A smart home system

Abstract

The present invention discloses a smart home system, relates to the field of smart control technology, and is used to solve the problem of poor control efficiency of smart homes. The system comprises a task allocation module, a rule predefinition module, an adaptation analysis module and an intelligent control module. The present invention first determines the performance index and the best task type of smart home equipment, allocates the tasks generated by the smart home system according to the task type and the performance index, combines a rule engine and machine learning to determine the best execution rules under the current network and device status as predefined rules when allocating tasks in the smart home, analyzes the matching situation between the trained model and the predefined rules, obtains the adaptation impact information generated in the process of rule matching, determines whether the model is adapted to the predefined rules, generates different signals for management and control of the smart home according to the adaptation situation of the predicted state of the model, thereby improving the data processing efficiency of the smart home system.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control technology, and more specifically, to a smart home system. Background Art

[0002] Smart home system is a solution that uses advanced automation technology to integrate home living environment with various smart devices and services. These smart systems connect home devices such as lighting, air conditioning, security systems, entertainment equipment, etc. to a common network to achieve remote control, monitoring and automation of these devices. Users can control smart devices in their home through smartphones, tablets or other network devices, no matter where they are, thus providing convenience, improving energy efficiency and enhancing home security.

[0003] Deficiencies of existing technologies:

[0004] In the current smart home system, most complex communication interactions, device control, model training and other tasks are handled by the cloud. Although this centralized task processing method has certain advantages in resource integration and management, it also brings many problems. First, it significantly increases the burden on the cloud, because all data processing and analysis tasks need to be performed through the cloud, which not only increases the traffic of data transmission, but may also cause overload of the cloud server. Secondly, this mode of relying on cloud processing introduces significant delays in device control, especially in poor network conditions. The delay problem is more serious, which directly affects the user experience and the real-time response capability of the system. Therefore, the current smart home system urgently needs to be improved to reduce the burden on the cloud and solve the delay and security problems caused by centralized processing. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, the present invention first determines the performance indicators and optimal task types of smart home devices, allocates the tasks generated by the smart home system according to the task types and performance indicators, combines the rule engine and machine learning to determine the optimal execution rules under the current network and device status as predefined rules, analyzes the matching situation between the trained model and the predefined rules, obtains the adaptation impact information generated in the rule matching process, determines whether the model is adapted to the predefined rules, and generates different signals for management and control of the smart home according to the adaptation of the predicted state of the model, so as to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A smart home system includes a task allocation module, a rule pre-definition module, an adaptation analysis module and an intelligent control module, and each module is connected through a signal;

[0008] The task allocation module is used to evaluate the processing capacity, storage space and network conditions of the cloud and each edge device, determine the performance indicators of the smart home device, determine the best task type for the smart home device through system requirements and capability evaluation, and allocate the tasks generated by the smart home system according to the determined task type and performance indicators;

[0009] The rule pre-definition module is used to combine the rule engine and machine learning to determine the optimal execution rules under the current network and device status when allocating tasks for smart homes. It also controls the execution rule strategy of smart homes as pre-defined rules and uses historical data to train the model.

[0010] The adaptation analysis module is used to analyze the matching between the prediction results of the trained model and the predefined rules, obtain the adaptation impact information generated during the rule matching process, and determine whether the prediction results of the model are adapted to the predefined rules;

[0011] The intelligent control module is used to analyze the adaptation of the model's prediction state to the predefined rules, generate analysis results, and execute corresponding rules based on the analysis results.

[0012] In a preferred embodiment, when allocating tasks in a smart home, a rule engine and machine learning are combined to determine the optimal execution rules under the current network and device status, control the execution rule strategy of the smart home, and use historical data to train the model. The specific steps include:

[0013] Collect requirements and define key tasks in smart home systems, including data collection, processing, storage and execution of control instructions, and determine smart home system requirements;

[0014] Evaluate the processing power, storage space, and network conditions of the cloud and each edge device, including CPU speed, memory size, and storage space; check the network speed and stability of the device connection; evaluate the energy efficiency and power requirements of the cloud and each edge device; and determine the capacity evaluation results of the cloud and each edge device;

[0015] According to the system requirements and capability evaluation results, the optimal task type for each smart home device is determined, and the tasks generated by the smart home system are allocated according to the determined task type and performance indicators.

