Cleaner and control method therefor

Abstract

This cleaner comprises: multiple sensors for detecting the operation of the cleaner; an input device for receiving a user command to adjust the suction strength of the cleaner; a memory for storing a pre-trained neural network model and user usage pattern information related to the operation of the cleaner; and a processor for controlling the cleaner by determining suction strength to be applied to the cleaner by using at least one among the usage pattern information, sensor information detected by multiple sensors, and the pre-trained neural network model.

Description

Vacuum cleaner and method of controlling the same

[0001] The present disclosure relates to a vacuum cleaner and a control method thereof, and more particularly, to a vacuum cleaner and a control method thereof capable of adjusting the suction power of a motor based on a learned neural network model and a user usage pattern.

[0002] A vacuum cleaner comprises a main body, which houses a vacuum suction device and a dust collector, and a suction module connected to the main body. Recently, rotating brushes have been installed in the suction module to facilitate the suction of foreign substances from the surface being cleaned.

[0003] Vacuum cleaners operate on a variety of surfaces to be cleaned, and different suction power and brush operation conditions need to be applied to each surface to be cleaned.

[0004] In order to achieve the above-described purpose, a vacuum cleaner according to the present disclosure includes a plurality of sensors for detecting the operation of the vacuum cleaner, a memory for storing a pre-learned neural network model and user usage pattern information related to the operation of the vacuum cleaner, and a processor for controlling the vacuum cleaner by determining a suction strength to be applied to the vacuum cleaner.

[0005] The processor may determine a suction strength to be applied to the vacuum cleaner by using at least one of the usage pattern information, sensor information detected by the plurality of sensors, and a pre-learned neural network model.

[0006] The above-mentioned pre-learned neural network model is a model that individually outputs probability values ​​for multiple floor types based on sensor information detected by at least one sensor among the plurality of sensors, and the processor inputs the sensor information detected by the plurality of sensors into the pre-learned neural network model to confirm the floor type, and can determine the suction strength to be applied to the vacuum cleaner using the confirmed floor type and the usage pattern information.

[0007] The vacuum cleaner further includes a driving device that controls a motor that provides driving force to a brush of the vacuum cleaner, and the processor can determine a rotation speed of the brush using the identified floor type and the usage pattern information, and control the driving device so that the brush rotates at the determined rotation speed.

[0008] When a user command for adjusting the suction strength of the cleaner is input while the cleaner is operating at the determined suction strength, the processor can adjust the current suction strength to a suction strength corresponding to the user command and store the adjusted suction strength and the identified floor type in the memory.

[0009] The processor can modify the usage pattern information using the suction strength stored in the memory and the identified floor type.

[0010] The above usage pattern information may include at least one of individual user preferred suction strength, brush speed information, and protection mode application information for multiple floor types.

[0011] The above-mentioned pre-learned neural network model is a first neural network model, the memory stores the first neural network model and the second neural network model that output probability values ​​for each of a plurality of floor types, the processor inputs sensor information detected by the plurality of sensors into the first neural network model to obtain first probability information, inputs sensor information detected by the plurality of sensors into the second neural network model to obtain second probability information, and determines the floor type based on the obtained first probability information and second probability information.

[0012] The processor can check the highest probability value among the first probability information and the second probability information, and determine the floor type corresponding to the checked highest probability value as the current floor type.

[0013] The above multiple floor types may include general floor, carpet short pile, carpet medium pile, and carpet long pile.

[0014] The processor may determine a suction strength having a lower suction power than a suction power when the identified floor type is a carpet medium pile or a carpet long pile, and the usage pattern information includes carpet protection information.

[0015] The above-mentioned pre-learned neural network model is a model that is learned using sensor information and usage pattern information detected from the plurality of sensors and outputs a suction strength to be applied to the vacuum cleaner, and the processor can input sensor information detected from the plurality of sensors into the pre-learned neural network model to determine a suction strength to be applied to the vacuum cleaner.

[0016] When a user command for adjusting the suction strength of the cleaner is input while the cleaner is operating at the determined suction strength, the processor adjusts the current suction strength to a suction strength corresponding to the user command, stores the user command and sensor information detected by the plurality of sensors in the memory, and retrains the pre-learned neural network model using the user command and the sensor information stored in the memory.

[0017] The above-mentioned pre-learned neural network model is a first neural network model, and the memory stores the first neural network model that receives sensor information and outputs floor type information and a second neural network model that outputs suction power based on the type information, and the processor inputs sensor information detected by the plurality of sensors into the first neural network model to confirm a floor type, and inputs the confirmed floor type into the second neural network model to confirm a suction strength to be applied to the vacuum cleaner.

[0018] The above plurality of sensors include an acceleration sensor that detects the movement state of the vacuum cleaner, and the processor can determine an increased suction strength compared to the current suction strength when it is determined that the vacuum cleaner is repeatedly moving in the same area.

[0019] The plurality of sensors may include at least one of a first sensor that detects a suction pressure of the cleaner, a second sensor that detects a current supplied to a motor that drives a brush, a third sensor that detects an output value of the motor that drives the brush, and a fourth sensor that detects a rotation speed of the brush.

[0020] A method for controlling a vacuum cleaner according to one embodiment of the present disclosure includes a step of detecting the operation of the vacuum cleaner using a plurality of sensors, a step of determining a suction strength to be applied to the vacuum cleaner using sensor information detected by the plurality of sensors, a pre-learned neural network model, and pre-stored usage pattern information, and a step of controlling a motor of the vacuum cleaner using the determined suction strength.

[0021] The above-mentioned pre-learned neural network model is a model that outputs a probability value for each of a plurality of floor types based on sensor information detected by a sensor, and the step of determining the suction strength may include a step of inputting sensor information detected by the plurality of sensors into the pre-learned neural network model to confirm a floor type, and a step of confirming a suction strength to be applied to the vacuum cleaner using the confirmed floor type and the usage pattern information.

[0022] The control method may further include a step of receiving a user command for adjusting the suction strength of the cleaner while it is operating at the determined suction strength, a step of adjusting the current suction strength to a suction strength corresponding to the user command based on the user command, a step of storing the adjusted suction strength and the identified floor type in a memory, and a step of modifying the usage pattern information using the suction strength and the identified floor type stored in the memory.

[0023] The step of determining the suction strength may include a step of inputting the detected sensor information into a first neural network model that receives sensor information and outputs floor type information to confirm the floor type, and a step of inputting the confirmed type information into a second neural network model that outputs suction power based on the type information to confirm the suction strength to be applied to the vacuum cleaner.

[0024] In a non-transitory computer-readable recording medium storing a program for executing a control method for a vacuum cleaner according to an embodiment of the present disclosure, the control method includes a step of detecting an operation of the vacuum cleaner using a plurality of sensors, a step of determining a suction strength to be applied to the vacuum cleaner using sensor information detected by the plurality of sensors, a pre-learned neural network model, and pre-stored usage pattern information, and a step of controlling a motor of the vacuum cleaner using the determined suction strength.

[0025] The above and other aspects, features, and advantages of the embodiments of the present disclosure will become more apparent from the following description taken in conjunction with the accompanying drawings. In the accompanying drawings:

[0026] FIG. 1 is a drawing for explaining the operation of a vacuum cleaner according to one embodiment.

[0027] Figure 2 is a block diagram showing the configuration of a vacuum cleaner according to one embodiment.

[0028] Figure 3 is a block diagram showing the configuration of a vacuum cleaner according to one embodiment.

[0029] Figure 4 is a diagram illustrating examples of information collected from sensors in various floor environments.

[0030] FIG. 5 is a diagram illustrating an example of a configuration of a neural network model according to an embodiment of the present disclosure;

[0031] FIG. 6 is a diagram for explaining an example of a configuration of a neural network model according to an embodiment of the present disclosure;

[0032] FIG. 7 is a diagram illustrating an example of a configuration of a neural network model according to an embodiment of the present disclosure;

[0033] FIG. 8 is a diagram illustrating an example of a configuration of a neural network model according to an embodiment of the present disclosure;

[0034] FIG. 9 is a diagram for explaining an example of a configuration of a neural network model according to an embodiment of the present disclosure;

[0035] FIG. 10 is a diagram for explaining an example of a configuration of a neural network model according to an embodiment of the present disclosure;

[0036] FIG. 11 is a diagram illustrating an example of a configuration of a neural network model according to an embodiment of the present disclosure;

[0037] FIG. 12 is a diagram for explaining an example of a configuration of a neural network model according to an embodiment of the present disclosure;

[0038] FIG. 13 is a drawing showing an example of a display on a vacuum cleaner according to an embodiment of the present disclosure;

[0039] FIG. 14 is a drawing showing an example of a display on a vacuum cleaner according to an embodiment of the present disclosure;

[0040] FIG. 15 is a drawing showing an example of changing the suction force according to one embodiment of the present disclosure;

[0041] FIG. 16 is a flowchart for explaining a method for controlling a vacuum cleaner according to an embodiment of the present disclosure;

[0042] FIG. 17 is a flowchart for explaining a method for controlling suction power and brush operation of a vacuum cleaner according to an embodiment of the present disclosure;

[0043] FIG. 18 is a flowchart for explaining the learning operation of a neural network model according to an embodiment of the present disclosure, and

[0044] FIG. 19 is a flowchart for explaining a control operation using information of an external device according to one embodiment of the present disclosure.

