Time-limited feeding method and system for animal experiment

Abstract

The invention provides an animal experiment time-limited feeding method and system. The animal experiment time-limited feeding method and system are applied to an intelligent door control system comprising a feeding box body, an electric control door control channel, a multi-mode sensor array and an intelligent control unit. According to the method, monitoring data streams of animal states and behaviors are obtained through a multi-modal sensor array; implementing execution control of the time-limited feeding mode based on the monitoring data stream and preset parameters; outputting a personalized time-limited feeding intervention strategy based on the execution data and the monitoring data stream; and management and analysis of experimental data are realized through the data management module. The invention further provides a corresponding time-limited feeding system, the system has dynamic adaptability, comprehensive monitoring dimension, high flexibility and deep data integration capacity, the problems that in the prior art, manual control accuracy is poor, intelligent adaptability and data deep mining are lacked are solved, and the system is suitable for large-scale popularization and application. And the accuracy, scientificity and animal welfare level of animal experiments are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of biomedical experimental technology, in particular to a time-limited feeding method and system for animal experiments, which is used for accurately controlling and monitoring the feeding time, behavior and physiological state of experimental animals, and provides technical support for studying the interaction between biological clock and metabolism and exploring intervention strategies for metabolic diseases. BACKGROUND

[0002] Time-restricted feeding (TRF or Time-Restricted Eating, TRE) as a new dietary intervention model has received extensive attention in the biomedical field in recent years. With the changes in modern lifestyle, factors such as irregular diet and lack of exercise have led to an increasing incidence of metabolic diseases such as obesity, diabetes and cardiovascular disease, which has brought a huge burden to global health. Studies have found that biological clock is closely related to metabolic rhythm, and biological clock disorder can directly lead to metabolic disorder. The core idea of time-restricted feeding is to set a fixed feeding time window (usually 4-12 hours) every day, extend the night fasting time, and not strictly limit the total calorie intake, aiming to improve metabolic health by coordinating the feeding time and biological clock rhythm.

[0003] The theoretical basis of time-restricted feeding believes that regular feeding and fasting cycles help maintain normal biological rhythms. The potential physiological mechanisms include: after a long period of fasting, the body's energy source shifts from glucose to fatty acids and ketone bodies, activating cellular adaptive responses; by adjusting the feeding time to synchronize with the circadian rhythm, optimizing the function of metabolic organs (such as liver, adipose tissue); regulating the structure and diversity of intestinal flora, and then affecting the host metabolism. Common time-restricted feeding patterns include 18:6, 16:8, 14:10 and other patterns, the core of which is "when to eat" rather than "how much to eat".

[0004] Both animal experiments and human studies have shown that time-restricted feeding has potential benefits in weight regulation, improving blood glucose control (such as reducing fasting insulin, glycosylated hemoglobin, diabetes conversion rate) and regulating lipid metabolism (such as reducing triglycerides, cholesterol). Therefore, time-restricted feeding is considered an important research tool for studying the interaction between biological clock and metabolism and exploring intervention strategies for metabolic diseases.

[0005] However, currently, in the implementation of time-restricted feeding in animal models, it is traditionally usually dependent on researchers manually controlling the feeding and removal of feed, or using simple timing devices for feed supply. These methods not only consume time and effort, but also are difficult to ensure the accuracy of time and operation.

[0006] The more advanced solutions in the prior art adopt a timing control-based automated feed dispensing system, which controls the opening and closing of the feed trough through a preset schedule, realizing the basic needs of animal time-limited feeding experiments. For example, there are mouse energy metabolism monitoring systems on the market, which are powerful and integrated with energy metabolism, behavior, physiological parameters, etc. and provide options for automated feeding control, but there is still room for improvement in terms of flexible, intelligent and adaptive time-limited feeding intervention.

[0007] The technical defects of the current commonly used time-limited feeding automation implementation solutions mainly include:

[0008] Lack of dynamic adaptability: Most systems are based on fixed schedules and lack the ability to dynamically adjust the feeding window according to the animal's real-time physiological state (such as body weight changes, activity level, actual intake) and environmental factors.

[0009] Insufficient monitoring dimensions: Simple timing switches cannot monitor the details of the animal's actual feeding behavior (such as feeding frequency, single feeding amount, micro-feeding events) and its immediate impact on physiological state in real time.

[0010] Lack of flexibility and individualization: Fixed schedules are difficult to adapt to complex and variable time-limited feeding experimental designs (such as different mode switching, individualized adjustment needs), and are also difficult to cope with individual differences commonly encountered in research.

[0011] Limited data integration and analysis: Existing systems often lack the ability to efficiently integrate, manage and deeply analyze multi-modal data such as feeding control, animal behavior and physiological state, limiting research efficiency and in-depth interpretation of results.

[0012] In addition, time-limited feeding research often has high requirements for animal housing environments, especially for long-term experiments, which need to ensure animal welfare and data reliability. However, existing solutions still lack intelligent adaptation, deep data mining, and integration with complex physiological monitoring (such as continuous blood glucose, heart rate variability, etc.). SUMMARY

[0013] To solve the above technical problems, the present application provides an animal experiment time-limited feeding method and system, which has intelligent self-adaptation, multi-dimensional monitoring and data deep analysis, etc. and can realize accurate control and comprehensive monitoring of animal time-limited feeding experiments.

[0014] The first aspect of the present application provides an animal experiment time-limited feeding method applied to an intelligent door control system, which includes an experimental animal housing box, an electrically controlled door control channel, a multi-modal sensor array and an intelligent control unit, and the method comprises:

[0015] A multi-modal sensor array containing food intake monitoring sensors, animal activity state monitoring sensors, physiological parameter monitoring sensors, and environmental parameter monitoring sensors is used to obtain multi-modal monitoring data streams of the experimental animal state and behavior.

[0016] Based on the multi-modal monitoring data streams and preset time-limited feeding parameters, an eating time window control algorithm, a dynamic adjustment algorithm, and a safety monitoring mechanism are used to implement execution control of the time-limited feeding mode, and execution data of the execution control is obtained.

[0017] Based on the execution data of the execution control and the multi-modal monitoring data streams, an animal state evaluation model, a diet mode optimization algorithm, and an intervention strategy generator are used to output personalized time-limited feeding intervention strategies.

[0018] Based on the multi-modal monitoring data streams, the execution data of the execution control, and the personalized time-limited feeding intervention strategies, a data acquisition and storage module, a data processing and analysis module, and a visualization interface are used to implement experimental data management and analysis.

[0019] Preferably, the multi-modal monitoring data streams of the experimental animal state and behavior include:

[0020] The food intake monitoring sensors containing weight sensors and miniature cameras are used to monitor food consumption in real time and filter out interference signals, obtaining food intake data.

[0021] The animal activity state monitoring sensors containing infrared activity detectors or acceleration sensors are used to identify and analyze animal movement patterns, obtaining activity classification data.

[0022] Based on the physiological parameter monitoring sensors, the physiological parameters of the experimental animal are monitored, obtaining physiological parameter data.

[0023] The environmental parameter monitoring sensors containing temperature and humidity pressure sensors are used to collect and process environmental parameters, obtaining environmental change data.

[0024] Based on the food intake data, the activity classification data, the physiological parameter data, and the environmental change data, time alignment and multi-source data fusion are performed to generate the multi-modal monitoring data streams.

[0025] Preferably, the implementation of the execution control of the time-limited feeding mode includes:

[0026] Based on the multi-modal monitoring data streams and the preset time-limited feeding parameters, a time trigger mechanism is used to determine the eating time window, and in combination with state monitoring feedback, an eating time window switch event table is generated.

[0027] controlling, based on the eating time window switch event table, a switch state of the electrically controlled gate channel through a state machine model;

[0028] based on the multi-modal monitoring data stream, an abnormal state detection algorithm constructed using a multi-level threshold model is used to identify potential health risks and abnormal behaviors, and when a potential health risk or abnormal behavior is identified, an emergency response mechanism is triggered;

[0029] based on the emergency response mechanism, in combination with the eating time window switch event table and the switch state of the electrically controlled gate channel, a gate execution instruction is generated and prioritized, and the execution control of the time-limited eating mode is realized.

[0030] Preferably, the abnormal state detection algorithm constructed using a multi-level threshold model to identify potential health risks and abnormal behaviors, and when a potential health risk or abnormal behavior is identified, an emergency response mechanism is triggered, including:

[0031] based on the multi-modal monitoring data stream, the multi-modal monitoring data stream is processed in multiple scales to construct a multi-scale time representation, and a cross-scale attention module is used to interact features of different scales of the multi-scale time representation to form multi-time scale features;

[0032] based on the multi-time scale features, a Bayesian neural network is used to calculate health status indicators and their uncertainties, and an abnormal type is identified; wherein the abnormal type includes high confidence anomaly, high uncertainty anomaly and multi-modal conflict;

[0033] based on the high confidence anomaly, high uncertainty anomaly and multi-modal conflict, different response strategies are selected according to the abnormal type, and the emergency response mechanism is triggered.

[0034] Preferably, the output of the personalized time-limited eating intervention strategy includes:

[0035] based on the multi-modal monitoring data stream, an animal health status evaluation model is constructed through a feature extraction algorithm and a machine learning model;

[0036] based on the animal health status evaluation model and the execution data of the execution control, a diet pattern influence analysis result is constructed through time series data analysis and correlation calculation;

[0037] based on the diet pattern influence analysis result and the preset research target, a personalized diet pattern optimization scheme is generated through a multi-objective optimization algorithm and dynamic sparse training;

[0038] based on the personalized diet pattern optimization scheme, a personalized time-limited eating intervention strategy is output through a rule conversion engine and parameter mapping.