[0016] In a preferred embodiment, when allocating tasks in a smart home, a rule engine and machine learning are combined to determine the optimal execution rules under the current network and device status, control the execution rule strategy of the smart home, and use historical data to train the model. The specific steps include:

[0017] Identify the characteristics of various tasks that need to be performed, including data processing, user request response, and video stream processing, and divide the tasks into routine tasks and complex tasks. Routine tasks are processed using rule engines, and complex tasks are predicted using machine learning models.

[0018] Collect historical data from the smart home system, including task type, execution location, network status, device load, and pre-process the historical data;

[0019] Use chi-square test, decision tree or recursive feature elimination to determine the information content characteristics of historical data and use them as the main features. Use long short-term memory neural network as the model and use the preprocessed historical data as training samples to train the model.

[0020] The main features and preprocessed historical data are input into the model as training data. The training data includes a training set and a test set. The training set is used to train the model, and the test set is used to evaluate the generalization ability of the model. After the model is trained, the prediction results of the smart home execution rules are obtained.

[0021] In a preferred embodiment, analyzing the matching between the trained model prediction results and the predefined rules to obtain the adaptation impact information generated during the rule matching process includes the following steps:

[0022] The adaptation impact information includes predicted fluctuation information and periodic change information. The predicted fluctuation information includes the cloud edge convergence stability index, and the periodic change information includes the change trend adaptation index.

[0023] The cloud edge convergence stability index in the forecast fluctuation information and the change trend adaptation index in the periodic change information are combined to generate the control stability coefficient;

[0024] The cloud edge convergence stability index is proportional to the control stability coefficient, and the change trend adaptation index is proportional to the control stability coefficient.

[0025] In a preferred embodiment, the cloud edge convergence stability index is obtained as follows:

[0026] The predicted data of the model in unit time is obtained as samples, the number of samples n, the model predicted value and the actual measured value are obtained, and the measurement error value is calculated. The calculation expression is: In the formula, sj i , yc i Represent the actual measured value and model predicted value of the i-th sample, and obtain the actual measured mean SJ avg and the model predicted mean YC avg , calculate the convergence determination value: Obtain the number of samples m required for the model output to reach a stable state from the initial fluctuation, and calculate the predicted stable value. The calculation expression is: Calculate the cloud edge convergence stability index, the calculation expression is:

[0027]

[0028] In a preferred embodiment, the change trend adaptation index is obtained in the following manner:

[0029] Get time series data, use linear regression to extract the change trend, determine the number of observation points N of the time series, define the time variable t, and get the trend value obtained by extracting the change trend. The trend value calculation expression is: y t =β 0 +β 1 t, where β 0 is the intercept, β 1 is the slope, use the model to predict future data points to get the predicted value, calculate the trend error value, and the calculation expression is: In the formula, is the predicted value of the model at time t, and the trend residual value is calculated. The calculation expression is: Get the trend average y avg , calculate the change trend adaptation index, the calculation expression is:

[0030] In a preferred embodiment, the cloud edge convergence stability index YBS and the change trend adaptation index BHQ are obtained and normalized to generate a control stability coefficient, which is calibrated as K x , the expression is: In the formula, K x is the control stability factor, z 1 、z 2 is the predefined proportional coefficient of the cloud edge convergence stability index YBS and the change trend adaptation index BHQ, and z 1 、z 2 Both are greater than 0.

[0031] In a preferred embodiment, the method for analyzing the adaptation of the predicted state of the model to the predefined rules, generating analysis results, and executing corresponding rules according to the analysis results includes the following steps:

[0032] The control stability coefficient obtained by analyzing the adaptation of the model's predicted state to the predefined rules is compared with the pre-set control threshold;

[0033] If the control stability coefficient is greater than or equal to the control threshold, a prediction accuracy signal is generated and the predefined rules are executed;

[0034] If the control stability coefficient is less than the control threshold, an abnormal prediction signal is generated and the control adjustment of the smart home rules is performed.

[0035] The technical effects and advantages of a smart home system of the present invention are as follows:

[0036] The present invention first determines the performance indicators and optimal task types of smart home devices, allocates tasks generated by the smart home system according to the determined task types and performance indicators, combines the rule engine and machine learning to determine the optimal execution rules under the current network and device status as predefined rules, controls the execution rule strategy of the smart home, and trains the model. The matching situation between the trained model prediction results and the predefined rules is analyzed, the adaptation impact information generated in the rule matching process is obtained, it is determined whether the prediction results of the model are adapted to the predefined rules, and different signals are generated according to the adaptation situation of the prediction status of the model to manage and control the smart home, thereby reducing the burden on the cloud and improving the data processing efficiency of the smart home system. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 The figure is a schematic diagram of the structure of a smart home system of the present invention. DETAILED DESCRIPTION

