Intelligent home management method and system based on artificial intelligence

Abstract

The invention relates to a smart home management method and system based on artificial intelligence, the system comprises a smart home platform and a plurality of smart home devices, and the smart home platform intelligently judges the working mode of each smart home device based on an artificial intelligence model in a multi-user scene. And requirements of different users can be met as far as possible.

Description

Technical Field

[0001] The present invention belongs to the field of smart home, and particularly relates to a smart home management method and system based on artificial intelligence.

Background Art

[0002] Smart home is a platform based on a residence, which integrates facilities related to home life by using technologies such as comprehensive wiring technology, network communication technology, security prevention technology, automatic control technology, and audio and video technology, constructs a management system for efficient residential facilities and family daily affairs, improves the safety, convenience, comfort, and artistry of the home, and realizes an environmentally friendly and energy-saving living environment. The existing smart home systems are not limited to the intelligence of a single home device, but adopt a unified smart home platform to control the working modes of various home devices.

[0003] In recent years, with the rise of artificial intelligence, the platform can adopt an artificial intelligence model to determine the next working mode of each device based on the states of various devices in the current residence, as well as other external states and user states. However, the existing artificial intelligence models usually only target a single user and consider the device working mode adapted to the needs of that user. When there are multiple users in a residence, the needs of different users may not be the same, and may even conflict. The existing artificial intelligence models have not considered how to control the device working mode when the needs of different users are different under the condition of multiple users.

Summary of the Invention

[0004] To solve the above problems in the prior art, the present invention provides a smart home management method and system based on artificial intelligence.

[0005] The technical solution adopted by the present invention is specifically as follows:

[0006] A smart home management method based on artificial intelligence, the method comprising:

[0007] Step 1: Collect historical data of the smart home management system in the residence; the historical data includes the working modes of various smart home devices, external data, and user states;

[0008] Step 2: For any smart home device D, construct a corresponding working mode prediction model M D , construct a training sample set based on the historical data, and train the model M D ; wherein, the model M D inputs the working modes of all smart home devices except the device D, external data, and a single user code, and predicts the working mode of the device D corresponding to the user code.

[0009] Step 3: When there are multiple users in the residence, the smart home platform identifies each user and determines the influence coefficient of device D on each user in the current residence based on the influence range of device D; the influence coefficient is used to determine the influence of device D on the location of the user.

[0010] Step 4: For any user with a non-zero influence coefficient, according to model M D Calculate the probabilities of all working modes of device D, and all the obtained probabilities form the probability matrix of device D.

[0011] Step 5: If the working mode of device D includes working parameters of non-continuous variables, determine the working parameter according to the weighting of the probability matrix by the influence coefficient.

[0012] Step 6: If the working mode of device D also includes working parameters of continuous variables, then process the working parameter predicted by model M D For each user into a normal distribution curve with the predicted value as the mean.

[0013] Step 7: Determine the working parameter of the continuous variable according to the weighted superposition of the normal distribution curve by the influence coefficient.

[0014] Further, step 4 specifically includes:

[0015] Suppose there are n users User1, User2,..., User n with non-zero influence coefficients, and device D has m working modes. Then for any user User i , obtain the probabilities of all working modes of device D output by model M D .

[0016] Let P ij be the probability of predicting the j-th working mode of device D for User i . Then obtain the probability matrix {P ij}, 1 ≤ i ≤ n, 1 ≤ j ≤ m.

[0017] Further, step 5 includes:

[0018] According to the weighting of the probability matrix by the influence coefficient, that is, calculate the weighted probability W j of the j-th working mode, that is:

[0019]

[0020] where A i is the user User iThe influence coefficient; determine the working mode ModeMax with the largest weighted probability, and use the working parameters of all non - continuous variables included in this working mode ModeMax as the working parameters of the determined device D.

