The present disclosure relates to a vacuum cleaner, and more particularly, to a vacuum cleaner suitable for calculating probability of a type of a currently cleaned floor using Artificial Intelligence (AI) technology and varying a suction force of the cleaner automatically depending on a situation.
A vacuum cleaner is a device for doing the cleaning by sucking dust or particles in a cleaning target area. Such cleaners may be classified into a manual cleaner for doing the cleaning in a manner that a user moves the cleaner in direct and an automatic cleaner for doing the cleaning in a manner of travelling by itself.
Manual cleaners may be classified into a canister type, an upright type, a handy type, a stick type and the like depending on cleaner styles.
A cleaner may clean a floor surface suing a nozzle. Generally, a nozzle is usable to suck air and dust. In addition, depending on the type of nozzle, a mop is attached to the nozzle, whereby a floor can be cleaned with the mop.
According to a related art, when the material of a floor changes in the course of cleaning the floor with a cleaner, a user may recognize or determine the change. If so, the user may apply an input for changing an output of the cleaner during the cleaning work. A controller receives the input for changing the output of the cleaner from the user and then varies an output of a suction motor in general.
In addition, recently developed Vacuum cleaners include a sensing unit capable of detecting a change of a floor material. If the sensing unit detects that the floor material has changed during the cleaning, a controller receives the detected change and may provide a function of varying a suction force automatically.
Generally, a vacuum cleaner determines a type of a floor with reference to a current flowing through a nozzle motor attached to a head and then varies a suction force. Particularly, Korean Patent No. 10-1411742 discloses a robot cleaner configured to vary a suction force depending on a floor detection result. The variation of the suction force is set to be changed depending on whether a current load measured at a nozzle part exceeds a threshold.
However, in case of predicting a type of a floor surface according to a single threshold like the related art, it may lead to unintended results due to a user's cleaning pattern change, a collision with an obstacle, and the like, thereby causing a problem of malfunction of a device.
One technical task of the present disclosure is to provide a vacuum cleaner capable of sensing a floor surface by applying a probability modeling scheme based on one or more composite data. Particularly, a vacuum cleaner according to embodiments may include an Artificial Intelligence (AI) unit capable of providing highly reliable sensing performance in various situations using an AI model configured using machine learning.
To achieve these objects and other advantages and in accordance with the purpose of the disclosure, as embodied and broadly described herein, a vacuum cleaner according to one embodiment of the present disclosure may include a suction motor providing a suction force, a nozzle unit having a rotating part and a dust inlet so as to suck dust on a floor surface by receiving the suction force from the suction motor, a nozzle motor provided to the nozzle unit to transfer a drive force to the rotating part, a nozzle shutter provided to the nozzle unit to adjust a size of the dust inlet, a battery providing power to the suction motor, the nozzle motor and the nozzle shutter, a model selecting unit generating information on an operating state corresponding to a drive mode of the suction motor and an open/closed state of the nozzle shutter, an artificial intelligence unit generating probability information through an artificial intelligence model using at least one of current information on a current value flowing through the nozzle motor, voltage information on a voltage value of the battery, and the information on the operating state, and a controller controlling the drive mode in response to the probability information.
Preferably, the probability information may include a first probability value indicating that the floor surface is a first type and a second probability value indicating that the floor surface is a second type.
More preferably, the drive mode may include a first mode and a second mode and the suction motor may have a higher suction force in the first mode rather than the second mode. If the first probability value is greater than the second probability value, the controller may drive the suction motor in the first mode. If the first probability value is not greater than the second probability value, the controller may drive the suction motor in the second mode.
In this case, the operating state may include one of a first state that the suction motor operates in the first mode while the nozzle shutter is in the open state, a second state that the suction motor operates in the second mode while the nozzle shutter is in the open state, and a third state that the suction motor operates in the first mode while the nozzle shutter is in the closed state. The artificial intelligence model may include information obtained through machine learning to determine the first probability value and the second probability value in response to the first to third states. The artificial intelligence unit may obtain the first probability value and the second probability value in one of the first to third states using the artificial intelligence model.
The artificial intelligence model may include a first model machine-learned in the first state, a second model machine-learned in the second state, and a third model machine-learned in the third state. The artificial intelligence unit may obtain the first probability value and the second probability value in a manner of using the first model when the operating state is the first state, using the second model when the operating state is the second state, and using the third model when the operating state is the third state.
