Edge computing methods and systems for pet physiological data in low-power IoT

Abstract

This invention relates to a pet physiological data edge computing method and system for low-power Internet of Things (IoT), belonging to the field of IoT technology. The method includes: S1, calculating a physiological fluctuation index to characterize the stability of the current physiological state based on a continuous physiological data sequence acquired by a data acquisition module; S2, generating a target computational load value matching the current physiological state using a preset nonlinear mapping function based on the physiological fluctuation index; S3, calculating a smoothed final computational load command value using a smoothing control model by combining the final computational load command value from the previous cycle with the target computational load value for the current cycle; S4, querying a preset multi-dimensional parameter configuration set based on the final computational load command value and determining specific operating parameters to update the operating configuration of the data acquisition module and the data processing module. This invention effectively extends the battery life of battery-powered devices.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) technology, specifically to a method and system for edge computing of pet physiological data for low-power IoT. Background Technology

[0002] In existing low-power IoT pet monitoring systems, edge computing units typically deployed on the device are responsible for processing physiological data collected by sensors in real time, such as heart rate, respiratory rate, body temperature, and activity level. These systems face a core challenge: the pet's physiological state can undergo drastic and unpredictable changes, such as suddenly switching from deep sleep to a stress state triggered by external stimuli.

[0003] Such state transitions require computing systems to rapidly enhance their data processing capabilities, including increasing data sampling frequency to capture more details, increasing the order of digital filtering to remove motion artifacts, and invoking deeper machine learning models for accurate anomaly identification. However, existing systems typically experience abrupt and rigid paradigm shifts when dealing with such sudden state changes. This approach triggers a sudden surge in computing resource requests, resulting in sharp peaks in the system power consumption curve and significantly shortening battery life. Furthermore, these sudden resource requests can cause nonlinear jitter in the data processing chain, potentially leading to delayed or missed detection of critical physiological anomalies. Therefore, achieving a smooth transition in system power consumption while ensuring no anomalies are missed, and optimizing the damping characteristics of the computational paradigm shift process, are pressing technical challenges in this field.

[0004] The information disclosed in the background section above is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for edge computing of pet physiological data for low-power Internet of Things (IoT) to solve the problems mentioned in the background art.

[0006] The technical solution of the present invention includes the following steps:

[0007] S1. Based on the continuous physiological data sequence acquired by the data acquisition module, calculate the physiological fluctuation index used to characterize the stability of the current physiological state;

[0008] S2. Based on the physiological fluctuation index, a target computational load value matching the current physiological state is generated through a preset nonlinear mapping function.

[0009] S3. Combining the final computational load instruction value of the previous cycle with the target computational load value of the current cycle, the smoothed final computational load instruction value is calculated using the smoothing control model.

[0010] S4. Based on the final calculated load instruction value, query the preset multi-dimensional parameter configuration set and determine the specific working parameters to update the running configuration of the data acquisition module and the data processing module.

[0011] Preferably, S1 specifically includes:

[0012] S11. Using a weighted variance operator, calculate the fluctuation variance of a continuous physiological data sequence within a preset time window;

[0013] S12. Normalize the fluctuation variance using the preset baseline physiological variance to generate the physiological fluctuation index.

[0014] Preferably, the baseline physiological variance is a baseline value obtained through statistical calibration based on the pet's long-term resting state.

[0015] Preferably, the nonlinear mapping function is a logistic function, and S2 specifically includes:

[0016] The physiological fluctuation index is input into the logistic function, which outputs the target computational load value based on the preset minimum system computational load, maximum computational load, gain coefficient, and fluctuation index reference threshold.

[0017] Preferably, the smooth control model calculates the final calculated load command value in the following manner:

[0018] S31. Based on the difference between the target calculated load value and the final calculated load command value of the previous cycle, the dynamic damping coefficient is calculated using the hyperbolic tangent function.