[0039] Preferably, the generating the personalized dietary pattern optimization scheme by the multi-objective optimization algorithm and the dynamic sparse training comprises:

[0040] Based on the dietary pattern influence analysis result, a joint feature representation containing time-related features, physiological response features and behavior pattern features is extracted to form an embedding representation of a high-dimensional vector space;

[0041] Based on the embedding representation, an importance score matrix is calculated by feature importance evaluation to identify key feature dimensions;

[0042] Based on the limitation of computing resources, the key feature dimensions are subjected to feature sparsification processing, and features with importance lower than a preset threshold are pruned to obtain pruned key feature dimensions; and the pruned key feature dimensions are subjected to computing level allocation to form a lightweight optimization model;

[0043] Based on the lightweight optimization model and guided by a preset research objective, a multi-objective optimization search is performed to generate a personalized dietary pattern optimization scheme.

[0044] Preferably, the method further comprises:

[0045] Based on the multi-modal monitoring data stream, the execution data of the execution control and the personalized time-limited eating intervention strategy, the traceable storage of multi-channel data is obtained by a time series database and a data synchronization mechanism.

[0046] For the stored multi-channel data, a pre-processing data set is generated by noise filtering, data standardization and missing value processing;

[0047] The pre-processing data set is processed by multiple data analysis methods to extract experimental results and pattern features; the multiple data analysis methods include statistical analysis, machine learning and time series analysis algorithms;

[0048] Based on the experimental results and pattern features, experimental data management and analysis are realized by interactive charts and multi-dimensional data visualization technology.

[0049] Preferably, the method further comprises:

[0050] Based on the execution result of the personalized time-limited eating intervention strategy and the multi-modal monitoring data stream, a comprehensive effect score is calculated by an effect evaluation algorithm to generate an intervention effect evaluation result;

[0051] When the intervention effect evaluation result is lower than a preset effect threshold, a feedback adjustment mechanism is started for multi-angle investigation to generate an adjustment suggestion;

[0052] Based on the adjustment suggestion, the animal state evaluation model and the optimization strategy are updated to realize iterative optimization of the intervention strategy.

[0053] Preferably, before the multi-modal monitoring data stream of the experimental animal state and behavior is acquired, the method further comprises:

[0054] Based on the experimental requirements and the animal type, a separate feeding space and a dedicated passage area are designed, and a feeding box basic structure meeting the SPF level cleaning standard is constructed;

[0055] Based on the feeding box basic structure, a sliding door grid driven by a micro servo motor and a position sensor are integrated to construct the electric control door passage;

[0056] Based on the feeding box basic structure and the electric control door passage, the food intake monitoring sensor, the animal activity state monitoring sensor, the physiological parameter monitoring sensor and the environmental parameter monitoring sensor are deployed to construct the multi-modal sensor array;

[0057] Based on the multi-modal sensor array and the electric control door passage, an embedded processor, a motor drive circuit and a high-precision analog-to-digital conversion module are integrated to construct the intelligent control unit.

[0058] The second aspect of the present application provides an animal experimental time-limited feeding system, comprising:

[0059] A feeding box for accommodating experimental animals;

[0060] An electric control door passage arranged on the feeding box for controlling the experimental animals to enter the food area;

[0061] A multi-modal sensor array comprising a food intake monitoring sensor, an animal activity state monitoring sensor, a physiological parameter monitoring sensor and an environmental parameter monitoring sensor for acquiring a multi-modal monitoring data stream of the experimental animal state and behavior;

[0062] An intelligent control unit connected with the electric control door passage and the multi-modal sensor array, comprising:

[0063] A data acquisition module for acquiring a multi-modal monitoring data stream of the experimental animal state and behavior through the multi-modal sensor array comprising the food intake monitoring sensor, the animal activity state monitoring sensor, the physiological parameter monitoring sensor and the environmental parameter monitoring sensor;

[0064] A feeding control module for realizing execution control of the time-limited feeding mode based on the multi-modal monitoring data stream and the preset time-limited feeding parameters through a feeding time window control algorithm, a dynamic adjustment algorithm and a safety monitoring mechanism to obtain execution data of the execution control;

[0065] an adaptive adjustment module for outputting a personalized time-limited feeding intervention strategy through an animal state evaluation model, a diet pattern optimization algorithm, and an intervention strategy generator based on the execution data of the execution control and the multi-modal monitoring data stream;

[0066] a data management module for realizing experimental data management and analysis through a data acquisition storage module, a data processing analysis module, and a visualization display interface based on the multi-modal monitoring data stream, the execution data of the execution control, and the personalized time-limited feeding intervention strategy.

[0067] The present application has the following beneficial effects:

[0068] 1. Through the hardware construction of the intelligent door control system, the high integration of the animal feeding box is realized, ensuring that the system meets the SPF clean standard, and the precise control of the animal feeding behavior is realized through the electric control door control channel.

[0069] 2. Through the multi-modal sensing and monitoring system, the all-round monitoring of the animal food intake, activity state, physiological parameters, and environmental parameters is realized, and rich multi-dimensional data stream is obtained, providing a solid foundation for subsequent analysis and adjustment.

[0070] 3. Through the programmed control of the time-limited feeding mode, the precise control of the feeding time window is realized, and through the dynamic adjustment algorithm and safety monitoring mechanism, the safety and reliability of the experimental process are ensured.

[0071] 4. Through the adaptive adjustment algorithm, the personalized intervention strategy generation based on the individual differences and state changes of animals is realized, greatly improving the scientificity and accuracy of the experiment.

[0072] 5. Through the data management and experimental analysis platform, the efficient management, processing, and visualization of experimental data are realized, improving the research efficiency and in-depth interpretation ability of the results. BRIEF DESCRIPTION OF DRAWINGS

[0073] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0074] Figure 1 is a flowchart of the animal experiment time-limited feeding method provided by the embodiments of the present application;

[0075] Figure 2 is an architectural schematic diagram of the intelligent door control system provided by the embodiments of the present application;

[0076] Figure 3 is a schematic diagram of a multi-modal sensing and monitoring system architecture provided by an embodiment of the present application;

[0077] Figure 4 is a structural schematic diagram of a time-limited feeding system for animal experiments provided by an embodiment of the present application;

[0078] Figure 5 is a structural schematic diagram of a time-limited feeding device for animal experiments provided by an embodiment of the present application. DETAILED DESCRIPTION

[0079] The embodiments of the present application will be described in detail with specific embodiments, and those skilled in the art can easily understand other advantages and effects of the present application from the disclosure of the present specification. The present application can also be implemented or applied by different specific embodiments, and the details in the specification can be modified or changed based on different views and applications without departing from the spirit of the present application. It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0080] Please refer to Figure 1 , Figure 1 is a flowchart of a time-limited feeding method for animal experiments provided by an embodiment of the present application. As shown in Figure 1 , the present embodiment provides a time-limited feeding method for animal experiments, applied to an intelligent door control system, the intelligent door control system comprising a feeding box of experimental animals, an electrically controlled door control channel, a multi-modal sensor array and an intelligent control unit, the method comprising:

[0081] S100, acquiring a multi-modal monitoring data stream of the state and behavior of experimental animals through the multi-modal sensor array comprising a food intake monitoring sensor, an animal activity state monitoring sensor, a physiological parameter monitoring sensor and an environmental parameter monitoring sensor;

[0082] S200, based on the multi-modal monitoring data stream and preset time-limited feeding parameters, implementing execution control of a time-limited feeding mode through a feeding time window control algorithm, a dynamic adjustment algorithm and a safety monitoring mechanism, and obtaining execution data of the execution control;

[0083] Based on the multi-modal monitoring data stream and the preset time-limited feeding parameters, the execution control of the time-limited feeding mode is realized through the feeding time window control algorithm, the dynamic adjustment algorithm and the safety monitoring mechanism, and the execution data of the execution control is obtained. This step first provides a variety of preset mode configuration options through a user-friendly graphical interface, including the classic 16 / 8 mode (16 hours of fasting and 8 hours of feeding), the 14 / 10 mode, the 5:2 intermittent fasting mode, etc. It also supports researchers to customize the feeding window according to specific research needs. The parameters set will be verified for scientificity and safety to ensure that the experimental animals will not be at risk due to too long fasting period.

[0084] After the configuration is completed, a time reference based on a real-time clock is established, and an accurate window opening and closing event table is generated. Unlike traditional systems, this system not only relies on a preset time table, but also monitors the animal state in real time through a state monitoring feedback mechanism. For example, when it is detected that the animal is feeding and the feeding window is about to close, the closing operation will be intelligently delayed to avoid the stress reaction that may be caused by sudden interruption of feeding. The system uses a state machine model to manage the gate control state, including standby, opening request, opening, opening completion, closing request and other states, and achieves millimeter-level accurate positioning of the gate control through a PID controller. At the same time, the current monitoring mechanism is used to prevent the gate control mechanism from being stuck or overloaded.

[0085] The safety monitoring mechanism is the core guarantee part of the S200 step, which uses a multi-level threshold model to continuously analyze the monitoring data stream, and divides the detected abnormalities into three levels of warning, moderate risk and high risk. When potential health risks are found, the corresponding emergency response mechanism is started according to the severity of the risk. For example, if it is detected that the animal has hardly taken any food in 24 hours and the body weight is rapidly decreasing, the food channel may be temporarily opened even during the original fasting period. All emergency events will be notified to the researchers through multiple channels, including mobile application push, SMS or email, to ensure that emergency situations are handled in a timely manner. The whole execution control process uses a priority sorting mechanism to ensure that safety-related instructions always have the highest execution priority, followed by dynamic adjustment instructions, and finally basic time window instructions.

[0086] S300, based on the execution data of the execution control and the multi-modal monitoring data stream, through an animal state evaluation model, a diet mode optimization algorithm and an intervention strategy generator, an individualized time-limited feeding intervention strategy is output.

[0087] Based on the execution data of the execution control and the multi-modal monitoring data stream, through the animal state evaluation model, the diet pattern optimization algorithm and the intervention strategy generator, the personalized time-limited feeding intervention strategy is output. This step first analyzes the key health indicators in the monitoring data stream through the animal health state evaluation model, including body weight change trend, activity frequency and intensity, feeding behavior pattern, etc. Machine learning models are used to quantitatively evaluate the current health status of animals, providing basic data support for subsequent decision-making.

