A method for intelligent control of a robot sweeper and related equipment

Abstract

The application belongs to the technical field of robot control, and discloses a robot intelligent control method and related equipment, which realizes home safety monitoring of an old person living alone through a non-vision sensor carried by a robot, and enables the old person to intelligently adjust a security strategy according to an activity space, has the advantages of giving consideration to safety monitoring and privacy protection, fully utilizing existing technical resources, reducing development cost, and providing a more friendly and natural intelligent security solution for the old person living alone.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of robot control, in particular to a robot intelligent control method and related equipment. BACKGROUND

[0002] Existing home security systems, such as intelligent camera monitoring systems, play an important role in ensuring home safety. However, for the special group of elderly people living alone, the traditional camera monitoring method has a significant risk of privacy leakage, which can easily cause resistance and discomfort of the elderly. In order to balance security monitoring and privacy protection, non-vision sensor technology has emerged. For example, infrared sensors can sense human heat sources, ultrasonic and laser radar sensors can detect object distance and spatial structure, and pressure sensors can sense object weight and position changes. These non-vision sensors have natural advantages in protecting user privacy.

[0003] On the other hand, as a popular smart home device, the robot vacuum cleaner already has mature navigation and obstacle avoidance functions based on non-vision sensors. By integrating infrared, ultrasonic or laser radar sensors, the robot vacuum cleaner can move autonomously and perceive environmental information in the home environment. Combining non-vision sensor technology with the robot vacuum cleaner platform and applying it to the home security monitoring of elderly people living alone can not only make full use of existing technology resources and reduce development costs, but also provide a more friendly and natural intelligent security solution for elderly people living alone. Considering the particularity of the home environment of elderly people living alone, such as the fact that the elderly mainly move in specific spaces such as bedrooms and living rooms, the robot vacuum cleaner needs to be able to learn and identify these main activity spaces and optimize energy use efficiency while ensuring safety and reducing potential interference with the daily life of the elderly.

[0004] Therefore, how to effectively use the non-vision sensors carried by the robot vacuum cleaner to realize home security monitoring for elderly people living alone and enable it to intelligently adjust the security strategy according to the activity space of the elderly has become a problem to be solved. Especially in terms of how to enable the robot vacuum cleaner to learn and identify the daily activity space of the elderly and optimize its security patrol strategy according to this information, so as to ensure safety while reducing interference with the daily life of the elderly and improving energy utilization efficiency, there are still many challenges.

[0005] In view of the above problems, the prior art needs to be improved. SUMMARY

[0006] The present application aims to provide a robot intelligent control method and related equipment, which has the advantages of balancing security monitoring and privacy protection, making full use of existing technology resources, and reducing development costs, and provides a more friendly and natural intelligent security solution for elderly people living alone.

[0007] In a first aspect, the application provides a method for intelligent control of a robot vacuum cleaner, for home safety monitoring of a single elderly person by the robot vacuum cleaner, the method comprising the steps of:

[0008] A1. controlling the robot vacuum cleaner to autonomously patrol in the home environment using non-vision sensors, record the activity frequency and residence time of the elderly person in different spatial regions, and add historical activity data;

[0009] A2. statistically analyzing the historical activity data, calculating the average residence time and average activity frequency of each spatial region, and identifying the main spatial regions of the elderly person's daily activities;

[0010] A3. For the identified main spatial regions, continuously monitor the activity state of the elderly person in each main spatial region, and obtain baseline data of the elderly person's daily activity pattern; the baseline data includes average residence time, average activity frequency, residence time variance and activity frequency variance;

[0011] A4. Control the robot vacuum cleaner to collect activity data of the elderly person in the main spatial region in real time using non-vision sensors, and compare it with the baseline data of the daily activity pattern to determine whether an abnormal state occurs;

[0012] A5. When an abnormal state occurs, interact with the elderly person through voice and confirm, and in the case of no response or confirmation in an emergency, send an alarm message to the pre-set emergency contact person;

[0013] A6. Adjust the security patrol strategy of the robot vacuum cleaner according to the identified main spatial regions; the security patrol strategy includes the patrol mode, patrol frequency and abnormal detection sensitivity in each spatial region.

[0014] Through the non-vision sensors carried by the robot vacuum cleaner, home safety monitoring of the single elderly person is realized, and the security strategy can be intelligently adjusted according to the activity space of the elderly person, which has the advantages of balancing safety monitoring and privacy protection, fully utilizing existing technical resources, reducing development cost, and providing a more friendly and natural intelligent security solution for single elderly people.

[0015] Preferably, step A1 comprises:

[0016] A101. Retrieve the security patrol strategy of the robot vacuum cleaner, and extract the patrol mode of the robot vacuum cleaner in the home environment from it; the patrol mode includes random walk mode, spiral cleaning mode and preset path mode; wherein in the random walk mode, the robot vacuum cleaner moves randomly in the spatial region at a preset speed, and adjusts the moving direction when encountering obstacles; in the spiral cleaning mode, the robot vacuum cleaner cleans the spatial region in a spiral trajectory, and the spiral radius gradually increases; in the preset path mode, the robot vacuum cleaner patrols according to the pre-set path;

[0017] A102. According to the extracted patrol mode, the robot is controlled to autonomously patrol in the home environment using non-vision sensors;

[0018] A103. During the patrol process, the non-vision sensors are used to detect whether there is an old person activity in the current space area in real time, and when an old person activity is detected, the space area position information, activity frequency and stay time of the current space area are recorded;

[0019] A104. The activity frequency, stay time and space area position information recorded in this patrol are added to the historical activity data, and the weight of early data in the historical activity data is reduced according to the time decay factor.

[0020] Through these steps, the scheme can make the patrol of the robot more intelligent and the data recording more accurate, thereby improving the effect of home safety monitoring. By using the time decay factor, the influence of early data on the current analysis result is reduced, so that the system can adapt to the change of the old person activity mode more quickly, and the sensitivity and accuracy of the monitoring are improved.

[0021] Preferably, step A104 comprises:

[0022] B1. The activity frequency, stay time and space area position information recorded in this patrol are stored in the newly added activity data buffer area according to the preset data structure, the activity data buffer area adopts a ring queue structure, and the activity data of the last N patrols is stored, N being a preset positive integer;

[0023] B2. The historical activity data is extracted from the activity data buffer area, and the weighted activity frequency and weighted stay time of each space area position information are calculated according to the following formula: weighted activity frequency = original activity frequency * (1-time decay factor) ^ time interval, weighted stay time = original stay time * (1-time decay factor) ^ time interval, time interval being the difference between the current time and the data recording time;

[0024] B3. The weighted activity frequency and weighted stay time are used to replace the activity frequency and stay time of the corresponding space area position information in the original historical activity data, so as to reduce the weight of early data.

[0025] This method can effectively reduce the influence of early data on the current activity mode recognition, improve the accuracy of the old person activity state evaluation, and solve the problem that using a fixed time decay factor cannot accurately reflect the change of the old person activity mode with time and season.