More preferably, the artificial intelligence model may include a learning model for the first probability value and the second probability value corresponding to a combination of the voltage information and the current information. The artificial intelligence unit may generate the first probability value and the second probability value corresponding to the current information and the voltage information through the learning model.
Preferably, the vacuum cleaner may further include a pre-processing unit generating the current information by processing the current value flowing through the nozzle motor. The pre-processing unit may generate the current information by calculating an arithmetic mean of the current value measured with a configured time length and a configured time interval.
In another aspect of the disclosure, as embodied and broadly described herein, a method of controlling a vacuum cleaner according to one embodiment of the present disclosure may include a first step of generating information on an operating state by receiving a drive mode of a suction motor and an open/closed state of a nozzle shutter, a second step of generating current information by receiving a current value flowing through a nozzle motor, a third step of generating voltage information by receiving a voltage value of a battery, a fourth step of generating probability information through an artificial intelligence model using at least one of the operating state, the current information and the voltage information, and a fifth step of controlling the drive mode of the suction motor in response to the probability information.
Preferably, the probability information may include a first probability value indicating that a floor surface currently cleaned is a first type and a second probability value indicating that the floor surface is a second type.
More preferably, the drive mode may include a first mode and a second mode. The suction motor may have a higher suction force in the first mode rather than the second mode. If the first probability value is greater than the second probability value, the fifth step may drive the suction motor in the first mode. If the first probability value is not greater than the second probability value, the fifth step may drive the suction motor in the second mode.
In this case, the operating state may include one of a first state that the suction motor operates in the first mode while the nozzle shutter is in the open state, a second state that the suction motor operates in the second mode while the nozzle shutter is in the open state, and a third state that the suction motor operates in the first mode while the nozzle shutter is in the closed state. The artificial intelligence model may include information obtained through machine learning to determine the first probability value and the second probability value in response to the first to third states. The fourth step may obtain the first probability value and the second probability value in one of the first to third states using the artificial intelligence model.
The artificial intelligence model may include a first model machine-learned in the first state, a second model machine-learned in the second state, and a third model machine-learned in the third state. The fourth step may obtain the first probability value and the second probability value in a manner of using the first model when the operating state is the first state, using the second model when the operating state is the second state, and using the third model when the operating state is the third state.
More preferably, the artificial intelligence model may include a learning model including the first probability value and the second probability value corresponding to a combination of the voltage information and the current information. The fourth step may generate the probability information by finding the first probability value and the second probability value corresponding to the current information and the voltage information from the learning model.
Preferably, the second step may generate the current information by calculating an arithmetic mean of the current value measured with a configured time length and a configured time interval.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory and are intended to provide further explanation of the disclosure as claimed.
Firstly, an AI unit included in a vacuum cleaner according to the present disclosure may generate probability information indicating a prescribed type of a floor surface currently cleaned by a user. A controller receives the probability information on the floor surface type from the AI unit, thereby controlling a drive mode of the vacuum cleaner.
Particularly, the AI unit may predict a type of a currently cleaned floor surface using a conventionally machine-learned neural network model. The AI unit may predict a type of a currently cleaned floor surface using one or more informations such as a value of a current flowing through a nozzle motor, a voltage value of a battery, an open/closed state of a nozzle shutter, and a drive mode of a suction motor. This increases a probability of predicting a floor surface type more accurately than a related art of using a current flowing through a nozzle motor only.
Reference will now be made in detail to the preferred embodiments of the present disclosure, examples of which are illustrated in the accompanying drawings, to facilitate those having ordinary skill in the art to implement the disclosure.
In this process, the size, shape, etc. of the components illustrated in the drawings may be exaggerated for clarity and convenience of description. In addition, the terms specifically defined in consideration of the configuration and action of the present disclosure may vary according to the intentions or practices of users and operators.
Terms including an ordinal number such as ‘first’, ‘and/or’, ‘second’, etc. may be used to describe various components, not limiting the components. The terms are used only for the purpose of distinguishing one component from another component. For example, within the scope of rights under the concept of the present disclosure, a first component may be named a second component, and similarly the second component may also be named the first component.
These terms shall be defined and understood based on the content throughout the present specification.