[0021] Further, the step 7 includes:

[0022] For the user User i , multiply the corresponding normal distribution curve by its influence coefficient A i , obtain the weighted normal distribution curve, add all n weighted normal distribution curves to get a final probability curve; take the point on this probability curve with the largest ordinate, and its corresponding abscissa is the parameter value with the largest probability, and use this parameter value as the working parameter of the determined device D.

[0023] Further, the external data includes meteorological data.

[0024] Further, the user status includes the user code in the current residence and the location where the user is located.

[0025] Further, according to the device type, the residential layout, and a predetermined influence range division method, determine the influence range of device D.

[0026] Further, through steps 6 - 7, determine all the working parameters of device D, and the combination of these working parameters constitutes the predicted working mode of device D.

[0027] Further, the smart home devices include televisions, air conditioners, lighting fixtures, curtains, and monitoring devices.

[0028] The present invention also provides an artificial - intelligence - based smart home management system, which includes a smart home platform and multiple smart home devices, and the smart home platform can manage the smart home devices according to the above - mentioned method.

[0029] The beneficial effects of the present invention are: in a multi - user scenario, it intelligently determines the working modes of each device in the smart home system to meet the needs of different users as much as possible, thereby reducing the possibility of multi - user demand conflicts.

BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The drawings described herein are used to provide a further understanding of the present invention, form a part of this application, but do not constitute an improper limitation to the present invention. In the drawings:

[0031] Figure 1 is the basic structure diagram of the smart home management system of the present invention.

DETAILED DESCRIPTION OF THE EMBODIMENTS

[0032] The present invention will be described in detail below in conjunction with the accompanying drawings and specific embodiments. The illustrative embodiments and descriptions are only used to explain the present invention, but not to limit the present invention.

[0033] Refer to the attached Figure 1 , which shows the basic structure diagram of the smart home management system of the present invention. The system includes a smart home platform and multiple smart home devices. The smart home platform and smart home devices are all arranged within the user's residence and communicate with each other through the home internal network.

[0034] The smart home devices can be various intelligent devices within the residence and support the unified management of the smart home system. For example, intelligent TVs, air conditioners, lighting fixtures, curtains, monitoring devices, etc. These devices can have different working modes. Taking the air conditioner as an example, the air conditioner can have a cooling mode, a heating mode, a dehumidifying mode, a sleep mode, a shutdown mode, etc. For the cooling mode of the air conditioner, each set cooling temperature can also be considered as a finer working mode. For example, in the cooling mode, the air conditioner can be set to 22 degrees or 23 degrees, which can also be regarded as two working modes. The working modes of the smart home devices can be actively adjusted according to the current residence state and user state, reflecting their intelligent features.

[0035] The smart home platform is a platform for unified management of each smart home device. It can be a server within the home or any other type of control device. With the user's authorization and consent, it can also be a remote server. In the present invention, the smart home platform runs an artificial intelligence model, and the model receives the current residence state and user state to determine the working modes of each smart home device.

[0036] Based on the above smart home management system, the specific steps of the smart home management method of the present invention will be described in detail below.

[0037] Step 1: Collect historical data of the smart home management system within the residence.

[0038] Specifically, for a residence installed with the smart home management system of the present invention, within a period of time after installation, collect system operation data, external data, and user status to form the historical data.

[0039] The system operation data includes the working modes of each smart home device within the system. In specific implementation, the device manufacturer can provide a unique code for any possible working mode of the device. Therefore, the smart home platform only needs to receive and record the unique code of the working mode executed by each device and the corresponding time.

[0040] The external data may include, for example, meteorological data. The meteorological data can be queried and downloaded from the Internet by the smart home platform. The external data may also include other types of relevant data, which are not limited in the present invention.

