Method and device for monitoring abnormal state of sweeper

Abstract

The invention provides a method and device for monitoring the abnormal state of a sweeper, and relates to the technical field of smart home. The method comprises the steps of collecting working parameters in real time; preprocessing the working parameters to obtain preprocessed data; constructing an anomaly detection model, optimizing hyper-parameters by adopting a snake-egret algorithm, inputting pre-processed data, and outputting a working state score; setting an abnormal state scoring threshold value and dynamically adjusting the abnormal state scoring threshold value; judging whether the sweeper is abnormal or not; and analyzing the abnormity, sending an alarm and taking control measures. According to the invention, by constructing the anomaly detection model and combining with the optimization algorithm, the precise judgment of the complex working state is ensured; threshold value adjustment enables anomaly detection to be more flexible and accurate, different working scenes and changes can be adapted, problems can be rapidly positioned and solved, valuable reference is provided for product optimization and improvement, meanwhile, losses and risks caused by anomaly are reduced to the maximum extent, safe and reliable operation of the sweeper is guaranteed, and the method is suitable for large-scale popularization and application. And the service life is prolonged.

Description

Technical Field

[0001] The present invention relates to the technical field of smart home, and particularly to a method and device for monitoring the abnormal state of a floor sweeper. Background Art

[0002] A floor sweeper refers to a floor cleaning robot, usually a wireless machine, mainly in a disc shape, generally operating with a rechargeable battery, and the operation method is through a remote control or an operation panel on the machine. It can set a time reservation for cleaning and charge itself; there are sensors installed in the front to detect obstacles. When it encounters a wall or other obstacles, it will turn automatically and follow different routes according to the settings of different manufacturers, and has a planned cleaning area; with the continuous development of smart home technology, floor cleaning robots have gradually become an important tool for household cleaning.

[0003] During the operation of a floor cleaning robot, various abnormal states may occur, such as collisions, freezes, insufficient battery power, poor cleaning effect, etc. The current monitoring methods are often not comprehensive and accurate enough to detect and handle these abnormalities in a timely and effective manner, which affects the user experience and the service life of the robot. Summary of the Invention

[0004] The present invention provides a method and device for monitoring the abnormal state of a floor sweeper to solve the defects in the prior art.

[0005] In a first aspect, the present invention provides a method for monitoring the abnormal state of a floor sweeper, including: S1. Real-time collect the working parameters of the floor sweeper.

[0006] S2. Preprocess the working parameters to obtain preprocessed data.

[0007] S3. Construct an anomaly detection model based on a convolutional neural network combined with a recurrent neural network, and optimize the hyperparameters of the anomaly detection model using the snake egret algorithm. Input the preprocessed data and output the working state score of the floor sweeper.

[0008] S4. Set the anomaly state score threshold when the floor sweeper is working, and construct an adaptive threshold adjustment method based on historical data combined with real-time data distribution to dynamically adjust the anomaly state score threshold.

[0009] S5. Compare the working state score with the anomaly state score threshold. If the working state score exceeds the anomaly state score threshold, it indicates that the working state of the floor sweeper is abnormal.

[0010] S6. Conduct an abnormal classification analysis on the floor sweeper according to the working state score to obtain abnormal data, where the abnormal data includes the location, cause, and degree of abnormality.

[0011] S7. Send out abnormal alerts in real time and take corresponding control measures based on the abnormal data.

[0012] According to a method for monitoring the abnormal state of a sweeping robot provided by the present invention, in step S1, the working parameters include the motion state data, position data, environmental data, motor parameter data, battery state data, and cleaning system parameters of the sweeping robot.

[0013] According to a method for monitoring the abnormal state of a sweeping robot provided by the present invention, in step S2, the process of preprocessing the working parameters includes: checking and clearing duplicate records in the working parameters. Filling in missing values in the working parameters using the interpolation method. Detecting and processing abnormal values in the working parameters using the Z-score method. Performing standardization processing on the working parameters to obtain standardized data as the preprocessed data.

[0014] According to a method for monitoring the abnormal state of a sweeping robot provided by the present invention, in step S3, the process of constructing an abnormal detection model based on a convolutional neural network combined with a recurrent neural network includes: S31. Collect the historical working parameters of the sweeping robot and preprocess the historical working parameters to obtain historical preprocessed data.

[0015] S32. Divide the historical preprocessed data into a training set, a validation set, and a test set according to a preset ratio.