[0088] Subsequently, the diet pattern influence analysis algorithm studies the impact of the current time-limited feeding pattern on the animal's health status through time series data analysis and correlation calculation. This algorithm not only focuses on simple direct correlations, but also attempts to separate the effects of time-limited feeding intervention and other environmental factors through causal inference methods. Based on these analysis results, a personalized diet pattern optimization scheme is generated through multi-objective optimization algorithms and dynamic sparse training techniques. This process takes into account research goals and individual animal characteristics, seeking the best balance between experimental scientific value and animal welfare.

[0089] The intervention strategy generator is the final link of step S300, which converts abstract optimization schemes into specific executable control instructions. The strategy generator uses a knowledge graph-based rule conversion engine that contains expert knowledge and experience in the field of time-limited feeding research. For example, when the optimization algorithm suggests "increase the morning feeding window to improve metabolic rate", the rule conversion engine will parse this abstract goal into a series of specific instructions such as "turn on the morning feeding window 30 minutes earlier". The parameter mapping module further quantifies these rules into precise system parameters, taking into account hardware limitations, safety boundaries and operation precision. Finally, the execution plan generation algorithm organizes these parameters into a time series execution plan to generate a sequence of instructions that the control system can directly understand, ensuring that the personalized intervention strategy can be accurately executed.

[0090] S400, based on the multi-modal monitoring data stream, the execution data of the execution control and the personalized time-limited feeding intervention strategy, through the data acquisition and storage module, the data processing and analysis module and the visualization interface, the experimental data management and analysis is realized.

[0091] Based on the multi-modal monitoring data stream, the execution data of the execution control and the personalized time-limited feeding intervention strategy, through the data acquisition and storage module, the data processing and analysis module and the visualization interface, the experimental data management and analysis is realized. This step first ensures that all sensor raw high-resolution data is saved completely, using a time series database for storage, establishing a multi-channel data synchronization mechanism to ensure data traceability and integrity. The data preprocessing module performs noise filtering, data standardization and missing value processing on the stored raw data to generate reliable preprocessed data sets for subsequent analysis.

[0092] A rich library of multi-dimensional analysis algorithms is provided, including statistical analysis tools, machine learning models, time series analysis methods, behavior pattern analysis techniques, and energy metabolism calculation formulas. Researchers can perform in-depth data mining based on these tools, support user-defined data grouping and standardized processing. The analysis results are presented through an intuitive visualization display system, providing interactive charts, heat maps, trajectory graphs, and other visualization methods to support comprehensive display of multi-dimensional data and customized report templates.

[0093] The data management module also integrates a research report and conclusion generator function, which assists researchers in summarizing experimental findings and forming preliminary conclusions through conclusion derivation algorithms and natural language generation techniques. It provides standardized data output formats compatible with common analysis software, facilitating more professional data analysis for researchers. In addition, the system integrates an alarm function based on user-defined thresholds, which can notify researchers in a timely manner when key indicators exceed the preset range. Through this comprehensive data management and analysis platform, researchers can fully grasp the experimental process and results, improving research efficiency and data mining depth, and providing strong data support for interpreting the scientific mechanisms of time-limited feeding.

[0094] In particular, it also supports a continuous improvement mechanism for intervention strategies, calculates the comprehensive effect score of intervention measures through an effect evaluation algorithm, and when the effect evaluation result is lower than expected, initiates a feedback adjustment mechanism for cause analysis and generates targeted adjustment suggestions. An incremental learning strategy is adopted to continuously update the health status evaluation model and optimize the strategy as experimental data accumulates, making the system more and more adaptable to specific experimental subjects. This self-adaptive closed-loop design is the key advantage of this system over traditional fixed parameter systems, enabling it to gradually optimize individualized intervention plans for different animal groups, significantly improving the scientific value and animal welfare level of experiments.

[0095] In one specific embodiment, before acquiring the multi-modal monitoring data stream of the experimental animal's state and behavior, it further includes:

[0096] Based on experimental requirements and animal types, a separate feeding space and dedicated passage area are designed, and a feeding box basic structure meeting SPF level cleaning standards is constructed;

[0097] Based on the feeding box basic structure, a sliding door grid driven by a micro servo motor and a position sensor are integrated to construct the electrically controlled door passage;

[0098] Based on the feeding box basic structure and the electrically controlled door passage, the food intake monitoring sensor, the animal activity state monitoring sensor, the physiological parameter monitoring sensor, and the environmental parameter monitoring sensor are deployed to construct the multi-modal sensor array;

[0099] Based on the multi-modal sensor array and the electrically controlled gate channel, an integrated embedded processor, motor drive circuit and high-precision analog-to-digital conversion module are constructed to build an intelligent control unit.

[0100] In the process of specific implementation, first, based on the standard experimental animal feeding box, materials resistant to high-pressure sterilization and conventional cleaning process are selected to meet the cleaning standards of SPF level or higher animal room. By adding partitions and special channel areas, a separate feeding space is constructed to form a controllable animal living area and a feed supply area. Good ventilation and gas exchange should be considered in the design, and the design of advanced metabolic cages should be used as a reference to ensure stable cage environment and reduce CO2 and ammonia accumulation. Transparent materials such as cage covers can be used to facilitate observation and maintain normal circadian rhythms of animals. The design should maximize the living space of animals and integrate environmental enrichment components to improve animal welfare.

[0101] Please refer to Figure 2 , Figure 2 is the architecture diagram of the intelligent gate control system provided by the embodiment of the application. As shown in Figure 2 , based on the separate feeding space, a sliding door grid or a rotating door mechanism driven by a miniature servo motor or a stepper motor is designed. It contains anti-stuck protection device and precise position sensor. The motor and control circuit should take appropriate waterproof and insulation measures to deal with possible humid environment and cleaning needs. The gate control can be designed for food or liquid, and even interchangeable.

[0102] Integrate high-precision weight sensors to monitor food and water consumption and animal body weight. Integrate infrared activity detectors or acceleration sensors to monitor animal activity status and position. Integrate miniature cameras to assist in monitoring feeding behavior, activity patterns and abnormal states. Integrate temperature and humidity sensors, light sensors and sound sensors to monitor the feeding environment parameters. Other sensors can be extended, such as H2 and CH4 sensors for microbiome research, or 13CO2 sensors for stable isotope tracing. The deployment of sensors should consider easy cleaning and maintenance.

[0103] Based on an embedded processor platform (such as ARM Cortex series or ESP32). Integrate motor drive circuit, high-precision analog-to-digital conversion module (ADC), data storage module (such as SD card or Flash). The control unit should be designed as low power consumption and consider electrical insulation requirements. Adopt modular design to facilitate future function expansion and maintenance. Design a low-power power management circuit to support mains or battery power supply scheme. Provide wired (such as Ethernet) / wireless (such as Wi-Fi, Bluetooth) communication interface, and USB interface for local configuration and data download.

[0104] In one specific embodiment, the multi-modal monitoring data stream of the experimental animal state and behavior includes:

[0105] The food intake monitoring sensor containing a weight sensor and a miniature camera monitors food consumption in real time and filters out interference signals to obtain food intake data;

[0106] The animal activity state monitoring sensor containing an infrared activity detector or an acceleration sensor identifies and analyzes animal movement patterns to obtain activity classification data;

[0107] Based on the physiological parameter monitoring sensor, the physiological parameters of the experimental animal are monitored to obtain physiological parameter data;

[0108] The environmental parameter monitoring sensor containing a temperature and humidity pressure sensor collects and processes environmental parameters to obtain environmental change data;

[0109] Based on the food intake data, the activity classification data, the physiological parameter data and the environmental change data, time alignment and multi-source data fusion are performed to generate the multi-modal monitoring data stream.

[0110] Please refer to Figure 3 , Figure 3 is a multi-modal sensing and monitoring system architecture provided by the embodiment of the present application. As shown in Figure 3 in the specific implementation process, first based on high-precision weight sensor and miniature camera. Through high-precision ADC acquisition, digital signal filtering processing. Realize high sampling rate, capture including micro food event in the eating behavior details. Combined with image recognition algorithm, analyze eating posture, time length, etc. Consider the food container with anti-splashing design.

[0111] Based on XYZ three-axis infrared activity detector or acceleration sensor. Through motion pattern recognition algorithm (distinguish coarse and fine motion, activity intensity calculation, position tracking and behavior classification model). The system should be able to distinguish between real activity and sensor static obstacles.

[0112] Non-invasive physiological parameter acquisition equipment can be integrated, or implantable telemetry equipment can be compatible. Through signal processing algorithm, abnormal state detection model for monitoring.

[0113] Based on temperature and humidity, light, air pressure sensor. Through environmental parameter acquisition processing, circadian rhythm recognition algorithm for monitoring. Record the external environmental factors that may affect the experiment.

[0114] Based on the above various types of monitoring data, through time alignment algorithm, multi-source data fusion technology and feature extraction method, the multi-modal monitoring data stream reflecting the animal state is generated. In the multi-modal data fusion processing, the system faces a core technical challenge: how to extract consistent feature representation from different dimensional, different distribution characteristics of sensor data. In view of this challenge, the system adopts the dimensionless convergence technology of the diffusion model of approximate Gaussian mixture.

[0115] Among them, the time-dependent noise scheduling function β(t) in the diffusion model adopts the cosine scheduling strategy:

[0116]

[0117] Among them, β min = 10 -4 , β max = 0.02, T is the total diffusion step, set to 1000. This scheduling method maintains a small noise in the early diffusion, which is beneficial to retain the detailed features of the data, and increases the noise intensity in the later period to accelerate the convergence to the standard distribution.