[0026] Preferably, step A2 comprises:

[0027] A201. Smoothing the historical activity data using Kalman filtering algorithm to eliminate the noise data introduced by sensor error, and obtaining smoothed activity data;

[0028] A202. Setting activity frequency threshold and stay duration threshold, and counting the number of times that the activity frequency of each spatial region exceeds the activity frequency threshold and the number of times that the stay duration of each spatial region exceeds the stay duration threshold in the smoothed activity data;

[0029] A203. Calculating the weighted sum of the number of times that the activity frequency of each spatial region exceeds the activity frequency threshold and the number of times that the stay duration of each spatial region exceeds the stay duration threshold, and determining the main spatial region of the old person's daily activities according to the weighted sum result.

[0030] Preferably, step A202 comprises:

[0031] According to the smoothed activity data, calculating the mean and variance of the activity frequency of each spatial region, and the mean and variance of the stay duration of each spatial region;

[0032] Based on the mean and variance of the activity frequency of each spatial region, an adaptive algorithm is used to calculate the activity frequency threshold, and based on the mean and variance of the stay duration of each spatial region, an adaptive algorithm is used to calculate the stay duration threshold; wherein the activity frequency threshold is positively correlated with the mean of the activity frequency and the variance of the activity frequency, and the stay duration threshold is positively correlated with the mean of the stay duration and the variance of the stay duration;

[0033] Counting the number of times that the activity frequency of each spatial region exceeds the activity frequency threshold and the number of times that the stay duration of each spatial region exceeds the stay duration threshold in the smoothed activity data.

[0034] Preferably, step A3 comprises:

[0035] A301. Constructing an activity pattern clustering model, the model input includes the average stay duration, the average activity frequency, the stay duration variance and the activity frequency variance in a preset time window, and the output is the activity pattern category;

[0036] A302. Based on the activity pattern clustering model, clustering analysis is performed on the historical activity data to obtain multiple activity pattern categories, and the center vector of each activity pattern category is determined, the center vector includes the average stay duration, the average activity frequency, the stay duration variance and the activity frequency variance;

[0037] A303. For the identified main spatial region, continuously monitoring the activity state of the old person in each main spatial region, obtaining the feature vector of the current activity pattern of the old person, the feature vector includes the average stay duration, the average activity frequency, the stay duration variance and the activity frequency variance;

[0038] A304. Calculate the distance between the feature vector of the current activity mode and the center vector of each activity mode category, and take the activity mode category corresponding to the center vector with the smallest distance as the current daily activity mode of the old person, and take the center vector as the baseline data of the daily activity mode of the old person.

[0039] Preferably, step A301 comprises:

[0040] Obtain historical activity data within a preset time window, extract the health status information of the old person corresponding to the historical activity data, and the health status information includes the duration of chronic disease and the medication situation;

[0041] Quantitatively process the health status information to obtain a health status score, and take the health status score as one of the input features of the activity mode clustering model, and construct a multi-dimensional input vector together with other activity features;

[0042] An activity mode clustering model is constructed by using an improved K-means clustering algorithm to cluster analyze the multi-dimensional input vector, and the activity mode category is determined according to the clustering result; in the iteration process of the improved K-means clustering algorithm, the initial position of the cluster center point is adjusted according to the health status score, and the lower the health status score, the more the initial position of the cluster center point deviates to the activity mode with weaker activity ability.

[0043] In a second aspect, the present application provides a smart control device for a sweeping robot, which is used for home safety monitoring of an old person living alone by using the sweeping robot, and the device comprises:

[0044] A data acquisition module is used for controlling the sweeping robot to autonomously patrol in a home environment by using a non-vision sensor, recording the activity frequency and residence time of the old person in different space regions, and adding historical activity data;

[0045] A space learning module is used for statistically analyzing the historical activity data, calculating the average residence time and average activity frequency of each space region, and identifying the main space region of the daily activity of the old person;

[0046] An activity mode recognition module is used for continuously monitoring the activity state of the old person in each main space region for the identified main space region, and obtaining baseline data of the daily activity mode of the old person; the baseline data includes average residence time, average activity frequency, residence time variance and activity frequency variance;

[0047] An abnormal state detection module is used for controlling the sweeping robot to collect activity data of the old person in the main space region in real time by using the non-vision sensor, and comparing the activity data with the baseline data of the daily activity mode to determine whether an abnormal state occurs;

[0048] The alarm response module is configured to interact with the old person through voice when an abnormal state occurs, and send an alarm message to the pre-set emergency contact person when no response is received or the situation is urgent.

[0049] The patrol strategy adjustment module is configured to adjust the security patrol strategy of the sweeping robot according to the identified main space area, wherein the security patrol strategy includes a patrol mode, a patrol frequency and an abnormality detection sensitivity in each space area.

[0050] In a third aspect, the present application provides an electronic device including a processor and a memory, wherein the memory stores a computer program executable by the processor, and the processor executes the computer program to perform the steps of the sweeping robot intelligent control method.

[0051] In a fourth aspect, the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to perform the steps of the sweeping robot intelligent control method.

[0052] Beneficial effects: The sweeping robot intelligent control method and related device provided by the present application realize the home safety monitoring of the old person living alone through the non-vision sensor carried by the sweeping robot, and enable the security strategy to be intelligently adjusted according to the activity space of the old person, have the advantages of giving consideration to safety monitoring and privacy protection, fully utilizing existing technical resources, reducing development cost, and providing a more friendly and natural intelligent security solution for the old person living alone. BRIEF DESCRIPTION OF DRAWINGS

[0053] Figure 1 The flowchart of the sweeping robot intelligent control method provided by the embodiment of the present application.

[0054] Figure 2 The structural schematic diagram of the sweeping robot intelligent control device provided by the embodiment of the present application.

[0055] Figure 3 The structural schematic diagram of the electronic device provided by the embodiment of the present application.

[0056] Label explanation: 1, data acquisition module; 2, space learning module; 3, activity mode identification module; 4, abnormal state detection module; 5, alarm response module; 6, patrol strategy adjustment module; 301, processor; 302, memory; 303, communication bus. DETAILED DESCRIPTION

[0057] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0058] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0059] refer to Figure 1 This application proposes an intelligent control method for a robotic vacuum cleaner, used to monitor the home safety of elderly people living alone. The method includes the following steps:

[0060] A1. Control the robot vacuum cleaner to autonomously patrol the home environment using non-visual sensors, record the frequency of the elderly’s activities and the time spent in different spatial areas, and add historical activity data;

[0061] A2. Statistically analyze historical activity data, calculate the average duration of stay and average activity frequency in each spatial area, and identify the main spatial areas of the elderly's daily activities;

[0062] A3. For the identified main spatial areas, continuously monitor the activity status of the elderly in each main spatial area to obtain baseline data of the elderly's daily activity patterns; the baseline data includes average stay duration, average activity frequency, variance of stay duration, and variance of activity frequency;

[0063] A4. Control the robot vacuum cleaner to collect real-time activity data of the elderly in the main space area using non-visual sensors, and compare it with the baseline data of daily activity patterns to determine whether an abnormal state has occurred.

[0064] A5. When an abnormal situation occurs, interact with the elderly via voice to confirm, and send an alarm message to the pre-set emergency contact if no response is received or if the situation is confirmed to be urgent;

[0065] A6. Adjust the security patrol strategy of the robot vacuum cleaner according to the identified main spatial areas; the security patrol strategy includes the patrol mode, patrol frequency and abnormality detection sensitivity in each spatial area.