A basic structure of a vacuum cleaner 1 included in embodiments will be described with reference to
Referring to
Referring to
The nozzle shutter 101 may stay in an open or closed state. When the nozzle shutter 101 is in the open state, since a size of the dust inlet (not shown) is increased, particles in relatively large size may be sucked in. When the nozzle shutter 101 is in the closed state, since a size of the dust inlet (not shown) is decreased, it brings an effect that a suction force of the vacuum cleaner 1 is increased.
Once the nozzle shutter 101 enters the open state, as a size of the dust inlet is increased, a suction force is decreased. In doing so, an adhesive force between the nozzle unit 100 and the floor is weakened owing to the decreased suction force. In this case, since the rotating part 103 is facilitated to move over the floor surface relatively, an amount of a current flowing through the nozzle motor 102 is reduced.
On the other hand, once the nozzle shutter 101 enters the closed state, as a size of the dust inlet is decreased, a suction force is increased. In doing so, an adhesive force between the nozzle unit 100 and the floor is strengthened owing to the increased suction force. In this case, in order to facilitate the rotating part 103 to move over the floor surface, an amount of a current flowing through the nozzle motor 102 is raised.
When a user does the cleaning of a floor surface using the vacuum cleaner 1, the floor surface may be assumed as having two kinds of types. In case of a hard type floor such as a paper floor, a wooden floor or the like, since dust and the like are exposed on a floor surface as they are, effective cleaning is available with a low suction force. Yet, in case of a carpet type floor such as Wilton, plush or the like, since dust is hidden in a floor surface, the cleaning needs to be done with a high suction force.
Assume a case that a user moves into an area of a carpet type floor in the course of cleaning a hard floor. According to the related art, when a user detects a change of a floor material, the vacuum cleaner 1 of an early type stops a cleaning operation of the vacuum cleaner 1 according to user's determination and provides a function of varying a suction force in response to a user's input of a button provided to the cleaner.
According to the related art, the vacuum cleaner 1 equipped with a function of varying a suction force automatically provides a function of varying the suction power in a manner that the sensing unit of the vacuum cleaner 1 self-determines a change of a floor material.
According to the related art of varying a suction force in a manner that a cleaner automatically determines a floor type, as the rotating part 103 receives a different rotation resistance depending on a different floor surface type, an amount of the current flowing through the nozzle motor 102 is changed.
According to the related arts, a vacuum cleaner is controlled on the assumption that a range of a current value flowing through the nozzle motor 102 differs depending on a type of a currently cleaned floor surface. Therefore, if the current flowing through the nozzle motor 102 exceeds a threshold, it is determined that a floor of a carpet type is being cleaned. If the current flowing through the nozzle motor 102 does not exceed the threshold, it is determined that a floor of a hard type is being cleaned. Yet, there are various limits and defects in the scheme of classifying a type of a currently cleaned floor surface using a current flowing through the nozzle motor 102.
The vacuum cleaner 1 according to the embodiments may include a suction motor 200 providing a suction force, a nozzle unit 100 sucking dust from a floor surface by receiving the suction force from the suction motor 200, a nozzle motor 102 provided to the nozzle unit 100 to transfer a drive force to a rotating part 103, and a nozzle shutter 101 provided to the nozzle unit 100 to adjust a size of a dust inlet. In addition, the vacuum cleaner 1 may include a battery 300 providing power to the suction motor 200, the nozzle motor 102 and the nozzle shutter 101.
The vacuum cleaner 1 according to the embodiments may include a model selecting unit S101 generating information on an operating state corresponding to a drive mode W of the suction motor 200 and an open/closed state 0 of the nozzle shutter 101 and an Artificial Intelligence (AI) unit S103 generating probability information on a type of a currently cleaned floor surface through an AI model. The AI model may generate probability information using current information A on a current value flowing through the nozzle motor 102, voltage information V on a voltage value of the battery 300, and information on an operating state of the vacuum cleaner 1.
A controller S104 included in the vacuum cleaner 1 may control the drive mode W of the suction motor 200 in response to the probability information generated from the AI unit S103.
Namely, the vacuum cleaner 1 according to the embodiments may use at least one of the information on the operating state, the current information A and the voltage information V in obtaining the type of the currently cleaned floor surface. This may have floor surface type prediction performance higher than that of the related art that uses the current information A only.