[0045] The present embodiments may be modified and have various embodiments, and specific embodiments are illustrated and described in detail in the drawings. However, this is not intended to limit the scope to specific embodiments, and it should be understood that various modifications, equivalents, and / or alternatives of the embodiments of the present disclosure are included. In connection with the description of the drawings, similar reference numerals may be used for similar components.

[0046] In describing the present disclosure, if it is determined that a specific description of a related known function or configuration may unnecessarily obscure the gist of the present disclosure, a detailed description thereof will be omitted.

[0047] Additionally, the following embodiments may be modified in various other forms, and the scope of the technical concepts of the present disclosure is not limited to the following embodiments. Rather, these embodiments are provided to further faithfully and completely convey the technical concepts of the present disclosure to those skilled in the art.

[0048] The terminology used in this disclosure is for the purpose of describing specific embodiments only and is not intended to limit the scope of the rights. Singular expressions include plural expressions unless the context clearly dictates otherwise.

[0049] In this disclosure, expressions such as “has,” “can have,” “includes,” or “may include” indicate the presence of a corresponding feature (e.g., a component such as a number, function, operation, or part), and do not exclude the presence of additional features.

[0050] In this disclosure, expressions such as “A or B,” “at least one of A and / or B,” or “one or more of A or / and B” can include all possible combinations of the listed items. For example, “A or B,” “at least one of A and B,” or “at least one of A or B” can all refer to instances where (1) at least one A is included, (2) at least one B is included, or (3) at least one A and at least one B are included.

[0051] The expressions “first,” “second,” “first,” or “second,” etc., used in this disclosure can describe various components, regardless of order and / or importance, and are only used to distinguish one component from another, but do not limit the components.

[0052] When it is said that a component (e.g., a first component) is “(operatively or communicatively) coupled with / to” or “connected to” another component (e.g., a second component), it should be understood that the component may be directly coupled to the other component, or may be coupled via another component (e.g., a third component).

[0053] On the other hand, when it is said that a component (e.g., a first component) is "directly connected" or "directly connected" to another component (e.g., a second component), it can be understood that no other component (e.g., a third component) exists between said component and said other component.

[0054] The expression "configured to" as used in the present disclosure may be used interchangeably with, for example, "suitable for," "having the capacity to," "designed to," "adapted to," "made to," or "capable of." The term "configured to" may not necessarily mean only "specifically designed to" in terms of hardware.

[0055] Instead, in some contexts, the phrase "a device configured to" may mean that the device, in conjunction with other devices or components, is "capable of" performing A, B, and C. For example, the phrase "a processor configured (or set) to perform A, B, and C" may refer to a dedicated processor (e.g., an embedded processor) for performing those operations, or a general-purpose processor (e.g., a CPU or application processor) that can perform those operations by executing one or more software programs stored in a memory device.

[0056] In the embodiments, a 'module' or 'part' performs at least one function or operation, and may be implemented as hardware or software, or as a combination of hardware and software. Furthermore, a plurality of 'modules' or 'parts' may be integrated into at least one module and implemented as at least one processor, except for a 'module' or 'part' that needs to be implemented as a specific hardware.

[0057] According to various embodiments, operations performed by a module, program or other component may be executed sequentially, in parallel, iteratively or heuristically, or at least some operations may be executed in a different order, omitted, or other operations may be added.

[0058] Meanwhile, the various elements and areas in the drawings are schematically drawn. Therefore, the technical concept of the present invention is not limited by the relative sizes or spacing depicted in the attached drawings.

[0059] Meanwhile, a cleaner according to various embodiments of the present disclosure may include at least one of, for example, a vacuum cleaner, a robot cleaner, a handy cleaner, a stick cleaner, a mop cleaner, etc.

[0060] Hereinafter, with reference to the attached drawings, embodiments according to the present disclosure will be described in detail so that a person having ordinary knowledge in the technical field to which the present disclosure pertains can easily implement the present disclosure.

[0061] FIG. 1 is a drawing for explaining the operation of a vacuum cleaner according to one embodiment of the present disclosure.

[0062] Referring to FIG. 1, a vacuum cleaner (100) according to an embodiment of the present disclosure has a stick-type vacuum cleaner (or upright type). In the illustrated example, an upright type shape in which a suction module is formed integrally with the main body is depicted. However, when implemented, a canister type in which the suction module is provided separately from the main body and connected by an extension tube may be used, and various types of vacuum cleaners such as a wireless control vacuum cleaner, a robot vacuum cleaner, and a handheld vacuum cleaner may be used.

[0063] A vacuum cleaner (100) may include a vacuum cleaner body (10) and a suction head (30). Here, the vacuum cleaner body (10) refers to a component including major components such as a motor and a processor. This concept of a body assumes the separation of components such as a suction head (30) or a stick (20). However, if the above-described components have an integrated form, these components may also be referred to as a body.

[0064] The vacuum cleaner (100) may include a stick (20) connecting the vacuum cleaner body (10) and the suction head (30) and a handle (40) connected to the vacuum cleaner body (10). The stick (20) connects the vacuum cleaner body (10) and the suction head (30), and, if necessary, the stick (20) and the vacuum cleaner body (10) may be detachable. In addition, the vacuum cleaner body (10) may have a cleaning tool other than the above-described stick (20) or suction head (30) attached to it.

[0065] The handle (40) is a part that is connected to the vacuum cleaner body (10) and can be provided so that the user can hold it and operate the vacuum cleaner (100).

[0066] The handle (40) may be provided with an operating section (not shown) to allow the user to control the vacuum cleaner (100). Alternatively, during implementation, a screen related to the operation of the vacuum cleaner (100) or an operating area for receiving user control commands may be formed on the upper portion of the vacuum cleaner body (10). Examples of this will be described later with reference to FIGS. 13 and 14.

[0067] The vacuum cleaner body (10) may include a dust collector (11) and a driving device (12) placed inside it. The dust collector (11) may perform the function of collecting dust by separating foreign substances from the air sucked in by the suction head (30).

[0068] The driving device (12) may include a motor assembly (50) that generates suction pressure of the cleaner (100). The motor assembly (50) may generate power to generate suction force inside the cleaner body (10).

[0069] The motor assembly (50) includes a first motor, and can generate suction pressure through the rotation of the first motor. Specifically, when a drive command for the first motor is input and power is supplied to the first motor, the impeller rotates by the driving of the first motor. Suction pressure is generated by the rotation of the impeller, and air containing foreign substances can be sucked into the suction port by this suction pressure. In addition, as the rotation speed of the first motor increases, the suction pressure increases.

[0070] When the vacuum cleaner (100) or the user sets (or determines) the suction strength, the vacuum cleaner (100) can control the first motor to rotate at a rotation speed corresponding to the set suction strength described above.

[0071] A suction head (30) may be provided at the lower portion of the cleaner body (10) and positioned so as to be in contact with the surface to be cleaned. The suction head (30) may be provided so as to be in contact with the surface to be cleaned and to draw dust or contaminants from the surface to be cleaned into the interior of the cleaner body (10) using suction force generated from the motor assembly (50).

[0072] A suction head (30) like this may include a brush and a second motor. Specifically, when a drive command for the second motor is input and power is supplied to the second motor, the brush may be rotated by the drive of the second motor. The second motor may be various motors such as a DC (Direct Current) motor, an AC (Alternating Current) motor, a BLDC (Brushless DC) motor, etc. Meanwhile, a drive device for providing drive power to the second motor described above may be provided on the suction head (30) side during implementation, and in some cases, a drive device for controlling the second motor may be provided on the cleaner body (10).

[0073] The brush is formed to extend a certain length outside the suction port, so that when the brush rotates, it can strike foreign substances such as dust, dirt, and hair stuck to the surface to be cleaned. This allows the foreign substances to be separated from the surface to be cleaned and easily sucked up by the suction port. Such brushes may be made of a material with a low coefficient of friction and good durability, such as natural hair or PA (polyamide), but are not necessarily limited thereto.

[0074] Meanwhile, it is desirable that the suction strength and the rotation speed of the brush described above be adjusted to corresponding values ​​depending on the type of surface to be cleaned.

[0075] For example, when operating at maximum suction power for all types of surfaces to be cleaned, the floor surface may stick to the suction head (30) on certain floor surfaces, making cleaning difficult. In addition, rapid brush rotation on floor surfaces such as carpets may cause damage to the carpet. In this regard, the vacuum cleaner needs to identify the type of surface to be cleaned and operate at an appropriate suction strength and brush rotation speed.

[0076] The commonly used cleaning surfaces can be categorized into general floors, raised floors, mats, carpet short pile (carpet pile length of 4 mm or less), carpet medium pile (carpet pile length of 4 to 16 mm), and carpet long pile (carpet pile length of 16 mm or more). Here, raised pile may include cases where the floor board can stick to the suction head (30) of the vacuum cleaner due to high suction strength, such as a floor board. A carpet is a thick woolen fabric with hairs such as wool woven to create a fluff on the surface. General floors may be cases other than the above-mentioned mats, carpets, and raised floors.