[0066] The method realizes the home safety monitoring of the elderly living alone through the non-vision sensor carried by the robot vacuum cleaner, and enables the elderly to intelligently adjust the security strategy according to the activity space of the elderly, has the advantages of balancing safety monitoring and privacy protection, fully utilizing existing technical resources, reducing development cost, and providing a more friendly and natural intelligent security solution for the elderly living alone.

[0067] In step A1, the robot vacuum cleaner autonomously patrols in the home environment through non-vision sensors (such as infrared sensors, ultrasonic sensors or laser radars, etc.), records the activity frequency and stay time of the elderly in different areas such as bedroom and living room. The activity frequency is defined as the number of times the elderly appears in the area per unit time, and the stay time is defined as the duration of the elderly staying in the area continuously. These data are stored in the memory of the robot vacuum cleaner and added to the historical activity data to form a data set containing time stamp, area location, activity frequency and stay time.

[0068] In step A2, the historical activity data is statistically analyzed to calculate the average stay time and average activity frequency of each spatial area. For example, the average stay time in the bedroom is 8 hours and the average activity frequency is 2 times / hour, and the average stay time in the living room is 4 hours and the average activity frequency is 5 times / hour. According to these data, the main spatial areas of the elderly's daily activities are identified, such as the bedroom and the living room.

[0069] In step A3, for the main spatial areas (such as the bedroom and the living room), the activity state of the elderly in these areas is continuously monitored to obtain the baseline data of the elderly's daily activity pattern. The baseline data includes average stay time, average activity frequency, stay time variance and activity frequency variance. These data are used to establish the activity model of the elderly in normal state.

[0070] In step A4, the robot vacuum cleaner collects the activity data of the elderly in the main spatial areas (such as the bedroom and the living room) in real time and compares it with the baseline data of the daily activity pattern. For example, if the stay time of the elderly in the bedroom suddenly exceeds 12 hours or the activity frequency drops to 0, it is determined that an abnormal state has occurred.

[0071] In step A5, when an abnormal state occurs, the robot vacuum cleaner first interacts with the elderly through voice to confirm whether the elderly need help. If no response is received within the preset time or the elderly confirms that the situation is urgent, the robot vacuum cleaner will immediately send an alarm information to the pre-set emergency contact, including the location information of the elderly and the description of the abnormal state.

[0072] In step A6, the security patrol strategy of the sweeping robot is adjusted according to the identified main space area. For example, the patrol frequency in the main space area (such as the bedroom and the living room) is increased, the abnormality detection sensitivity in the main space area (such as the bedroom and the living room) is increased to discover potential security risks more timely, and the patrol frequency in the non-main space area and the abnormality detection sensitivity in the main space area (such as the bedroom and the living room) can be reduced.

[0073] Specifically, the technical solution utilizes the non-vision sensor carried by the sweeping robot to autonomously patrol in the home environment and collect activity data of the old person in different space areas. By analyzing these data, the main space area of the old person's daily activities is identified, and the baseline data of the daily activity pattern is established. In real-time monitoring, the collected activity data is compared with the baseline data to determine whether an abnormal state occurs. When an abnormal state is detected, the old person is ensured to receive timely assistance through voice interaction and emergency contact alarm. At the same time, the security strategy is intelligently adjusted according to the activity space of the old person, the energy use efficiency is optimized, and the potential interference to the daily life of the old person is reduced.

[0074] In some preferred embodiments, step A1 comprises:

[0075] A101. The security patrol strategy of the sweeping robot is retrieved, and the patrol mode of the sweeping robot in the home environment is extracted from the security patrol strategy; the patrol mode includes a random walk mode, a spiral cleaning mode, and a preset path mode; in the random walk mode, the sweeping robot moves randomly in the space area at a preset speed, and adjusts the moving direction when encountering an obstacle; in the spiral cleaning mode, the sweeping robot cleans the space area in a spiral trajectory, and the spiral radius gradually increases; in the preset path mode, the sweeping robot patrols according to a preset path;

[0076] A102. According to the extracted patrol mode, the sweeping robot autonomously patrols in the home environment by using the non-vision sensor;

[0077] A103. In the patrol process, the non-vision sensor is used to detect whether there is old person activity in the current space area in real time, and the space area position information, activity frequency, and stay time of the current space area are recorded when old person activity is detected;

[0078] A104. The activity frequency, stay time, and space area position information recorded in this patrol are added to the historical activity data, and the weight of the early data in the historical activity data is reduced according to a time decay factor.

[0079] In step A101, the security patrol strategy can be retrieved by the robot from the local memory or through a cloud server. The robot connects to the cloud through a wireless network to obtain the latest patrol strategy.

[0080] In step A102, the robot is controlled to patrol. If the patrol mode is random walk mode, the robot is set to a preset moving speed, and a random number generator is started to generate a random moving direction. When the robot detects an obstacle, an obstacle avoidance algorithm is triggered to adjust the moving direction. If the patrol mode is spiral cleaning mode, the robot is set to move in a spiral trajectory. If the patrol mode is preset path mode, the robot is controlled to move to each path point in sequence according to the path point coordinates read from the patrol strategy file.

[0081] In step A103, the non-vision sensor is used to detect whether there is an old person activity in the current space area in real time. For example, when the non-vision sensor is an infrared sensor, it is used to detect whether there is an old person activity by detecting human heat sources. When the infrared sensor detects a human heat source, the current space area position information is recorded, the activity frequency is incremented by 1, and the stay time is recorded. The space area position information is obtained by a positioning module of the robot. The positioning module can use a SLAM algorithm to construct a home environment map in real time and determine the position of the robot.

[0082] In step A104, the time decay factor can be a preset value.

[0083] Specifically, the scheme extracts the security patrol strategy of the robot, determines the current patrol mode, including random walk, spiral cleaning, and preset path. Then, according to the determined patrol mode, the robot is controlled to patrol autonomously, and the old person activity is detected in real time during the patrol to record the activity information. Finally, the recorded activity information is added to the historical activity data, and the time decay factor is used to reduce the weight of early data, so as to pay more attention to recent activity data and improve the accuracy of data analysis. Through these steps, the scheme can make the patrol of the robot more intelligent and the data recording more accurate, thereby improving the effect of home security monitoring. By using the time decay factor, the influence of early data on the current analysis result is reduced, so that the system can adapt to the change of the old person activity mode more quickly, and the sensitivity and accuracy of the monitoring are improved.

[0084] In some possible implementations, step A104 includes:

[0085] B1. record the activity frequency, stay time and space area position information of the current patrol according to a preset data structure, and store the information in a newly added activity data buffer area, the activity data buffer area adopts a ring queue structure, and the activity data of the last N patrol records is stored, N is a preset positive integer;

[0086] B2. extract the historical activity data from the activity data buffer area, and calculate the weighted activity frequency and weighted stay time of each space area position information according to the following formula: weighted activity frequency = original activity frequency * (1-time attenuation factor) ^ time interval, weighted stay time = original stay time * (1-time attenuation factor) ^ time interval, the time interval is the difference between the current time and the data record time (referring to the data record time of the historical activity data);

[0087] B3. replace the activity frequency and stay time of the corresponding space area position information in the original historical activity data with the weighted activity frequency and weighted stay time, and reduce the weight of the early data.