The probability information according to the embodiments may include a first probability value PC of a probability that the currently cleaned floor surface is a first type and a second probability value PH of a probability that the currently cleaned floor surface is a second type. In this case, the floor surface of the first type may include a floor surface of a carpet type that should be cleaned with a relatively higher suction force, and the floor surface of the second type may include a floor surface of a hard type that should be cleaned with a relatively lower suction force.
The drive mode W according to the embodiments may include a first mode M1 and a second mode M2. The suction motor 200 may have a higher suction force in the first mode M1 rather than the second mode M2. If the first probability value PC is greater than the second probability value PH, the controller S104 may drive the suction motor 200 in the first mode M1. If the first probability value PC is not greater than the second probability value PH, the controller S104 may drive the suction motor 200 in the second mode M2.
Particularly, if it is determined that the probability that the currently cleaned floor surface is the carpet type is higher than the probability that the currently cleaned floor surface is the hard type, the controller S104 may drive the suction motor 200 in the first mode M1. If it is determined that the probability that the currently cleaned floor surface is the hard type is higher than the probability that the currently cleaned floor surface is the carpet, the controller S104 may drive the suction motor 200 in the second mode M2.
The operating state of the vacuum cleaner 1 according to the embodiments may include a first state that the suction motor 200 is operating in the first mode M1 while the nozzle shutter 101 is in the open state, a second state that the suction motor 200 is operating in the second mode M2 while the nozzle shutter 101 is in the open state, or a third state that the suction motor 200 is operating in the first mode M1 while the nozzle shutter 101 is in the closed state.
The AI model according to the embodiments may include information, which may be obtained by a machine learning method, for determining the first probability value PC and the second probability value PH in response to the first to third states. The AI unit S103 may obtain the first probability value PC and the second probability value PH in one of the first to third states using the AI model.
The logic for a method of controlling the vacuum cleaner 1 according to the embodiments is described with reference to
Referring to
The controller S104 of the vacuum cleaner 1 according to the embodiments may use the relative comparison between estimated probability values per type instead of determining a type of a floor surface with the absolute sizes of the first probability value PC and the second probability value PH like the related art. Through this, it is possible to implement a composite determining method capable of including floor surfaces of various types.
For example, although both of the first probability value PC and the second probability value PH are low, it is able to determine a type of a floor surface by selecting a greater one of the first probability value PC and the second probability value PH.
If the first probability value PC is higher than the second probability value PH, the controller S104 according to the embodiments may increase the output of the suction motor 200 or drive the suction motor 200 in the first mode M1. In doing so, a model parameter used by the controller S104 in a next period may be the first or third state. If the first probability value PC is not higher than the second probability value PH, the controller S104 may decrease the output of the suction motor 200 or drive the suction motor 200 in the second mode M2. In doing so, a model parameter used by the controller S104 in a next period may be the second state.
Referring to
A method of obtaining probability information using MLP according to embodiments is described as follows. An AI model may receive an input of pre-processed current information Ad and an input of pre-processed voltage information V, respectively. In this case, by operating ω={a1,1,1, a1,2,1, b1,1, . . . , al,m,n, bl,n}, which is a parameter for an operating state information, it is able to obtain an intermediate result value Y={y1,1, y1,2, . . . , yl,n}. By repeating the above process, it is able to obtain P={y1,1, y1,2, . . . , yl,n}={PC, PH} that is a result value of a final layer l. An operation expression used to obtain the result value may include
y
1,1
=a
1,1,1
×x
1
+a
1,2,1
×x
2
+b
1,1.
The AI model may be learned by cleaning floor surfaces of different types for the different operating states. Hence, it is unnecessary to perform modeling manually. And, the AI model can be derived by being optimized for the corresponding situation data. This may facilitate high-dimensional modeling. Since the AI model according to embodiments uses high-dimensional modeling, complexity of the model may increase. And, it is possible to finely analyze the complicated casualty among composite input data items.
The AI model according to the embodiments may be designed as a high-dimensional model that can cope with various situations in comparison to the conventional single threshold determining method. In addition, it is possible to determine a type of a floor surface with accuracy higher than the related art.