[0118] The number of components M of the Gaussian mixture model is adaptively determined according to the data complexity, initially set to 5, and dynamically adjusted by the Bayesian information criterion (BIC) during training, and finally generally converges to between 3-8. The mixing weight πj is iteratively optimized by the expectation maximization algorithm (EM algorithm), and the iteration stopping condition is that the change of the log-likelihood function is less than 10 -6 or reaches the maximum iteration number 100.

[0119] The adaptive scaling factor λ i (t) in the dimensionless convergence mechanism is initially set to 1.0, and the Wasserstein distance objective function is optimized by gradient descent method:

[0120]

[0121] Among them, the target distribution q(t) is selected as the standard normal distribution N(0, I). The optimization process adopts Adam optimizer, the learning rate is 0.01, and the total iteration number is 500. In order to improve the calculation efficiency, the Wasserstein distance is calculated by Sinkhorn algorithm, and the regularization coefficient ε is set to 0.01.

[0122] First, the original data collected by each sensor is regarded as a distribution in a high-dimensional space. These distributions usually exhibit complex non-Gaussian characteristics. Traditional methods often use simple standardization to solve the problem of inconsistent dimensions, but this method cannot handle the inherent distribution differences of the data.

[0123] The present technology introduces a diffusion model as the theoretical basis for data representation. The diffusion model can map any complex distribution to a standard space by defining a continuous transformation process from the complex distribution to a simple distribution (usually a standard Gaussian distribution). Specifically, for data xi from sensor i, the system defines a diffusion process:

[0124]

[0125] where β(t) is a predefined time-dependent noise schedule function and dw is a standard Wiener process. This stochastic differential equation describes how data is gradually "polluted" by noise until it approaches a Gaussian distribution.

[0126] The innovation of the present system lies in approximating the intermediate process distribution by a Gaussian Mixture Model (GMM) and introducing a dimensionless convergence mechanism. Specifically, the system discretizes the diffusion process into K time steps, and at each time step t, the data distribution is approximated as a Gaussian Mixture Model:

[0127]

[0128] where M is the number of mixture components, π j are the mixture weights.

[0129] The dimensionless convergence mechanism handles the dimensionality problem of different sensor data by introducing an adaptive scaling factor λ i (t):

[0130]

[0131] The value of λ i (t) is automatically determined by minimizing the distribution difference of different sensor data in the latent space. The Wasserstein distance is used as the measure of distribution difference:

[0132]

[0133] where q(t) is a predefined target distribution.

[0134] The advantages of this method are: (1) no prior knowledge of the specific form of the data distribution is required; (2) it can handle complex distributions such as multi-modal and skewed; (3) it naturally realizes data denoising and anomaly detection through the diffusion process.

[0135] In animal experiment monitoring, this technology enables the system to extract a unified feature representation from data with different characteristics such as food consumption, activity intensity, and body temperature changes, significantly improving the accuracy and robustness of subsequent analysis. For example, when food consumption data shows periodic fluctuations and activity data shows sudden changes, traditional methods have difficulty effectively integrating them, while the present technology can capture the internal relationship between these different patterns.

[0136] In one specific embodiment, the implementation of the time-limited feeding mode execution control includes:

[0137] Based on the multi-modal monitoring data stream and the preset time-limited feeding parameters, a time trigger mechanism is used to determine the feeding time window, and a state monitoring feedback is used to generate a feeding time window switch event table;

[0138] Based on the feeding time window switch event table, a state machine model is used to control the on-off state of the electrically controlled gate channel;

[0139] Based on the multi-modal monitoring data stream, an abnormal state detection algorithm constructed by a multi-level threshold model is used to identify potential health risks and abnormal behaviors, and when a potential health risk or abnormal behavior is identified, an emergency response mechanism is triggered;

[0140] Based on the emergency response mechanism, in combination with the feeding time window switch event table and the on-off state of the electrically controlled gate channel, a gate execution instruction is generated and prioritized to realize the execution control of the time-limited feeding mode.

[0141] Specifically, the implementation of the time-limited feeding mode execution control is the key link of the core function of the system, which converts the experimental parameters set by researchers into precise physical control actions. First, the system determines the feeding time window through a time trigger mechanism. This mechanism is based on a high-precision real-time clock chip, which can provide millisecond-level time accuracy, ensuring that the opening and closing of the feeding window are strictly executed according to the preset time. The system supports a variety of complex time strategies, including fixed time windows (such as allowing feeding from 8 to 16 hours every day), segmented time windows (such as allowing feeding from 9 to 11 hours in the morning and from 17 to 19 hours in the evening), and periodic change windows (such as using different feeding times on weekdays and weekends). The time trigger mechanism not only considers absolute time, but also combines the biological rhythm of experimental animals, allowing the setting of relative time windows based on the light cycle, better simulating foraging behavior in natural environments.

[0142] On the basis of time triggering, the state monitoring feedback mechanism is also integrated to dynamically adjust the door control behavior. For example, when it is detected that the animal is in an active feeding state and the time window is about to close, the system will appropriately extend the window time according to the preset rules to avoid the stress reaction caused by the sudden interruption of feeding behavior. Similarly, if it is detected that the animal is in a sleep state at the beginning of the window opening, the system may delay the actual opening time of the feeding window and wait for the animal to wake up naturally. This feedback adjustment based on the behavior state makes the feeding window control more in line with the natural behavior pattern of the animal and reduces the disturbance of experimental intervention on the normal physiological rhythm of the animal. By integrating time triggering and state feedback, a detailed feeding time window opening and closing event table is generated. The event table not only contains basic time information, but also integrates key parameters such as triggering conditions, execution actions, and priority. Based on the event table, the gate control operation is performed to achieve precise time-limited feeding control. The event table supports multiple time modes (such as absolute time and relative time) and condition expressions and can be dynamically adjusted according to the animal state, combining strict time management and flexible adaptability, which is a technical guarantee for the accurate implementation of experimental schemes.

[0143] Based on the generated feeding time window opening and closing event table, a perfect state machine model is used to control the opening and closing state of the electric control gate channel. The state machine model divides the gate control operation into multiple refined states, including standby state, opening preparation state, opening execution state, opening completion state, closing preparation state, closing execution state, and closing completion state. Each state has clear entry conditions, exit conditions, and fault handling mechanisms. For example, before starting the opening execution, the system will first check the motor drive circuit state and the feedback of the gate position sensor to confirm that the system is in a controllable state; during the opening process, the motor current and gate position change are monitored in real time to ensure smooth completion of the action; if abnormal resistance or position deviation is detected, the preset abnormal handling process such as pause, retry, or alarm will be triggered. The application of the state machine model makes the gate control execution process more stable and reliable and can cope with various expected and unexpected situations.

[0144] The safety monitoring mechanism is an indispensable guarantee system in the time-limited feeding execution control. Based on multi-modal monitoring data streams, the mechanism uses an abnormal state detection algorithm constructed with multiple threshold models to continuously monitor the health status and behavior performance of animals. The algorithm sets three levels of safety thresholds: warning level, moderate risk level, and high risk level. For example, when the animal's food intake is detected to be less than 30% of the expected value within 24 hours, it may trigger a warning level reminder; when food intake is insufficient for 48 consecutive hours and body weight decreases by more than 5%, it may be upgraded to a moderate risk; if further findings show a significant decrease in activity level or abnormal behavior patterns, it may be determined as a high-risk state. Different risk levels correspond to different response strategies: the warning level mainly records and notifies researchers; the moderate risk may lead to automatic adjustment of the feeding window parameters; the high-risk state will immediately trigger emergency measures, such as temporarily opening the food passage, and sending emergency notifications through various channels (mobile app push, SMS, email, etc.).

[0145] When the safety monitoring mechanism identifies potential health risks or abnormal behaviors, it triggers an emergency response mechanism. This mechanism first assesses the urgency of the risk and the type of intervention required, then combines the current feeding window switch event table and the state of the electrically controlled door passage to generate appropriate response instructions. For example, for detected severe dehydration risk, the system may temporarily open the drinking water passage without changing the state of the food passage; for detected severe energy deficiency risk, the system may temporarily open the food passage for a limited time during the scheduled fasting period. All emergency response operations are recorded in detail, including trigger conditions, execution time, intervention type, and duration, to facilitate subsequent analysis and adjustment of experimental protocols by researchers.

[0146] After considering the time event table, safety monitoring results, and emergency response requirements, a special priority processing mechanism is used to generate the final gate execution instruction sequence. This mechanism ensures that safety-related instructions always have the highest execution priority, followed by emergency response instructions, then dynamic adjustment instructions, and finally basic time window instructions. This hierarchical priority design ensures that while meeting basic research needs, animal health and welfare are always given the highest level of protection. The final gate execution instructions are processed by a motor control signal conversion module to convert them into specific motor drive signals, including precise position control parameters, speed curves, and torque limits, to ensure smooth, accurate, and reliable gate operation. The entire execution process uses a closed-loop control design, with real-time feedback from position sensors and current monitors to ensure that each control instruction is accurately executed, and unexpected situations can be adjusted or terminated in a timely manner.

[0147] It is particularly emphasized that in the process of implementing the time-limited feeding mode execution control, a comprehensive fault-tolerant design is adopted. This includes hardware redundancy (such as backup power supply, double sensors), software fault tolerance (such as exception capture, safety mode switching), and communication backup (multi-channel alarm mechanism). Even in the case of partial hardware failure or software exception, the basic safety monitoring and control functions can still be maintained to ensure that experimental animals are not in danger due to system failure. At the same time, all execution actions and abnormal events are recorded in real time to the system log, providing complete event tracking information for fault diagnosis and experimental data analysis.

[0148] Through the above complex and precise control mechanism, highly precise execution control of the time-limited feeding mode is achieved. Compared with traditional simple timing switches, this system not only strictly controls the feeding time window according to the preset scheme, but also intelligently adjusts according to the real-time state of the animals and provides timely protection in potential risk situations, greatly improving the accuracy, reliability and safety of animal time-limited feeding experiments, providing an ideal experimental platform for researchers.

