JP4821536B2 - Vehicle air conditioner and control method for vehicle air conditioner - Google Patents

Description

  TECHNICAL FIELD The present invention relates to a vehicle air conditioner, a vehicle air conditioner control method, and a control device, and more particularly to a vehicle air conditioner that automatically optimizes an air condition according to a passenger's sense of temperature or situation and the vehicle. The present invention relates to a method for controlling an air conditioner.

  In general, in a vehicle air conditioner, the temperature, air volume, etc. of conditioned air sent from each outlet are automatically determined according to various parameters such as a set temperature, an outside air temperature, an inside air temperature, and an amount of solar radiation. However, there are individual differences in the sensible temperature and warmth (hot, cold, etc.) of the passenger. For this reason, the automatically determined temperature and air volume of the conditioned air may not be optimal values. In such a case, the passenger operates the operation panel as necessary to adjust the air conditioner so as to increase or decrease the set temperature, or increase or decrease the air volume. Therefore, when the passenger operates the operation panel and changes the set temperature, air volume, etc., learning control is incorporated to correct the control formula for determining the temperature and air volume of the conditioned air using various parameters at that time. An air conditioning control device has been developed (see Patent Document 1).

  By the way, the passenger changing the setting of the air conditioner is not necessarily due to a difference in warmth or the like, but may be due to an external environmental factor under a specific situation. For example, when the passenger is exercising immediately before driving, the set temperature may be lower than usual. In addition, in order to prevent the exhaust gas of the vehicle from filling the interior of the vehicle when a traffic jam is always reached, the inside air circulation mode may be set. However, the air conditioning control device described in Patent Document 1 cannot distinguish whether the setting of the air conditioning device has been changed due to an external environmental factor under a specific situation or whether the setting has been changed due to a difference in warmth or the like. Therefore, there is a possibility that learning may be performed using information that is not suitable for use in the learning control. When learning is performed using information that is not suitable for such learning, the above-described control expression may not always be corrected to match the passenger's feeling of warmth. Therefore, in order to avoid such inconvenience, it is considered to provide an operation switch that explicitly and manually instructs the air conditioner to perform learning, such as a “learning button”, on the operation panel of the air conditioner. It is done. However, learning cannot be performed without a passenger's operation, and it is difficult to say that it is completely automatic learning. Furthermore, it has been difficult to automatically optimize the air conditioning temperature or the like in accordance with the specific situation as described above.

  On the other hand, an air conditioner for automobiles has been developed in which data indicating the position of a traveling vehicle is added to learning data so that temperature adjustment learning and other learning can be distinguished (see Patent Document 2). However, the automotive air conditioner described in Patent Literature 2 refers to the position and date / time of the host vehicle and determines whether or not to perform temperature adjustment learning. When air conditioning is set for a cause, for example, it is not possible to cope with a case where the set temperature is lowered because the passenger has just exercised. Further, it has not been considered until the air conditioning temperature and the like are optimized in accordance with the specific situation as described above.

JP 2000-293204 A JP 2000-62431 A

  The objective of this invention is providing the control method and control apparatus of a vehicle air conditioner which can eliminate the trouble by the prior art mentioned above.

  Another object of the present invention is to provide a vehicle air conditioner and a control method therefor that are capable of learning air conditioning settings in accordance with a passenger's sense of temperature without being affected by a specific situation.

  Still another object of the present invention is to provide a vehicle air conditioner capable of automatically performing optimum air conditioning settings under a specific situation as well as optimizing according to a passenger's sense of temperature, and a control method therefor There is to do.

  Still another object of the present invention is to provide a vehicle air conditioner capable of automatically learning an optimal air conditioning setting for a passenger's feeling of warmth or a specific situation, and a control method thereof.