Problems of a method of obtaining a floor surface type using current information A only according to the related art are described with reference to
The first type floor surface may include one of various types such as Wilton, plush, rug and the like. In addition, the first type floor surface may include a floor surface of a first case T1a, a second case T1b and/or a third case T1c, which have different features.
Referring to
Particularly, if an uneven surface exists on a floor surface or a tape, a drink, a sticky thing or the like exists on the floor surface, a frictional force between the nozzle unit 100 and the floor surface may increase. In case of turning a direction or stroking in a curved line during the cleaning, a force misaligned with a rotation direction of the rotating part 103 is generated, whereby a frictional force may increase. In case that the vacuum cleaner 1 is bumped into an obstacle, as the rotation of the rotating part 103 is interrupted, a frictional force may increase. Once the frictional force increases, a current flowing through the nozzle motor 102 increases. Hence, it may raise the possibility that a currently cleaned floor surface is misjudged as a first type floor surface.
Looking into the current value T3 in case of possibly causing the determination error with reference to
Particularly, if an object such as a wire, a rod, a notebook or the like exists on a floor surface, the nozzle unit 100 may be lifted in the air. In addition, in case that a floor surface is a carpet or the like, as the carpet is overage or folded, a portion of the nozzle unit 100 may be lifted in the air. When a user pulls the vacuum cleaner 1 abruptly during the cleaning or the vacuum cleaner 1 is lifted during the cleaning by a tall user, a frictional force may decrease. If the frictional force decreases, the current flowing through the nozzle motor 102 is reduced, whereby the possibility of misjudging a currently cleaned floor surface as a second type floor surface is raised.
Regarding the current value T3 in case of possibly causing the determination error, the current value instantly decreases in the above-listed situations A4, A5 and A6, the current value T3 has a value below the threshold. According to the related art, in the above situation, it may be misjudged that the second type floor surface is being cleaned despite that the first type floor surface is being cleaned actually.
When the vacuum cleaner 1 including the pre-processing unit S102 according to embodiments determine a floor surface of a first type and a floor surface of a second type, which are different from each other, the advantages of the vacuum cleaner 1 are described with reference to
Regarding the current value flowing through the nozzle motor 102,
The first type floor surface may include one of various types such as Wilton, plush, rug and the like. In addition, the first type floor surface may include a floor surface of a first case T1a, a second case T1b and/or a third case T1c, which have different features. With respect to the first type floor surface of different types, each current value range may be different. In this case, the range of the current value for the first type floor surface is increased and may have a portion overlapping with the range of the current value of the second floor surface type.
In case of determining a floor surface using an unprocessed current value, the current range (T1 region) on the first type floor surface and the current range (T2 region) on the second type floor surface are formed wide. As described above, this is the effect caused because a non-general current value is included in the current range. Yet, when an arithmetic mean is calculated by measuring a current flowing through the nozzle motor 102 with a given time length and a given time interval, an effect of the non-general current value can be minimized. In case of pre-processing a current value by setting a time length to 400 ms, an overlapping portion (Overlap) of a current range for the floor surfaces of both types is not reduced considerably in comparison with the data that is not pre-processed. Yet, in case of pre-processing a current value by setting a time length to 800 ms or 1200 ms, it is confirmed that an overlapping portion (Overlap) of a current range for the floor surfaces of both types is reduced considerably. If the overlapping portion (Overlap) between the current ranges is reduced, setting a threshold is facilitated and accuracy in determining a floor surface type can be raised.
The time length of 800 ms may correspond to a time in which a stroke of a cleaner is performed once in consideration of a user's cleaning pattern. For a pre-processing result having a time length greater than this time, data on a level similar to the pre-processing result having the time length of 800 ms may be obtained. A result from performing a pre-processing by accumulating data for the time length of 800 ms in consideration of a response speed and accuracy of algorithm can be utilized for the floor surface type determination.
Particularly, a pre-processing method of current information A may utilize a method of obtaining an arithmetic mean for a current value flowing through the nozzle motor 102. For example, if a configured time length is 800 ms and a time interval for measuring a current value is 40 ms, 20 cumulative current values may be obtained. The pre-processing unit S102 may provide an arithmetic mean of the 20 cumulative current values to the AI unit S103.