[0149] The time-limited feeding mode programmed control process provided by the embodiment of the present application provides a user graphical interface in the specific implementation process. It supports multiple preset modes (16 / 8, 14 / 10, 5:2, Eat-Stop-Eat, etc.) and custom mode configuration. Time window-based or intake-based access control rules can be set. It includes parameter verification and safety prompt mechanism.

[0150] Based on the configured time-limited feeding parameters and monitoring data flow, the time trigger mechanism, state monitoring feedback and gate precise control algorithm are used to achieve accurate control of the time window in different feeding modes. The algorithm first establishes an absolute time reference based on the real-time clock (RTC), and generates an accurate time event table according to the configured parameters. The time trigger mechanism triggers the gate state transition at the time point specified in the event table, for example, opening the feed channel at the start of the feeding window and closing it at the end. The algorithm uses a state machine model to manage the gate state, including standby, open request, open in progress, open complete, close request, etc. It is particularly worth noting that the algorithm not only relies on time triggering, but also integrates monitoring data flow for state monitoring feedback, for example, when the animal is detected to be feeding and the window is about to close, the system will intelligently delay the closing operation to avoid stress reactions caused by sudden interruptions. The algorithm also realizes gate precise control, which realizes millimeter-level precise positioning of the door opening and closing position through a PID controller, and prevents the gate mechanism from being stuck or overloaded through current monitoring.

[0151] Based on the monitoring data stream and the execution, through the animal state evaluation algorithm, the adaptive rule engine and the adjustment decision model, the dynamic fine-tuning of the feeding time window is realized. The algorithm design solves the problem that the static time table cannot adapt to the individual differences and state changes of animals. The algorithm first analyzes the monitoring data stream in real time through the animal state evaluation module, calculates key health indicators including body weight change rate, activity level, feeding pattern, etc. The adaptive rule engine generates adjustment suggestions based on these indicators and preset safety thresholds, such as expanding the feeding window when detecting that the animal's body weight has decreased beyond the safety threshold. The adjustment decision model prioritizes and conflict resolves the adjustment suggestions under the premise of balancing experimental goals and animal welfare, and finally forms a specific time window adjustment strategy. This dynamic adjustment capability enables the system to better adapt to individual differences while maintaining the experimental purpose, improving the scientificity and humanity of the experiment.

[0152] In one specific embodiment, the abnormal state detection algorithm constructed using the multi-level threshold model identifies potential health risks and abnormal behaviors, and when a potential health risk or abnormal behavior is identified, an emergency response mechanism is triggered, including:

[0153] Based on the multi-modal monitoring data stream, the multi-scale processing is performed on the multi-modal monitoring data stream to construct a multi-scale time representation, and a cross-scale attention module is used to interact the features of different scales of the multi-scale time representation to form multi-time scale features.

[0154] Specifically, the construction of the multi-scale time representation uses a multi-level sliding window method, and the original data is collected at a frequency of 10Hz, and then minute-level and hour-level data representations are constructed through mean pooling operations. Specifically, for each sensor data stream X, a time pyramid structure X1 (original 10Hz sampling), X2 (1 minute mean) and X3 (1 hour mean) is constructed. Each scale of data is processed through a 3-layer convolutional neural network for feature extraction, with convolution kernel sizes of [3, 5, 7] and channel numbers of [32, 64, 128], and LeakyReLU is used as the activation function.

[0155] The implementation of the cross-scale attention module uses a multi-head attention mechanism, with 4 heads and an embedding dimension of 128. For the feature representation of each scale i, a query matrix Q_i is generated through linear projection, while the features of other scales j generate corresponding key-value matrices K_j and V_j. The attention calculation uses scaled dot-product attention:

[0156]

[0157] where d is the dimension of the attention module (set to 128), and the temperature coefficient is used for scaling to avoid gradient vanishing problems.

[0158] Based on the multi-time scale characteristics, the health state index and its uncertainty are calculated by a Bayesian neural network to identify the abnormal type; wherein the abnormal type includes high confidence anomaly, high uncertainty anomaly and multi-modal conflict;

[0159] Specifically, the Bayesian neural network model is trained based on historical experimental data, and the data sources include the behavior and physiological index records of healthy animals in a standard experimental environment and a labeled abnormal state sample library. The model architecture includes 4 fully connected layers with node numbers of [256, 128, 64, 32], ReLU is used as the activation function in the middle layer, and Sigmoid function is used in the output layer to map to the abnormal probability space.

[0160] The training process uses a stochastic gradient descent optimizer with a learning rate of 0.001, a batch size of 64, and 200 training rounds. To achieve Bayesian inference capability, Monte Carlo Dropout is introduced after each fully connected layer with a dropout rate of 0.2. The training process ensures the model's generalization ability through five-fold cross-validation, and F1 score and AUC value are used as evaluation indicators.

[0161] In the inference stage, 20 forward propagations (using different dropout masks) are performed to obtain the distribution characteristics of the health state index. By calculating the mean and variance σ2 of the index, the system can obtain the health assessment result and its confidence level at the same time:

[0162]

[0163] where T is the number of forward propagations (20), f(x, w i ) is the single inference result, w i is the i-th randomly sampled model parameter.

[0164] After the model is trained, it is optimized to a lightweight model through model quantization and parameter pruning techniques, and converted to TensorFlow Lite format for deployment on embedded processors. The final model size is controlled within 5MB, and the single inference time does not exceed 100ms, meeting the real-time monitoring requirements.

[0165] Based on the high confidence anomaly, high uncertainty anomaly and multi-modal conflict, different response strategies are selected according to the abnormal type to trigger the emergency response mechanism.

[0166] In the implementation process, the safety monitoring mechanism is the key guarantee for ensuring animal health and experimental reliability. This mechanism continuously analyzes and monitors data streams and identifies potential health risks and abnormal behaviors through an abnormal state detection algorithm. The algorithm uses a multi-level threshold model to classify abnormalities into three levels: warning, moderate risk, and high risk. For example, when the system detects that an animal's food intake is close to zero within 24 hours or its weight is rapidly decreasing, it will trigger an abnormal alarm of the corresponding level. The emergency response mechanism automatically initiates corresponding intervention measures based on the type and severity of the anomaly, such as temporarily opening the feed channel when a severe hunger risk is detected or adjusting environmental parameters when abnormal behavior is found. The alarm triggering system notifies researchers through multiple channels, including mobile app push, SMS, and email, ensuring that emergency situations can be handled in a timely manner. The entire safety monitoring mechanism uses a redundant design, allowing it to maintain basic monitoring functions even if some sensors fail.

[0167] The safety monitoring mechanism is a key link in ensuring the health of subjects in animal experiments, requiring rapid and accurate identification of potential health risks and system failures. However, traditional monitoring methods face two main challenges: 1) different time scales of anomalies are difficult to capture simultaneously (from second-level fluctuations to daily periodic changes); 2) it is difficult to distinguish between sensor failures and real physiological abnormalities.

[0168] To address these issues, the system uses a multi-scale time fusion network technology based on an attention mechanism, specifically designed for fault diagnosis of uncertain patterns in multi-modal data.

[0169] First, the system organizes each sensor data stream into a multi-scale time representation. For data stream X, the system constructs a time pyramid structure:

[0170] X1: raw sampling frequency data (e.g., 10 Hz);

[0171] X2: down-sampled medium time scale data (e.g., 1-minute average);

[0172] X3: long time scale data (e.g., hourly trends).

[0173] Each scale of data is processed by a specific encoder E1, E2, E3 to generate corresponding feature representations.

[0174] Multi-scale time fusion is achieved through an attention mechanism. A cross-scale attention module is introduced to allow features of different time scales to interact with each other:

[0175]

[0176] where Q i , V_j are the query matrix from scale i and the key-value matrix from scale j, respectively. This allows the system to adaptively focus on abnormal patterns at different time scales, e.g., linking long-term trend changes with short-term surges.

[0177] Uncertain pattern diagnosis is the core innovation of this technology. The system not only predicts abnormal states, but also estimates the uncertainty of the prediction, which is achieved through Bayesian neural networks:

[0178] p(y|x, θ) = ∫p(y|x, w)·p(w|θ, D)dw

[0179] where y is the health status indicator, x is the multi-modal input, w is the model parameter, θ is the distribution hyperparameter, and D is the historical data.

[0180] The Monte Carlo random dropout (MC Dropout) technique is used to approximate the parameter posterior distribution, and the prediction distribution is obtained through multiple forward propagation instead of single point estimation:

[0181]

[0182] This method allows the system to distinguish between several key situations:

[0183] High confidence anomaly: clear abnormality in physiological indicators, requiring immediate intervention;

[0184] High uncertainty anomaly: may be a sensor failure or unknown pattern, requiring further verification;

[0185] Multi-modal conflict: different sensor data indicates contradictory state, may exist partial failure.

[0186] In practical applications, when the system detects an anomaly, different response strategies are taken according to the level of uncertainty:

[0187] For high confidence anomaly, directly trigger safety intervention (such as opening the feed channel);

[0188] For high uncertainty anomaly, increase sampling frequency and start cross-validation procedure;

[0189] For modal conflict, isolate suspicious sensors and continue monitoring based on redundant data.

[0190] This attention-based multi-scale time fusion network significantly improves the reliability of the system, enabling it to accurately distinguish between real physiological abnormalities, environmental disturbances and equipment failures in complex biological experimental environments, thereby ensuring the health and safety of experimental animals and the quality of data. For example, the system can accurately distinguish between normal short-term fasting behavior of animals and feed supply interruption due to equipment failure, avoiding unnecessary intervention or false negative risk situations.