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

[0039] In order to achieve the above objectives, Figure 1 A schematic diagram of the structure of a smart home system of the present invention is given, which specifically includes a task allocation module, a rule pre-definition module, an adaptation analysis module and an intelligent control module, and each module is connected by a signal;

[0040] The task allocation module is used to evaluate the processing capacity, storage space and network conditions of the cloud and each edge device, determine the performance indicators of the smart home device, determine the best task type for the smart home device through system requirements and capability evaluation, and allocate the tasks generated by the smart home system according to the determined task type and performance indicators;

[0041] The rule pre-definition module is used to combine the rule engine and machine learning to determine the optimal execution rules under the current network and device status when allocating tasks for smart homes. It also controls the execution rule strategy of smart homes as pre-defined rules and uses historical data to train the model.

[0042] The adaptation analysis module is used to analyze the matching between the prediction results of the trained model and the predefined rules, obtain the adaptation impact information generated during the rule matching process, and determine whether the prediction results of the model are adapted to the predefined rules;

[0043] The intelligent control module is used to analyze the adaptation of the model's prediction state to the predefined rules, generate analysis results, and execute corresponding rules based on the analysis results.

[0044] The cloud-edge collaborative smart home system refers to an intelligent environment that combines cloud computing and edge computing technologies, allowing data processing, storage, and analysis to be closer to the data source, that is, the user's home, while also using cloud resources for deeper data processing and intelligent decision-making. This collaborative model aims to improve response speed, reduce network latency and bandwidth usage, and increase the reliability and security of data processing. The application scenarios of the smart home system include multiple components and steps, as follows:

[0045] Edge devices include smart home devices such as smart light bulbs, thermostats, security cameras, etc., which all have certain computing capabilities and can perform preliminary processing of data;

[0046] There are also more powerful devices in the edge devices, namely edge node devices, which can perform more complex data processing and storage tasks, as well as temporary data caching, such as routers, home servers or dedicated edge servers;

[0047] Cloud services usually refer to cloud platforms, which provide large-scale data storage, analysis, and advanced intelligent functions, such as artificial intelligence-driven data analysis and machine learning models;

[0048] Design the smart home system and determine the specific requirements of the system, including functional requirements, performance indicators and security requirements, so as to design a network topology suitable for the home environment, including selecting appropriate edge devices and nodes, connecting various smart home devices (edge ​​devices) with edge nodes, and configuring the network to ensure stable and reliable data transmission.

[0049] Formulate a data processing strategy to determine whether data should be processed on the edge device or uploaded to the cloud for processing, that is, optimize the allocation of tasks on the cloud and edge devices to determine a dynamic task allocation strategy. The specific steps are as follows:

[0050] Evaluate system requirements and capabilities. First, collect requirements and define key tasks in smart home systems, such as data collection, processing, storage, and execution of control commands. Key tasks during system operation, such as data collection (e.g., temperature and humidity sensor data), data processing (e.g., analyzing consumer electricity usage habits), storage requirements (e.g., historical data storage), and execution of control commands (e.g., automatically adjusting air conditioning temperature);

[0051] Evaluate the capabilities of the cloud and edge devices, such as device processing power, storage space, and network conditions. Understand the performance indicators of each smart home device, including CPU speed, memory size, and storage space. Check the network speed and stability of the device connection, especially for tasks that require real-time processing. Evaluate the energy efficiency and power requirements of the device, especially for devices operating in environments with unstable power supplies. For example, evaluate and analyze the camera to see whether the processor equipped with the camera can quickly analyze video data or needs to send data to the cloud for processing. The bandwidth required for the camera to upload a high-definition video stream and whether the local network supports such bandwidth. When the camera has no power backup, power fluctuations may affect the video recording quality and data integrity.

[0052] By evaluating system requirements and capabilities, the optimal task type for each smart home device is determined, and tasks are appropriately assigned based on the capabilities of the device. This not only improves the efficiency and responsiveness of the system, but also ensures the stability and reliability of the system when critical tasks are performed.

[0053] Assign tasks generated by smart home systems based on identified task types and performance metrics. For example, real-time data processing and emergency response may be more suitable for execution at the edge, while big data analysis and long-term storage may be more suitable for the cloud.