[0088] In step B1, the activity data buffer area adopts a ring queue structure, and one implementation way is to pre-allocate a fixed size of memory space for storing activity data records, each record containing activity frequency, stay time and space area position information. When a new patrol record is generated, it is written to the tail of the ring queue, and if the queue is full, the earliest record at the head is overwritten, so that only the last N patrol records are retained.

[0089] In step B2, the value range of the time attenuation factor can be set to 0 to 1, for example 0.01, and the unit of the time interval can be day or other time units. The calculation of the weighted activity frequency and the weighted stay time reflects that the farther the activity data is from the current time, the smaller the weight is.

[0090] In step B3, the weighted activity frequency and stay time are used to update the historical activity data, and one implementation way is to directly replace the original data with the weighted data, or to use a sliding average method to weight the weighted data and the original data to obtain new historical activity data.

[0091] Specifically, by step B1, the latest patrol data is stored in the active data buffer, and the old data is automatically removed, ensuring the freshness of the data. Step B2 weights the historical data using a time decay factor, and the larger the time decay factor, the smaller the influence of the historical data. Step B3 updates the weighted data to the historical activity data, realizing the dynamic update of the historical data, so that the method can effectively reduce the influence of early data on current activity pattern recognition, improve the accuracy of the evaluation of the activity state of the old people, and solve the problem that only using a fixed time decay factor cannot accurately reflect the changes of the activity pattern of the old people over time and season.

[0092] Further, before step B2, it can also include the following steps:

[0093] B4. Construct a time decay factor adjustment model, the model input is the activity frequency change rate, the stay duration change rate and the season information in a preset time period, and the output is the adjusted time decay factor;

[0094] B5. Based on the time decay factor adjustment model, calculate the adjusted time decay factor according to the activity frequency change rate, the stay duration change rate and the season information, to replace the initial time decay factor (for example, the initial time decay factor is a preset value).

[0095] Wherein, the construction of the time decay factor adjustment model involves three input features: activity frequency change rate, stay duration change rate and season information. The calculation of the activity frequency change rate and the stay duration change rate can use time series analysis methods such as linear regression and exponential smoothing to analyze the change trend of the activity frequency and the stay duration in a preset time period. Seasonal information can be encoded as a discrete variable, for example, spring is 1, summer is 2, autumn is 3, and winter is 4. The time decay factor adjustment model can be constructed using machine learning models such as neural networks, support vector machines or regression trees, and by training historical data, a mapping relationship between input features and time decay factors is established. The calculation of the time decay factor is by inputting the activity frequency change rate, the stay duration change rate and the season information into the trained time decay factor adjustment model to obtain the adjusted time decay factor.

[0096] Specifically, by constructing the time decay factor adjustment model and dynamically adjusting the time decay factor using the activity frequency change rate, the stay duration change rate and the season information, the historical activity data can more accurately reflect the current activity pattern of the old people. The activity frequency change rate and the stay duration change rate reflect the short-term change trend of the activity pattern of the old people, and the season information considers the influence of seasonal factors on the activity of the old people. By considering these factors comprehensively, the time decay factor can adaptively adjust the weight of the historical activity data, thereby improving the accuracy of subsequent activity pattern recognition and abnormal state detection.

[0097] In some embodiments, the time decay factor adjustment model employs a three-layer fully connected neural network. The activity frequency change rate and the stay duration change rate are calculated by computing the linear regression slope of the activity frequency and the stay duration in the past 7 days, respectively. The season information is encoded as an integer from 1 to 4, representing spring, summer, autumn, and winter, respectively. The input layer of the neural network has 3 neurons, corresponding to the activity frequency change rate, the stay duration change rate, and the season information, respectively. The hidden layer has 10 neurons with a ReLU activation function. The output layer has 1 neuron, which outputs the adjusted time decay factor. The neural network is trained using historical activity data, and the optimization goal is to minimize the mean squared error between the predicted time decay factor and the actual optimal time decay factor. The actual optimal time decay factor is determined by a grid search method, i.e., evaluating the accuracy of activity pattern recognition and abnormal state detection at different time decay factor values, and selecting the time decay factor with the highest accuracy as the actual optimal value.

[0098] In some embodiments, step A2 comprises:

[0099] A201. Smoothing the historical activity data using a Kalman filter algorithm to eliminate noise data introduced by sensor errors and obtain smoothed activity data;

[0100] A202. Setting activity frequency threshold and stay duration threshold, and counting the number of times the activity frequency of each spatial region exceeds the activity frequency threshold and the number of times the stay duration exceeds the stay duration threshold in the smoothed activity data;

[0101] A203. Calculating the weighted sum of the number of times the activity frequency of each spatial region exceeds the activity frequency threshold and the number of times the stay duration exceeds the stay duration threshold, and determining the main spatial region of the elderly's daily activities based on the weighted sum result.

[0102] In step A201, the Kalman filter algorithm is used to smooth the historical activity data and eliminate sensor errors. Kalman filtering includes prediction and update stages. In the prediction stage, the state at the current time is estimated based on the state at the previous time. In the update stage, the predicted state is corrected using the measurement value at the current time to obtain the optimal state estimation at the current time.

[0103] In step A202, the activity frequency threshold and the stay duration threshold are used to screen spatial regions with frequent activities and long stay durations. The activity frequency threshold and the stay duration threshold can be pre-set or adaptively adjusted according to actual conditions.

[0104] In step A203, the weighted sum is used to comprehensively consider the activity frequency and the stay duration to determine the main activity region (i.e., the main spatial region, and the main activity region hereinafter also refers to the main spatial region). The calculation formula of the weighted sum is: weighted sum = w1*the number of times of activity frequency exceeding the threshold value + w2*the number of times of stay duration exceeding the threshold value, wherein w1 and w2 are weight coefficients, which can be adjusted according to actual conditions. A classification threshold value can be set. When the weighted sum of the number of times of activity frequency exceeding the activity frequency threshold value and the number of times of stay duration exceeding the stay duration threshold value of a spatial region is not less than the classification threshold value, the corresponding spatial region is determined as a main spatial region, otherwise, the corresponding spatial region is determined as a non-main spatial region.

[0105] Specifically, the historical activity data is smoothed by the Kalman filtering algorithm to reduce the influence of noise on subsequent analysis and improve data accuracy. Then, the activity frequency threshold value and the stay duration threshold value are set to filter out the spatial regions where the old people frequently and stay for a long time. Finally, the weighted sum is calculated to comprehensively consider the activity frequency and the stay duration to more accurately identify the main spatial region of the old people's daily activities. Thus, the problem of noise introduced by sensor errors can be effectively solved, and the accuracy of identifying the main spatial region of the old people's daily activities can be improved.

[0106] In some preferred embodiments, step A202 comprises:

[0107] The mean and variance of the activity frequency of each spatial region, and the mean and variance of the stay duration of each spatial region are calculated.

[0108] Based on the mean and variance of the activity frequency of each spatial region, an adaptive algorithm is used to calculate the activity frequency threshold value, and based on the mean and variance of the stay duration of each spatial region, an adaptive algorithm is used to calculate the stay duration threshold value; wherein the activity frequency threshold value is positively correlated with the activity frequency mean and the activity frequency variance, and the stay duration threshold value is positively correlated with the stay duration mean and the stay duration variance.

