock handling, data recording and traceability, and equipment operation and maintenance and calibration management. The process is divided into four phases: a safe steady-state phase, a treatment execution phase, an optimization and enhancement phase, and an abnormal / termination phase. The abnormal / termination phase is identified by issuing a control boundary table, setting the infrared output power to zero, locking the infrared mode switch, reverting the environmental settings to a safe combination, and increasing ventilation and airflow to above the guaranteed lower limit while maintaining stable airflow. The safe steady-state phase is identified by issuing a low-disturbance phase control boundary table. This low-disturbance phase control boundary table provides: upper and lower limits for infrared output power and duty cycle, specifying the use of either a low-power continuous or intermittent low duty cycle template; upper limits for power increase rate, power decrease rate, duty cycle rate, and minimum mode switching interval; and upper limits for environmental temperature setting change rate, environmental humidity setting change rate, and airflow rate change rate. The ventilation guarantee lower limit is fixed and activated; the stage is marked as the treatment execution stage, and the execution control boundary table is issued. Closed-loop tracking adjustment is carried out within the infrared output and environmental regulation control domain specified in the boundary table to achieve the treatment goal; the stage is marked as the optimization and enhancement stage, and the enhancement control boundary table is issued; the boundary table makes a small enhancement to the infrared output or environmental settings without exceeding the infrared power safety limit and the environmental safety limit, while having stricter disturbance constraints, namely a smaller upper limit for power change rate, a smaller upper limit for duty cycle change rate, a longer minimum interval for mode switching, and a smaller upper limit for temperature setting change rate, humidity setting change rate, and air volume change rate. The treatment target set outputs the state target item by item in the form of engineering parameters. Each item includes the target range or threshold, allowable deviation, upper limit of allowable rate of change, recovery time window, applicable stage range, triggering condition and rollback condition.

2. The infrared physiotherapy chamber system based on the synergistic regulation of animals and the environment according to claim 1, characterized in that, The animal status and cabin environment monitoring module includes an animal physiological monitoring unit, an animal activity monitoring unit, and a cabin environment monitoring unit. The animal activity monitoring unit includes a camera acquisition component and a millimeter-wave radar acquisition component, used to output animal behavioral characteristics and animal position or movement parameters. The cabin environment monitoring unit includes a temperature and humidity sensor and an airflow sensor, used to output cabin temperature, humidity, airflow, or ventilation-related parameters. The animal physiological monitoring unit includes at least one or a combination of a photoplethysmography (PPG) sensor for acquiring heart rate and blood oxygen saturation, a chest strap sensor for acquiring respiratory rate, and a non-contact body temperature sensor for acquiring body temperature. After the system self-test passes, the cabin environment monitoring unit of the animal status and cabin environment monitoring module enters a stable sampling phase, where it performs smoothing processing on the cabin temperature, humidity, and airflow data within a preset time window to form pre-entry environmental reference parameters. These pre-entry environmental reference parameters serve as a benchmark for subsequent environmental control and anomaly detection.

3. The infrared therapy chamber system based on the synergistic regulation of animals and the environment according to claim 1, characterized in that, The infrared physiotherapy and cabin environment control module includes an infrared control unit and a cabin environment control unit; the infrared control unit is used to execute infrared physiotherapy output control commands and read back the drive and equipment status; the cabin environment control unit is used to execute temperature regulation, humidity regulation, and airflow and ventilation control commands and read back the working status of the environmental actuator.

4. The infrared therapy chamber system based on the synergistic regulation of animals and the environment according to claim 1, characterized in that, The animal health assessment and efficacy assessment module includes a data quality assessment and credibility management unit, a health status assessment and risk prediction unit, and a cycle-end efficacy assessment unit. The data quality assessment and credibility management unit uses the master clock as a unified time reference to synchronize and align the dispersed time-series data, and performs resampling, interpolation, or sliding window registration on data with different sampling rates to form a multimodal aligned time-series data stream. Online quality assessment and credibility update are performed on each channel. The credibility update includes at least: basic availability determination based on data integrity and timing characteristics, quality assessment based on channel signal quality, and fusion weighting processing based on cross-channel consistency. The credibility results are then updated with time smoothing to form a channel credibility sequence and overall credibility. The data quality assessment and credibility management unit is used to identify and handle at least one type of abnormality mode among transient artifacts, short-term missing data, persistent offsets, and systematic anomalies. Specifically, robust filtering or outlier suppression is performed on transient artifacts and replaced with neighborhood estimation. For short-term missing data, interpolation or prediction completion is performed when recoverable conditions are met. For persistent offsets, online bias compensation is performed when correctable, and when uncorrectable, handling events such as re-wearing, re-alignment, calibration, or clearing obstructions are generated. When a systematic anomaly meets the triggering conditions, a stop interlock trigger record is generated and output.

5. The infrared therapy chamber system based on the synergistic regulation of animals and the environment according to claim 4, characterized in that, The health status assessment and risk prediction unit performs individualized baseline processing on current observations based on pre-entry environmental reference parameters, animal physiological baselines, or historical health baselines. It constructs a unified health status vector based on the multimodal aligned time-series data stream, outputting the contributions of thermal steady-state, cardiovascular load, oxygenation and ventilation, stress and comfort, and environmental adaptation components. It also performs gating and fusion adjustment on the comprehensive health score by combining channel confidence sequences or overall confidence. Based on the prediction model, it generates future multimodal states and outputs future comprehensive health scores and score trend indicators, ultimately determining the risk level. The risk levels include none, warning, high risk, and emergency shutdown; when the overall credibility is insufficient, a conservative strategy is adopted to output suspected risk warnings.