[0191] Based on the time window control, dynamic adjustment and safety monitoring results, precise gating action instructions are generated and executed through instruction sequence generation algorithms, priority processing and motor control signal conversion. This module will convert research parameters into specific physical control actions and ensure the accurate implementation of experimental strategies and animal safety. Based on the basic time window plan, dynamic adjustment results and safety monitoring output, the instruction sequence generation algorithm will produce a complete gating action plan. This algorithm uses an event-driven design pattern to integrate control requirements from different sources into a unified timeline. The priority processing module ensures that safety-related instructions always have the highest execution priority, followed by dynamic adjustment instructions and finally basic time window instructions. The motor control signal conversion module translates high-level instructions into specific motor drive signals, taking into account acceleration and deceleration curves, position feedback and torque limits to ensure the smoothness and reliability of gating actions. During execution, the system continuously monitors the gating execution status and applies closed-loop control to ensure accurate action completion. If unexpected resistance or position deviation occurs, the control parameters will be automatically adjusted or an abnormal processing procedure will be triggered.

[0192] These sub-steps collectively form a complete closed-loop control system, achieving full-process management from research parameter setting to physical execution, to state monitoring and dynamic adjustment. This system not only accurately executes various complex time-limited feeding patterns, but also has intelligent adaptation and safety guarantee capabilities, significantly improving the accuracy, safety and scientific value of animal experiments.

[0193] In a specific embodiment, the output of the individualized time-limited feeding intervention strategy includes:

[0194] Based on the multi-modal monitoring data stream, an animal health status evaluation model is constructed through feature extraction algorithms and machine learning models;

[0195] Based on the animal health status evaluation model and the execution data of the execution control, a diet pattern influence analysis result is constructed through time series data analysis and correlation calculation;

[0196] Based on the diet pattern influence analysis result and the preset research target, an individualized diet pattern optimization scheme is generated through multi-objective optimization algorithms and dynamic sparse training;

[0197] Based on the individualized diet pattern optimization scheme, an individualized time-limited feeding intervention strategy is output through a rule conversion engine and parameter mapping.

[0198] In the specific implementation process, first, based on the multi-modal monitoring data stream, an evaluation model for quantitatively evaluating the health status of animals is constructed through feature extraction algorithms, health index calculation and machine learning models.

[0199] Based on the health status evaluation results and the time-limited eating execution data, an algorithm for analyzing the impact of the dietary pattern on health is constructed through time series data analysis, correlation calculation and causal inference method.

[0200] In one specific embodiment, the personalized dietary pattern optimization scheme is generated by a multi-objective optimization algorithm and dynamic sparse training, including:

[0201] Based on the dietary pattern impact analysis results, a joint feature representation containing time-related features, physiological response features and behavior pattern features is extracted to form an embedded representation of a high-dimensional vector space;

[0202] Based on the embedded representation, an importance score matrix is calculated by feature importance evaluation to identify key feature dimensions;

[0203] Based on the limitation of computing resources, the key feature dimensions are subjected to feature sparsification processing, and features with importance lower than a preset threshold are pruned to obtain pruned key feature dimensions; and the pruned key feature dimensions are subjected to computing level allocation to form a lightweight optimization model;

[0204] Based on the lightweight optimization model, and guided by a preset research goal, a multi-objective optimization search is performed to generate a personalized dietary pattern optimization scheme.

[0205] In the design of the personalized dietary pattern optimization algorithm, the system needs to process a large amount of multi-dimensional time series data to extract individual animal features and develop optimal intervention strategies. Traditional methods face two major challenges: first, the limited computing resources in embedded environments make it difficult to support complex models; second, the characteristics of different individuals differ significantly, and static models are difficult to adapt.

[0206] Among them, the dynamic sparse training process first calculates the feature importance score matrix from the joint feature representation. The feature importance evaluation adopts a gradient-based method:

[0207]

[0208] Where L is a composite objective function, including the weighted sum of health status prediction loss (mean square error) and behavior pattern preservation loss (KL divergence):

[0209]

[0210] The weight coefficients α=0.7, β=0.3 reflect the balance requirement between the main goal of health status prediction and the stability of behavior.

[0211] The pruning threshold τ in feature sparsification is initially set as the lower quartile of the importance score distribution and dynamically adjusted during the experiment. When the system resource occupancy exceeds 80%, τ is raised to the lower tertile; when the occupancy is below 40%, τ is lowered to the lower quintile, achieving a dynamic balance between computational resources and model accuracy.

[0212] The hierarchical distribution assigns features into three levels: features with importance scores in the top 30% are processed locally and in real-time at the edge device; features with scores between 30% and 70% are processed in batches at a low frequency (every 10 minutes); the remaining features are processed every hour or uploaded to the server for in-depth analysis.

[0213] The multi-objective optimization search employs an improved version of the NSGA-II algorithm, with optimization objectives including: (1) weight maintenance index; (2) activity balance score; (3) food utilization efficiency. Constraints include minimum daily energy intake requirements and maximum fasting duration limits. The population size is set to 100, the number of evolutionary generations is 50, the crossover probability is 0.8, and the mutation probability is 0.1. The final optimal solution is selected through the Pareto frontier and the decision-maker preference function.

[0214] To address these challenges, the system employs a dynamic sparse training embedded joint feature extraction technique. The core idea of this technique is to dynamically adjust the model complexity and feature space dimension, significantly reducing the computational burden while maintaining expressive ability.

[0215] First, a joint feature representation is constructed for each individual animal. This representation integrates information from multiple dimensions:

[0216] Time-dependent features: capture the temporal patterns of animal feeding and activity.

[0217] Physiological response features: including weight changes, metabolic indicators, and responses to dietary interventions.

[0218] Behavioral pattern features: reflect the habits and preferences of animals.

[0219] These features are encoded as an embedded representation E ∈ R^d in a high-dimensional vector space, where d is the initial dimension of the feature space.

[0220] Dynamic sparse training is achieved through two key mechanisms: feature importance evaluation and dynamic structure adjustment.

[0221] Feature importance evaluation quantifies the contribution of each feature dimension by introducing an importance score matrix S:

[0222]

[0223] where L is the objective function (e.g., health state prediction error). The importance score reflects the degree of influence of a particular feature dimension on the model performance.

[0224] Dynamic structure adjustment adaptively adjusts the model structure according to the computing environment and task requirements:

[0225] Feature sparsification: pruning feature dimensions with importance lower than threshold τ:

[0226]

[0227] where

[0228] Computing level allocation: dynamically allocate computing resources to different processing levels according to feature importance:

[0229] High importance features: real-time processing locally on the device;

[0230] Medium importance features: batch processing at low frequency;

[0231] Low importance features: optionally uploaded to the cloud for periodic analysis.

[0232] Knowledge distillation: by distilling the knowledge of complex models into lightweight models, the performance is maintained while reducing the number of parameters:

[0233] min θ KL(p T (y|x)‖p S (y|x;θ))

[0234] where p T and p S are the output probability distributions of the teacher model and the student model, respectively.

[0235] This dynamic sparse training method enables the system to run efficiently on embedded devices while maintaining the model's adaptability to individual differences. In practical applications, the system can dynamically adjust the model complexity for different individuals and different periods, such as retaining more feature dimensions in the early stages of the experiment to capture baseline behavior, and reducing dimensions in the stable stage to improve efficiency.

[0236] Through this method, the system can extract unique features of each animal individual from limited observation data, and generate more accurate personalized dietary intervention strategies accordingly, significantly improving the experimental effect and research value.

[0237] where the dynamic sparse training process first calculates the feature importance score matrix from the joint feature representation. The feature importance evaluation uses a gradient-based method:

[0238]

[0239] where L is the composite objective function, including the weighted sum of health state prediction loss (mean square error) and behavior pattern maintenance loss (KL divergence):

[0240]

[0241] The weight coefficients α = 0.7, β = 0.3 reflect the balance requirement between the primary goal of health state prediction and the stability of behavior.

[0242] The clipping threshold τ in feature sparsification is initially set to the lower quartile of the importance score distribution and dynamically adjusted during the experiment. When the system resource occupancy rate exceeds 80%, τ is raised to the lower tertile; when the occupancy rate is lower than 40%, τ is lowered to the lower quintile, achieving a dynamic balance between computing resources and model accuracy.

[0243] The hierarchical distribution of computing assigns features into three levels: features with importance scores in the top 30% are processed locally and in real-time at the edge device; features with scores between 30% and 70% are processed in batches at a low frequency (every 10 minutes); the remaining features are processed once an hour or uploaded to the server for deep analysis.

[0244] The multi-objective optimization search uses an improved version based on the NSGA-II algorithm, with optimization objectives including: (1) weight maintenance index; (2) activity balance score; (3) food utilization efficiency. The constraint conditions include the minimum daily energy intake requirement and the maximum fasting time limit. The population size is set to 100, the evolution generation is 50, the crossover probability is 0.8, and the mutation probability is 0.1. Finally, the optimal solution is selected through the Pareto front and the decision maker preference function.

[0245] Based on the optimization results, an intervention strategy generator that can output specific gate adjustment parameters is developed through a rule conversion engine, parameter mapping, and an execution plan generation algorithm. The intervention strategy generator is developed to convert the generated optimization scheme into specific executable control instructions. This process first parses the abstract optimization goals into specific operation rules through the rule conversion engine. The rule conversion engine uses a knowledge graph-based reasoning model that contains expert knowledge and experimental experience in the time-restricted feeding research field. For example, when the optimization algorithm suggests the abstract goal of "increasing the length of the morning feeding window to improve metabolic rate", the rule conversion engine will convert it into a series of specific rules such as "moving the start time of the morning feeding window forward by 30 minutes", "adjusting the feed delivery rate to 1.2 times the standard value", etc. The parameter mapping module is responsible for quantifying these rules into precise system parameters, taking into account hardware limitations, safety boundaries, and operation precision. For example, the gate opening time will not only be mapped to a specific clock time, but also consider the acceleration characteristics and position accuracy of the gate mechanism. The execution plan generation algorithm is the last link of the intervention strategy generator, which organizes the quantified parameters into a time sequence execution plan and generates an instruction sequence that can be directly understood by the control system. This algorithm uses an event-driven architecture that can handle complex conditional branches and timing dependencies, ensuring that intervention measures are executed in the correct order and at the right time. The output of the entire intervention strategy generator is a complete execution plan document containing specific time points, gate parameters, monitoring thresholds, and exception handling strategies, which can be directly executed by the control system.