[0019] S32. Using a dynamic damping coefficient, the target calculated load value is weighted and averaged with the final calculated load command value of the previous cycle to generate the final calculated load command value.

[0020] Preferably, the calculation process of the dynamic damping coefficient is constrained by the preset upper and lower bounds of the damping coefficient.

[0021] Preferably, S4 determines the specific operating parameters, including:

[0022] S41. Map the final computational load instruction value to the computational energy level in the multi-dimensional parameter configuration set;

[0023] S42. Extract the working parameter set bound to the computational energy level as the specific working parameters; the working parameter set shall include at least the data sampling frequency, the digital filter order, and the model inference depth.

[0024] An edge computing system for pet physiological data for low-power Internet of Things, comprising:

[0025] The physiological state fluctuation quantification module is used to calculate the physiological fluctuation index, which characterizes the stability of the current physiological state, based on the continuous physiological data sequence acquired by the data acquisition module.

[0026] The target computational load generation module is used to generate a target computational load value that matches the current physiological state based on the physiological fluctuation index and through a preset nonlinear mapping function.

[0027] The dynamic damping adjustment module for the calculated load is used to combine the final calculated load command value of the previous cycle with the target calculated load value of the current cycle, and use a smooth control model to calculate the smoothed final calculated load command value.

[0028] The multi-parameter collaborative configuration module is used to query the preset multi-dimensional parameter configuration set based on the final calculated load command value, and determine the specific working parameters to update the running configuration of the data acquisition module and the data processing module.

[0029] This invention provides an improved edge computing method and system for pet physiological data in the context of low-power Internet of Things (IoT). Compared with existing technologies, it has the following improvements and advantages:

[0030] 1. This invention does not directly respond to changes in physiological state, but buffers and adjusts through a series of interconnected calculation steps. Compared with the instantaneous impact of computing resource requests in the prior art, this invention, through this dynamic damping adjustment mechanism, transforms the resource request process from a step function to a ramp function, significantly smoothing the overall power consumption curve of the system, avoiding current spikes, and thus effectively extending the battery life of battery-powered devices.

[0031] 2. This invention ensures the stability of processing latency and improves system reliability. Because the final computational load instruction value changes smoothly, resource scheduling can proceed smoothly, avoiding congestion and interruption of computational tasks. This allows the latency on the data processing link to remain within a relatively stable and predictable range, reducing nonlinear jitter. As a result, the real-time performance and reliability of the system are significantly improved, and the timeliness and accuracy of judgment for critical physiological abnormalities requiring high-timeliness monitoring, such as cardiac arrest and epileptic premonitory events, are effectively guaranteed.

[0032] 3. Enhanced system robustness and overall optimal resource allocation; this invention, through a multi-parameter collaborative configuration step, abstracts the final computational load command value. The system maps to a pre-defined multi-dimensional parameter configuration set. This configuration set is determined by offline benchmarking of the hardware platform, comprehensively balancing different parameter combinations, data sampling frequency, digital filter order, power consumption, and processing performance under model inference depth. This is fundamentally different from the existing technology that may adjust various parameters in isolation. This invention ensures that multiple parameters affecting power consumption and performance are linked and coordinated, realizing the overall optimized configuration of system resources in different dimensions. The dynamically changing damping coefficient further enhances the system's robustness, enabling the system to have both stability and responsiveness. It can maintain extremely low energy consumption when the state is stable, and can quickly improve processing capabilities at a controllable power consumption ramp-up rate when an emergency occurs, ensuring that no monitoring function is missed. Attached Figure Description

[0033] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0034] Figure 1 This is a flowchart of the system of the present invention. Detailed Implementation

[0035] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0036] Example 1:

[0037] Please see Figure 1 This invention provides a pet physiological data edge computing method for low-power Internet of Things, including the following steps: S1, calculating a physiological fluctuation index to characterize the stability of the current physiological state based on a continuous physiological data sequence acquired by a data acquisition module;

[0038] S2. Based on the physiological fluctuation index, a target computational load value matching the current physiological state is generated through a preset nonlinear mapping function.