[0077] This disclosure assumes and describes the use of the six types of cleaning surfaces described above. However, during implementation, only some of the types described above may be utilized, and other types of flooring may be utilized in addition to the examples described above. For example, general flooring may be classified by type of flooring material (marble, wood, etc.).

[0078] As described above, even if the type of surface to be cleaned is accurately identified, if cleaning is not performed with the suction strength and brush rotation speed that match the user's intention, the user may feel uncomfortable with the vacuum cleaner.

[0079] For example, when cleaning a carpet, some users may want to clean it in a way that minimizes damage to the carpet, while other users may want to clean it thoroughly without causing any damage to the carpet.

[0080] Therefore, it is desirable to use a vacuum cleaner by determining the suction strength and brush rotation speed by considering not only the type of floor surface but also the user's usage intention (or the user's usage pattern).

[0081] Below, we explain how to determine the suction strength (or brush rotation speed) of a vacuum cleaner by taking into account both the type of floor and the user's usage pattern.

[0082] FIG. 2 is a drawing for explaining the configuration of a vacuum cleaner according to an embodiment of the present disclosure.

[0083] Referring to FIG. 2, the vacuum cleaner (100) may include a plurality of sensors (110), an input device (120), a memory (130), and a processor (140).

[0084] A plurality of sensors (110) can detect the operation of the cleaner (100) and generate detection information. For example, the plurality of sensors (110) may include a first sensor that detects the suction pressure of the cleaner, a second sensor that detects the current supplied to the motor that drives the brush, a third sensor that detects the output value of the motor that drives the brush, and a fourth sensor that detects the rotation speed of the brush. In addition, the plurality of sensors (110) may also include an acceleration sensor (or gyro sensor) that detects the movement of the cleaner.

[0085] The sensing operation of the plurality of sensors (110) can be performed based on the control command of the processor (140) described below, and can be automatically measured in preset cycle units and provided to the processor (140). At this time, the information of each sensor (or sensing value) can be used as information measured at the corresponding moment, or the average value of the corresponding cycle unit can be used. In addition, the preset cycle unit described above can be 20 ms, but is not limited thereto.

[0086] Meanwhile, the first sensor, fourth sensor, etc. described above may be placed within the suction head (30) of Fig. 1, and the second sensor, third sensor, acceleration sensor, etc. may be placed on the side of the cleaner body (10). Such an arrangement example is merely an example and may be changed upon implementation.

[0087] In addition, although five sensors were used as examples above, other sensors (e.g., lidar sensors, ultrasonic sensors, etc.) may be additionally used in addition to the sensors described above during implementation, and some of the sensors described above may be omitted.

[0088] The input device (120) receives user commands. These user commands may be commands for turning the vacuum cleaner on / off, commands for adjusting the suction strength of the vacuum cleaner, cleaning modes of the vacuum cleaner (e.g., AI mode, AI protection mode, AI clean mode, protection mode, or protection mode release), etc.

[0089] Meanwhile, although the cleaning command and suction strength, etc. are described as being input directly from the vacuum cleaner (100) in the above, they may be input through a separate external device (e.g., a user terminal device, etc.) during implementation.

[0090] The memory (130) may store at least one instruction regarding the vacuum cleaner (100). In addition, the memory (130) may store an O / S (Operating System) for driving the vacuum cleaner (100). Such instructions may store instructions for identifying the floor type described below, instructions for determining the cleaning suction strength, instructions for controlling various components of the vacuum cleaner, instructions for retraining a pre-stored neural network model, etc.

[0091] The memory (130) may include a semiconductor memory such as a flash memory or a magnetic storage medium such as a hard disk. For example, various software modules for operating the cleaner (100) according to various embodiments of the present disclosure may be stored in the memory (130), and the processor (140) may execute various software modules stored in the memory (130) to control the operation of the cleaner (100). That is, the memory (130) is accessed by the processor (140), and data reading / writing / modifying / deleting / updating, etc. may be performed by the processor (140).

[0092] Meanwhile, in the present disclosure, the term memory (130) may be used to mean a memory (130), a ROM, a RAM in a processor (140), or a memory card (e.g., a micro SD card, a memory stick) mounted in a cleaner (100).

[0093] The memory (130) can store a pre-trained neural network model. For example, the neural network model can be implemented as a CNN (Convolutional Neural Network), LSTM (Long Short-Term Memory), DNN (Deep Neural Network), RNN (Recurrent Neural Network), RBM (Restricted Boltzmann Machine), DBN (Deep Belief Network), BRDNN (Bidirectional Recurrent Deep Neural Network), etc., but is not limited to these examples.

[0094] These neural network models are computing systems implemented based on the neural networks of human or animal brains, and may also be referred to as learning models, machine learning models, artificial intelligence models, or deep learning models.

[0095] Meanwhile, the neural network model used in the present disclosure can be implemented in various forms, which are described in detail in FIGS. 5 to 12.

[0096] And the memory (130) may include user usage pattern information. Such user pattern information may include user preferred suction strength, brush speed information, and protection mode application information for each of a plurality of floor types.

[0097] For example, the user pattern information may store information about the suction strength, brush speed, etc. preferred by the user for each floor type, such as information such as suction strength 3, brush speed 4, and no protection mode applied on a normal floor, or suction strength 2, brush speed 2, and protection mode applied on a single-pile carpet. Meanwhile, such user pattern information may initially use the default values ​​provided by the manufacturer, and may be updated through the process described below according to the user's use.

[0098] Additionally, the user preference information described above can be stored separately for each user. For example, in an environment where users can be identified, user pattern information corresponding to the identified user can be utilized in the process described below.

[0099] Meanwhile, the above described case where the user pattern information described above is stored in the form of a lookup table has been described, but when implemented, the user pattern information described above may be stored in the form of a rule base, or may be stored as a neural network model.

[0100] And the memory (130) can store learning data for retraining the neural network model or modifying usage pattern information. For example, if a change in the user's suction strength or information from an acceleration sensor that repeatedly cleans the same area is confirmed while using the floor type or suction strength determined using the neural network model, the vacuum cleaner's status, sensor information, user operation commands, etc. at that time can be stored as learning data. Meanwhile, the retraining of the neural network model described above or modification of the usage pattern information can be performed internally in the vacuum cleaner (100), or related information can be transmitted from a separate external device and retrained in the external device.

[0101] The processor (140) controls the overall operation of the vacuum cleaner (100). Specifically, the processor (140) can control the overall operation of the vacuum cleaner (100) by executing at least one instruction stored in the memory (130) as described above.

[0102] The processor (140) may be composed of one or more processors. In this case, the one or more processors (140) may include at least one of a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), and an NPU (Neural Processing Unit), but is not limited to the examples of the processor (140) described above.

[0103] CPUs are general-purpose processors capable of performing not only general calculations but also artificial intelligence calculations. Their multi-layered cache structure allows for the efficient execution of complex programs. CPUs are advantageous for serial processing, enabling organic linking of previous and subsequent calculation results through sequential calculations. General-purpose processors are not limited to the examples described above, except where specifically identified as CPUs.

[0104] A GPU is a processor for large-scale calculations, such as floating-point operations used in graphics processing. It integrates a large number of cores to perform large-scale calculations in parallel. In particular, GPUs may be advantageous over CPUs in parallel processing methods, such as convolution operations. Furthermore, GPUs may be utilized as a coprocessor (140) to supplement the functions of a CPU. Processors for large-scale calculations are not limited to the examples described above, except in cases where they are specifically referred to as GPUs.

[0105] An NPU is a processor specialized in artificial intelligence computation using artificial neural networks, and each layer of the artificial neural network can be implemented in hardware (e.g., silicon). Since NPUs are designed specifically according to the company's specifications, they have less freedom than CPUs or GPUs, but can efficiently process the AI ​​computations requested by the company. Meanwhile, as a processor specialized in AI computation, an NPU can be implemented in various forms, such as a Tensor Processing Unit (TPU), an Intelligence Processing Unit (IPU), or a Vision Processing Unit (VPU). Except as specifically designated as an NPU, an AI processor is not limited to the examples described above.

[0106] Additionally, one or more processors (140) may be implemented as a SoC (System on Chip). In this case, the SoC may further include, in addition to one or more processors (140), a memory (130), and a network interface such as a bus for data communication between the processor (140) and the memory (130).

[0107] When a plurality of processors (140) are included in a SoC (System on Chip) included in a vacuum cleaner (100), the vacuum cleaner (100) can perform operations related to artificial intelligence (e.g., operations related to learning or inference of an artificial intelligence model) by using some of the plurality of processors (140). For example, the vacuum cleaner (100) can perform operations related to artificial intelligence by using at least one of a GPU, an NPU, a VPU, a TPU, and a hardware accelerator specialized in artificial intelligence operations such as convolution operations and matrix multiplication operations among the plurality of processors (140). However, this is merely an example, and it is of course possible to process operations related to artificial intelligence by using a CPU or a general-purpose processor (140).

[0108] In addition, the vacuum cleaner (100) can perform operations related to functions related to artificial intelligence by utilizing multiple cores (e.g., dual cores, quad cores, etc.) included in a single processor (140). In particular, the vacuum cleaner (100) can perform artificial intelligence operations such as convolution operations, matrix multiplication operations, etc. in parallel by utilizing multiple cores included in the processor (140).