[0054] For example, using rules to quickly determine task allocation, that is, by setting a series of logical rules, the system can automatically decide whether the task is processed in the cloud or executed on the edge device based on the real-time monitored environment and device status. First, it is necessary to define the key factors that affect the task allocation decision, such as: device load status including CPU usage, memory usage, etc., network conditions including latency, bandwidth usage, etc., task characteristics including computing intensive or data intensive, urgency, expected response time, etc.;

[0055] Based on the key decision factors that affect task allocation, a set of rules are formulated for task allocation. If the device CPU usage exceeds 75%, the data processing task is transferred to the cloud. If the network delay is less than 100ms, the real-time response task is performed on the edge device first. For tasks with a high data security level, they are always processed on the edge device regardless of the device load.

[0056] The rules are encoded and integrated into the smart home system. According to the operation status and user feedback, the rules can be adjusted and optimized. For example, the threshold of CPU usage can be adjusted, or the response strategy to network delay can be improved to adapt to special situations encountered in actual operation.

[0057] The task allocation of smart homes is to combine the rule engine and machine learning to determine the optimal execution rules under the current network and device status as predefined rules, control the execution rule strategy of smart homes, and use historical data to train the model;

[0058] Clarify the characteristics of various tasks that need to be performed, such as data processing, user request response, video stream processing, etc., and divide the tasks into routine tasks and complex tasks. Routine tasks are suitable for rule engine processing, and complex tasks are suitable for machine learning model prediction. Fixed rules are determined based on experience to quickly process routine tasks. For example, for tasks with low latency requirements, such as real-time video processing of doorbell cameras, rules can be directly specified to be executed on edge devices. The rule engine can be integrated into the smart home system, such as using tools such as Droo ls or Node-RED to realize the automatic execution of rules.

[0059] Collect historical data from the smart home system, including task type, execution location (cloud or edge), network status (latency, bandwidth), device load, etc., extract useful features from the collected data for training machine learning models, including preprocessing of historical data, such as normalization, encoding, and selection of key features:

[0060] Normalize numerical data such as CPU usage, memory usage, latency, etc. to ensure that these features are evaluated by the model on the same scale. Common methods include minimum-maximum normalization or Z-score normalization (standardization);

[0061] Encode categorical data. For example, task type and network type are categorical variables, which can be converted into numerical data using One-Hot Encoding. For example, for task types (data analysis, video processing, etc.), if there are three types, data analysis can be encoded as 1, 0, 0, and video processing can be encoded as 0, 1, 0.

[0062] Use statistical methods (such as chi-squared test), model-based feature selection (such as using decision trees), or iterative methods (such as recursive feature elimination) to select the most informative features. For example, you may find that network latency has the greatest impact on the choice of task execution location, so you choose it as the main feature.

[0063] A long short-term memory neural network (LSTM) is used as the model, and historical data is used as training samples for model training. The processed features and historical task execution results (such as whether the task is successfully executed on the edge device) are used as training data. The historical data is divided into a training set for model learning, and the test set is used to evaluate the generalization ability of the model. 80% of the data is used as a training set and 20% of the data is used as a test set. The model is trained using the divided training set, and the test set is used to evaluate the model's performance indicators such as accuracy, recall rate, and F1 score. The trained model is used to predict new data, and the matching with the predefined rules is determined based on the prediction results of the training model.

[0064] The specific model inputs include task type, execution location, network status, and device load, and the output is the adaptation of predefined rules;

[0065] Assume that a smart home system includes a smart temperature control system. The purpose of the system is to automatically adjust the working state of the air conditioner according to the preferences of family members and the current room temperature. The predefined rules are as follows: if the room temperature exceeds 25°C and someone is in the room, the air conditioner should be adjusted to cooling mode; if the room temperature is below 18°C ​​and someone is in the room, the air conditioner should be adjusted to heating mode; if no one is in the room, the air conditioner should remain off;

[0066] The model receives inputs such as the current room temperature, detection of occupancy in the room (via sensors or cameras), and set preferences from historical data.

[0067] The model analyzes the current input data and predicts the most suitable air conditioning setting state;

[0068] Assume that the model predicts that the room temperature is 26°C and there is someone in the room. According to predefined rule 1, the model output should be "turn on the air conditioner and adjust it to cooling mode". Match the model output with the predefined rule to check whether the model output is compatible with the rule.

[0069] In smart home systems, matching predefined rules usually involves uploading data to the cloud for processing. However, in some cases, edge device processing may be more effective with higher accuracy and efficiency. In this case, edge data processing can bring more advantages. Edge devices can process and analyze data immediately without waiting for data to be uploaded to the cloud and then return the processing results, thereby achieving faster real-time response. Completing some data processing tasks at the edge can reduce the burden on cloud servers, reduce data transmission volume and processing delays, and improve the performance and stability of the entire system. Therefore, when designing a smart home system, you can reasonably select the location of data processing for analysis based on specific circumstances and needs, and then select a better control strategy based on the analysis results, thereby improving the management efficiency of the smart home.