[0109] The number of times of activity frequency exceeding the activity frequency threshold value and the number of times of stay duration exceeding the stay duration threshold value in the smoothed activity data are counted.

[0110] The mean and variance of the activity frequency can be calculated by using a sliding window method, for example, the mean and variance of the activity frequency of each spatial region in the past 24 hours are calculated every 1 hour.

[0111] The adaptive algorithm can calculate the activity frequency threshold value using the following formula: activity frequency threshold value = activity frequency mean + K1*activity frequency variance, where K1 is an adjustable parameter used to control the sensitivity of the threshold value to the activity frequency variance. The calculation method of the stay duration threshold value is similar: stay duration threshold value = stay duration mean + K2*stay duration variance, where K2 is an adjustable parameter used to control the sensitivity of the threshold value to the stay duration variance. The value range of K1 and K2 can be adjusted according to the actual application scenario, for example, K1 and K2 can be set to a value between 0.5 and 1.5.

[0112] Specifically, by calculating the mean of activity frequency and stay duration, the average activity level of the elderly in each spatial region can be reflected. By calculating the variance of activity frequency and stay duration, the fluctuation degree of the elderly activity can be reflected. The activity frequency threshold value and the stay duration threshold value are adaptively adjusted, so that the threshold value can change with the change of the elderly activity mode, avoiding the false judgment caused by the fixed threshold value. The activity frequency threshold value is positively correlated with the activity frequency mean, ensuring that the threshold value can change with the change of the overall activity level of the elderly. The activity frequency threshold value is positively correlated with the activity frequency variance, ensuring that when the elderly activity fluctuates greatly, the threshold value can be appropriately relaxed to avoid false judgment. The relationship between the stay duration threshold value and the stay duration mean and variance is similar to that of the activity frequency threshold value. By counting the number of times the activity frequency exceeds the activity frequency threshold value and the number of times the stay duration exceeds the stay duration threshold value, the main activity area of the elderly can be more accurately identified, providing a basis for subsequent security patrol strategy adjustment.

[0113] In some preferred embodiments, step A3 comprises:

[0114] A301. Construct an activity pattern clustering model, the model input includes the average stay duration, the average activity frequency, the stay duration variance and the activity frequency variance in the preset time window, and the output is the activity pattern category;

[0115] A302. Based on the activity pattern clustering model, perform clustering analysis on the historical activity data (preferably smoothed activity data) to obtain multiple activity pattern categories, and determine the center vector of each activity pattern category, the center vector including the average stay duration, the average activity frequency, the stay duration variance and the activity frequency variance;

[0116] A303. For the identified main spatial region, continuously monitor the activity state of the elderly in each main spatial region, obtain the feature vector of the current activity pattern of the elderly, and the feature vector includes the average stay duration, the average activity frequency, the stay duration variance and the activity frequency variance;

[0117] A304. Calculate the distance between the feature vector of the current activity pattern and the center vector of each activity pattern category, and determine the activity pattern category corresponding to the center vector with the smallest distance as the current daily activity pattern of the elderly, and take the center vector as the baseline data of the daily activity pattern of the elderly.

[0118] In step A301, the construction of the activity pattern clustering model involves determining the input and output of the model. The input includes the average stay duration, the average activity frequency, the stay duration variance and the activity frequency variance in the preset time window. These parameters are collected and calculated by non-visual sensors on the sweeping robot. The output is the activity pattern category, such as "normal activity", "rest", "abnormal stay" and the like. The implementation of clustering analysis can adopt K-means algorithm, which divides the historical activity data with similar activity characteristics into the same category through iterative optimization.

[0119] In step A302, the center vector of each activity pattern category is obtained by calculating the average value of all data points in the category, which represents the typical activity pattern of the category.

[0120] In step A303, the feature vector of the current activity pattern of the elderly is obtained by real-time monitoring of the activity state of the elderly in the main space area, and calculating the average stay duration, the average activity frequency, the stay duration variance and the activity frequency variance.

[0121] In step A304, the distance between the feature vector of the current activity pattern and the center vector of each activity pattern category is calculated, which can adopt the Euclidean distance formula to measure the similarity between the current activity pattern and the historical activity pattern.

[0122] Specifically, by constructing an activity pattern clustering model, clustering analysis is performed on historical activity data to obtain multiple activity pattern categories. When the sweeping robot real-time monitors the activity state of the elderly in the main space area, the feature vector of the current activity pattern of the elderly is obtained, and the distance between the feature vector and the center vector of each activity pattern category is calculated. The activity pattern category corresponding to the center vector with the smallest distance is determined as the current daily activity pattern of the elderly, and the center vector is taken as the baseline data of the daily activity pattern of the elderly. Therefore, the daily activity pattern of the elderly can be more accurately identified, and more reliable baseline data can be provided for subsequent abnormal state detection, thereby improving the accuracy and reliability of home safety monitoring.

[0123] In some possible implementations, step A301 includes:

[0124] Obtain historical activity data in a preset time window, extract the health status information of the elderly corresponding to the historical activity data, and the health status information includes the duration of chronic disease and medication.

[0125] The health status information is quantitatively processed to obtain a health status score, and the health status score is taken as one of input features of an activity pattern clustering model to construct a multi-dimensional input vector together with other activity features;

[0126] An activity pattern clustering model is constructed by using an improved K-means clustering algorithm to perform clustering analysis on the multi-dimensional input vector, and the activity pattern category is determined according to the clustering result; in the iteration process of the improved K-means clustering algorithm, the initial position of the cluster center point is adjusted according to the health status score, and the lower the health status score, the more the initial position of the corresponding cluster center point deviates from the activity pattern with weaker activity ability (the strength of the activity ability can be represented by the average stay duration and the average activity frequency, and the smaller the average stay duration and the average activity frequency, the weaker the activity ability).

[0127] The medication condition can include the drug type, the dose and the taking frequency. The chronic disease duration includes the duration of each chronic disease suffered by the old person. The health status information can be recorded in the local storage of the sweeping robot or in the cloud server, so that the health status information can be extracted from the local storage of the sweeping robot or the cloud server.

[0128] The health status information is quantitatively processed to obtain a health status score can be realized as follows:

[0129] The chronic disease durations of each chronic disease suffered by the old person are weighted and calculated according to the types of the chronic diseases to obtain a preliminary disease duration score; the weights of the chronic disease durations of various chronic diseases are determined according to the types of the diseases, for example, a mapping table of the types of the diseases and the weights can be queried (the mapping table can be pre-stored in the local storage of the sweeping robot or in the cloud server and called when needed) to obtain the weights;

[0130] The equivalent dosages of the same drug are obtained by multiplying the dose and the taking frequency, and the equivalent dosages of various drugs taken by the old person are weighted and calculated according to the types of the drugs to obtain a preliminary medication condition score; the weights of the equivalent dosages of various drugs are determined according to the types of the drugs, for example, a mapping table of the types of the drugs and the weights can be queried (the mapping table can be pre-stored in the local storage of the sweeping robot or in the cloud server and called when needed) to obtain the weights;

[0131] The preliminary disease duration score and the preliminary medication condition score are normalized (for example, normalized calculation is performed according to the maximum and minimum values of the respective preset ranges) to obtain a normalized disease duration score and a normalized medication condition score;

[0132] The weighted sum of the normalized illness duration score and the normalized medication score (the sum of the weights of the two is 1) is calculated to obtain the health status score.