  According to the first aspect of the present invention, the vehicle air conditioner according to the present invention calculates the control information based on the air conditioning information which is one of the state information and the setting information input through the operation unit (59). And an air conditioning control unit (65) for controlling the air conditioning unit (10) for supplying the conditioned air to the vehicle interior of the vehicle according to the control information calculated by the control formula, and a predetermined unit through the operation unit (59). Whenever a setting operation is performed, the storage unit (61) that stores state information at the time of the setting operation and when the number of operations for performing a predetermined setting operation exceeds a predetermined number, the storage unit (61) stores A learning unit (66) that corrects the control formula when the state information related to the setting operation is selected based on the state information thus selected, and the selected state information does not include information other than the air conditioning information. In the specific situation. Sarezu, it is possible to perform the learning of the combined air-conditioning set to warming of the passenger. The state information includes at least one of air conditioning information inside and outside the vehicle, vehicle position information, vehicle behavior information, time information, or vehicle passenger biometric information. In addition, the air conditioning information includes the inside temperature inside the vehicle, the outside temperature outside the vehicle, and the amount of solar radiation. The predetermined setting operation refers to an operation for changing the operating state of the vehicle air conditioner, such as changing the set temperature, changing the air flow, setting the inside air circulation mode, or operating or stopping the defroster. The setting information refers to information that defines the operation of the vehicle air conditioner, such as the set temperature, the air volume, the intake ratio of the inside and outside air, and the air volume ratio of the conditioned air sent from each outlet. Further, the control information refers to information that is obtained based on setting information such as the temperature of the air-conditioned air, the rotation speed of the blower fan, and the opening degree of the air mix door, and controls the operation of each part of the air-conditioning unit. Furthermore, the control expression is not limited to a function described in the form of a function, and may be a reference table or a map representing an output value with respect to an input value.

Also, the vehicle air conditioner according to the present invention has at least one probabilistic model associated with a given setting operation, by inputting the status information on at least one probabilistic model, a predetermined setting operation line A control information correction unit (64) that corrects control information or setting information using correction information associated with at least one probability model when the appearance probability is calculated and the appearance probability satisfies a predetermined condition. Is preferred. In this case, it is preferable that the learning unit (66) constructs a probability model related to a predetermined setting operation when the selected state information includes information other than the air conditioning information.
With such a configuration, it is possible to automatically perform optimal air conditioning control corresponding to a specific situation as well as a passenger's sense of warmth. The correction information associated with the probability model is used to change the setting information or the control information after the correction or the setting information or the control information to a desired correction value in the correction specified by the probability model. The amount of correction added to or multiplied by the setting information or control information.

Furthermore, as described in claim 2 , the learning unit (66) uses the state information stored in the storage unit (61) in association with a predetermined operation, and the conditions of the nodes included in the graph structure. By determining the attached probability, a temporary probability model that calculates the probability of occurrence of a predetermined setting operation is constructed, and the most appropriate temporary probability model is selected from the temporary probability models according to predetermined determination criteria Then, it is preferable to select the state information as input information of the selected temporary probability model as the state information related to the predetermined setting operation.

Alternatively, as described in claim 3 , the learning unit (66) includes a plurality of standard models having a predetermined graph structure, and when the number of operations is equal to or greater than a predetermined number, A temporary probability model for calculating an appearance probability that a predetermined setting operation is performed is determined by determining conditional probabilities of nodes included in a predetermined graph structure using the state information stored in the storage unit (61) In the temporary probability model, the most appropriate temporary probability model is selected according to a predetermined criterion, and the state information that is input information of the selected temporary probability model is the state related to the predetermined setting operation. It is preferable to select as information.
According to the configuration of claim 2 or 3 , state information related to a predetermined setting operation among a large number of state information, that is, a state that is considered as a factor for determining whether or not the predetermined setting operation is performed Information can be selected appropriately.

Further, as described in claim 4 , the predetermined determination criterion is an information amount criterion, and the learning unit (66) has a minimum or maximum value of the information amount criterion calculated for each provisional probability model. It is preferable to select a temporary probability model.
By obtaining a probability model for calculating the appearance probability of performing a predetermined setting operation and examining the input information, it is possible to select information that is truly related to the setting operation.

Furthermore, as described in claim 5 , when the selected state information includes information other than the air conditioning information, the learning unit (66) preferably uses the provisional probability model as a probability model related to a predetermined setting operation. .