[0246] In one specific embodiment, the method further comprises:

[0247] Based on the execution results of the personalized time-restricted feeding intervention strategy and the multi-modal monitoring data stream, a comprehensive effect score is calculated through an effect evaluation algorithm to generate an intervention effect evaluation result;

[0248] When the intervention effect evaluation result is lower than the preset effect threshold, a feedback adjustment mechanism is started to conduct multi-angle investigation and generate adjustment suggestions;

[0249] Based on the adjustment suggestions, the animal state assessment model and the optimization strategy are updated to realize iterative optimization of the intervention strategy.

[0250] The intervention effect evaluation and iterative optimization realize the self-learning ability of the system, enabling the intervention strategy to be continuously improved based on actual effects. The effect evaluation algorithm quantitatively evaluates the effectiveness of the intervention measures by comparing the changes in the animal's state before and after the intervention. This algorithm uses a multi-index comprehensive scoring system to compare the changes in body weight, food intake, activity level, physiological indicators, and other dimensional data with the expected target to calculate the comprehensive effect score. To handle random fluctuations and individual differences during the experiment, the algorithm incorporates Bayesian statistical methods to quantify uncertainty, enabling it to distinguish between deterministic changes and random fluctuations in intervention effects. The feedback adjustment mechanism generates targeted adjustment suggestions based on the evaluation results. When the intervention effect is found to be less than expected, this mechanism analyzes possible causes such as insufficient dosage, improper timing, or individual sensitivity differences, and proposes corresponding adjustment directions. It is particularly noteworthy that the feedback mechanism not only focuses on short-term effects but also tracks long-term trends to avoid the system falling into a local optimization trap. The iterative optimization method applies adjustment suggestions to the health status evaluation model, influence analysis algorithm, and optimization strategy, forming a complete optimization closed loop. This method uses an incremental learning strategy, continuously updating the health status evaluation model, influence analysis algorithm, and optimization strategy as experimental data accumulates, making the system increasingly adaptable to specific experimental subjects. This adaptive learning process enables the system to gradually develop individualized intervention strategies tailored to the characteristics of specific animal groups, even starting from the same initial parameters.

[0251] The synergistic work of these steps enables the system to have true intelligent adaptive ability. Unlike traditional pre-programmed control systems, this system can continuously learn and improve from the experimental process, adapting to individual differences and state changes of different experimental objects. For example, when the system finds that a group of animals does not adapt well to the standard 16 / 8 time-limited feeding scheme, it may eventually develop a more suitable individualized scheme such as 14 / 10 or segmented feeding through multiple rounds of iterative adjustment. This ability greatly improves the scientific value and animal welfare level of the experiment, while reducing the workload of researchers. Through the closed-loop design of the intervention strategy generator and the iterative optimization of the effect evaluation, the system realizes the complete intelligent conversion from data to knowledge and from monitoring to intervention, marking an important breakthrough in the precision and intelligence of animal experiment technology.

[0252] In one specific embodiment, the implementation of experimental data management and analysis includes:

[0253] Based on the multi-modal monitoring data stream, the execution data of the execution control, and the personalized time-limited feeding intervention strategy, through a time-series database and a data synchronization mechanism, the multi-channel data is stored in a traceable manner, obtaining stored multi-channel data.

[0254] For the stored multi-channel data, through noise filtering, data standardization and missing value processing, a pre-processed data set is generated;

[0255] The pre-processed data set is processed by multiple data analysis methods to extract experimental results and pattern characteristics; the multiple data analysis methods include statistical analysis, machine learning and time series analysis algorithm;

[0256] Based on the experimental results and pattern characteristics, interactive charts and multi-dimensional data visualization technology are used to realize experimental data management and analysis.

[0257] In the specific implementation process, the experimental data acquisition and storage system development ensures the traceable storage of all sensor raw data. A reliable database is used for storage. Synchronous acquisition and display of multi-channel data are supported.

[0258] The data preprocessing and feature extraction module realizes noise filtering, data standardization / calibration, missing value processing, feature extraction, etc.

[0259] The multi-dimensional analysis algorithm library construction includes statistical analysis, machine learning, time series analysis, behavior pattern analysis, energy metabolism calculation, etc. User-defined data grouping and standardization are supported.

[0260] The visualization display system realizes the provision of interactive charts. Multi-dimensional data visualization and customized report templates are supported.

[0261] The research report and conclusion generator is developed based on the analysis results and visualization content, through conclusion derivation algorithm, natural language generation technology and scientific report formatting tool, to develop a report generator to assist researchers in forming research conclusions. A standardized data output format is provided, which may be compatible with common analysis software or format. The alarm function is integrated, which can perform remote notification based on user-defined threshold.

[0262] Among them, the experimental data management adopts a hierarchical storage architecture combined with a time series database (InfluxDB) and a relational database (PostgreSQL). The original sensor data is collected at a frequency of 10Hz, filtered in real time and stored in the time series database at a frequency of 1Hz, retaining 30 days; the minute-level data after downsampling processing is retained for 180 days; the hour-level summary data and experimental event data are stored in the relational database for a long time.

[0263] The data preprocessing process includes: (1) median filtering to remove outliers, window size 5; (2) Kalman filter to smooth time series data, process noise covariance Q=0.01, measurement noise covariance R=0.1; (3) noise removal based on wavelet transform, using Daubechies wavelet, decomposition level 4; (4) missing value interpolation using a combination of forward filling and linear interpolation method.

[0264] Multidimensional data analysis employs multiple algorithms: statistical analysis includes descriptive statistics, hypothesis testing (t-test, ANOVA), and non-parametric statistics; time series analysis includes autoregressive models, fast Fourier transform, and wavelet analysis; machine learning methods include random forests (100 trees, depth limit 12), support vector machines (RBF kernel, C = 1.0), and deep learning models.

[0265] Visualization techniques implement interactive charts based on D3.js and Plotly, supporting multiple visualization forms: time series plots, heatmaps, correlation matrices, behavioral state transition diagrams, and eating pattern radar charts. Multidimensional data filtering, scaling, and drilling are supported, as well as custom comparison views, facilitating in-depth analysis and hypothesis verification by researchers.

[0266] And the experimental data management adopts hierarchical storage architecture, combined with time series database (InfluxDB) and relational database (PostgreSQL). Raw sensor data is collected at a frequency of 10Hz, filtered in real time and stored in a time series database at a frequency of 1Hz, retaining 30 days; minute-level data after downsampling processing retains 180 days; hour-level summary data and experimental event data are permanently stored in the relational database.

[0267] The data preprocessing process includes: (1) median filtering to remove outliers, window size 5; (2) Kalman filter to smooth time series data, process noise covariance Q = 0.01, measurement noise covariance R = 0.1; (3) noise removal based on wavelet transform, using Daubechies wavelet, decomposition level 4; (4) missing value interpolation using a combination of forward filling and linear interpolation methods.

[0268] Multidimensional data analysis employs multiple algorithms: statistical analysis includes descriptive statistics, hypothesis testing (t-test, ANOVA), and non-parametric statistics; time series analysis includes autoregressive models, fast Fourier transform, and wavelet analysis; machine learning methods include random forests (100 trees, depth limit 12), support vector machines (RBF kernel, C = 1.0), and deep learning models.

[0269] Visualization techniques implement interactive charts based on D3.js and Plotly, supporting multiple visualization forms: time series plots, heatmaps, correlation matrices, behavioral state transition diagrams, and eating pattern radar charts. Multidimensional data filtering, scaling, and drilling are supported, as well as custom comparison views, facilitating in-depth analysis and hypothesis verification by researchers.

[0270] Please refer to Figure 2 and Figure 4 , Figure 2 is the architecture diagram of the intelligent gate control system provided by the embodiments of the present application,Figure 4 is a structural schematic diagram of an animal experiment time-limited feeding system provided by an embodiment of the present application. As shown in Figure 2 , the architecture schematic diagram of the intelligent door control system includes:

[0271] a feeding cage 10 for accommodating experimental animals;

[0272] an electrically controlled door passage 20 arranged on the feeding cage for controlling the experimental animals to enter a food area;

[0273] a multi-modal sensor array 30 including food intake monitoring sensors, animal activity state monitoring sensors, physiological parameter monitoring sensors, and environmental parameter monitoring sensors for acquiring multi-modal monitoring data streams of the states and behaviors of the experimental animals;

[0274] an intelligent control unit 40 connected with the electrically controlled door passage and the multi-modal sensor array, as shown in Figure 4 , including:

[0275] a data acquisition module 41 for acquiring multi-modal monitoring data streams of the states and behaviors of the experimental animals through the multi-modal sensor array including food intake monitoring sensors, animal activity state monitoring sensors, physiological parameter monitoring sensors, and environmental parameter monitoring sensors;

[0276] a feeding control module 42 for implementing execution control of a time-limited feeding mode based on the multi-modal monitoring data streams and preset time-limited feeding parameters through a feeding time window control algorithm, a dynamic adjustment algorithm, and a safety monitoring mechanism to obtain execution data of the execution control;

[0277] an adaptive adjustment module 43 for outputting individualized time-limited feeding intervention strategies based on the execution data of the execution control and the multi-modal monitoring data streams through an animal state evaluation model, a diet mode optimization algorithm, and an intervention strategy generator;

[0278] a data management module 44 for implementing experimental data management and analysis based on the multi-modal monitoring data streams, the execution data of the execution control, and the individualized time-limited feeding intervention strategies through a data acquisition and storage module, a data processing and analysis module, and a visual display interface.