[0133] The multi-dimensional input vector construction process can be implemented as follows: the health status score, the average stay duration, the average activity frequency, the stay duration variance, and the activity frequency variance are combined into a five-dimensional vector as the input of the K-means clustering algorithm.

[0134] The improved K-means clustering algorithm can be implemented as follows: when initializing the cluster center points, the initial values of the average stay duration and the average activity frequency are adjusted according to the health status score. For example, if the health status score is low, the initial values of the average stay duration and the average activity frequency of the corresponding cluster center points are set to low values, for example, 10% to 30% of the historical data.

[0135] Specifically, by extracting the health status information of the elderly and quantifying it into a health status score, the score is integrated into the multi-dimensional input vector of the activity pattern clustering model. In the clustering analysis process, the improved K-means clustering algorithm adjusts the initial position of the cluster center points according to the health status score, so that the activity patterns of the elderly with lower health status scores are more likely to be classified into the category with weaker activity ability. Thus, the clustering result can more accurately reflect the actual activity patterns of the elderly, providing a more reliable basis for subsequent abnormal state detection.

[0136] In some embodiments, when the activity pattern clustering model performs clustering analysis on the multi-dimensional input vectors using the improved K-means clustering algorithm and determines the activity pattern categories according to the clustering results, it performs:

[0137] C1. Initialize the cluster center points of the K-means clustering algorithm, and adjust the initial values of the average stay duration and the average activity frequency in the cluster center points according to the health status score. The lower the health status score, the smaller the initial values of the average stay duration and the average activity frequency.

[0138] C2. Calculate the distance between each multi-dimensional input vector and each cluster center point, divide the multi-dimensional input vector into the nearest cluster, and update the cluster center point according to the division result. The updated cluster center point is the mean value of all multi-dimensional input vectors in the cluster.

[0139] C3. Repeat step C2 until the cluster center points no longer change or the maximum number of iterations is reached, obtain the clustering result, and determine the activity pattern categories according to the clustering result.

[0140] In the C1 step, the adjustment of the initial position of the cluster center point can be achieved by the following formula: adjusted average stay duration = initial average stay duration * (1-health status score), and adjusted average activity frequency = initial average activity frequency * (1-health status score). Thus, the lower the health status score, the smaller the initial values of the average stay duration and the average activity frequency, so that the clustering algorithm pays more attention to the elderly with weaker activity ability.

[0141] In the C2 step, the distance between each multi-dimensional input vector and each cluster center point is calculated, the multi-dimensional input vector is divided into the nearest cluster, and the cluster center point is updated according to the division result. As a preferred embodiment, the distance calculation can use the Euclidean distance. The update of the cluster center point can be achieved by calculating the mean of all multi-dimensional input vectors in the cluster.

[0142] In the C3 step, the step C2 is repeatedly executed until the cluster center point no longer changes or the maximum iteration number is reached, and the clustering result is obtained. The maximum iteration number can be set according to actual needs, and as a preferred embodiment, the maximum iteration number can be set to 100 times.

[0143] Specifically, by introducing the health status score, the initial position of the cluster center point can be adjusted, so that the K-means algorithm can consider the individual differences of the elderly at the beginning of iteration, avoiding the deviation caused by random initialization. By calculating the distance between each data point and the cluster center point, and dividing the data point into the nearest cluster, the clustering analysis of the activity pattern of the elderly can be achieved. By repeatedly calculating the distance and updating the cluster center point, the clustering result can gradually converge, so that a more accurate activity pattern category can be obtained. Thus, the technical scheme can improve the accuracy and stability of the clustering result, so as to more accurately identify the activity pattern of the elderly, and solve the problem of sensitivity of the K-means algorithm to the initial cluster center point.

[0144] In some embodiments, step A4 comprises:

[0145] A401. The control floor cleaning robot collects the stay duration data and activity frequency data of the elderly in the main space area in real time using non-vision sensors;

[0146] A402. Calculate the first deviation value of the average stay duration of the real-time collected stay duration data and the daily activity pattern baseline data, and calculate the second deviation value of the average activity frequency of the real-time collected activity frequency data and the daily activity pattern baseline data;

[0147] A403. Determine whether the first deviation value exceeds the preset stay duration deviation threshold, and whether the second deviation value exceeds the preset activity frequency deviation threshold, if both exceed, it is determined that an abnormal state occurs.

[0148] Only when the deviation values of the staying time and the activity frequency both exceed the corresponding threshold values, it is determined that the abnormal state occurs, avoiding false positives caused by single data anomaly, and improving the reliability of the system. The reason for judging the two deviation values at the same time is that single data anomaly may be caused by multiple reasons, and only when both data are abnormal at the same time, it can be more accurately judged whether the old person is in an abnormal state.

[0149] It should be noted that when patrolling in the non-main space area, if the old person is detected, the staying time data and the activity frequency data of the old person in the non-main space area can be collected, and compared with the preset staying time threshold value and the preset activity frequency threshold value respectively, so as to determine whether an abnormal state occurs.

[0150] In some embodiments, step A5 comprises:

[0151] A501. When it is determined that an abnormal state occurs, the robot controls the voice module to play a preset voice prompt and starts a timer;

[0152] A502. Determine whether the old person sends a confirmation instruction through the voice module before the timer expires;

[0153] A503. If the confirmation instruction is received before the timer expires, cancel the alarm and record the interaction information this time;

[0154] A504. If the confirmation instruction is not received after the timer expires, or an urgent voice instruction is received, the robot sends alarm information containing the current location information of the old person and the type of abnormal state to the pre-set emergency contact through the wireless communication module.

[0155] After detecting the abnormal state, the method first interacts with the old person through voice to confirm, sets a timeout time, and if no response is received or an urgent instruction is received, sends an alarm information, thereby reducing the false positive rate and ensuring timely response in emergency situations.

[0156] In step A6, the patrol frequency and the abnormal detection sensitivity of the main space area can be improved, and the patrol frequency and the abnormal detection sensitivity of the non-main space area can be reduced. Improving the patrol frequency and the abnormal detection sensitivity of the main activity space can ensure timely, safe and reliable monitoring of the old person in the main activity area, and reducing the patrol frequency and the abnormal detection sensitivity of the non-main space area can reduce unnecessary patrol work, thereby realizing safe, energy-saving and low-interference monitoring.

[0157] In addition, the obstacle avoidance ability and the patrol efficiency of various patrol modes can be evaluated (which can be determined through comparison tests), for the main space area, the patrol mode with stronger obstacle avoidance ability is preferred to reduce the interference to the activities of the old people and improve the safety, and for the non-main space area, the patrol mode with higher patrol efficiency is preferred to improve the patrol efficiency; therefore, the safety and the efficiency can be considered.

[0158] Reference Figure 2 The application further provides a robot intelligent control device for home safety monitoring of the old people by using the robot, which comprises:

[0159] A data acquisition module 1 is configured to control the robot to autonomously patrol in the home environment by using the non-vision sensor, record the activity frequency and the residence time of the old people in different space areas, and add the historical activity data (for the specific process, refer to the step A1 in the foregoing description).