[0109] One or more processors (140) are controlled to process input data according to predefined operation rules or artificial intelligence models stored in the memory (130). The predefined operation rules or artificial intelligence models are characterized by being created through learning.

[0110] Here, "created through learning" means that a predefined set of behavioral rules or an AI model with desired characteristics is created by applying a learning algorithm to a large number of learning data. This learning may be performed on the device itself, where the AI ​​according to the present disclosure is implemented, or through a separate server / system.

[0111] An artificial intelligence model may be composed of multiple neural network layers. At least one layer has at least one weight value and performs its operation through the operation result of the previous layer and at least one defined operation. Examples of neural networks include a convolutional neural network (CNN), a deep neural network (DNN), a recurrent neural network (RNN), a restricted boltzmann machine (RBM), a deep belief network (DBN), a bidirectional recurrent deep neural network (BRDNN), deep Q-networks, and a transformer. The neural networks in the present disclosure are not limited to the above-described examples unless otherwise specified.

[0112] A learning algorithm is a method for training a target device (e.g., a robot) using a large amount of learning data, enabling the target device to make decisions or predictions on its own. Examples of learning algorithms include supervised learning, unsupervised learning, semi-supervised learning, or reinforcement learning. Unless otherwise specified, the learning algorithms in this disclosure are not limited to the aforementioned examples.

[0113] In particular, in one or more embodiments, the processor (140) can accurately identify the type of floor surface and perform an operation appropriate to the identified type of floor surface. Specifically, the processor (140) can perform processes related to various embodiments according to the present disclosure using a plurality of modules.

[0114] The plurality of modules may be implemented as hardware modules or software modules, and at least some of the modules may include a neural network model. For convenience of explanation, the following description will assume that all of the plurality of modules are implemented through the memory (130) and processor (140) of the vacuum cleaner (100). However, depending on the embodiment, at least some of the modules may be implemented by an external device or server.

[0115] The processor (140) determines the suction strength to be applied to the vacuum cleaner. For example, the suction strength may be determined using information on the user's usage pattern, sensor information detected by multiple sensors, etc., and the methods of utilizing the above-described information may vary.

[0116] In this regard, an embodiment of first identifying a floor type using a neural network model and then determining suction strength using the identified floor type and user pattern information is first described.

[0117] When using a neural network model that outputs a probability value for each of a plurality of floor types based on sensor information detected from a plurality of sensors, the processor (140) can input the sensor information detected from the plurality of sensors into a pre-trained neural network model to confirm the floor type.

[0118] The processor (140) can then determine the suction strength to be applied to the vacuum cleaner using the identified floor type and usage pattern information. For example, if the identified floor type is a medium-pile carpet or a long-pile carpet and the usage pattern information includes carpet protection information, the processor (140) can determine a suction strength with a lower suction strength than that of a floor identified as a normal floor.

[0119] Here, the suction strength indicates the degree of suction of the vacuum cleaner. If the vacuum cleaner is divided into multiple stages, from minimum to maximum suction power, it may be information indicating one of the stages. For example, if three suction intensities (or suction stages) are supported, the determined suction strength may be one of the three suction intensities described above. This suction strength may also be referred to as suction power, suction level, suction intensity, cleaning intensity, or cleaning power.

[0120] In addition, the processor (140) can determine the rotation speed of the brush not only by determining the suction strength, but also by using the identified floor type and usage pattern information. Once the rotation speed of the brush is determined, the processor (140) can control a driving device that drives a second motor so that the brush rotates at the determined rotation speed.

[0121] Meanwhile, when a user command to adjust the suction strength of the vacuum cleaner during operation with the suction strength determined through the above-described process is input, the processor (140) can adjust the current suction strength to a suction strength corresponding to the user command.

[0122] At this time, the processor (140) can store the changed suction strength and the confirmed floor type in the memory (130). Then, the processor (140) can modify the usage pattern information using the changed suction strength and the confirmed floor type stored in the memory.

[0123] Meanwhile, although the above description uses one neural network model, multiple neural network models may be used during implementation.

[0124] For example, when using a first neural network model and a second neural network model that output probability values ​​for each of a plurality of floor types, the processor (140) can input sensor information detected by a plurality of sensors into the first neural network model to obtain first probability information, and input sensor information detected by a plurality of sensors into the second neural network model to obtain second probability information.

[0125] Here, the two neural network models may be of different types, each with distinct characteristics. For example, the first neural network model may be a CNN model, and the second neural network model may be an LSTM model. The reason for using these different models is that the classification accuracy of each type of information may vary depending on the model. For example, a CNN model can classify floors and carpets with high accuracy using sensor information, but may have low accuracy in distinguishing carpet length. On the other hand, an LSTM model may have somewhat lower accuracy in distinguishing between regular floors and carpets, but may have high accuracy in distinguishing carpet lengths.

[0126] By using multiple neural network models with different characteristics, the type of floor can be identified with greater accuracy.

[0127] And the processor (140) can determine the floor type based on the acquired first probability information and second probability information. For example, the processor (140) can check the highest probability value among the first probability information and the second probability information, and determine the floor type corresponding to the checked probability value as the current floor type. Meanwhile, during implementation, instead of using the highest probability value, the estimated probability values ​​from each neural network model can be averaged, and the floor type corresponding to the value with the highest average probability can be determined as the current floor type.

[0128] Alternatively, when using a first neural network model that receives sensor information and outputs floor type information and a second neural network model that outputs suction power based on the type information, the processor (140) can input sensor information detected from multiple sensors into the first neural network model to confirm the floor type, and input the confirmed floor type into the second neural network model to confirm the suction strength to be applied to the vacuum cleaner.

[0129] Meanwhile, when the suction power is adjusted through user operation while operating with the suction strength determined through the above-described operation, the processor (140) can store the adjusted information and the information sensed by the sensor, etc. in the memory (130) and use them for re-learning the second neural network model.

[0130] Meanwhile, although the method of sequentially determining the floor type and suction power was described above, the suction power can also be determined directly using a single neural network model during implementation.

[0131] For example, if a pre-learned neural network model is used to output a suction strength to be applied to a vacuum cleaner by learning using sensor information and usage pattern information detected by a sensor, the processor (140) can input the sensor information detected by the sensor into the pre-learned neural network model to determine the suction strength to be applied to the vacuum cleaner.

[0132] At this time, when a user command to adjust the suction strength of the vacuum cleaner while operating at the determined suction strength is input, the processor (140) adjusts the current suction strength to a suction strength corresponding to the user command, stores the user command and sensor information detected by the sensor in memory, and retrains the neural network model using the user command and sensor information stored in the memory.

[0133] The processor (140) may periodically perform the determination of the above-described suction intensity. This period may correspond to the measurement period of the sensor described above, but a period of time twice the measurement period may also be used.

[0134] And the processor (140) controls the vacuum cleaner (100) using the determined suction strength. For example, the processor (140) can control a driving device that drives the motor so that the first motor rotates at a rotation speed corresponding to the determined suction strength.

[0135] To this end, the memory (130) can store a lookup table that stores suction strength and a motor speed corresponding to each suction strength, and the processor (140) can control the driving device using the motor speed determined based on the above-described lookup table and the determined suction strength. Meanwhile, in the above description, the suction strength is determined in advance and the motor speed is determined later using the determined suction strength, but when implemented, the motor speed may be determined directly in the preceding determination step.

[0136] When a suction power adjustment command is input through the input device (120), the processor (140) can control the vacuum cleaner (100) to operate with a suction power corresponding to the input adjustment command.

[0137] Meanwhile, although the suction power is described and illustrated as changing according to the user's control command in the above, the suction power may be changed by a specific user action rather than the user's control command during implementation. For example, if one of the multiple sensors is an acceleration sensor, and the sensing information input through the acceleration sensor confirms a user action (or gesture) of repeatedly moving a specific area, the processor (140) may determine an increased suction power compared to the current suction power.

[0138] As described above, the vacuum cleaner according to the present embodiment determines suction power using a neural network model and user pattern information, and uses the determined suction power, that is, it can consider not only the condition of the floor surface to be cleaned but also the user's usage pattern, so that cleaning can be performed in a manner suitable to the cleaning environment and user intention.

[0139] Meanwhile, although only a simple configuration of the vacuum cleaner (100) is illustrated above, various other configurations may be included during implementation. This will be described below with reference to FIG. 3.

[0140] FIG. 3 is a drawing for explaining the configuration of a vacuum cleaner according to one embodiment of the present disclosure.

[0141] Referring to FIG. 3, the vacuum cleaner (100) may include a plurality of sensors (110), an input device (120), a memory (130), a processor (140), a display (150), a communication device (160), and a driving device (170).

[0142] The plurality of sensors (110), input devices (1200, memory (130), and processor (140) have been previously described in FIG. 2, and only operations different from the operations described above will be described below.

[0143] The display (150) can display various types of information supported by the vacuum cleaner (100). This display (150) may be a display such as an LCD, and may also be implemented as a touch screen that can perform the functions of the input device described above.