[0039] S3. Combining the final computational load instruction value of the previous cycle with the target computational load value of the current cycle, the smoothed final computational load instruction value is calculated using the smoothing control model.

[0040] S4. Based on the final calculated load instruction value, query the preset multi-dimensional parameter configuration set and determine the specific working parameters to update the running configuration of the data acquisition module and the data processing module.

[0041] This embodiment provides an edge computing method for pet physiological data in low-power IoT, aiming to solve the power consumption shock and latency jitter problems caused by rigid shifts in computing paradigms at edge nodes. The method constructs a closed-loop adjustment process, starting with the precise quantification of the pet's physiological state, then generating an ideal computing target, and subsequently using a core smoothing control model to dynamically dampen the target, transforming the smoothed instructions into a coordinated configuration of multi-dimensional operating parameters of the system. This process transforms drastic changes in external physiological state into a smooth transition in internal computing resource requirements, ensuring that key physiological abnormalities are not missed while achieving continuous smoothing of the device's power consumption curve, thereby extending the battery life of wearable monitoring devices.

[0042] Example 2

[0043] S1 specifically includes:

[0044] S11. Using a weighted variance operator, calculate the fluctuation variance of a continuous physiological data sequence within a preset time window;

[0045] S12. Normalize the fluctuation variance using a preset baseline physiological variance to generate a physiological fluctuation index.

[0046] The baseline physiological variance is a baseline value obtained through statistical calibration based on pets in a long-term resting state.

[0047] In this embodiment, the calculation of the physiological fluctuation index aims to translate abstract physiological states into dimensionless mathematical scalars. This calculation draws on the statistical process control concept in the field of signal processing, judging process stability by evaluating the statistical changes of the signal within a specific time window; the weighted average signal value within the window is defined as... :

[0048]

[0049] The physiological fluctuation index is calculated by using a weighted variance operator and incorporating a baseline physiological variance for normalization. The computational model is defined as follows:

[0050]

[0051] in, for The dimensionless physiological fluctuation index at time; This represents the weighted average physiological signal value within the current time window. for Physiological signal values ​​at any given time; The baseline physiological variance is a core calibration value determined by statistically calibrating physiological data collected from specific pets during long-term resting states such as deep sleep, during device initialization or offline training. It represents the inherent baseline fluctuation level of an individual's physiological signals. This calibration process can be performed during device initialization by continuously recording the pet's heart rate or triaxial acceleration modulus data for one hour during deep sleep at night, and directly calculating the overall variance of this dataset. This variance is then used as the baseline physiological variance. ; The size of the time window for evaluating stability; An index for historical data points; : Subscript, representing the baseline; To indicate time or moment;

[0052] In this embodiment, the size of the time window It can be set to 10. If the data acquisition frequency is 1Hz, it corresponds to a 10-second evaluation cycle, which is sufficient for capturing physiological changes in most pets.

[0053] For the first The weights of historical data points are set in an exponentially decaying form to give more importance to recent data; for example, this exponential decay form can be specifically defined as... ,in It is the attenuation constant. The base of the natural logarithm can be 0.5 in this embodiment to ensure that the most recent physiological data has a decisive weight in the calculation;

[0054] This formula introduces a benchmark physiological variance with the same dimensions as the fluctuation variance. As the denominator, it ensures the physiological fluctuation index of the final output. It is a dimensionless relative metric; this design provides a unified and standard input for subsequent calculation modules with different physical units. The higher the value, the more drastic the fluctuation of the current physiological state compared to its baseline state, providing an accurate and reliable basis for the adaptive adjustment of the system.

[0055] Example 3

[0056] The nonlinear mapping function is a logistic function, and S2 specifically includes:

[0057] The physiological fluctuation index is input into the logistic function, which outputs the target computational load value based on the preset minimum system computational load, maximum computational load, gain coefficient, and fluctuation index reference threshold.