[0160] A space learning module 2 is configured to statistically analyze the historical activity data, calculate the average residence time and the average activity frequency of each space area, and identify the main space area of the daily activities of the old people (for the specific process, refer to the step A2 in the foregoing description).

[0161] An activity mode recognition module 3 is configured to continuously monitor the activity state of the old people in each main space area for the identified main space area, and acquire the baseline data of the daily activity mode of the old people; the baseline data comprises the average residence time, the average activity frequency, the residence time variance and the activity frequency variance (for the specific process, refer to the step A3 in the foregoing description).

[0162] An abnormal state detection module 4 is configured to control the robot to collect the activity data of the old people in the main space area by using the non-vision sensor in real time, compare the activity data with the baseline data of the daily activity mode, and determine whether an abnormal state occurs (for the specific process, refer to the step A4 in the foregoing description).

[0163] An alarm response module 5 is configured to, when the abnormal state occurs, interact with the old people through voice and confirm, and when no response is received or the situation is urgent, send an alarm information to the pre-set emergency contact person (for the specific process, refer to the step A5 in the foregoing description).

[0164] A patrol strategy adjustment module 6 is configured to adjust the security patrol strategy of the robot according to the identified main space area; the security patrol strategy comprises the patrol mode, the patrol frequency and the abnormal detection sensitivity in each space area (for the specific process, refer to the step A6 in the foregoing description).

[0165] It should be noted that the abnormal state detection module 4 can also be used to collect the stay duration data and the activity frequency data of the old person in the non-main space area when the robot is patrolling in the non-main space area, and compare the stay duration data and the activity frequency data with the preset stay duration threshold and the preset activity frequency threshold, respectively, to determine whether an abnormal state occurs.

[0166] Please refer to Figure 3 , Figure 3 A structural schematic diagram of an electronic device provided by an embodiment of the present application, the present application provides an electronic device, comprising: a processor 301 and a memory 302, the processor 301 and the memory 302 are interconnected and communicate with each other through a communication bus 303 and / or other forms of connection mechanism (not marked), the memory 302 stores a computer program executable by the processor 301, when the electronic device runs, the processor 301 executes the computer program to execute the intelligent control of the robot in any optional implementation manner of the above-mentioned embodiment of the robot, to realize the following functions: controlling the robot to patrol in the home environment by using the non-vision sensor, recording the activity frequency and the stay time of the old person in different space areas, and adding the historical activity data; statistical analysis of the historical activity data, calculating the average stay duration and the average activity frequency of each space area, identifying the main space area of the old person's daily activity; for the identified main space area, continuously monitoring the activity state of the old person in each main space area, obtaining the baseline data of the old person's daily activity mode; the baseline data includes the average stay duration, the average activity frequency, the stay duration variance and the activity frequency variance; controlling the robot to collect the activity data of the old person in the main space area by using the non-vision sensor, for comparison with the baseline data of the daily activity mode, to determine whether an abnormal state occurs; when an abnormal state occurs, interacting with the old person through voice and confirming, and sending alarm information to the pre-set emergency contact person in the case of no response or confirmation in an emergency; according to the identified main space area, adjusting the security patrol strategy of the robot; the security patrol strategy includes the patrol mode, the patrol frequency and the abnormal detection sensitivity in each space area.

[0167] The embodiment of the application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to perform the intelligent control of the sweeping robot in any optional implementation manner of the above embodiment, so as to realize the following functions: controlling the sweeping robot to perform autonomous patrol in a home environment by using a non-vision sensor, recording activity frequency and stay time of an old person in different space regions, and adding historical activity data; statistically analyzing the historical activity data, calculating average stay time and average activity frequency of each space region, and identifying main space regions of daily activities of the old person; continuously monitoring activity states of the old person in each main space region for the identified main space regions, and obtaining baseline data of a daily activity mode of the old person; the baseline data includes average stay time, average activity frequency, stay time variance and activity frequency variance; controlling the sweeping robot to collect activity data of the old person in the main space region by using the non-vision sensor in real time, so as to compare with baseline data of the daily activity mode, and determine whether an abnormal state occurs; when the abnormal state occurs, interacting with the old person through voice, and sending alarm information to a pre-set emergency contact person when no response or confirmation is received or in an emergency situation; adjusting a security patrol strategy of the sweeping robot according to the identified main space regions; the security patrol strategy includes a patrol mode, a patrol frequency and an abnormality detection sensitivity in each space region.

[0168] The computer readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk or an optical disk.

[0169] In the embodiments of the present application, it should be understood that the disclosed apparatus and method can be implemented in other manners. The embodiments described above are merely exemplary, for example, the division of the units is only a logical function division, and there can be another division manner in actual implementation; for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections can be indirect couplings or communication connections through some interfaces, devices or units, and can be in electrical, mechanical or other forms.

[0170] In addition, the units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, and can be located in one position, or can be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purposes of the embodiments of the present application.

[0171] In addition, the functional modules in each embodiment of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0172] In this document, the terms such as first and second are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations.