[0070] The adaptation analysis module analyzes the prediction results of the training model and the rule matching situation, and obtains the adaptation impact information generated in the rule matching process. The adaptation impact information includes prediction fluctuation information and periodic change information. The adaptation impact information is used to analyze whether the smart home devices are adapted to the predefined rules when they are working, and then adjust the strategy according to the adaptation situation.

[0071] The predicted fluctuation information includes the cloud edge convergence stability index and is calibrated as YBS, and the periodic change information includes the change trend adaptation index and is calibrated as BHQ;

[0072] The cloud-edge convergence stability index in the prediction fluctuation information indicates the fluctuation of the prediction results of task allocation between cloud computing and edge computing in the smart home system. It is an indicator used to measure the stability and consistency of the model prediction results. The cloud-edge convergence stability index mainly reflects the fluctuation degree and convergence speed of the model's prediction of task allocation between cloud and edge devices within a given time period. The cloud-edge convergence stability index measures the change range of the model's task allocation decision in multiple consecutive prediction cycles. The greater the volatility, the worse the stability of the model output, which may be greatly affected by data input fluctuations, model parameter settings or external environment.

[0073] The cloud edge convergence stability index will have an effect on the following aspects:

[0074] Performance monitoring: By calculating the cloud-edge convergence stability index in real time, system administrators and maintenance personnel can quickly identify possible problems with model performance, such as improper parameter settings, model overfitting or underfitting, etc.

[0075] Tuning basis: The cloud edge convergence stability index provides a quantitative indicator to help the technical team fine-tune the model and optimize the model's learning rate, regularization parameters, or other parameters that affect the model's stability.

[0076] As a key indicator for evaluating the prediction stability of models in smart home systems, the cloud-edge convergence stability index helps system designers and maintainers better understand and optimize the performance of models in actual operations, especially in complex environments involving cloud computing and edge computing. By monitoring and optimizing the cloud-edge convergence stability index, the reliability of the system and user satisfaction can be significantly improved.

[0077] The cloud edge convergence stability index is obtained as follows:

[0078] The predicted data of the model in unit time is obtained as samples, the number of samples n, the model predicted value and the actual measured value are obtained, and the measurement error value is calculated. The calculation expression is: In the formula, sj i , yc i Represent the actual measured value and model predicted value of the i-th sample, and obtain the actual measured mean SJ avg and the model predicted mean YC avg , calculate the convergence determination value: Obtain the number of samples m required for the model output to reach a stable state from the initial fluctuation, and calculate the predicted stable value. The calculation expression is: Calculate the cloud edge convergence stability index, the calculation expression is:

[0079]

[0080] It should be noted that the model's prediction data may include the task allocation decisions within each prediction cycle, as well as the corresponding actual system load, network status and other related parameters; when the model makes a prediction, the output of each prediction is recorded as the prediction value through the normalization method, and the actual measurement value is the actual measurement value corresponding to the prediction time or prediction cycle; the number of samples required for the model output to reach a stable state from the initial fluctuation is obtained.

[0081] The change trend adaptation index in the periodic change information is used to indicate the model's adaptation to the long-term trend changes in the time series data generated by each device in the smart home, helping to analyze and understand the performance of the model in the face of data trend changes. The change trend adaptation index measures the degree of consistency between the model prediction results and the actual data trend, reflecting the model's ability to capture and follow the medium- and long-term change trends of the data. The change trend adaptation index will have an impact on the following aspects:

[0082] Structural adjustment: The feedback of the changing trend adaptation index can directly affect the model's architecture design. If the index is small, it means that the model is not adaptable enough to the trend, and it may be necessary to select or design a model that is more suitable for capturing the trend, such as a neural network that integrates more long-term and short-term memory units, or a composite model that adds autoregressive components.

[0083] Performance evaluation: The trend adaptation index can measure the model's ability to effectively understand and predict data trends in practical applications, which is crucial for prediction accuracy. A high trend adaptation index usually means that the model can maintain good performance under different test conditions.