6. The infrared physiotherapy chamber system based on the synergistic regulation of animals and the environment according to claim 4, characterized in that, The cycle-end efficacy assessment unit is used to summarize and generalize process data, control execution records, and assessment results after a treatment cycle ends. Based on the physiological baseline upon entry and the environmental reference parameters before entry, a baseline comparison is performed, and cycle-level efficacy quantitative conclusions and traceable cycle-level assessment records are output. The online fine-tuning records generated by the online strategy update and intelligent fine-tuning unit are time-registered with the health status and risk prediction results output by the health status assessment and risk prediction unit under a unified time reference to form an action-state response sequence, and a structured sample is generated based on this sequence for subsequent strategy assessment or parameter optimization.

7. The infrared physiotherapy chamber system based on the synergistic regulation of animals and the environment according to claim 5, characterized in that, The personalized treatment and intelligent decision-making module includes a stage management and treatment goal generation unit, a personalized treatment plan generation unit, and an online strategy update and intelligent fine-tuning unit. The stage management and treatment goal generation unit generates stage identifiers, stage control boundary tables, and treatment goal sets based on the risk level, comprehensive health score, and score trend indicators, combined with the actuator working status read back from the infrared physiotherapy and cabin environment control modules. The stages include at least a safe steady-state stage, a treatment execution stage, an optimization and enhancement stage, and an abnormal or termination stage, and corresponding stage control boundary tables and backoff constraints are issued for different stages. The personalized treatment plan generation unit generates infrared physiotherapy control parameters and cabin environment control parameters under the hard constraints of the stage control boundary tables and treatment goal sets, and implements cross-channel collaborative control of the two types of control parameters. The system employs control and linkage constraints to ensure that changes in infrared heat input are consistent with environmental temperature, humidity, airflow, and ventilation responses. The infrared therapy control parameters and cabin environment control parameters generated by the personalized treatment plan generation unit are submitted to the human-computer interaction and safety monitoring module for manual correction before being sent to the infrared therapy and cabin environment control module for execution, thus forming an effective version. The human-computer interaction and safety monitoring module records the parameters before and after correction, as well as their effective versions, and archives them. The online strategy update and intelligent fine-tuning unit, without exceeding the hard constraints of the stage control boundary table and treatment target set, uses reinforcement learning to perform small-scale online fine-tuning of continuously adjustable parameters, and gates the fine-tuning amplitude based on overall credibility or key component credibility. When the risk level increases or credibility decreases, the fine-tuning amplitude is tightened or online updates are frozen and a rollback is triggered.

8. The infrared physiotherapy chamber system based on the synergistic regulation of animals and the environment according to claim 1, characterized in that, The human-computer interaction and safety supervision module includes a user interface interaction unit, a manual intervention and control unit, an alarm and notification unit, a data management and reporting unit, and an equipment operation and calibration unit. The user interface interaction unit is used for basic animal information input, threshold and configuration loading, operation status presentation, and parameter verification or correction of control parameters. The manual intervention and control unit is used to provide manual intervention, interlock permission, or manual control channels in specific scenarios. The alarm and notification unit is used to provide alerts for changes in risk level, credibility anomalies, channel anomalies, and safety events, and to output handling suggestions or interlock handling prompts. The data management and reporting unit is used to associate and archive multimodal data, evaluation data packages, control parameter outputs, online fine-tuning records, and interlock alarm events according to a unified time base, and output traceable reports. The equipment operation and calibration unit is used to perform self-checks on the door lock, emergency stop switch, power supply and grounding status before entering the physiotherapy execution state, and to verify the connection status, response status and calibration validity of sensors and actuators. When the verification is passed, calibration parameters are loaded and reset processing is performed when necessary. The door lock status, emergency stop status and system availability verification are used as prerequisites for entering the physiotherapy execution state.

9. The infrared physiotherapy chamber system based on the synergistic regulation of animals and the environment according to claim 2, characterized in that, Animal physiological monitoring units are used to collect and output one or more of the following: heart rate and blood oxygen saturation, respiratory rate, and body temperature.

10. A method for an infrared therapy chamber based on the synergistic regulation of animals and the environment, applied to the device described in any one of claims 1-9, characterized in that, include: Pre-inspection of the cabin and system initialization, self-test of power supply and grounding status, and verification of sensors and actuators; Before entering the cabin, environmental reference parameters are established, and cabin temperature, humidity and airflow data are collected and processed by moving average. Animal basic information entry; Sensor wearing and multimodal monitoring are established; data is collected and a multimodal aligned timing data stream is generated. The scattered timing data is synchronously aligned with the master control clock as a reference to form a multimodal aligned timing data stream. Online data quality assessment and credibility management: online quality assessment and credibility update of multimodal aligned time-series data streams, identification and handling of anomalies, and output of data quality and credibility data packets; Health status assessment and risk prediction: Based on multimodal aligned time-series data streams and credibility data packets, construct a health status vector, calculate the contribution values ​​of each component and the comprehensive health score, construct a scoring trend index, and output the risk level; Phase management and treatment goal generation: Generate phase identifiers, phase control boundary tables, and treatment goal sets based on health status, risk level, and actuator status; Personalized treatment plans are generated by generating a sequence of control parameters for infrared physiotherapy and cabin environment regulation under the constraints of the stage control boundary table and the treatment target set. The actuator executes the effective control parameters, and triggers a rollback when the risk is increased or the credibility is decreased, with the stage control boundary table and the treatment target set as hard constraints. At the end of the cycle, the efficacy was evaluated, and baseline control was implemented by comparing the physiological information upon entry into the cabin with the environmental reference data before entry. Cycle statistics and efficacy patterns were calculated.

Images