[0016] S33. Design the network architecture, including a convolutional neural network part and a recurrent neural network part. The convolutional neural network part includes a data input layer, a convolutional layer, and a pooling layer. The data input layer is used to receive the historical preprocessed data. The convolutional layer performs convolutional operations through convolutional kernels to extract the spatial features of the historical preprocessed data. The pooling layer is used to reduce the feature dimension of the spatial features and retain the key features. The recurrent neural network part includes a feature input layer, long short-term memory units, and a fully connected layer. The feature input layer is used to receive the key features and convert the key features into corresponding sequence data. The long short-term memory units are used to process the sequence data to obtain the current hidden state. The fully connected layer is used to map the current hidden state to the abnormal detection scoring task.

[0017] S34. Use the Sigmoid function to output the working state score.

[0018] According to a method for monitoring the abnormal state of a sweeping robot provided by the present invention, in step S3, the process of optimizing the hyperparameters of the abnormal detection model using the snake heron algorithm includes: Set the hyperparameter combination, and the hyperparameter combination includes the learning rate, the number of convolutional layers, the convolutional kernel size, the number of long short-term memory units, and the regularization parameter.

[0019] Randomly generate a group of secretarybird individuals, where each secretarybird individual represents a set of hyperparameter combinations, and the position of each secretarybird individual represents the value of the hyperparameters.

[0020] Input each set of hyperparameter combinations into the anomaly detection model, and when calculating each set of hyperparameter combinations on the validation set, calculate the accuracy of the working state score output by the model, and use the accuracy as the fitness value.

[0021] Identify the individual with the highest fitness value from all secretarybird individuals as the optimal individual, and record the position and fitness value of the optimal individual.

[0022] For each secretarybird individual, calculate the distance between it and the optimal individual.

[0023] Set the position boundaries, and update the position of each secretarybird individual within the position boundaries according to the hunting behavior.

[0024] Use the secretarybird individuals with updated positions to train the anomaly detection model, and evaluate the new fitness values of the secretarybird individuals with updated positions on the validation set.

[0025] Loop through the calculation of the fitness values of the secretarybird individuals, the update of the positions, and the evaluation of the new fitness values until the preset number of iterations is met.

[0026] Select the secretarybird individual with the highest fitness value and its corresponding position as the hyperparameter combination of the anomaly detection model.

[0027] According to a method for monitoring the abnormal state of a sweeping robot provided by the present invention, in step S4, the process of constructing an adaptive threshold adjustment method based on historical data combined with real-time data distribution includes: Collect historical data on the working state scores of the sweeping robot, and set an initial abnormal state score threshold by analyzing the score distributions in the normal state and abnormal state in the historical data.

[0028] Real-time monitor the working state of the sweeping robot, and obtain the real-time working state score of the anomaly detection model according to the working parameters of the sweeping robot.

[0029] Perform statistical analysis on the real-time working state scores at preset time intervals, and calculate the mean and standard deviation of the real-time working state scores.

[0030] Compare the mean and standard deviation with the scores in the normal state in the historical data to evaluate the current change in the working state.

[0031] Formulate a threshold adjustment strategy according to the current change in the working state.

[0032] According to a method for monitoring the abnormal state of a sweeping robot provided by the present invention, the process of evaluating the current change in the working state includes: If the mean value exceeds the preset value compared to the scores in the normal state of historical data and the standard deviation remains unchanged, it indicates that the current working state has improved.

[0033] If the mean value is lower than the preset value compared to the scores in the normal state of historical data and the standard deviation remains unchanged, it indicates that the current working state has deteriorated.

[0034] According to a method for monitoring the abnormal state of a floor sweeper provided by the present invention, the process of formulating an adjustment strategy includes: When the current working state improves, lower the abnormal state score threshold. When the current working state deteriorates, raise the abnormal state score threshold.

[0035] Introduce a smoothing factor and set the upper and lower limits of the abnormal state score threshold.

[0036] According to a method for monitoring the abnormal state of a floor sweeper provided by the present invention, in step S7, the corresponding control measures taken include automatic shutdown and enabling the safety mode. Enabling the safety mode includes decelerated operation and function limitation.

[0037] On the other hand, the present invention also provides a device for monitoring the abnormal state of a floor sweeper, including: A data acquisition module for real-time acquisition of the working parameters of the floor sweeper.

[0038] A data preprocessing module for cleaning and standardizing the working parameters to obtain preprocessed data.

[0039] An anomaly detection module for constructing an anomaly detection model based on a convolutional neural network combined with a recurrent neural network, and optimizing the hyperparameters of the anomaly detection model using the snake heron algorithm, inputting the preprocessed data, and outputting the working state score of the floor sweeper.