[0144] The display (150) can display information such as the operating status of the vacuum cleaner (100) (clean mode, AI mode, manual mode), the suction power of the vacuum cleaner, and the battery status.

[0145] And, when the suction strength is changed, the processor (140) can control the display (150) so that the changed suction strength is displayed. In addition, when the operation mode of the vacuum cleaner is changed, the processor (140) can control the display (150) so that the changed operation mode is displayed.

[0146] The communication device (160) is formed to connect the vacuum cleaner (100) to an external device (specifically, a terminal device, a home server, an external server, etc.), and can be connected by a short-range wireless communication method (e.g., Bluetooth, WiFi, WiFi Direct) as well as a long-range wireless communication method (e.g., wireless communication such as GSM, UMTS, LTE, WiBRO, etc.).

[0147] The communication device (160) may include at least one of a WiFi module, a Bluetooth module, a wireless communication module, an NFC module, and a UWB (Ultra-Wide Band) module. Specifically, the WiFi module and the Bluetooth module may each perform communication in the WiFi or Bluetooth manner. When using a WiFi module or a Bluetooth module, various connection information, such as an SSID, may be first transmitted and received, and then communication may be established using this, after which various pieces of information may be transmitted and received.

[0148] In addition, the wireless communication module can perform communication according to various communication standards such as IEEE, Zigbee, 3G (3rd Generation), 3GPP (3rd Generation Partnership Project), LTE (Long Term Evolution), 5G (5th Generation), etc. And the NFC module can perform communication using NFC (Near Field Communication) method using 13.56MHz band among various RF-ID frequency bands such as 135kHz, 13.56MHz, 433MHz, 860~960MHz, 2.45GHz, etc.

[0149] The communication device (160) can receive a neural network model learned from an external device, receive learning data necessary for learning the neural network model, or transmit learning data collected from the vacuum cleaner (100) to the external device.

[0150] Alternatively, when utilizing a neural network model stored in an external server (not shown), the communication device (160) may transmit sensing information collected from multiple sensors and / or user operation commands to the external server. In response, the communication device (160) may receive information on the type of floor identified or the suction power to be used from the external server.

[0151] The driving device (170) controls the motor. Specifically, the vacuum cleaner according to the present embodiment includes two motors (e.g., a first motor and a second motor), and can provide driving power and / or a control signal corresponding to the rotation speed determined by the processor (140) to each motor so that each motor can rotate at the rotation speed determined by the processor (140).

[0152] Meanwhile, if the vacuum cleaner (100) is implemented as a robot vacuum cleaner, the driving device (170) may additionally be equipped with a motor for moving the vacuum cleaner.

[0153] While various configurations that the vacuum cleaner (100) may include are illustrated and described in FIG. 3, some of the aforementioned configurations may be omitted during implementation, and other configurations not illustrated may be additionally provided. For example, a microphone and a speaker may be further included to receive voice input of control commands from a user, or to output various information about the vacuum cleaner as audio.

[0154] Figure 4 is a diagram illustrating examples of information collected from sensors in various floor environments.

[0155] Referring to FIG. 4, the first data (410) is an example of information measured from a first sensor (specifically, suction pressure of the vacuum cleaner), and the second data (420) is an example of information measured from a second sensor (specifically, current supplied to a motor that drives a brush).

[0156] Referring to the first data (410) and second data (420) described above, it can be confirmed that different patterns exist depending on the type of floor surface. However, comparing the data measured on a regular floor and a carpet reveals some similarity. In other words, if only the measurement values ​​of the data described above are used, it can be confirmed that it is difficult to distinguish between a regular floor and a carpet.

[0157] Therefore, the present disclosure utilizes sensing information measured by four sensors and inputs this information into a neural network model to more accurately distinguish floor types. Specifically, the frictional level of the brush may differ between regular floors and carpets.

[0158] In order to identify these differences, the present disclosure utilizes the brush current, the motor output value, and the motor rotation speed, and as a result, it is possible to distinguish not only between a normal floor and a carpet, but also whether the carpet is short, medium, or long.

[0159] Below, the form in which the above-described sensing information is applied to the neural network model is described in detail with reference to FIGS. 5 to 12.

[0160] FIG. 5 is a diagram illustrating an example of a configuration of a neural network model according to an embodiment of the present disclosure.

[0161] As illustrated in FIG. 5, when first to fourth data are received from a plurality of sensors (111, 112, 113, 114), the preprocessing module (510) can process (or process) at least a portion of the entire data including the first to fourth data to obtain a data set to be input into a neural network model.

[0162] In the above, it was explained that the information provided to the preprocessing module (510) is input through sensors, but during implementation, some of the information may not be hardware sensors, but may be information stored in memory (130), etc., or preset data values.

[0163] The preprocessing module (510) can combine sensing values ​​of a certain period (e.g., 100 ms when the data collection cycle is 20 ms) acquired through multiple sensors (111, 112, 113, 114), determine which sensing values ​​among the sensing values ​​acquired through multiple sensors (111, 112, 113, 114) to include in a data set, and can also acquire a data set by assigning weights to the sensing values. In addition, data can be encoded or embedded according to the purpose of the neural network model (520).

[0164] The 'neural network model (520)' illustrated in FIG. 5 refers to an artificial intelligence model including a neural network trained to acquire information indicating the type of floor surface based on sensing information. As illustrated in FIG. 5, when a data set is received from the preprocessing module (510), the neural network model (520) can identify the type of floor surface corresponding to the data set and output a probability value indicating which type of floor surface the data set corresponds to.

[0165] For example, the neural network model (520) can be implemented as a CNN (Convolutional Neural Network), LSTM (Long Short-Term Memory), DNN (Deep Neural Network), RNN (Recurrent Neural Network), RBM (Restricted Boltzmann Machine), DBN (Deep Belief Network), BRDNN (Bidirectional Recurrent Deep Neural Network), etc., but is not limited to these examples.

[0166] Here, "type information" is used as a general term to refer to information indicating the type of floor surface. Specifically, type information may be expressed as probability values ​​for multiple floor types, but may also include information regarding the presence of carpet on the floor surface, the material of the floor surface, and at least one of the material of the carpet.

[0167] The control module (530) can determine the floor type using the type information (i.e., probability values ​​for each of multiple floor surfaces) output from the neural network model (520). For example, the type with the highest probability value among the output type information can be determined as the current floor type.

[0168] Once the floor type is determined in this way, the control module (530) can determine a suction power corresponding to the floor type using pre-stored user preference information (usage pattern information) and generate control information to cause the vacuum cleaner to operate with the determined suction power.

[0169] Additionally, the control module (530) can determine not only the suction power but also the rotation speed of the brush.

[0170] Meanwhile, in the above, the control module (530) is shown as outputting suction force as control information, but when implemented, it may output the rotation speed of the first motor or command information (e.g., voltage value, etc.) to be used for the first motor rather than the suction force.

[0171] Meanwhile, in the above, it has been described and illustrated that the output of the neural network model (520) is directly inputted and utilized by the control module (530), that is, it has been described that the control module (530) performs the determination of the floor type by inputting probability information for multiple floor types, but when implemented, the determination of the floor type described above can be performed by the post-processing module. That is, the control module (530) can also receive the determined floor type as input and perform only the control operation accordingly.

[0172] Below, the relearning operation when using the neural network model described above is explained with reference to Fig. 6.

[0173] FIG. 6 is a diagram illustrating an example configuration of a neural network model according to an embodiment of the present disclosure. Specifically, FIG. 6 is a diagram illustrating an example configuration in a case where the neural network model illustrated in FIG. 5 can be retrained.

[0174] As described in Fig. 5, the control module (530) controls the vacuum cleaner (100) by ultimately determining the suction strength, etc. based on the type information output from the neural network model (520).

[0175] Using the suction strength described above, the user can adjust the suction strength using the input device (120) while the cleaner is in operation.

[0176] When such an adjustment of suction strength is input, the control module (530) can change the control information to operate with a suction strength corresponding to the user operation.

[0177] For example, if a user inputs a control command to increase the suction strength, the control module (530) can change the suction strength to a suction strength higher than the current suction strength.

[0178] If the current suction strength is at its maximum and the user inputs a control command to increase the suction strength, the control module (530) can generate control information to maintain the current suction strength and increase the rotation speed of the brush by a certain speed.

[0179] If a command requesting an increase in suction strength is input when the suction strength and brush rotation speed are at their maximum, the control module (530) can display a message indicating that an increase in suction strength and brush speed is not possible without a separate change in suction strength and brush rotation speed.

[0180] Meanwhile, when the motor speed is at its maximum during implementation, if a command to increase the suction strength is input, the control module (530) displays a message to the user before increasing the rotation speed of the brush, informing that the suction strength cannot be increased and only the rotation speed of the brush can be increased, and can further increase the rotation speed of the brush based on the user's confirmation.

[0181] And the control module (530) can store the above-described information and sensing information measured at that point in time in the memory (130). Such information can be used as learning data.

[0182] Meanwhile, the control module (530) can selectively store the aforementioned information. For example, if the change in suction intensity is due to a user error, and the changed suction intensity is not maintained for several seconds, but rather the original suction intensity is restored, there may be no need to use the aforementioned change in suction intensity as learning data.