[0173] The above only describes the embodiments of the present application, and is not used to limit the protection scope of the present application. For those skilled in the art, various modifications and changes can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for intelligent control of a robotic vacuum cleaner, used to monitor the home safety of elderly people living alone using a robotic vacuum cleaner, characterized in that, Steps of the method include: A1. Control the floor cleaning robot to autonomously patrol in the home environment using non-vision sensors, record the activity frequency and stay time of the old person in different space areas, and add historical activity data; A2. Statistically analyze the historical activity data, calculate the average stay time and average activity frequency of each space area, and identify the main space area of the old person's daily activities; A3. For the identified main space area, continuously monitor the activity state of the old person in each main space area to obtain baseline data of the old person's daily activity pattern; the baseline data includes average stay time, average activity frequency, stay time variance and activity frequency variance; A4. Control the floor cleaning robot to collect activity data of the old person in the main space area in real time using non-vision sensors, and compare it with the baseline data of the daily activity pattern to determine whether an abnormal state occurs; A5. When an abnormal state occurs, interact with the old person through voice and confirm, and in the case of no response or confirmation in an emergency, send an alarm message to the pre-set emergency contact person; A6. According to the identified main space area, adjust the security patrol strategy of the floor cleaning robot; the security patrol strategy includes the patrol mode, patrol frequency and abnormal detection sensitivity in each space area; Step A3 includes: A301. Build an activity pattern clustering model, the model input includes the average stay time, average activity frequency, stay time variance and activity frequency variance in the preset time window, and the output is the activity pattern category; A302. Based on the activity pattern clustering model, cluster analysis is performed on the historical activity data to obtain multiple activity pattern categories, and the center vector of each activity pattern category is determined, including average stay time, average activity frequency, stay time variance and activity frequency variance; A303. For the identified main space area, continuously monitor the activity state of the old person in each main space area to obtain the feature vector of the current activity pattern of the old person, including average stay time, average activity frequency, stay time variance and activity frequency variance; A304. Calculate the distance between the feature vector of the current activity pattern and the center vector of each activity pattern category, and the activity pattern category corresponding to the center vector with the smallest distance is taken as the current daily activity pattern of the old person, and the center vector is taken as the baseline data of the old person's daily activity pattern; Step A301 includes: Obtain the historical activity data in the preset time window, extract the old person's health status information corresponding to the historical activity data, and the health status information includes the duration of chronic disease and medication; Quantitatively process the health status information to obtain a health status score, and take the health status score as one of the input features of the activity pattern clustering model, and construct a multi-dimensional input vector together with other activity features; The improved K-means clustering algorithm adjusts the initial position of the cluster center point according to the health state score in the iteration process. The lower the health state score, the more the initial position of the cluster center point deviates from the activity mode with weaker activity ability. 2.The intelligent control method of the sweeping robot according to claim 1, characterized in that, Step A1 comprises: A101. Call the security patrol strategy of the sweeping robot, and extract the patrol mode of the sweeping robot in the home environment from the security patrol strategy; the patrol mode includes a random walk mode, a spiral cleaning mode and a preset path mode; in the random walk mode, the sweeping robot moves randomly in the space region at a preset speed, and adjusts the moving direction when encountering an obstacle; in the spiral cleaning mode, the sweeping robot cleans the space region in a spiral trajectory, and the spiral radius gradually increases; in the preset path mode, the sweeping robot patrols according to the preset path; A102. According to the extracted patrol mode, control the sweeping robot to autonomously patrol in the home environment by using the non-vision sensor; A103. In the process of patrolling, the non-vision sensor is used to detect whether there is an old person activity in the current space region in real time, and the space region position information, the activity frequency and the stay time of the current space region are recorded when the old person activity is detected; A104. Add the activity frequency, the stay time and the space region position information recorded in this patrol to the historical activity data, and reduce the weight of the early data in the historical activity data according to the time decay factor. 3.The intelligent control method of the sweeping robot according to claim 2, characterized in that, Step A104 comprises: B1. Store the activity frequency, the stay time and the space region position information recorded in this patrol to the newly added activity data buffer area according to the preset data structure, the activity data buffer area adopts a ring queue structure, and the activity data of the last N times of patrol records are stored, N is a preset positive integer; B2. Extract the historical activity data from the activity data buffer area, and calculate the weighted activity frequency and the weighted stay time of each space region position information according to the following formula: weighted activity frequency = original activity frequency * (1-time decay factor)^time interval, weighted stay time = original stay time * (1-time decay factor)^time interval, the time interval is the difference between the current time and the data recording time; B3. Replace the activity frequency and the stay time of the corresponding space region position information in the original historical activity data with the weighted activity frequency and the weighted stay time, and reduce the weight of the early data. 4.The intelligent control method of the sweeping robot according to claim 1, characterized in that, Step A2 comprises: A201. Smooth the historical activity data by using the Kalman filtering algorithm to eliminate the noise data introduced by the sensor error, and obtain the smoothed activity data; A202. Set an activity frequency threshold and a stay time threshold, and count the number of times that the activity frequency of each space region in the smoothed activity data exceeds the activity frequency threshold and the number of times that the stay time exceeds the stay time threshold; A203. Calculate the weighted sum of the number of times that the activity frequency of each space region exceeds the activity frequency threshold and the number of times that the stay time exceeds the stay time threshold, and determine the main space region of the old person's daily activity according to the weighted sum result.

5. The intelligent control method of the sweeping robot according to claim 4, characterized in that, Step A202 comprises: According to the smoothed activity data, the mean and variance of the activity frequency of each spatial region, and the mean and variance of the stay duration of each spatial region are calculated; Based on the mean and variance of the activity frequency of each spatial region, an adaptive algorithm is used to calculate the activity frequency threshold, and based on the mean and variance of the stay duration of each spatial region, an adaptive algorithm is used to calculate the stay duration threshold; wherein the activity frequency threshold is positively correlated with the activity frequency mean and the activity frequency variance, and the stay duration threshold is positively correlated with the stay duration mean and the stay duration variance; The number of times the activity frequency of each spatial region in the smoothed activity data exceeds the activity frequency threshold and the number of times the stay duration exceeds the stay duration threshold are counted.

6. A smart control device for a sweeping robot for home safety monitoring of a solitary elderly person by a sweeping robot, characterized in that, The device comprises: The data acquisition module is configured to control the sweeping robot to perform autonomous patrol in the home environment using the non-vision sensor, record the activity frequency and stay time of the old person in different spatial regions, and add historical activity data; The spatial learning module is configured to statistically analyze the historical activity data, calculate the average stay duration and average activity frequency of each spatial region, and identify the main spatial regions of the old person's daily activities; The activity pattern recognition module is configured to continuously monitor the activity state of the old person in each main spatial region for the identified main spatial regions, and obtain baseline data of the old person's daily activity pattern; the baseline data includes the average stay duration, the average activity frequency, the stay duration variance, and the activity frequency variance; The abnormal state detection module is configured to control the sweeping robot to collect activity data of the old person in the main spatial region in real time using the non-vision sensor, and compare the activity data with the baseline data of the daily activity pattern to determine whether an abnormal state occurs; The alarm response module is configured to interact with the old person through voice when an abnormal state occurs, and send an alarm message to the pre-set emergency contact person in an emergency situation when no response or confirmation is received; The patrol strategy adjustment module is configured to adjust the security patrol strategy of the sweeping robot according to the identified main spatial regions; the security patrol strategy includes the patrol mode, the patrol frequency, and the abnormal detection sensitivity in each spatial region; When the activity pattern recognition module continuously monitors the activity state of the old person in each main spatial region for the identified main spatial regions, and obtains baseline data of the old person's daily activity pattern, the following steps are performed: A301. An activity pattern clustering model is constructed, the model input includes the average stay duration, the average activity frequency, the stay duration variance, and the activity frequency variance within a preset time window, and the output is an activity pattern category; A302. Based on the activity pattern clustering model, the historical activity data is clustered and analyzed to obtain multiple activity pattern categories, and the center vector of each activity pattern category is determined, the center vector includes the average stay duration, the average activity frequency, the stay duration variance, and the activity frequency variance; A303. For the identified main spatial regions, the activity state of the old person in each main spatial region is continuously monitored, and the feature vector of the current activity pattern of the old person is obtained, the feature vector includes the average stay duration, the average activity frequency, the stay duration variance, and the activity frequency variance; A304. Calculate the distance between the feature vector of the current activity mode and the center vector of each activity mode category, and take the activity mode category corresponding to the center vector with the smallest distance as the current daily activity mode of the old person, and take the center vector as the baseline data of the daily activity mode of the old person; Step A301 comprises: Obtaining historical activity data in a preset time window, extracting the health status information of the old person corresponding to the historical activity data, and the health status information including the duration of chronic disease and the medication condition; Quantitatively processing the health status information to obtain a health status score, and taking the health status score as one of the input features of the activity mode clustering model, and constructing a multi-dimensional input vector together with other activity features; An activity mode clustering model is constructed by using an improved K-means clustering algorithm to perform clustering analysis on the multi-dimensional input vector, and the activity mode category is determined according to the clustering result; in the iteration process of the improved K-means clustering algorithm, the initial position of the cluster center point is adjusted according to the health status score, and the lower the health status score, the more the initial position of the cluster center point deviates to the activity mode with weaker activity ability.

7. An electronic device, comprising: The computer program is executed by the processor to run the steps of the intelligent control method of the sweeping robot in any one of claims 1-5.

8. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to run the steps of the intelligent control method of the sweeping robot in any one of claims 1-5.

Images