[0040] An adaptive threshold adjustment module for setting the abnormal state score threshold according to the distribution of historical data and real-time data, and implementing an adaptive adjustment mechanism.

[0041] A state comparison module for comparing the working state score with the abnormal state score threshold. If the working state score exceeds the abnormal state score threshold, it indicates that the working state of the floor sweeper is abnormal.

[0042] An anomaly classification and analysis module for performing anomaly classification and analysis on the floor sweeper according to the working state score to obtain anomaly data.

[0043] An anomaly alarm and control module for real-time sending of anomaly alarms and taking corresponding control measures according to the anomaly data.

[0044] A method and device for monitoring the abnormal state of a sweeping robot provided by the present invention lay a solid data foundation for accurately evaluating the working state of the sweeping robot by collecting and preprocessing working parameters in real time. The advanced abnormal detection model combined with the optimization algorithm ensures accurate judgment of complex working states. The adaptive threshold adjustment method makes the abnormal detection more flexible and accurate, adapting to different working scenarios and changes. The abnormal classification analysis and obtaining detailed abnormal data not only help quickly locate and solve problems, but also provide valuable references for the optimization and improvement of products. Sending alerts and taking control measures in real time minimizes the losses and risks brought by abnormalities, ensures the safe and reliable operation of the sweeping robot, and extends its service life. For users, they can timely understand the working conditions of the sweeping robot, improving the convenience and satisfaction of use. For manufacturers, it helps improve product quality and brand image and reduce after-sales service costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0046] Figure 1 It is a schematic flowchart of a method for monitoring the abnormal state of a sweeping robot provided by an embodiment of the present invention; Figure 2 It is a schematic flowchart of constructing an adaptive threshold adjustment method in an embodiment of the present invention; Figure 3 It is a schematic structural diagram of a device for monitoring the abnormal state of a sweeping robot provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] To make the objectives, technical solutions, and advantages of the present invention clearer, the following will clearly and completely describe the technical solutions in the present invention in conjunction with the drawings in the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

[0048] The following will describe Figures 1 - 3 a method and device for monitoring the abnormal state of a sweeping robot of the present invention.

[0049] Figure 1 It is a schematic flowchart of a method for monitoring the abnormal state of a sweeping robot provided by an embodiment of the present invention; Figure 2It is a schematic flowchart of constructing an adaptive threshold adjustment method in an embodiment of the present invention.

[0050] As Figures 1 - 2 shown, a method and device for monitoring the abnormal state of a sweeping robot provided by an embodiment of the present invention, the execution subject may be a method for monitoring the abnormal state of a sweeping robot, and the method includes: S1. Collect the working parameters of the sweeping robot in real time.

[0051] The working parameters include the motion state data, position data, environmental data, motor parameter data, battery state data, and cleaning system parameters of the sweeping robot.

[0052] In this embodiment, in order to collect the motion state data of the sweeping robot, a series of sensors need to be installed inside the sweeping robot, including an acceleration sensor, a gyroscope, etc. These sensors can sense information such as the speed, acceleration, and steering angle of the sweeping robot in real time. For the collection of position data, the global positioning system (GPS) or indoor positioning technology can be used, including Bluetooth positioning, ultra-wideband positioning, etc. GPS is suitable for outdoor environments and can obtain the longitude and latitude coordinates of the sweeping robot more accurately. In an indoor environment, by deploying Bluetooth beacons or ultra-wideband base stations, the sweeping robot can receive corresponding signals to determine its position indoors. The collection of environmental data relies on various environmental sensors, including dust sensors, humidity sensors, temperature sensors, etc. The dust sensor can detect the dust concentration in the air to judge the cleaning effect. The humidity and temperature sensors can sense the temperature and humidity conditions of the surrounding environment and provide a reference for the working strategy of the sweeping robot. The collection of motor parameter data is achieved by installing current sensors, voltage sensors, and speed sensors on the motor. The current and voltage sensors can monitor the working current and voltage of the motor to understand the energy consumption and load conditions of the motor. The speed sensor can feedback the rotation speed of the motor to control the traveling speed and cleaning intensity of the sweeping robot. The collection of battery state data is completed by the battery management system. This system can monitor parameters such as the voltage, current, remaining power, and charging state of the battery in real time to ensure that the sweeping robot will not suddenly stop working due to insufficient power during operation. The collection of cleaning system parameters includes the roller brush speed, suction power, etc. By installing sensors on the corresponding components, these parameters can be obtained to evaluate the working efficiency and effect of the cleaning system. By comprehensively and timely obtaining various key data of the sweeping robot during operation, it can provide a rich information basis for subsequent analysis and processing, and help to more accurately understand the real-time working state of the sweeping robot.