[0058] In this embodiment, the target computational load value is generated using the logistic function as a nonlinear mapping function. The underlying logic of this choice is that the relationship between computational resource demand and physiological state fluctuations is not linear, but rather a scenario where demand increases sharply only after crossing a certain state threshold. The S-shaped curve of the logistic function can accurately simulate this reality. This function receives the physiological fluctuation index generated in the previous step and outputs a target computational load value that matches it, the mathematical expression of which is:

[0059]

[0060] in for The target computational load at any given time; and These are the normalized minimum and maximum computing power levels determined by the system hardware platform, both of which are dimensionless parameters. This is the gain coefficient; The input is the physiological fluctuation index; This serves as a reference threshold for the volatility index. : Subscript, representing the smallest; : Subscript, representing the largest; Subscript: indicates reference;

[0061] In this model, the core adjustable parameters and The method for determining this is through analysis of a large number of known physiological states. The data is scientifically determined through statistical cluster analysis, which ensures that the mapping relationship is both theoretically based and consistent with the actual data distribution.

[0062] In a tested implementation, the volatility index reference threshold is... It can be set to 2.5, which means that when the physiological fluctuation index exceeds 2.5 times the baseline steady state, the system should begin to prepare for a significant increase in computational load; at the same time, the gain coefficient... It can be set to 1.5 to control the steepness of the S-curve, ensuring that the transition of the load target is not too abrupt;

[0063] This function will handle the continuous fluctuations generated by the preceding steps. The value is transformed into an ideal computational load target that may experience a step change. This provides a clear direction and goal for subsequent smooth adjustment processes, ensuring that the final direction of system resource scheduling accurately corresponds to the pet's actual physiological needs.

[0064] Example 4

[0065] The smooth control model calculates the final calculated load command value in the following way:

[0066] S31. Based on the difference between the target calculated load value and the final calculated load command value of the previous cycle, the dynamic damping coefficient is calculated using the hyperbolic tangent function.

[0067] S32. Using a dynamic damping coefficient, the target calculated load value and the final calculated load command value of the previous cycle are weighted and averaged to generate the final calculated load command value.

[0068] The calculation process of the dynamic damping coefficient is constrained by the preset upper and lower bounds of the damping coefficient.

[0069] In this embodiment, the smooth control model aims to eliminate potential instantaneous jumps in the target computational load value, generating a smooth final computational load command value. This model introduces a dynamic damping coefficient, enabling the system's response speed to adaptively adjust. The dynamic damping coefficient is calculated based on the difference between the target and the actual load, and this process is constrained by preset upper and lower bounds for the damping coefficient.

[0070]

[0071] in, It is a constant greater than 1, for example, c=3; when c=3, if the load change reaches its maximum, The input is 3. (3)≈0.995, very close to 1; for The dynamic damping coefficient at time t; : Subscript, representing the lower bound or minimum; and These are the preset upper and lower bounds of the damping coefficient. The method for determining them is based on a trade-off test between the system power consumption ramp-up rate and the response delay of abnormal events. The aim is to ensure that the system does not respond too slowly, resulting in missed detections, nor does it respond too quickly, causing an impact. : Subscript, representing the upper bound or maximum; Hyperbolic tangent function;

[0072] For example, in this embodiment, a lower bound for the damping coefficient can be set. This is used for fine-tuning when the system tends to stabilize, to avoid jitter; it sets the upper limit of the damping coefficient. It is used to achieve rapid tracking when there is a large difference between the target load and the actual load, ensuring timely response to emergencies; and The definitions are the same as above, representing the maximum and minimum computing power levels of the system normalization, respectively;