[0084] The change trend adaptation index is obtained as follows:

[0085] Get time series data, use linear regression to extract the change trend, determine the number of observation points N of the time series, define the time variable t, and get the trend value obtained by extracting the change trend. The trend value calculation expression is: y t =β 0 +β 1 t, where β 0 is the intercept, β 1 is the slope, use the model to predict future data points to get the predicted value, calculate the trend error value, and the calculation expression is: In the formula, is the predicted value of the model at time t, and the trend residual value is calculated. The calculation expression is: Get the trend average y avg , calculate the change trend adaptation index, the calculation expression is:

[0086] It should be noted that time series data is arranged in chronological order and is continuous data; define the time variable to represent each observation point of the time series. For example, if the data is collected daily, t can be an integer from 1 to N, where N is the total number of days in the observation period; use statistical software or programming libraries (such as Python's statsmode ls) to perform linear regression and obtain β 0 , β 1 , β 0 is the intercept, which indicates the predicted value when t=0, β 1 is the slope, which represents the rate of change of the predicted value over time t.

[0087] The predicted fluctuation information and periodic variation information are combined to generate the control stability coefficient;

[0088] The cloud edge convergence stability index YBS and the change trend adaptation index BHQ are obtained and normalized to generate the control stability coefficient, which is calibrated as K x , the expression is: In the formula, K x is the control stability factor, z 1 、z 2 is the predefined proportional coefficient of the cloud edge convergence stability index YBS and the change trend adaptation index BHQ, and z 1 、z2 Both are greater than 0.

[0089] This embodiment uses a weighted summation method to combine the cloud-edge convergence stability index and the change trend adaptation index to generate a comprehensive control stability coefficient. This control stability coefficient can be used as an input parameter of the smart home system to determine the execution of smart home rules.

[0090] It should be noted that the size of the predefined proportional coefficient is to quantify each parameter to obtain a specific numerical value. In order to facilitate subsequent comparison, the size of the coefficient depends on the amount of sample data and the preliminary setting of the corresponding predefined proportional coefficient for each group of sample data by technical personnel in this field. It is not unique, as long as it does not affect the proportional relationship between the parameter and the quantized numerical value. For example, the cloud edge convergence stability index is proportional to the control stability coefficient. The cloud edge convergence stability index and the change trend adaptation index are normalized to have the same dimension and range. This can be achieved by subtracting the mean from the original data and dividing it by the standard deviation, or mapping the data to the range of [0, 1].

[0091] The larger the cloud-edge convergence stability index and the change trend adaptation index, the larger the generated control stability coefficient, which indicates that the model has a higher accuracy when making rule predictions, is less susceptible to fluctuations when making smart home control predictions, and has a higher degree of adaptation to predefined matching rules. The consistency between model predictions and predefined rules improves the credibility of decisions, and decision makers can rely on model outputs to make more accurate business decisions.

[0092] The smaller the cloud-edge convergence stability index and the smaller the change trend adaptation index, the smaller the generated control stability coefficient, and the more likely the predicted smart home control rules are to be incompatible with the predefined matching rules. The model is not sufficiently adapted to the predefined rules during prediction and has weak adaptability to trends. Frequent external adjustment rules may be required to correct the model.

[0093] Compare the generated control stability coefficient with the preset control threshold to generate a prediction accuracy signal and an abnormal prediction signal;

[0094] After obtaining the control stability coefficient, the control stability coefficient is compared with the control threshold;

[0095] If the control stability coefficient is greater than or equal to the control threshold, it indicates that the prediction results of the model have high accuracy and stability, generate accurate prediction signals, the model prediction performance is good, the model prediction results are highly compatible with the predefined rules, and the smart home system can more reliably execute the automation rules, thereby ensuring the comfort, safety and energy efficiency of the home environment;

[0096] If the control stability coefficient is less than the control threshold, this may indicate that there is a certain degree of uncertainty or volatility between the model's prediction results and the predefined matching rules, generating an abnormal prediction signal. In this case, it may be necessary to further adjust or modify the model or change the predefined rules;

[0097] One possible approach is to increase the amount of data or improve the model algorithm to improve the model's prediction accuracy. This may involve collecting more training data, optimizing feature selection, adjusting model parameters, and other methods to enable the model to better capture patterns and trends in the data. You can also consider introducing more external information or context to enhance the model's predictive ability. For example, combining weather data, user behavior patterns, or family activity plans can improve the model's adaptability to environmental changes, thereby improving prediction accuracy.