[0053] S2. Preprocess the working parameters to obtain preprocessed data.

[0054] The process of preprocessing the working parameters includes: checking and clearing duplicate records in the working parameters. Filling in missing values in the working parameters using interpolation. Detecting and processing outliers in the working parameters using Z-score. Standardizing the working parameters to obtain standardized data as the preprocessed data.

[0055] In this embodiment, first, check for duplicate records in the working parameters. Through a data comparison algorithm, each collected record is compared one by one to identify exactly the same records and clear them to ensure data uniqueness. For the handling of missing values, interpolation is used. Based on the adjacent valid data points before and after, possible values are inferred through mathematical calculations to fill in the missing part. For example, for linear missing values, linear interpolation can be used. For non-linear cases, more complex interpolation methods may be used. The Z-score method is used for outlier detection. Calculate the deviation degree of each data point from the mean of the data set, and data points exceeding a certain threshold are regarded as outliers. For the detected outliers, they are corrected or deleted according to the specific situation. Finally, standardization is performed. Map the values of the working parameters to a standard range, usually a distribution with a mean of 0 and a standard deviation of 1. By subtracting the mean and then dividing by the standard deviation, data of different magnitudes and units are made comparable, and the standardized data is obtained as the preprocessed data. The entire preprocessing process is implemented by writing corresponding program code. By clearing duplicate records, filling in missing values, handling outliers, and standardizing, the quality and consistency of the data are improved, the interference of data noise and deviation on subsequent analysis is reduced, and the accuracy and stability of the subsequent model are guaranteed.

[0056] S3. Construct an anomaly detection model based on a convolutional neural network combined with a recurrent neural network, and optimize the hyperparameters of the anomaly detection model using the snake heron algorithm. Input the preprocessed data and output the working state score of the sweeper.

[0057] The process of constructing an anomaly detection model based on a convolutional neural network combined with a recurrent neural network includes: S31. Collect the historical working parameters of the sweeper and preprocess the historical working parameters to obtain historical preprocessed data.

[0058] S32. Divide the historical preprocessed data into a training set, a validation set, and a test set according to a preset ratio.

[0059] S33. Design a network architecture, including a convolutional neural network part and a recurrent neural network part. The convolutional neural network part includes a data input layer, a convolutional layer, and a pooling layer. The data input layer is used to receive historical preprocessed data. The convolutional layer performs convolutional operations through convolutional kernels to extract the spatial features of the historical preprocessed data. The pooling layer is used to reduce the feature dimension of the spatial features and retain key features. The recurrent neural network part includes a feature input layer, long short-term memory units, and a fully connected layer. The feature input layer is used to receive the key features and convert the key features into corresponding sequence data. The long short-term memory units are used to process the sequence data to obtain the current hidden state. The fully connected layer is used to map the current hidden state to the anomaly detection scoring task.

[0060] S34. Use the Sigmoid function to output the working status score.

[0061] The process of optimizing the hyperparameters of the anomaly detection model using the secretary bird algorithm includes: Set a combination of hyperparameters, which includes a learning rate, the number of convolutional layers, the size of convolutional kernels, the number of long short-term memory units, and a regularization parameter.

[0062] Randomly generate a group of secretary bird individuals. Each secretary bird individual represents a combination of hyperparameters, and the position of each secretary bird individual represents the value of the hyperparameters.

[0063] Input each combination of hyperparameters into the anomaly detection model, and when calculating each combination of hyperparameters on the validation set, calculate the accuracy of the working status score output by the model, and use the accuracy as the fitness value.

[0064] Identify the individual with the highest fitness value from all the secretary bird individuals as the optimal individual, and record the position and fitness value of the optimal individual.

[0065] For each secretary bird individual, calculate the distance between it and the optimal individual.

[0066] Set the position boundaries and update the position of each secretary bird individual within the position boundaries according to the hunting behavior.

[0067] Use the secretary bird individuals with updated positions to train the anomaly detection model and evaluate the new fitness values of the secretary bird individuals with updated positions on the validation set.

[0068] Loop through the calculation of the fitness values of the secretary bird individuals, the position update, and the evaluation of the new fitness values until the preset number of iterations is met.