[0279] In addition, as Figure 5As shown, the embodiment of the present application also provides an animal experiment time-limited feeding device, which comprises a cage cover 51 and a supporting frame 52, the bottom of the cage cover is provided with a food container 53 and a rotating wheel device 54, and the top of the cage cover is provided with a weight sensor 55 and a temperature and humidity and air pressure sensor 56. The rotating wheel device comprises a function of enabling / disabling for voluntary, workload and motor skill running. The food intake monitoring sensor of the system comprises a weight sensor and a miniature camera, which is used for monitoring food consumption in real time and filtering out interference signals to obtain food intake data. The animal activity state monitoring sensor comprises an infrared activity detector or an acceleration sensor, which is used for identifying and analyzing animal movement patterns to obtain activity classification data. The physiological parameter monitoring sensor is used for monitoring physiological parameters of experimental animals to obtain physiological parameter data. The environmental parameter monitoring sensor comprises a temperature and humidity and air pressure sensor, which is used for collecting and processing environmental parameters to obtain environmental change data.

[0280] The embodiments of the present application are only the preferred embodiments of the present application, and are not used to limit the protection scope of the present application. Any changes, modifications or equivalent replacements made within the protection scope required by the present application shall be included in the protection scope of the present application.

Claims

1. A time-restricted feeding method for animal experiments, characterized in that: Applied to an intelligent gate control system, the intelligent gate control system includes a feeding box for experimental animals, an electrically controlled gate control channel, a multimodal sensor array, and an intelligent control unit. The method includes: Acquiring a multimodal monitoring data stream of the state and behavior of the experimental animal through the multimodal sensor array comprising a food intake monitoring sensor, an animal activity status monitoring sensor, a physiological parameter monitoring sensor, and an environmental parameter monitoring sensor; Based on the multimodal monitoring data stream and the preset time-restricted eating parameters, the execution control of the time-restricted eating mode is realized through the eating time window control algorithm, the dynamic adjustment algorithm and the safety monitoring mechanism, and the execution data of the execution control is obtained; Based on the execution data of the execution control and the multimodal monitoring data stream, outputting a personalized time-restricted feeding intervention strategy through an animal state assessment model, a diet pattern optimization algorithm, and an intervention strategy generator; Based on the multimodal monitoring data stream, the execution data of the execution control and the personalized time-restricted feeding intervention strategy, experimental data management and analysis are achieved through a data acquisition and storage module, a data processing and analysis module and a visual display interface.

2. The method according to claim 1, characterized in that The multimodal monitoring data stream for obtaining the state and behavior of the experimental animal includes: The food intake monitoring sensor comprising a weight sensor and a micro camera is used to monitor food consumption in real time and filter out interference signals to obtain food intake data; Identifying and analyzing animal movement patterns to obtain activity classification data using the animal activity status monitoring sensor comprising an infrared activity detector or an acceleration sensor; Based on the physiological parameter monitoring sensor, the physiological parameters of the experimental animal are monitored to obtain physiological parameter data; The environmental parameter monitoring sensor including the temperature, humidity and air pressure sensor is used to collect and process environmental parameters to obtain environmental change data; Based on the food intake data, the activity classification data, the physiological parameter data and the environmental change data, time alignment and multi-source data fusion are performed to generate the multimodal monitoring data stream.

3. The method according to claim 1, characterized in that The execution control of implementing the time-restricted eating mode includes: Based on the multimodal monitoring data stream and the preset time-limited eating parameters, the eating time window is determined through a time trigger mechanism, and combined with the status monitoring feedback, an eating time window switch event table is generated; Based on the eating time window switch event table, controlling the switch state of the electrically controlled gated channel through a state machine model; Based on the multimodal monitoring data stream, an abnormal state detection algorithm constructed using a multi-level threshold model is used to identify potential health risks and abnormal behaviors, and when potential health risks or abnormal behaviors are identified, an emergency response mechanism is triggered; Based on the emergency response mechanism, combined with the eating time window switch event table and the switch status of the electronically controlled gating channel, gated execution instructions are generated and prioritized to achieve execution control of the time-limited eating mode.

4. The method according to claim 3, characterized in that The abnormal state detection algorithm constructed using the multi-level threshold model identifies potential health risks and abnormal behaviors. When potential health risks or abnormal behaviors are identified, an emergency response mechanism is triggered, including: Based on the multimodal monitoring data stream, multi-scale processing is performed on the multimodal monitoring data stream to construct a multi-scale time representation, and a cross-scale attention module is used to interact features of different scales of the multi-scale time representation to form a multi-time scale feature; Based on the multi-timescale features, the health status indicator and its uncertainty are calculated by a Bayesian neural network to identify the abnormality type; wherein the abnormality type includes high confidence abnormality, high uncertainty abnormality and multimodal conflict; Based on the high-confidence anomalies, high-uncertainty anomalies and multimodal conflicts, different response strategies are selected according to the anomaly types to trigger an emergency response mechanism.

5. The method according to claim 1, wherein The outputting of a personalized time-restricted feeding intervention strategy includes: constructing an animal health status assessment model based on the multimodal monitoring data stream through a feature extraction algorithm and a machine learning model; Based on the animal health status assessment model and the execution data of the execution control, constructing a dietary pattern impact analysis result through time series data analysis and correlation calculation; Based on the dietary pattern impact analysis results and the preset research objectives, a personalized dietary pattern optimization plan is generated through a multi-objective optimization algorithm and dynamic sparse training; Based on the personalized dietary pattern optimization scheme, a personalized time-restricted eating intervention strategy is output through a rule conversion engine and parameter mapping.

6. The method according to claim 5, characterized in that The personalized diet pattern optimization scheme is generated by using a multi-objective optimization algorithm and dynamic sparse training, including: Based on the dietary pattern impact analysis results, extracting a joint feature representation including time-related features, physiological response features, and behavioral pattern features to form an embedding representation in a high-dimensional vector space; Based on the embedding representation, an importance score matrix is ​​calculated through feature importance evaluation to identify key feature dimensions; Based on the limitation of computing resources, the key feature dimensions are subjected to feature sparsification processing, and the features of the key feature dimensions whose importance is lower than a preset threshold are clipped to obtain clipped key feature dimensions; and the clipped key feature dimensions are subjected to computational hierarchical allocation to form a lightweight optimization model; Based on the lightweight optimization model and guided by the preset research objectives, a multi-objective optimization search is performed to generate a personalized dietary pattern optimization plan.

7. The method according to claim 1, characterized in that The experimental data management and analysis are implemented, including: Based on the multimodal monitoring data stream, the execution data of the execution control, and the personalized time-restricted feeding intervention strategy, the multi-channel data is traceably stored through a time series database and a data synchronization mechanism to obtain the stored multi-channel data; For the stored multi-channel data, generating a preprocessed data set by performing noise filtering, data standardization, and missing value processing; Processing the preprocessed data set using a variety of data analysis methods to extract experimental results and pattern features; the various data analysis methods include statistical analysis, machine learning, and time series analysis algorithms; Based on the experimental results and pattern characteristics, experimental data management and analysis are achieved through interactive charts and multidimensional data visualization technology.

8. The method according to claim 1, characterized in that Also includes: Based on the execution results of the personalized time-restricted feeding intervention strategy and the multimodal monitoring data stream, a comprehensive effect score is calculated using an effect evaluation algorithm to generate an intervention effect evaluation result; When the intervention effect evaluation result is lower than the preset effect threshold, the feedback adjustment mechanism is activated to conduct multi-angle investigation and generate adjustment suggestions; Based on the adjustment suggestions, the animal status assessment model and optimization strategy are updated to achieve iterative optimization of the intervention strategy.

9. The method according to claim 1, characterized in that Before obtaining the multimodal monitoring data stream of the experimental animal status and behavior, the method further includes: Based on experimental requirements and animal types, separate feeding spaces and dedicated access areas are designed, and the basic structure of the feeding box is constructed to meet SPF-level cleanliness standards; Based on the basic structure of the breeding box, a sliding gate driven by a micro servo motor and a position sensor are integrated to construct the electrically controlled gated channel; Based on the feeding box basic structure and the electrically controlled gated channel, the food intake monitoring sensor, the animal activity status monitoring sensor, the physiological parameter monitoring sensor and the environmental parameter monitoring sensor are deployed to construct the multimodal sensor array; Based on the multimodal sensor array and the electrically controlled gating channel, an embedded processor, a motor drive circuit and a high-precision analog-to-digital conversion module are integrated to construct an intelligent control unit.

10. A time-restricted feeding system for animal experiments, characterized in that: include: A breeding box for accommodating experimental animals; An electrically controlled gated channel is provided on the feeding box and is used to control the experimental animals from entering the food area; A multimodal sensor array, including sensors for monitoring food intake, animal activity status, physiological parameters, and environmental parameters, is used to obtain multimodal monitoring data streams of the status and behavior of experimental animals; An intelligent control unit, connected to the electronically controlled gating channel and the multimodal sensor array, comprising: a data acquisition module for acquiring a multimodal monitoring data stream of the state and behavior of the experimental animal through the multimodal sensor array comprising a food intake monitoring sensor, an animal activity status monitoring sensor, a physiological parameter monitoring sensor, and an environmental parameter monitoring sensor; An eating control module is used to implement execution control of the time-limited eating mode based on the multimodal monitoring data stream and preset time-limited eating parameters through an eating time window control algorithm, a dynamic adjustment algorithm, and a safety monitoring mechanism, and obtain execution data of the execution control; an adaptive adjustment module for outputting a personalized time-restricted feeding intervention strategy based on the execution data of the execution control and the multimodal monitoring data stream through an animal state assessment model, a dietary pattern optimization algorithm, and an intervention strategy generator; The data management module is used to manage and analyze experimental data based on the multimodal monitoring data stream, the execution data of the execution control, and the personalized time-restricted feeding intervention strategy through a data acquisition and storage module, a data processing and analysis module, and a visual display interface.

Images