[0098] According to the generated accurate prediction signal or abnormal prediction signal, corresponding signal processing and feedback control can be performed. If an accurate prediction signal is generated, the predefined rule control strategy of the corresponding smart home can be executed according to the prediction result;

[0099] If an abnormal prediction signal is generated, it indicates that there is a certain degree of uncertainty or volatility in the prediction results of the model. Especially in the data processing process of smart home devices involving cloud-edge collaboration, it may be necessary to adjust the data processing rules, that is, some data processing and model prediction can be implemented on the edge device to reduce dependence on the cloud. At the same time, a backup system can be established in the cloud. When an abnormal situation occurs on the edge device, it can quickly switch to the cloud for processing to ensure the continuity and stability of the system, thereby reducing the computing load on the cloud.

[0100] For example, a temperature monitoring and control device is involved in a smart home system, which is responsible for monitoring the indoor temperature and controlling the on / off status of the air conditioning system according to predefined rules. Normally, temperature data is collected by the device and sent to the cloud for processing and analysis to determine whether the air conditioning status needs to be adjusted. In some cases, due to network problems or device failures, the edge device may not be able to upload data to the cloud in time, resulting in data processing delays or abnormal situations. At this time, the edge device can process data to deal with abnormal situations;

[0101] Temperature monitoring equipment has certain processing capabilities and can pre-process and analyze the collected temperature data locally. For example, the device can monitor temperature data in real time and determine whether the current indoor temperature exceeds the set range based on predefined rules. If the edge device detects an abnormal indoor temperature, for example, the temperature exceeds the predefined range, an abnormal prediction signal will be generated, indicating that there may be a problem with the system or that corresponding measures need to be taken;

[0102] When an abnormal situation occurs on the edge, the system can automatically switch to the cloud for processing. The cloud backup system can receive the abnormal prediction signal sent by the edge device and perform further data processing and analysis. For example, the cloud system can re-evaluate the current temperature, formulate a strategy for adjusting the air conditioning status, and send instructions to the edge device to perform corresponding operations.

[0103] In this way, the predefined rules are controlled and adjusted, so that the smart home system can process part of the data on the edge device side, reducing dependence on the cloud. At the same time, when abnormal situations occur, it can quickly switch to the cloud for processing, ensuring the continuity and stability of the system, thereby reducing the computing load on the cloud.

[0104] To sum up, when an abnormal prediction signal is generated, control measures need to be taken in smart home devices involving cloud-edge collaboration to deal with abnormal situations and ensure the stability and reliability of the system. These measures include local processing and backup, anomaly detection and correction, dynamic adjustment strategies, and user prompts and feedback, which can help the system effectively deal with abnormal situations and minimize the impact of anomalies on system performance, thereby improving the efficiency of smart home use.

[0105] It should be noted that the relevant threshold information in this embodiment is pre-set by professionals and will not be explained in detail here. For example, the control of the stability threshold is determined based on historical data analysis, expected model performance and business needs. Some parameters in the embodiments have the same English letters, but have different meanings when used, which will not be explained one by one here.

[0106] The present invention first determines the performance indicators and optimal task types of smart home devices, allocates tasks generated by the smart home system according to the determined task types and performance indicators, combines rule engines and machine learning to determine the optimal execution rules under the current network and device states as predefined rules when allocating tasks in the smart home, controls the execution rule strategy of the smart home, trains the model, analyzes the matching situation between the trained model prediction results and the predefined rules, obtains the adaptation impact information generated in the rule matching process, determines whether the prediction results of the model are adapted to the predefined rules, and generates different signals for management and control of the smart home according to the adaptation situation of the prediction state of the model, thereby reducing the burden on the cloud and improving the data processing efficiency of the smart home system.

[0107] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The predefined parameters in the formula are set by technicians in this field according to actual conditions.