[0069] Select the secretary bird individual with the highest fitness value and its corresponding position as the combination of hyperparameters for the anomaly detection model.

[0070] In this embodiment, by combining the advantages of convolutional neural networks and recurrent neural networks, complex patterns and time-series features in the working parameters can be effectively captured, and the hyperparameters optimized by the snake egret algorithm further improve the performance and accuracy of the model, thereby more precisely evaluating the working state of the sweeping robot.

[0071] S4. Set the abnormal state score threshold when the sweeping robot is working, and construct an adaptive threshold adjustment method based on historical data combined with real-time data distribution to dynamically adjust the abnormal state score threshold.

[0072] Figure 2 It is a schematic flowchart of constructing an adaptive threshold adjustment method in an embodiment of the present invention.

[0073] As Figure 2 shown, the process of constructing an adaptive threshold adjustment method based on historical data combined with real-time data distribution includes: Collect historical data on the working state scores of the sweeping robot, and set an initial abnormal state score threshold by analyzing the score distributions in normal and abnormal states in the historical data.

[0074] Real-time monitor the working state of the sweeping robot, and obtain the real-time working state score of the anomaly detection model according to the working parameters of the sweeping robot.

[0075] Statistically analyze the real-time working state scores at preset time intervals, and calculate the mean and standard deviation of the real-time working state scores.

[0076] Compare the mean and standard deviation with the scores in the normal state in the historical data to evaluate the current change in the working state.

[0077] Formulate a threshold adjustment strategy according to the current change in the working state.

[0078] The process of evaluating the current change in the working state includes: If the mean exceeds the preset value compared with the scores in the normal state in the historical data and the standard deviation remains unchanged, it indicates that the current working state has improved.

[0079] If the mean is lower than the preset value compared with the scores in the normal state in the historical data and the standard deviation remains unchanged, it indicates that the current working state has deteriorated.

[0080] The process of formulating an adjustment strategy includes: When the current working state improves, lower the abnormal state score threshold. When the current working state deteriorates, raise the abnormal state score threshold.

[0081] Introduce a smoothing factor and set the upper and lower limits of the abnormal state score threshold.

[0082] In this embodiment, first, a large amount of historical data on the working state scores of the sweeping robot is collected. Using data analysis tools, the score distributions in the normal state and abnormal state in these historical data are deeply analyzed, and a relatively reasonable initial abnormal state score threshold is set based on this. During the working process of the sweeping robot, its working state is monitored in real time. Through the anomaly detection model, a real-time working state score is generated based on the working parameters obtained in real time. At preset time intervals, for example, every 5 minutes, statistical analysis is performed on the real-time working state scores during this period. Using mathematical calculation methods, the mean and standard deviation of the real-time working state scores are calculated. The calculated mean and standard deviation are compared with the scores in the normal state in the historical data to evaluate the change in the current working state. If the mean exceeds the preset value compared with the scores in the normal state in the historical data and the standard deviation remains unchanged, it means that the current working state has been optimized. On the contrary, if the mean is lower than the preset value and the standard deviation remains unchanged, it indicates that the current working state tends to deteriorate. Then a threshold adjustment strategy is formulated. When it is determined that the current working state has improved, the abnormal state score threshold is appropriately lowered to adapt to the better working conditions. When it is determined that the current working state has deteriorated, the abnormal state score threshold is increased to enhance the sensitivity to the abnormal state. In order to make the threshold adjustment smoother and more stable, a smoothing factor is introduced. For example, each time the threshold is adjusted, instead of directly changing it significantly, it is gradually adjusted according to the proportion of the smoothing factor. At the same time, upper and lower limits of the abnormal state score threshold are set to avoid excessive adjustment of the threshold and losing its rationality. Through the above continuous monitoring, analysis, and adjustment process, adaptive threshold adjustment based on historical data combined with real-time data distribution is achieved, enabling the abnormal state score threshold to more accurately reflect the actual working state of the sweeping robot. Adaptive adjustment based on historical data and real-time data distribution makes the threshold more in line with the actual working conditions, avoiding misjudgment or missed judgment that may be caused by a fixed threshold, and improving the accuracy and reliability of anomaly detection.

[0083] S5. Compare the working state score with the abnormal state score threshold. If the working state score exceeds the abnormal state score threshold, it indicates that the working state of the sweeping robot is abnormal.