[0041] The user status includes the users in the current residence and their locations. To obtain the user status, multiple monitoring devices can be installed in the residence. The smart home platform obtains monitoring images from each monitoring device and determines each user in the residence and their locations through face recognition. The smart home platform can encode each different face recognized as a user code, and store the user code, the user's location, and the current time correspondingly as a user status data. It should be noted that the smart home platform performs face recognition and stores the user status, all of which should be completed under the premise of the user's authorization and consent.

[0042] By collecting historical data for a period of time, the smart home platform can obtain multiple historical data. Each piece of historical data includes time, the working modes of each device at that time, the external data at that time, and the user status at that time.

[0043] Step 2: For any smart home device D, construct a corresponding working mode prediction model M D , construct a training sample set based on the historical data, and train the model M D for training.

[0044] Specifically, after the smart home platform has collected sufficient historical data related to device D, based on a pre-provided artificial intelligence model, construct a corresponding working mode prediction model M D . The model M D inputs the working modes, external data, and a single user code of all current smart home devices (except device D), and predicts the working mode of device D corresponding to this user code. In fact, the model M D outputs the probabilities of all working modes corresponding to this user code, and takes the working mode with the maximum probability as the prediction result. Simply put, this model M D can, for any user, based on the current system status and external data, determine the working mode of device D. As a single-user model, its construction and training are relatively simple, which can improve the overall efficiency.

[0045] To train this model M D, historical data related to device D needs to be collected as training samples. As mentioned above, each piece of historical data includes time, the working modes of each device at that time, the external data at that time, and the user status at that time; based on this, if a piece of historical data includes the working mode data of device D, then obtain the working mode data DevicesData of other devices except device D, the working mode ModeD of device D, the external data OtherData, and then based on the user status in this piece of historical data, obtain the user code UserID of the user who is most relevant to device D at that time. Thus, a training sample <DevicesData, OtherData, UserID, ModeD> is obtained, where ModeD is used as the sample label.

[0046] Among them, in order to determine the user who is most relevant to device D at that time, based on the positions of each user in the user status, the user with the closest distance to device D can be determined as the user who is most relevant to it. If there are multiple users with the closest distance to device D, each of them can be regarded as the user who is most relevant, and a corresponding training sample is constructed for each user who is most relevant, so as to obtain multiple corresponding training samples. The distance here can be only a rough estimate and does not need to be too precise. For example, users in the same room as device D can be regarded as users with the closest distance.

[0047] After obtaining a sufficient training sample set, the smart home platform trains the model M D to obtain a trained working mode prediction model. Based on the same method, for each smart home device, a corresponding working mode prediction model can be trained.

[0048] Step 3: When there are multiple users in the residence, the smart home platform identifies each user and determines the influence coefficient of device D on each user in the current residence based on the influence range of device D.

[0049] The present invention mainly solves the problem of smart home management in a multi-user scenario. If it is a single-user scenario, the model M can be directly used DTo determine the working mode of device D, which will not be elaborated here. In a multi-user scenario, the smart home platform first performs face recognition on each user to determine the user code of each user. Then, the smart home platform determines the influence range of device D according to the location where device D is located. For example, if device D is a TV, its influence range is mainly in the room where device D is located, with little influence on adjacent rooms and no influence on non-adjacent rooms (i.e., not within the influence range of device D). If device D is an air conditioner, its main influence range is in the room where device D is located, with a certain influence on adjacent rooms, and the influence on farther rooms attenuates with distance. More fine-grained range division can also be adopted. For example, the main influence range of the TV can be restricted to a fan-shaped area in front of it, and the influence on other areas of the room where it is located is smaller than that of this fan-shaped area. In short, the smart home platform can determine the influence range of device D according to the device type, residential layout, and predetermined influence range division method.

[0050] Then, the smart home platform can determine the influence coefficient of device D on each user in the current residence according to the influence range of device D. The influence coefficient is used to determine the influence of device D on the location where the user is located. For example, if a user is not within the influence range of device D, the influence coefficient of device D on this user is 0; if a user is within the main influence range of device D, the influence coefficient of device D on this user is 1; if a user is within the range where device D has a small influence, an influence coefficient between 0 and 1 can be determined according to the magnitude of the influence. The influence coefficient can be calculated according to the management strategy predetermined by the device manufacturer, or the user can set the calculation strategy of the influence coefficient by themselves.