[0108] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0109] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0110] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0111] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0112] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A smart home system, characterized in that: It includes a task allocation module, a rule pre-definition module, an adaptation analysis module and an intelligent control module, and each module is connected through signals; The task allocation module is used to evaluate the processing capacity, storage space and network conditions of the cloud and each edge device, determine the performance indicators of the smart home device, determine the best task type for the smart home device through system requirements and capability evaluation, and allocate the tasks generated by the smart home system according to the determined task type and performance indicators; The rule pre-definition module is used to combine the rule engine and machine learning to determine the optimal execution rules under the current network and device status when allocating tasks for smart homes. It also controls the execution rule strategy of smart homes as pre-defined rules and uses historical data to train the model. The adaptation analysis module is used to analyze the matching between the prediction results of the trained model and the predefined rules, obtain the adaptation impact information generated during the rule matching process, and determine whether the prediction results of the model are adapted to the predefined rules; The intelligent control module is used to analyze the adaptation of the model's prediction status to the predefined rules, generate analysis results, and execute corresponding rules according to the analysis results; Analyze the matching between the trained model prediction results and the predefined rules to obtain the adaptation impact information generated during the rule matching process, including the following steps: The adaptation impact information includes predicted fluctuation information and periodic change information. The predicted fluctuation information includes the cloud edge convergence stability index, and the periodic change information includes the change trend adaptation index. The cloud edge convergence stability index in the forecast fluctuation information and the change trend adaptation index in the periodic change information are combined to generate the control stability coefficient; The cloud edge convergence stability index is proportional to the control stability coefficient, and the change trend adaptation index is proportional to the control stability coefficient; The cloud edge convergence stability index is obtained as follows: The predicted data of the model in unit time is obtained as samples, the number of samples n, the model predicted value and the actual measured value are obtained, and the measurement error value is calculated. The calculation expression is: , where , Represent the actual measured value and model predicted value of the i-th sample, and obtain the actual measured mean The model predicts the mean , calculate the convergence determination value: , obtain the number of samples m required for the model output to reach a stable state from the initial fluctuation, and calculate the predicted stable value. The calculation expression is: , calculate the cloud edge convergence stability index, the calculation expression is: ; The change trend adaptation index is obtained as follows: Get time series data, use linear regression to extract the change trend, determine the number of observation points N of the time series, define the time variable t, and get the trend value extracted by the change trend. The trend value calculation expression is: , where is the intercept, is the slope, use the model to predict future data points to get the predicted value, calculate the trend error value, and the calculation expression is: , where is the predicted value of the model at time t, and the trend residual value is calculated. The calculation expression is: , get the trend average , calculate the change trend adaptation index, the calculation expression is: .

2. A smart home system according to claim 1, characterized in that: When allocating tasks for smart homes, the rule engine and machine learning are combined to determine the optimal execution rules under the current network and device status, control the execution rule strategy of the smart home, and use historical data to train the model. The specific steps include: Collect requirements and define key tasks in smart home systems, including data collection, processing, storage and execution of control instructions, and determine smart home system requirements; Evaluate the processing power, storage space, and network conditions of the cloud and each edge device, including CPU speed, memory size, and storage space; check the network speed and stability of the device connection; evaluate the energy efficiency and power requirements of the cloud and each edge device; and determine the capacity evaluation results of the cloud and each edge device; Based on the system requirements and capability evaluation results, the optimal task type for each smart home device is determined, and the tasks generated by the smart home system are allocated according to the determined task type and performance indicators.

3. A smart home system according to claim 2, characterized in that: When allocating tasks for smart homes, the rule engine and machine learning are combined to determine the optimal execution rules under the current network and device status, control the execution rule strategy of the smart home, and use historical data to train the model. The specific steps include: Identify the characteristics of various tasks that need to be performed, including data processing, user request response, and video stream processing, and divide the tasks into routine tasks and complex tasks. Routine tasks are processed using rule engines, and complex tasks are predicted using machine learning models. Collect historical data from the smart home system, including task type, execution location, network status, device load, and pre-process the historical data; Use chi-square test, decision tree or recursive feature elimination to determine the information content characteristics of historical data and use them as the main features. Use long short-term memory neural network as the model and use the preprocessed historical data as training samples to train the model. The main features and preprocessed historical data are input into the model as training data. The training data includes a training set and a test set. The training set is used to train the model, and the test set is used to evaluate the generalization ability of the model. After the model is trained, the prediction results of the smart home execution rules are obtained.

4. A smart home system according to claim 3, characterized in that: The cloud edge convergence stability index in the prediction fluctuation information and the change trend adaptation index in the periodic change information are combined to generate the control stability coefficient, including the following steps: The cloud edge convergence stability index YBS and the change trend adaptation index BHQ are obtained and normalized to generate the control stability coefficient, which is calibrated as , the expression is: , where To control the stability factor, , is the predefined proportional coefficient of the cloud edge convergence stability index YBS and the change trend adaptation index BHQ, and , Both are greater than 0.

5. A smart home system according to claim 4, characterized in that: It is used to analyze the adaptation of the model's prediction status to the predefined rules, generate analysis results, and execute corresponding rules according to the analysis results, including the following steps: The control stability coefficient obtained by analyzing the adaptation of the model's predicted state to the predefined rules is compared with the pre-set control threshold; If the control stability coefficient is greater than or equal to the control threshold, a prediction accuracy signal is generated and the predefined rules are executed; If the control stability coefficient is less than the control threshold, an abnormal prediction signal is generated and the control adjustment of the smart home rules is performed.

Images