[0084] In this embodiment, during the operation of the floor sweeper, the working status score calculated in the previous steps is obtained in real time. At the same time, the currently set abnormal status score threshold is obtained. Through the programmed code, the real-time working status score is directly compared numerically with the abnormal status score threshold. If the working status score is less than or equal to the abnormal status score threshold, it indicates that the working status of the floor sweeper is within the normal range, and the system will continue to monitor and record the working status score normally. However, once the working status score exceeds the abnormal status score threshold, the system will immediately determine that the working status of the floor sweeper is abnormal. To ensure the accuracy and timeliness of the comparison, the comparison process will continue at a high frequency, usually completed once per second or at shorter time intervals to achieve real-time monitoring and abnormal judgment of the working status of the floor sweeper.

[0085] S6. Perform abnormal classification and analysis on the floor sweeper according to the working status score, and obtain abnormal data, where the abnormal data includes the location, cause, and degree of abnormality.

[0086] In this embodiment, a database and rule system for abnormal classification need to be established. This system contains various reasons that may cause the floor sweeper to work abnormally, corresponding abnormal characteristics, and corresponding location association information. When it is determined that the working status of the floor sweeper is abnormal, that is, when the working status score exceeds the abnormal status score threshold, the system will start the abnormal classification analysis program. The program will extract detailed information related to the moment of abnormality from the stored working parameter data. This includes the location data of the floor sweeper when the abnormality occurs; motor parameter data, including abnormal changes in current, voltage, and speed; cleaning system parameters, including abnormal changes in the rotation speed of the roller brush, suction power, etc.; environmental data, including encountering overly high obstacles, abnormal floor materials, etc., and battery status data, including a sharp drop in battery power, charging failure, etc. Then, the extracted abnormal-related data is matched and compared with the rules in the abnormal classification database. For example, if the motor current suddenly increases and the speed decreases, it may be judged that the motor is overloaded. If the battery power is quickly depleted and the charging is abnormal, it may be a battery failure. According to the matching results, the cause of the abnormality is determined. At the same time, combined with the specific value of the working status score and the deviation degree of the abnormal parameters, the degree of abnormality is evaluated. For example, if the working status score only slightly exceeds the threshold, it may be a mild abnormality. While a large excess may be a serious abnormality. Finally, information such as the location where the abnormality occurs, the determined cause, and the evaluated degree of abnormality is integrated into abnormal data for storage and output, so as to perform subsequent fault handling, maintenance decision-making, or feedback detailed abnormal situations to the user.

[0087] S7. Send abnormal alarms in real time and take corresponding control measures according to the abnormal data.

[0088] Take corresponding control measures including automatic shutdown and enabling the safety mode. Enabling the safety mode includes reducing the running speed and restricting functions. Reducing the running speed means controlling the running speed of the sweeper within a preset range. Restricting functions means restricting the non-essential functions of the sweeper.

[0089] In this embodiment, when it is determined that the working state of the sweeper is abnormal, the system will immediately trigger an abnormal alarm sending mechanism. Through the network communication module, the alarm information containing abnormal data will be sent to the user in real time by means of text messages, application pushes, etc. At the same time, the system will automatically take corresponding control measures according to the specific situation of the abnormal data. If the degree of abnormality is relatively serious, the system will directly control the sweeper to automatically shut down to avoid further damage or danger. If the degree of abnormality is relatively light, the system will enable the safety mode. In the safety mode, first is to reduce the running speed. By controlling the power supply parameters of the motor, the running speed of the sweeper is reduced to a preset safe range. Second is to restrict functions. The system will restrict the non-essential functions of the sweeper. For example, some advanced cleaning modes or intelligent navigation functions will be temporarily turned off, and only the basic cleaning and moving functions will be retained to reduce the burden and possible risks of the system. Throughout the process, the system will continuously monitor the state of the sweeper. Once the abnormal situation is resolved or improved, the normal working mode and functions will be gradually restored to ensure that the sweeper continues to complete the cleaning task on the premise of safety and reliability.

[0090] To sum up, this embodiment provides a method for monitoring the abnormal state of a sweeper. By collecting and preprocessing working parameters in real time, a solid data foundation is laid for accurately evaluating the working state of the sweeper. The advanced abnormal detection model combined with the optimization algorithm ensures accurate judgment of complex working states. The adaptive threshold adjustment method makes the abnormal detection more flexible and accurate, adapting to different working scenarios and changes. Abnormal classification analysis and obtaining detailed abnormal data not only help to quickly locate and solve problems, but also provide valuable references for product optimization and improvement. Real-time sending of alarms and taking control measures minimize the losses and risks brought by abnormalities, ensure the safe and reliable operation of the sweeper, and extend its service life. For users, they can timely understand the working conditions of the sweeper, improving the convenience and satisfaction of use. For manufacturers, it helps to improve product quality and brand image and reduce after-sales service costs.