A method for the vacuum cleaner 1 according to the embodiments determines a floor surface using a current information A and a drive mode W as parameters is described with reference to
When a drive mode W of the suction motor 200 includes both a first mode M1 and a second mode M2,
For the floor surface of the same type, current information A in the first mode M1 having a relatively high output of the suction motor 200 and current information A in the second mode M2 having a relatively low output may be different from each other. In this case, the current information A may be independently determined in the first mode M1 and the second mode M2.
Particularly, since the adhesion between the nozzle unit 100 and the floor surface is high in the first mode M1, the current value flowing through the nozzle motor 102 may have a relatively high size. On the other hand, since the adhesion between the nozzle unit 100 and the floor surface is low in the second mode M2, the current value flowing through the nozzle motor 102 may have a relatively low size.
In this case, when the floor surface of the second type is cleaned in the first mode M1, as a current value is increased relatively, an upper limit of the current range (T2 region) on the second type floor surface is raised. On the other hand, when the floor surface of the first type is cleaned in the second mode M1, as a current value is decreased relatively, a lower limit of the current range (T1 region) on the first type floor surface is lowered. Therefore, an overlapping portion (Overlap) between the current range values on the floor surfaces of the two types is increased.
An embodiment of the vacuum cleaner generating probability information on each floor surface type using the combination of different voltage information V and different current information A is described with reference to
An AI model according to embodiments may include a learning model for a first probability value PC and a second probability value PH corresponding to the combination of voltage information V and current information A. The AI unit S103 may generate the first probability value PC and the second probability value PH corresponding to voltage information V and current information A through the learning model.
Particularly, when a floor is cleaned with the cleaner 1, the nozzle motor 102 operates to enable the rotating part 103 to rotate at a predetermined speed. The rotational speed of the rotating part 103 may be determined by the amount of power consumed by the nozzle motor 102. The amount of the consumed power may be expressed as a product of voltage and current. Therefore, in a motor rotating at the same rotational speed, the current flowing through the nozzle motor 102 at low voltage may have a value greater than the current flowing through the nozzle motor 102 at high voltage.
Particularly, after the vacuum cleaner 1 has been fully charged, the voltage of the battery 300 gradually decreases as a user proceeds with the cleaning process, and finally the battery 300 discharges.
On the other hand, the rotational speed of the rotating part 103 should be maintained constant during the cleaning process, so the current value flowing through the nozzle motor 102 increases relatively. Even if a floor surface of the same type is cleaned, the current value flowing through the nozzle motor 102 may vary depending on the capacity of the battery 300, which makes it difficult to determine a floor surface type.
An embodiment of
The AI unit S103 and AI model of the vacuum cleaner 1, which generates probability information on each floor surface type in consideration of voltage information V together despite obtaining the same current value, according to an embodiment are described with reference to
Particularly, in case of a floor surface of high adhesion, an average current value may be high relatively despite a floor surface of a second type. On the other hand, in case of a carpet such as Plush of a smooth type, an average current value may be low relatively despite a floor surface of a first type. The vacuum cleaner 1 according to the embodiments may determine a type of a floor surface more accurately by considering a voltage value of the battery 300 together although the current flowing through the nozzle motor 102 is same.
With reference to
A test is performed actually in a manner of changing a cleaning area from a second type floor surface into a first type floor surface.
Additionally, in order to simulate a situation caught on an obstacle, a case that an average current value is increased more than the first type floor surface is randomly inserted for 120 ms during the cleaning on the second type floor surface, which appears in a region E1.
Moreover, a case that a current value is decreased to a level similar to a case of the second type is inserted when the first type floor surface is cleaned according to user's manipulation, which appears in a region E2.
Referring to
In situations that it is difficult to solve by a floor surface type determination scheme using current information A of the related art, the reliability of the floor surface sensing performance according to embodiments of the present disclosure can be substantiated.
It will be apparent to those skilled in the art that various modifications and variations can be made in the present invention without departing from the spirit or scope of the inventions. Thus, it is intended that the present invention covers the modifications and variations of this invention provided they come within the scope of the appended claims and their equivalents.
Number | Date | Country | Kind |
---|---|---|---|
10-2020-0125958 | Sep 2020 | KR | national |
Filing Document | Filing Date | Country | Kind |
---|---|---|---|
PCT/KR2020/015886 | 11/12/2020 | WO |