[0073] It is the hyperbolic tangent function; Calculate the load value for the target of the current period; This is the final calculated load command value from the previous cycle; the underlying logic of this formula is that when the system state needs to change drastically, Larger This increases to achieve a rapid convergence; conversely, when the system tends to stabilize, Reduce the damping to keep the system in a highly damped state to avoid fine-tuning;

[0074] The obtained dynamic damping coefficient It is immediately applied to the weighted average model to generate the final calculated load instruction value:

[0075]

[0076] in for The final calculated load command value at time t; these two formulas together constitute a complete dynamic damping regulator, the former determining the step size towards the target, and the latter executing that step; this smooth control model will handle target load sequences that may have sharp peaks. This is transformed into a continuous, smoothly changing final instruction sequence. This fundamentally avoids the instantaneous impact of computing resource requests, allowing the power consumption curve of wearable devices to transition smoothly when dealing with pet stress.

[0077] This invention smooths the power consumption curve, effectively extending battery life. Instead of directly responding to changes in physiological state, this invention buffers and regulates through a series of interconnected calculation steps. The physiological state fluctuation quantification step produces a dimensionless physiological fluctuation index. This index is used in the target computational load generation step, through a pre-defined logistic function calibrated based on a large amount of experimental data. Generate a target computational load value that precisely matches the current physiological state. Up to this point, the system response may still be step-like; the key advancement lies in calculating the dynamic damping adjustment step of the load, which introduces a core smooth control model. This model first calculates a dynamically changing damping coefficient using a hyperbolic tangent function based on the difference between the target load and the actual load at the previous moment. The formula is The practical significance of the formula lies in enabling the system's response speed, i.e., the magnitude of damping, to be adaptive. Subsequently, this dynamic damping coefficient is used to perform a weighted average of the target calculated load value and the actual calculated load value of the previous cycle to generate the final calculated load command value. Mathematically, this formula is an exponential smoothing filter, and its internal logic determines the output value. It must be for the input value The smooth following, rather than instantaneous jumps, of this system transforms a potentially drastically changing input signal into a continuous and smooth output command. Compared to the instantaneous impact of computing resource requests in existing technologies, this invention, through its dynamic damping adjustment mechanism, transforms the resource request process from a step function to a ramp function, significantly smoothing the overall power consumption curve of the system, avoiding current spikes, and thus effectively extending the battery life of battery-powered devices.

[0078] Example 5

[0079] S4 determines the specific operating parameters, including:

[0080] S41. Map the final computational load instruction value to the computational energy level in the multi-dimensional parameter configuration set;

[0081] S42. Extract the working parameter set bound to the computational energy level as the specific working parameters; the working parameter set shall include at least the data sampling frequency, the digital filter order, and the model inference depth.

[0082] In this embodiment, the step of determining specific operating parameters aims to abstract and smooth the final computational load instruction value calculated in the preceding steps. This translates into specific control actions on the system hardware and software. This process relies on a preset multi-dimensional parameter configuration set, which efficiently maps continuous load commands to discrete, optimized combinations of hardware parameters. The method for formulating this configuration set is as follows: by conducting offline benchmark tests on the hardware platform, systematically analyzing the power consumption and processing performance under different parameter combinations, and combining the diagnostic accuracy requirements under different physiological scenarios, the configuration set is obtained after weighing and calibrating.

[0083] For clarity, an example of a multi-dimensional parameter configuration set containing three computational energy levels is provided below:

[0084] Energy Level 1, Low Power Mode: When the final calculation of the load command value In the range At that time, the system queries and applies the operating parameter set for this energy level: data sampling frequency. Digital filter order Model inference depth Only basic heart rate statistics and abnormality detection are performed;

[0085] Level 2, Normal Mode: When the final load command value is calculated In the range At this time, the system applies the following operating parameter set for this energy level: data sampling frequency. Digital filter order Model inference depth Perform multi-dimensional physiological state analysis, including activity level;