[0183] Accordingly, the control module (530) can store the above-described information as learning data when the user maintains the changed suction intensity for a preset period of time. Meanwhile, during implementation, it is also possible to store all data and selectively use the stored data during the learning process described below.

[0184] The learning module (540) can retrain the neural network model (520) using pre-stored learning data. Preliminarily, the learning module (540) can determine whether the pre-stored learning data should be used for retraining the neural network model or for modifying user pattern information.

[0185] For example, as explained above, the operation of the neural network model that performs floor surface classification is not abnormal, but if the user's usage pattern on a specific floor surface has changed, there is no need to retrain the neural network model.

[0186] Accordingly, the learning module (540) can update (or modify) the user pattern information using information that can be used to modify the user pattern information among the pre-stored learning data. For example, if the user repeatedly issues a control command to increase the suction power several times while the carpet is identified as a long-haired carpet, the learning module (540) can determine that the user's usage pattern in the long-haired carpet has changed and modify the suction power in the carpet information within the pre-stored usage pattern information.

[0187] Alternatively, the learning module (540) may retrain the neural network model using information that can be used to modify the neural network model among the pre-stored learning data. For example, the learning module (540) may train the neural network model (520) using a supervised learning method based on a learning data set including labels. In addition, the neural network model (1000) may also be trained using an unsupervised learning method, a semi-supervised learning method, or a reinforcement learning method.

[0188] Here, the learning data set refers to a collection of data used for training the neural network model (520). Like the data set, the learning data set may include at least some of the data acquired through the first to fourth sensors.

[0189] Meanwhile, in Fig. 6, retraining of the neural network model (520) is illustrated as being performed in the vacuum cleaner (100), but in the implementation, the retraining described above may be performed in an external server, and the vacuum cleaner (100) may receive and use the retrained neural network model from an external device.

[0190] FIG. 7 is a diagram illustrating an example of a neural network model configuration according to an embodiment of the present disclosure. Specifically, FIG. 7 is a diagram illustrating an example of using a neural network model that receives information from multiple sensors and outputs suction force information.

[0191] As illustrated in FIG. 7, when first to fourth data are received from a plurality of sensors (111, 112, 113, 114), the preprocessing module (710) can process (or process) at least a portion of the entire data including the first to fourth data to obtain a data set to be input into a neural network model.

[0192] The 'neural network model (720)' illustrated in FIG. 7 refers to an artificial intelligence model that includes a neural network trained to acquire suction force information to be used based on sensing information and a user's usage pattern. As illustrated in FIG. 7, when a data set is received from the preprocessing module (710), the neural network model (720) can output suction force information. Here, the suction force information can be output as probability values ​​for multiple suction intensities, or a single suction intensity can also be output.

[0193] The control module (730) can generate control information to cause the vacuum cleaner to operate with a suction force corresponding to the suction force information output from the neural network model (720).

[0194] Additionally, the control module (730) can determine a brush rotation speed corresponding to the determined suction force and generate control information including the determined rotation speed.

[0195] Below, the relearning operation when using the neural network model (720) is described with reference to FIG. 8.

[0196] FIG. 8 is a diagram illustrating an example of a configuration of a neural network model according to an embodiment of the present disclosure.

[0197] As described in FIG. 7, the control module (730) controls the vacuum cleaner (100) using the suction power information output from the neural network model (720).

[0198] While the vacuum cleaner is operating using the suction strength described above, the user can adjust the suction strength using the input device (120).

[0199] When such an adjustment of suction strength is input, the control module (730) can change the control information to operate with a suction strength corresponding to the user operation.

[0200] And the control module (730) can store user control commands and sensing information measured at that point in time in the memory (130). Such information can be used as learning data.

[0201] Meanwhile, unlike FIG. 6, the neural network model (720) described above is a model learned using not only the type of floor surface but also the user's usage pattern as learning information, and the control module (530) can retrain the neural network model (720) using the learning information stored in the memory (130).

[0202] Such re-learning can be performed after the cleaning operation of the vacuum cleaner (100) is completed, and can also be performed when a certain amount of learning information has been collected or when collection has occurred for a predetermined number of days or more.

[0203] FIG. 9 is a diagram illustrating an example configuration of a neural network model according to an embodiment of the present disclosure. Specifically, FIG. 9 is a diagram illustrating an example using two neural network models.

[0204] As illustrated in FIG. 9, when first to fourth data are received from a plurality of sensors (111, 112, 113, 114), the preprocessing module (910) can process (or process) at least a portion of the entire data including the first to fourth data to obtain a data set to be input into a neural network model.

[0205] The first neural network model (920) refers to an artificial intelligence model that includes a neural network trained to acquire information indicating the type of floor surface based on sensing information. The operation of the first neural network model (920) is identical to the neural network model illustrated in FIG. 5, and thus, a redundant description will be omitted.

[0206] The second neural network model (930) refers to an artificial intelligence model that includes a neural network trained to acquire suction power information based on the determined floor type. This second neural network model may be trained using data on the floor type determined by the first neural network model and the suction strength used by the user on that floor type.

[0207] In this way, when the second neural network model (930) outputs suction power information, the control module (940) can generate control information that causes the vacuum cleaner to operate with a suction strength corresponding to the suction power information.

[0208] Additionally, the control module (940) can determine the rotation speed of the brush in response to the determined suction force information, or can determine the rotation speed of the brush in response to the type information output from the first neural network model (920).

[0209] The retraining operation when using the two neural network models described above is described below with reference to Fig. 10.

[0210] FIG. 10 is a diagram illustrating an example of a configuration of a neural network model according to an embodiment of the present disclosure.

[0211] As described in FIG. 9, the control module (940) controls the vacuum cleaner (100) using the suction power information output from the second neural network model (930).

[0212] When the user adjusts the suction strength using the input device (120) during the operation of the vacuum cleaner using the above-described suction strength, the control module (940) can change the control information to operate with a suction strength corresponding to the user operation.

[0213] When a user operation that causes such adjustment of suction strength is input, the control module (940) can store the user control command and sensing information measured at that point in time in the memory (130). Such information can be used as learning data.

[0214] When a certain amount of such learning data is collected or a certain amount of time has passed, the learning module (950) can retrain the second neural network model (930) using the previously stored learning data.

[0215] Meanwhile, in the illustrated example, the operation of retraining only the second neural network model is described, but as described in FIG. 6, if the learning data is distinguished and there is learning data to be used for retraining the neural network model of FIG. 6, the first neural network model (920) may also be retrained. Alternatively, the second neural network model may be retrained on its own in the vacuum cleaner (100), and the first neural network model may be retrained in an external device and then provided to the vacuum cleaner (100).

[0216] FIG. 11 is a diagram illustrating an example configuration of a neural network model according to an embodiment of the present disclosure. Specifically, FIG. 11 is a diagram illustrating an example in which two neural network models are used to identify the type of floor surface.

[0217] As illustrated in FIG. 11, when first to fourth data are received from a plurality of sensors (111, 112, 113, 114), the preprocessing module (1110) can process (or process) at least a portion of the entire data including the first to fourth data to obtain a data set to be input into a neural network model.

[0218] The first neural network model (1120) and the second neural network model (1130) illustrated in FIG. 11 refer to artificial intelligence models including a neural network trained to acquire information indicating the type of a floor surface based on sensing information. As illustrated in FIG. 11, when a data set is received from the preprocessing module (1110), the data set is input to each of the first neural network model (1120) and the second neural network model (1130), and each neural network model (1120, 1130) outputs individual type information.

[0219] For example, the first neural network model (1120) may be a CNN (Convolutional Neural Network) model, and the second neural network model (1130) may be a LSTM (Long Short-Term Memory) model.

[0220] When type information is output from each neural network model in this way, the control module (1140) can determine the floor type using the type information output from each neural network model. For example, the type information described above can be expressed as probability values ​​for multiple floor types, and the control module (1140) can check the highest probability value among two types of information and determine the type corresponding to the checked probability value as the floor type. This operation can be referred to as a voting technique, but other techniques may also be used in addition to the voting technique.

[0221] Once the floor type is determined in this way, the control module (1140) can determine a suction power corresponding to the floor type using pre-stored user preference information (usage pattern information) and generate control information to cause the vacuum cleaner to operate with the determined suction power.

[0222] Meanwhile, Fig. 11 has the same form as Fig. 5 except that multiple neural network models are used. Therefore, the relearning operation in the case of Fig. 11 can be performed similarly to Fig. 6, and a duplicate description thereof is omitted.

[0223] FIG. 12 is a diagram illustrating an example configuration of a neural network model according to an embodiment of the present disclosure. Specifically, FIG. 12 is a diagram illustrating an example utilizing a neural network model provided on an external server.

[0224] Referring to FIG. 12, when first to fourth data are received from multiple sensors (111, 112, 113, 114), the preprocessing module (1210) can process (or process) at least a portion of the entire data including the first to fourth data to obtain a data set to be input into a neural network model.

[0225] The data set can be transmitted to an external device (200) via a communication device (160).

[0226] When a data set is received, the processor (220) of the external device can control the communication device (210) to generate type information using the pre-learned neural network model and transmit the generated type information to the cleaner (100).