[0051] Step 4: For any user with a non-zero influence coefficient, according to model M D Calculate the probabilities of all working modes of this device D.

[0052] As mentioned above, model M D Actually outputs the probabilities of all working modes of device D for a user, and then takes the working mode with the maximum probability as the prediction result. Then, assume there are n users User1, User2,..., User n , and device D has m working modes. Then for any user User i , model M D Can output the probabilities of all working modes of device D (a total of m probabilities, and their sum is 1), thus obtaining a total of mn probability values.

[0053] Let P ij Be the probability of predicting the jth working mode of device D for User i , then the probability matrix {P ij}(1 ≤ i ≤ n, 1 ≤ j ≤ m).

[0054] Step 5: If the operating mode of device D includes operating parameters of discontinuous variables, determine the operating parameter according to the weighting of the probability matrix by the influence coefficient.

[0055] The operating mode described in the present invention is a combination of multiple operating parameters. Taking an air conditioner as an example, cooling at 22 degrees, cooling at 23 degrees, heating at 26 degrees, and heating at 27 degrees are four different operating modes. Among them, there are two operating parameters. The first operating parameter is to select one of the cooling mode, heating mode, dehumidification mode, sleep mode, etc., and the second operating parameter is to select a specific temperature. Among them, the second operating parameter temperature is a continuously variable, usually continuously variable and adjustable between 16 - 30 degrees, that is, the second operating parameter is a continuous variable. And the first operating parameter is not a continuously variable, so it is an operating parameter of a discontinuous variable.

[0056] First, it is necessary to determine the operating parameters of discontinuous variables. According to the weighting of the probability matrix by the influence coefficient, that is, calculate the weighted probability W of the jth operating mode j , that is:

[0057]

[0058] where A i is the influence coefficient of user User i . Thus, the operating mode ModeMax with the largest weighted probability can be determined, and all the operating parameters of discontinuous variables included in the operating mode ModeMax are used as the determined operating parameters of device D.

[0059] Step 6: If the operating mode of device D also includes operating parameters of continuous variables, then process the operating parameter predicted by the model M D for each user into a normal distribution curve with the predicted value as the mean.

[0060] As mentioned above, the model M D actually outputs the probabilities of all operating modes of device D for a user, and then takes the operating mode with the maximum probability as the prediction result. The operating mode with the maximum probability also includes the operating parameters of the continuous variable, which is the predicted value of the operating parameter of the model M D for this user. For example, for an air conditioner, if the predicted operating mode for a certain user is cooling at 22 degrees, then 22 is the predicted value for this user and is an operating parameter of a continuous variable. Taking this predicted value (such as 22) as the mean (i.e., the mean μ = 22 of the normal distribution), and a predetermined variance, a normal distribution curve is established.

[0061] In this way, for the above n users, n normal distribution curves can be established. Through the normal distribution curves, the probability of each user choosing a temperature is described.

[0062] Step 7: Determine the working parameters of the continuous variable according to the weighted superposition of the normal distribution curves by the influence coefficient.

[0063] Specifically, for user User i , multiply the corresponding normal distribution curve by its influence coefficient A i , to obtain a weighted normal distribution curve. Add up all the n weighted normal distribution curves to get a final probability curve. Take the point on this probability curve with the largest ordinate, and its corresponding abscissa is the parameter value with the highest probability. Use this parameter value as the working parameter of device D determined.

[0064] Through the above Steps 6 - 7, all the working parameters of device D can be determined. The combination of these working parameters constitutes a working mode of device D. Take this working mode as the predicted working mode of device D in the multi - user scenario of the present invention. Thus, the smart home platform can notify device D to automatically switch to this working mode, or switch to this working mode after obtaining the consent of the user.