[0086] Energy Level 3, High-Performance Mode: When the final calculation load command value is... In the range At this time, the system applies the following operating parameter set for this energy level: data sampling frequency. Digital filter order Model inference depth Perform refined anomaly pattern recognition and diagnosis;

[0087] The final load command value will be calculated. Mapped to a discrete computational energy level in this configuration set Extract the calculated energy level The bound working parameter group { As specific operating parameters, the operating parameter set includes at least the data sampling frequency. Digital filter order Model inference depth and the index of discrete computational energy levels ;

[0088] This step completes the final stage of the entire closed-loop control, effectively applying smooth computational instructions to the physical system. Its technical effect is to achieve coordinated optimization and linkage adjustment of multiple core parameters. When the system updates the sampling frequency, filter order, and model depth according to the instructions, these changes will affect the characteristics of the physiological signals collected in the next cycle, thereby starting a new round of perception-decision-adjustment cycle, forming a complete, self-consistent edge computing load adjustment system with adaptive damping characteristics.

[0089] This invention ensures the stability of processing latency and improves system reliability. In existing technologies, momentary congestion of computing tasks can cause nonlinear jitter in the latency of the data processing link, which may lead to delays or even omissions in the judgment of critical physiological abnormalities. In this invention, since the final computing load instruction value changes smoothly, resource scheduling can be carried out smoothly, avoiding congestion and interruption of computing tasks. This allows the latency on the data processing link to be kept within a relatively stable and predictable range, reducing nonlinear jitter. As a result, the real-time performance and reliability of the system are significantly improved, and the timeliness and accuracy of the judgment of critical physiological abnormalities requiring high-time-sensitivity monitoring, such as cardiac arrest and epileptic premonitory symptoms, are effectively guaranteed.

[0090] Example 6

[0091] A pet physiological data edge computing system for low-power Internet of Things includes: a physiological state volatility quantification module, used to calculate a physiological volatility index to characterize the stability of the current physiological state based on a continuous physiological data sequence acquired by a data acquisition module;

[0092] The target computational load generation module is used to generate a target computational load value that matches the current physiological state based on the physiological fluctuation index and through a preset nonlinear mapping function.

[0093] The dynamic damping adjustment module for the calculated load is used to combine the final calculated load command value of the previous cycle with the target calculated load value of the current cycle, and use a smooth control model to calculate the smoothed final calculated load command value.

[0094] The multi-parameter collaborative configuration module is used to query the preset multi-dimensional parameter configuration set based on the final calculated load command value, and determine the specific working parameters to update the running configuration of the data acquisition module and the data processing module.

[0095] This embodiment also provides a pet physiological data edge computing system for low-power IoT, serving as a hardware and functional modular implementation of the aforementioned method. The system includes a physiological state fluctuation quantification module, a target computing load generation module, a computing load dynamic damping adjustment module, and a multi-parameter collaborative configuration module. The functions of these four modules correspond to steps S1 to S4 in the aforementioned method, and they are coupled together to form a closed-loop adaptive adjustment system. The physiological state fluctuation quantification module is responsible for sensing and quantification; the target computing load generation module is responsible for deciding the ideal target; the computing load dynamic damping adjustment module, as the core, is responsible for smooth transitions; and the multi-parameter collaborative configuration module is responsible for execution and feedback. This modular system architecture enables refined and smooth management of computing resources. When applied to pet wearable monitoring devices, it can significantly smooth the power consumption curve and stabilize data processing latency without sacrificing the ability to monitor key physiological abnormalities, thereby enhancing the overall robustness and battery life of the system.