[0227] When type information is received through the communication device (160), the control module (1220) can determine the floor type based on the received type information.

[0228] In addition, the control module (1140) can determine a suction power corresponding to the floor type using pre-stored user preference information (usage pattern information) and generate control information to cause the vacuum cleaner to operate with the determined suction power.

[0229] Meanwhile, in Fig. 12, the neural network model described in Fig. 5 is described assuming that it is deployed on an external device, but at the time of implementation, at least one neural network model of the embodiment described in Figs. 7 to 11 may also be implemented in a form in which it is deployed on an external server.

[0230] FIG. 13 is a diagram illustrating an example of a display on a vacuum cleaner according to an embodiment of the present disclosure. Specifically, FIG. 13 illustrates various UI screens that may be displayed on the display of the vacuum cleaner.

[0231] Referring to FIG. 13, when a user turns on the power button of the vacuum cleaner, one of the illustrated UI screens (e.g., AI mode (1330)) may be displayed. In this display state, when the user selects a button for adjusting the suction strength (e.g., increasing or decreasing), the suction strength may be adjusted stepwise as illustrated.

[0232] For example, when the "-" button (or decrease button) is selected in AI mode (1330), the suction intensity can be changed to strong mode (or suction intensity 2). When the "-" button is selected again in this state, the suction intensity can be changed to normal mode (or suction intensity 1). Conversely, when the "+" button (or increase button) is selected in normal mode, the suction intensity can be changed to strong mode (or suction intensity 2).

[0233] Additionally, when the "+" button is selected in AI mode (1330), the suction strength can be changed to super strong mode (or suction strength 3). When the "+" button is selected again in this state, the suction strength can be changed to maximum mode (or suction strength 4).

[0234] Meanwhile, although the illustrated example shows the display only showing the current suction strength (or the mode name corresponding to the suction strength), one side of the display may also show the current battery status, or information about the type of floor surface currently detected.

[0235] The above shows an example of adjusting the suction strength using two buttons, but a case of changing operations other than the suction strength is described below with reference to FIG. 14.

[0236] FIG. 14 is a drawing showing an example of a display on a vacuum cleaner according to an embodiment of the present disclosure.

[0237] First, the first UI screen (1410) illustrates an example of a display UI that may be displayed when set to AI clean mode. For example, in the state of the third UI screen (1330) of FIG. 13, if the user presses the "+" button for a preset period of time (i.e., long press), the AI ​​clean mode may be operated.

[0238] "AI Clean Mode" automatically adjusts the vacuum's suction power. Unlike the standard AI mode, it operates with higher suction power on certain floor surfaces than in AI mode. This specific floor surface may be carpet, but is not limited thereto. For example, in carpet mode, the vacuum may operate with higher suction power or higher brush rotation speed than in the AI ​​Protection mode described below.

[0239] The second UI screen (1420) illustrates an example of a display UI displayed when the AI ​​ambiguous mode is set. For example, in the state of the third UI screen (1330) of FIG. 13, if the user presses the "-" button for a preset period of time (i.e., long press), the AI ​​protection mode may be operated. Alternatively, in the state of the first UI screen (1410), the AI ​​protection mode may also be switched when the "-" button is pressed for a preset period of time.

[0240] This AI protection mode allows the vacuum cleaner to automatically and adaptively adjust its suction power depending on the type of floor surface, but when a carpet is identified, it operates with lower suction power or lower brush rotation speed to minimize damage to the carpet during the cleaning process.

[0241] Meanwhile, while the above description explained that suction power changes only depending on the type of floor surface in AI mode, in implementation, suction power can automatically change based on the amount of dirt and debris being sucked. In other words, suction power can automatically increase and operate on floors with a lot of dust.

[0242] FIG. 15 is a drawing illustrating an example of changing suction power according to one embodiment of the present disclosure.

[0243] In the above, it has been explained that the suction power of the vacuum cleaner is changed by the user's button operation or by a change in the detected floor surface. However, during implementation, the suction power can also be changed based on the user's cleaning method, etc.

[0244] Referring to FIG. 15, when a user repeatedly cleans the same area as shown, the vacuum cleaner can increase suction power.

[0245] For example, if the x-axis value of the gyro sensor changes + / - values ​​more than n times in n seconds, it can be recognized as repetitive cleaning of the same area. When such an output of the gyro sensor is confirmed, the cleaner (100) can increase the suction power. Alternatively, the cleaner (100) can maintain the suction power and only increase the rotation speed of the brush, or increase the suction power and rotation speed mode of the brush.

[0246] FIG. 16 is a flowchart for explaining a method for controlling a vacuum cleaner according to an embodiment of the present disclosure.

[0247] Referring to FIG. 16, the operation of the vacuum cleaner can be detected using multiple sensors (S1610). For example, the operation of the vacuum cleaner can be detected using a first sensor that detects the suction pressure of the vacuum cleaner, a second sensor that detects the current supplied to the motor that drives the brush, a third sensor that detects the output value of the motor that drives the brush, and a fourth sensor that detects the rotation speed of the brush.

[0248] And, the suction strength to be applied to the vacuum cleaner is determined using the sensor information detected from multiple sensors, the pre-learned neural network model, and the pre-stored usage pattern information (S1620). For example, in the case of using a neural network model that outputs a probability value for each of multiple floor types based on the sensor information detected from the sensors (i.e., the embodiment of FIG. 5), the sensor information detected from multiple sensors is input into the pre-learned neural network model to confirm the floor type, and the suction strength to be applied to the vacuum cleaner can be confirmed using the confirmed floor type and usage pattern information.

[0249] And the motor of the vacuum cleaner is controlled using the determined suction strength (S1630). Specifically, the first motor or the driving device that drives the first motor can be controlled so that the first motor rotates at a motor rotation speed corresponding to the suction strength.

[0250] As described above, the control method according to the present embodiment determines the suction power using a neural network model and user pattern information, and uses the determined suction power, that is, it can consider not only the condition of the floor surface to be cleaned but also the user's usage pattern, so that cleaning can be performed in a manner suitable to the cleaning environment and user intention.

[0251] FIG. 17 is a flowchart for explaining a method for controlling suction power and brush operation of a vacuum cleaner according to an embodiment of the present disclosure.

[0252] Referring to Figure 17, it can be determined whether the user has set the AI ​​mode (S1710). If the user does not use the AI ​​mode and instead selects the manual mode, the vacuum cleaner can operate at the suction strength selected by the user (S1715). At this time, the vacuum cleaner can continuously store sensing information detected by multiple sensors and information on the suction strength used by the user, and use this as learning data.

[0253] If set to AI mode, the vacuum cleaner can be controlled to operate with suction power corresponding to the type of floor surface identified (S1720). However, if sufficient sensing data to identify the current floor surface type has not been collected, the vacuum cleaner may operate with the default suction power.

[0254] At this time, the vacuum cleaner can perform different actions depending on whether the user selects carpet protection mode (or AI protection mode) or carpet clean mode (or AI clean mode).

[0255] If the user selects only the AI ​​mode without selecting a separate additional mode, the general operation is performed and the suction strength and brush rotation speed corresponding to the user pattern can be operated (S1725).

[0256] If the carpet protection mode is selected, the vacuum cleaner performs normal operations when the surface is not carpeted, and operates with lower suction power and brush rotation speed than in the normal mode when a carpet is detected (S1730).

[0257] Meanwhile, in such a case, if the user adjusts the suction intensity (S1470), specifically, if a command to lower the suction intensity is input, an operation to lower the suction intensity can be performed (S1740-Lower).

[0258] Conversely, if an action is performed to increase the suction strength (S1740-increase), and the current suction strength is higher than that in normal mode (S1750), the AI ​​protection mode can be changed to AI clean mode.

[0259] Meanwhile, if the AI ​​clean mode is selected and a carpet is detected, it can operate with higher suction power and faster brush rotation speed than in the normal mode (S1760).

[0260] In this case, if the user lowers or raises the suction strength, the suction strength can be raised or lowered accordingly, and if the suction strength lowered by the user corresponds to the AI ​​protection mode, the AI ​​clean mode can be switched to the AI ​​protection mode.

[0261] Meanwhile, in the above, it was explained that the operation is performed in response to the user additionally selecting the AI ​​protection mode or AI clean mode, but at the time of implementation, the usage pattern information may store information on whether the AI ​​protection mode or AI clean mode to be applied to a specific floor type is used, and in this case, the above-described operation may be performed using the pre-stored usage pattern information without a separate user operation.

[0262] FIG. 18 is a flowchart for explaining the learning operation of a neural network model according to one embodiment of the present disclosure.

[0263] Referring to Figure 18, sensing information is collected using multiple sensors (S1810). Specifically, the above-described collection may be information collected during the process of utilizing a neural network model to determine the floor type or suction power, as described previously in Figure 15.

[0264] When a control command for adjusting the suction strength is input from the user during cleaning using the suction strength determined through the above-described operation, the input control command and the collected sensing information at that point in time can be stored in the memory (S1820).

[0265] Furthermore, retraining of a previously stored neural network model can be performed using the training data stored in memory (S1830). The specific retraining method may vary depending on the configuration of the neural network model used, and is described in detail in FIGS. 6, 8, and 10, and thus, a duplicate description is omitted.