[0091] Based on the same general inventive concept, the present invention also protects a device for monitoring the abnormal state of a sweeper. The following describes a device for monitoring the abnormal state of a sweeper provided by the present invention. The device for monitoring the abnormal state of a sweeper described below can be mutually corresponding and referred to with the method for monitoring the abnormal state of a sweeper described above.

[0092] Figure 3It is a schematic structural diagram of a device for monitoring the abnormal state of a sweeping robot provided by an embodiment of the present invention.

[0093] As Figure 3 shown, a device for monitoring the abnormal state of a sweeping robot includes: A data acquisition module for real-time acquisition of the working parameters of the sweeping robot.

[0094] A data preprocessing module for cleaning and standardizing the working parameters to obtain preprocessed data.

[0095] An anomaly detection module for constructing an anomaly detection model based on a convolutional neural network combined with a recurrent neural network, and optimizing the hyperparameters of the anomaly detection model using the snake heron algorithm, inputting the preprocessed data, and outputting the working state score of the sweeping robot.

[0096] An adaptive threshold adjustment module for setting the anomaly state score threshold according to the distribution of historical data and real-time data, and implementing an adaptive adjustment mechanism.

[0097] A state comparison module for comparing the working state score with the anomaly state score threshold. If the working state score exceeds the anomaly state score threshold, it indicates that the working state of the sweeping robot is abnormal.

[0098] An anomaly classification and analysis module for performing anomaly classification and analysis on the sweeping robot according to the working state score to obtain anomaly data.

[0099] An anomaly alarm and control module for real-time sending of anomaly alarms and taking corresponding control measures according to the anomaly data.

[0100] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative labor.

[0101] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0102] Finally, 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 them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for monitoring abnormal state of a sweeping machine, characterized in that: include: S1. Real-time collection of sweeper working parameters; S2. Preprocessing the working parameters to obtain preprocessing data; S3, constructing an anomaly detection model based on a convolutional neural network combined with a recurrent neural network, and using the snake heron algorithm to optimize the hyperparameters of the anomaly detection model, inputting the preprocessed data, and outputting the working status score of the sweeper; S4. Setting an abnormal state scoring threshold when the sweeper is working, and constructing an adaptive threshold adjustment method based on historical data combined with real-time data distribution to dynamically adjust the abnormal state scoring threshold; S5. Compare the working state score with the abnormal state score threshold. If the working state score exceeds the abnormal state score threshold, it indicates that the working state of the sweeping machine is abnormal. S6. Perform abnormal classification analysis on the sweeping machine according to the working status score to obtain abnormal data, where the abnormal data includes the location, cause and degree of the abnormality; S7. Send abnormal alarms in real time and take corresponding control measures based on the abnormal data.

2. The method for monitoring abnormal state of a sweeping machine according to claim 1, characterized in that: In step S1, the working parameters include motion state data, position data, environmental data, motor parameter data, battery state data and cleaning system parameters of the sweeping machine.

3. The method for monitoring abnormal state of a sweeping machine according to claim 1, characterized in that: In step S2, the process of preprocessing the working parameters includes: checking and clearing duplicate records in the working parameters; filling missing values ​​in the working parameters by interpolation; detecting and processing abnormal values ​​in the working parameters by Z-score; and standardizing the working parameters to obtain standardized data as preprocessed data.

4. The method for monitoring abnormal state of a sweeping machine according to claim 1, characterized in that: In step S3, the process of constructing an anomaly detection model based on a convolutional neural network combined with a recurrent neural network includes: S31, collecting historical working parameters of the sweeping machine, and preprocessing the historical working parameters to obtain historical preprocessing data; S32, dividing the historical preprocessed data into a training set, a validation set and a test set according to a preset ratio; S33. Design a network architecture, including a convolutional neural network part and a recurrent neural network part; the convolutional neural network part includes a data input layer, a convolution layer and a pooling layer; the data input layer is used to receive the historical preprocessed data; the convolution layer performs a convolution operation through a convolution kernel to extract the spatial features of the historical preprocessed data; the pooling layer is used to reduce the feature dimension of the spatial features and retain key features; the recurrent neural network part includes a feature input layer, a long short-term memory unit and a fully connected layer; the feature input layer is used to receive the key features and convert the key features into corresponding sequence data; the long short-term memory unit is used to process the sequence data to obtain the current hidden state; the fully connected layer is used to map the current hidden state to the anomaly detection scoring task; S34. Use the Sigmoid function to output the working status score.