[0065] The above steps have detailed the smart home management method based on artificial intelligence of the present invention. This method can intelligently determine the working modes of each device in the smart home system in a multi - user scenario to meet the needs of different users as much as possible.

[0066] The above description is only a preferred embodiment of the present invention. Therefore, any equivalent changes or modifications made according to the structures, features, and principles described in the scope of the patent application of the present invention are included in the scope of the patent application of the present invention.

Claims

1. A smart home management method based on artificial intelligence, characterized in that, The method includes: Step 1: Collect historical data of the smart home management system in the residence; the historical data includes the working modes of various smart home devices, external data, and user status; Step 2: For any smart home device D, construct a corresponding working mode prediction model M D , construct a training sample set based on the historical data, and train the model M D ; wherein, the model M D inputs the working modes of all smart home devices other than this device D, external data, and a single user code, and predicts the working mode of this device D corresponding to this user code; Step 3: When there are multiple users in the residence, the smart home platform identifies each user and determines the influence coefficient of device D on each user in the current residence based on the influence range of device D; the influence coefficient is used to determine the influence of device D on the location of the user; Step 4: For any user with a non-zero influence coefficient, according to model M D Calculate the probabilities of all working modes of the device D, and all the obtained probabilities form the probability matrix of the device D; Step 5: If the working mode of device D includes working parameters of non - continuous variables, determine the working parameter according to the weighting of the probability matrix by the influence coefficient; Step 6: If the working mode of device D also includes working parameters of continuous variables, then model M D The working parameter predicted for each user is processed into a normal distribution curve with the predicted value as the mean value. Step 7: Determine the working parameter of the continuous variable according to the weighted superposition of the normal distribution curve by the influence coefficient.

2. The method according to claim 1, wherein The specific content of step 4 includes: Suppose there are n users User1, User2, ……, User with non-zero influence coefficients n , and the device D has m working modes. Then for any user User i , obtain the probabilities of all working modes of the device D output by the model M D ; Let P ij be the probability of the j-th working mode of the prediction device D for User i . Then the probability matrix {P ij} of device D is obtained, where 1 ≤ i ≤ n and 1 ≤ j ≤ m.

3. The method according to claim 2, wherein Step 5 includes: Weighting the probability matrix according to the influence coefficient, that is, calculating the weighted probability W of the j-th working mode j , that is: Among which A i is the influence coefficient of user User i ; determine the working mode ModeMax with the maximum weighted probability, and use the working parameters of all non - continuous variables included in this working mode ModeMax as the working parameters of the determined device D.

4. The method according to claim 3, characterized in that, Step 7 includes: For user User i , multiply the corresponding normal distribution curve by its influence coefficient A i , to obtain a weighted normal distribution curve. Add all n weighted normal distribution curves to obtain a final probability curve; take the point on this probability curve with the largest ordinate, and the corresponding abscissa is the parameter value with the highest probability. Take this parameter value as the operating parameter of the determined device D.

5. The method according to any one of claims 1-4, characterized in that, The external data includes meteorological data.

6. The method according to any one of claims 1-5, characterized in that, The user status includes the user code in the current residence and the location of the user.

7. The method according to any one of claims 1-6, characterized in that, Determine the influence range of device D according to the device type, residence layout, and a predetermined influence range division method.

8. The method according to claim 1, characterized in that Through steps 6 - 7, determine all the working parameters of device D, and the combination of these working parameters constitutes the predicted working mode of device D.

9. The method according to claim 1, wherein The smart home devices include televisions, air conditioners, lighting fixtures, curtains, and monitoring devices.

10. A smart home management system based on artificial intelligence, characterized in that, The system includes a smart home platform and multiple smart home devices, and the smart home platform manages the smart home devices according to the method described in any one of claims 1 - 9.

Images