[0266] FIG. 19 is a flowchart for explaining a control operation using information of an external device according to one embodiment of the present disclosure.

[0267] Referring to FIG. 19, the operation of the vacuum cleaner can be detected using multiple sensors (S1910). For example, the operation of the vacuum cleaner can be detected using a first sensor that detects the suction pressure of the vacuum cleaner, a second sensor that detects the current supplied to the motor that drives the brush, a third sensor that detects the output value of the motor that drives the brush, and a fourth sensor that detects the rotation speed of the brush.

[0268] And, information collected from multiple sensors can be transmitted to an external device (S1920). This operation can be performed periodically.

[0269] In response to the transmission of the collected information, when the output result of the neural network model is received from an external device (S1930), the motor can be controlled using the received result (S1940).

[0270] For example, if an external server is a neural network model that only determines floor type, the received information is related to the floor type. Using this type information and user usage patterns, the suction power to be used during the cleaning process can be determined. Furthermore, the first motor can be controlled to operate at a speed corresponding to the determined suction power.

[0271] Meanwhile, the methods according to at least some of the various embodiments of the present disclosure described above may be implemented in the form of an application that can be installed on an existing electronic device.

[0272] Additionally, the methods according to at least some of the various embodiments of the present disclosure described above can be implemented with only a software upgrade or a hardware upgrade for an existing electronic device.

[0273] Additionally, the methods according to at least some of the various embodiments of the present disclosure described above may also be performed through an embedded server provided in an electronic device, or an external server of at least one of the electronic devices.

[0274] Meanwhile, according to one embodiment of the present disclosure, the various embodiments described above can be implemented as software including instructions stored in a machine-readable storage medium that can be read by a machine (e.g., a computer). The machine is a device that can call instructions stored from the storage medium and operate according to the called instructions, and may include an electronic device (e.g., electronic device (A)) according to the disclosed embodiments. When an instruction is executed by a processor, the processor can perform a function corresponding to the instruction directly or by using other components under the control of the processor. The instruction may include code generated or executed by a compiler or an interpreter. The machine-readable storage medium may be provided in the form of a non-transitory storage medium. Here, the 'non-transitory storage medium' only means that it is a tangible device and does not include a signal (e.g., an electromagnetic wave), and this term does not distinguish between cases where data is stored semi-permanently and cases where it is temporarily stored in the storage medium. No. For example, a 'non-transitory storage medium' may include a buffer in which data is temporarily stored. According to one embodiment, the method according to various embodiments disclosed in the present document may be provided as a computer program product. The computer program product may be traded between sellers and buyers as a commodity. The computer program product may be distributed in the form of a machine-readable storage medium (e.g., compact disc read only memory (CD-ROM)), or may be distributed online (e.g., downloaded or uploaded) via an application store (e.g., Play Store™) or directly between two user devices (e.g., smartphones).In the case of online distribution, at least a portion of the computer program product (e.g., a downloadable app) may be temporarily stored or temporarily created in a device-readable storage medium, such as the memory of a manufacturer's server, an application store's server, or an intermediary server.

[0275] Various embodiments of the present disclosure may be implemented as software including commands stored in a machine-readable storage medium that can be read by a machine (e.g., a computer). The device is a device that can call commands stored in the storage medium and operate according to the called commands, and may include an electronic device (e.g., a vacuum cleaner (100)) according to the disclosed embodiments.

[0276] When the above-described instruction is executed by the processor, the processor may perform the function corresponding to the instruction directly or by utilizing other components under the control of the processor. The instruction may include code generated or executed by a compiler or interpreter.

[0277] Although the preferred embodiments of the present disclosure have been illustrated and described above, the present disclosure is not limited to the specific embodiments described above, and various modifications may be made by a person having ordinary skill in the art to which the present disclosure pertains without departing from the gist of the present disclosure as claimed in the claims, and such modifications should not be understood individually from the technical idea or prospect of the present disclosure.

Claims

1. In the vacuum cleaner, A plurality of sensors for detecting the operation of the vacuum cleaner; An input device for receiving a user command to adjust the suction strength of the vacuum cleaner; A memory storing a pre-learned neural network model and user usage pattern information related to the operation of the vacuum cleaner; and A vacuum cleaner including a processor that controls the vacuum cleaner by determining a suction strength to be applied to the vacuum cleaner using at least one of the above-mentioned usage pattern information, sensor information detected by the plurality of sensors, and a pre-learned neural network model.

2. In paragraph 1, The above-mentioned pre-learned neural network model is a model that individually outputs probability values ​​for multiple floor types based on sensor information detected from a sensor among the multiple sensors. The above processor, A vacuum cleaner that inputs sensor information detected from the plurality of sensors into the learned neural network model to identify a floor type, and determines a suction strength to be applied to the vacuum cleaner using the identified floor type and the usage pattern information.

3. In paragraph 2, Further comprising a driving device for controlling a motor that provides driving force to the brush of the above cleaner; The above processor, A cleaner that determines the rotation speed of the brush by using the confirmed floor type and the usage pattern information, and controls the driving device so that the brush rotates at the determined rotation speed.

4. In paragraph 2, further comprising an input device for receiving a user command for adjusting the suction strength of the vacuum cleaner; The above processor, A cleaner, wherein when a user command for adjusting the suction strength of the cleaner is received while the cleaner is operating at the determined suction strength, the cleaner adjusts the current suction strength to a suction strength corresponding to the user command and stores the adjusted suction strength and the identified floor type in the memory.

5. In paragraph 4, The above processor, A vacuum cleaner that modifies the usage pattern information by using the suction strength corresponding to the user command stored in the memory and the confirmed floor type.

6. In paragraph 2, The above usage pattern information is, A vacuum cleaner comprising at least one of user preferred suction strength, brush speed information and protection mode application information for each of a plurality of floor types.

7. In paragraph 2, The above-mentioned pre-trained neural network model is the first neural network model, The above memory is, Store the first neural network model and the second neural network model that individually output probability values ​​for multiple floor types, The above processor, The sensor information detected from the above plurality of sensors is input into the first neural network model to obtain first probability information, The sensor information detected from the above plurality of sensors is input into the second neural network model to obtain second probability information, A vacuum cleaner that determines a floor type based on the first probability information and the second probability information obtained above.

8. In paragraph 7, The above processor, A vacuum cleaner that checks the highest probability value among the first probability information and the second probability information, and determines the floor type corresponding to the highest probability value checked as the current floor type.

9. In paragraph 2, The above multiple floor types are: Includes general flooring, raised flooring, mats, carpet short pile, carpet medium pile, and carpet long pile. The above processor, A vacuum cleaner that determines a suction strength having a lower suction power than when the identified floor type is a normal floor, when the identified floor type is a carpet medium pile or a carpet long pile, and the above usage pattern information includes carpet protection information.

10. In paragraph 1, The above-mentioned pre-learned neural network model is a model that learns by using sensor information and usage pattern information detected from one of the plurality of sensors and outputs a suction strength to be applied to the vacuum cleaner. The above processor, A vacuum cleaner that inputs sensor information detected from the plurality of sensors into the learned neural network model to determine the suction strength to be applied to the vacuum cleaner.

11. In paragraph 10, further comprising an input device for receiving a user command for adjusting the suction strength of the vacuum cleaner; The above processor, A cleaner, wherein when a user command for adjusting the suction strength of the cleaner is input while the cleaner is operating at the determined suction strength, the cleaner adjusts the current suction strength to a suction strength corresponding to the user command, stores the user command and sensor information detected by the plurality of sensors in the memory, and retrains the pre-learned neural network model using the user command and the sensor information stored in the memory.

12. In paragraph 1, The above-mentioned pre-trained neural network model is the first neural network model, The above memory is, Store the first neural network model that receives sensor information and outputs floor type information and the second neural network model that outputs suction power based on the type information, The above processor, A vacuum cleaner that inputs sensor information detected from the plurality of sensors into the first neural network model to identify a floor type, and inputs the identified floor type into the second neural network model to identify a suction strength to be applied to the vacuum cleaner.

13. In paragraph 1, The above multiple sensors are, An acceleration sensor for detecting the movement status of the vacuum cleaner is included; The above processor, A cleaner that determines a suction strength that is increased from the current suction strength when it is confirmed that the cleaner is moving repeatedly over the same area.

14. In the method of controlling a vacuum cleaner, A step of detecting the operation of the vacuum cleaner using multiple sensors; A step of determining a suction strength to be applied to the vacuum cleaner by using one or more sensor information detected from the plurality of sensors, a pre-learned neural network model, and pre-stored usage pattern information; and A control method comprising: a step of controlling a motor of the vacuum cleaner using the determined suction strength; 15. In a non-transitory computer-readable recording medium storing a program for executing a control method in a vacuum cleaner, The above control method is, A step of detecting the operation of the vacuum cleaner using multiple sensors; A step of determining a suction strength to be applied to the vacuum cleaner by using one or more sensor information detected from the plurality of sensors, a pre-learned neural network model, and pre-stored usage pattern information; and A computer-readable recording medium comprising: a step of controlling a motor of the vacuum cleaner using the determined suction strength;

Images