5. The method for monitoring abnormal state of a sweeping machine according to claim 1, characterized in that: In step S3, the process of optimizing the hyperparameters of the anomaly detection model using the snake heron algorithm includes: Setting a hyperparameter combination, wherein the hyperparameter combination includes a learning rate, a number of convolutional layers, a convolutional kernel size, a number of long short-term memory units, and a regularization parameter; Randomly generate a group of snake heron individuals, each of the snake heron individuals represents a group of hyperparameter combinations, and the position of each snake heron individual represents the value of the hyperparameter; Input each of the hyperparameter combinations into the anomaly detection model, and when calculating each of the hyperparameter combinations on a validation set, the model outputs the accuracy of the working status score, and uses the accuracy as the fitness value; Identify the individual with the highest fitness value from all the snake heron individuals as the best individual, and record the position and fitness value of the best individual; For each snake heron individual, calculate the distance between it and the optimal individual; Setting a location boundary, and updating the location of each snake heron individual within the location boundary according to the hunting behavior; The anomaly detection model is trained using the snake heron individuals whose positions are updated, and the new fitness values ​​of the snake heron individuals whose positions are updated are evaluated on the validation set; The fitness value calculation, position update and new fitness value evaluation of the snake heron individuals are executed cyclically until the preset number of iterations is met; The snake heron individual with the highest fitness value and its corresponding position are selected as the hyperparameter combination of the anomaly detection model.

6. The method for monitoring abnormal status of a sweeping machine according to claim 1, characterized in that: In step S4, the process of constructing an adaptive threshold adjustment method based on historical data combined with real-time data distribution includes: Collect historical data of the working status scores of the sweeper, and set an initial abnormal status score threshold by analyzing the score distribution of normal status and abnormal status in the historical data; Monitor the working status of the sweeper in real time, and obtain the real-time working status score of the anomaly detection model according to the working parameters of the sweeper; Performing statistical analysis on the real-time work status score within a preset time interval, and calculating the mean and standard deviation of the real-time work status score; Compare the mean and standard deviation with the normal status scores in historical data to evaluate the change in current working status; According to the change of the current working state, a threshold adjustment strategy is formulated.

7. The method for monitoring abnormal state of a sweeping machine according to claim 6, characterized in that: The process of evaluating the change of the current working status includes: If the mean value exceeds the preset value compared with the score of the normal state in the historical data, and the standard deviation remains unchanged, it means that the current working state has improved; If the mean is lower than a preset value compared to the score of the normal state in the historical data, and the standard deviation remains unchanged, it means that the current working state has deteriorated.

8. The method for monitoring abnormal state of a sweeping machine according to claim 6, characterized in that: The process of developing the adjustment strategy includes: When the current working state becomes better, the abnormal state scoring threshold is lowered; when the current working state becomes worse, the abnormal state scoring threshold is increased; A smoothing factor is introduced, and the upper and lower limits of the abnormal state scoring threshold are set.

9. The method for monitoring abnormal status of a sweeping machine according to claim 1, characterized in that: In step S7, taking the corresponding control measures includes automatic shutdown and enabling a safety mode; enabling the safety mode includes deceleration operation and function restriction.

10. A device for monitoring abnormal status of a sweeping machine, characterized in that: The device is used to execute a method for monitoring an abnormal state of a sweeping machine as claimed in any one of claims 1 to 9, comprising: Data acquisition module, used to collect the working parameters of the sweeper in real time; A data preprocessing module, used for cleaning and standardizing the working parameters to obtain preprocessed data; An anomaly detection module is used to build an anomaly detection model based on a convolutional neural network combined with a recurrent neural network, and use a snake heron algorithm to optimize the hyperparameters of the anomaly detection model, input the preprocessed data, and output a working status score of the sweeper; The adaptive threshold adjustment module is used to set the abnormal state scoring threshold according to the distribution of historical data and real-time data, and implement the adaptive adjustment mechanism; A state comparison module compares the working state score with the abnormal state score threshold. If the working state score exceeds the abnormal state score threshold, it indicates that the working state of the sweeping machine is abnormal. An abnormality classification and analysis module, used to perform abnormality classification and analysis on the sweeping machine according to the working status score and obtain abnormal data; The abnormal alarm and control module is used to send abnormal alarms in real time and take corresponding control measures according to the abnormal data.

Images