[0096] This invention, through a multi-parameter collaborative configuration step, abstracts the final calculated load command value. The system maps to a pre-defined multi-dimensional parameter configuration set. This configuration set is determined by offline benchmarking of the hardware platform, comprehensively balancing different parameter combinations, data sampling frequency, digital filter order, power consumption, and processing performance under model inference depth. This is fundamentally different from the existing technology that may adjust various parameters in isolation. This invention ensures that multiple parameters affecting power consumption and performance are linked and coordinated, realizing the overall optimized configuration of system resources in different dimensions. The dynamically changing damping coefficient further enhances the system's robustness, enabling the system to have both stability and responsiveness. It can maintain extremely low energy consumption when the state is stable, and can quickly improve processing capabilities at a controllable power consumption ramp-up rate when an emergency occurs, ensuring that no monitoring function is missed.

[0097] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A pet physiological data edge computing method for low-power Internet of Things, characterized in that, The method comprises the following steps: S1, calculating a physiological fluctuation index for representing stability of a current physiological state based on a continuous physiological data sequence acquired by a data acquisition module; S2, generating a target calculation load value matched with the current physiological state by a preset nonlinear mapping function according to the physiological fluctuation index; S3, combining a final calculation load instruction value of a previous period with the target calculation load value of the current period, and calculating a final calculation load instruction value after smoothing by using a smoothing control model; S4, querying a preset multi-dimensional parameter configuration set according to the final calculation load instruction value, and determining specific working parameters for updating running configurations of the data acquisition module and the data processing module; The smoothing control model calculates the final calculation load instruction value in the following manner: S31, calculating a dynamic damping coefficient by a hyperbolic tangent function based on a difference between the target calculation load value and the final calculation load instruction value of the previous period; S32, performing weighted average on the target calculation load value and the final calculation load instruction value of the previous period by using the dynamic damping coefficient to generate the final calculation load instruction value.

2. The pet physiological data edge computing method for low-power Internet of Things according to claim 1, characterized in that, S1 specifically comprises: S11, calculating a fluctuation variance of the continuous physiological data sequence within a preset time window by using a weighted variance operator; S12, performing normalization processing on the fluctuation variance by using a preset reference physiological variance to generate the physiological fluctuation index.

3. The pet physiological data edge computing method for low-power Internet of Things according to claim 2, characterized in that, The reference physiological variance is a reference value acquired by statistical calibration according to a pet in a long-term resting state.

4. The pet physiological data edge computing method for low-power Internet of Things according to claim 1, characterized in that, The nonlinear mapping function is a logistic function, and S2 specifically comprises: The physiological fluctuation index is input into the logistic function, and the logistic function outputs the target calculation load value based on a preset minimum calculation load, a maximum calculation load, a gain coefficient and a fluctuation index reference threshold of the system.

5. The pet physiological data edge computing method for low-power Internet of Things according to claim 1, characterized in that, The calculation process of the dynamic damping coefficient is constrained by preset upper and lower boundaries of the damping coefficient.

6. The pet physiological data edge computing method for low-power Internet of Things according to claim 1, characterized in that, S4 specifically comprises: S41, mapping the final calculation load instruction value to a calculation energy level in the multi-dimensional parameter configuration set; S42, extracting a working parameter group bound to the calculation energy level as the specific working parameters; the working parameter group at least includes a data sampling frequency, a digital filter order and a model reasoning depth.

7. A pet physiological data edge computing system for low-power Internet of Things, applied to the pet physiological data edge computing method for low-power Internet of Things in any one of claims 1-6, characterized in that, The method comprises: a physiological state fluctuation quantification module configured to calculate a physiological fluctuation index for representing stability of a current physiological state based on a continuous physiological data sequence acquired by a data acquisition module; a target calculation load generation module configured to generate a target calculation load value matched with the current physiological state by a preset nonlinear mapping function according to the physiological fluctuation index; a calculation load dynamic damping adjustment module configured to combine a final calculation load instruction value of a previous period with the target calculation load value of the current period, and calculate a final calculation load instruction value after smoothing by using a smoothing control model; a multi-parameter collaborative configuration module configured to query a preset multi-dimensional parameter configuration set according to the final calculation load instruction value, and determine specific working parameters for updating running configurations of the data acquisition module and the data processing module.

Images