According to the sixth aspect of the present invention, the setting information input through the air conditioning unit (10) for supplying the conditioned air to the passenger compartment of the vehicle and the air conditioning information as one of the state information and the operation unit (59). Provided is a control method for a vehicle air conditioner having an air conditioning control unit (65) for controlling an air conditioning unit (10) according to the control information calculated based on the control formula. Is done. In this control method, every time a predetermined setting operation is performed through the operation unit (59), the step of storing the state information at the time of the setting operation, and the number of operations for performing the predetermined setting operation has become a predetermined number of times In this case, based on the state information stored in the storage unit (61), a step of selecting state information related to a predetermined setting operation and whether or not the selected state information includes information other than air conditioning information is determined. And a step of correcting the control expression when it is determined in the determining step that the selected state information does not include information other than the air conditioning information.

Also, the control method according to the present invention, in the above determination step, if the status information selected is determined to include information other than the air conditioning information, calculates the probability of performing a predetermined setting operation based on the state information It is preferable to have a step of constructing a probability model to do.

Furthermore, according to another aspect , the control device according to the present invention includes an information acquisition unit that acquires state information including the first information and the second information, and a setting corresponding to a setting operation of the device to be controlled. An operation unit (59) for outputting information, and a control unit (65) for calculating control information by inputting the first information and setting information into a predetermined control equation and controlling the control target device according to the control information Each time a predetermined setting operation is performed through the operation unit (59), the storage unit (61) stores state information at the time of the predetermined setting operation, and the state information stored in the storage unit (61). A learning unit (66) that selects state information related to a predetermined setting operation and corrects a predetermined control expression when the selected state information does not include information other than the first information. It is characterized by.

  In addition, the code | symbol in the parenthesis attached | subjected to each said means is an example which shows a corresponding relationship with the specific means as described in embodiment mentioned later.

Hereinafter, a vehicle air conditioner to which the present invention is applied will be described.
The vehicle air conditioner to which the present invention is applied estimates the passenger's air conditioning setting operation based on at least one probability model learned according to the passenger's sense of temperature or specific situation, and automatically sets the air conditioning. Is what you do. In particular, learning can be performed separately for the passenger's sense of temperature and for the specific situation, so not only can the air conditioning be set automatically according to the passenger's sense of temperature, but also depending on the particular situation. The optimum air conditioning setting can be performed automatically.

  FIG. 1 is a configuration diagram showing the overall configuration of the vehicle air conditioner 1. As shown in FIG. 1, the vehicle air conditioner 1 includes an air conditioning unit 10 mainly composed of a mechanical configuration, and a control unit 60 that controls the air conditioning unit 10.

  First, the configuration of the refrigeration cycle R of the air conditioning unit 10 will be described. The refrigeration cycle R of the vehicle air conditioner 1 is configured as a closed circuit, and the closed circuit includes a condenser 15, a receiver 16, an expansion valve 17, and an evaporator 18 in a clockwise direction from the compressor 11. Then, the compressor 11 compresses the refrigerant into high pressure gas. Further, the compressor 11 includes an electromagnetic clutch 14 for power interruption that is transmitted from the vehicle-mounted engine 13 via the belt 12. The condenser 15 cools and liquefies the high-temperature and high-pressure refrigerant gas sent from the compressor 11. The receiver 16 stores the liquefied refrigerant gas. Further, in order to prevent the cooling performance from being deteriorated, gaseous bubbles contained in the liquefied refrigerant are removed, and only the completely liquefied refrigerant is sent to the expansion valve 17. The expansion valve 17 adiabatically expands the liquefied refrigerant to lower the temperature and pressure, and sends the refrigerant to the evaporator 18. The evaporator 18 performs heat exchange between the low-temperature and low-pressure refrigerant and the air sent to the evaporator 18 to cool the air.

  Next, the configuration in the air conditioning case 20 of the air conditioning unit 10 will be described. A blower fan 21 is disposed on the upstream side of the evaporator 18. The blower fan 21 is a centrifugal blower fan, and is driven to rotate by a drive motor 22. An inside / outside air switching box 23 is arranged on the suction side of the blower fan 21. Inside the inside / outside air switching box 23, an inside / outside air switching door 25 driven by an inside / outside air servomotor 24 is arranged. The inside / outside air switching door 25 opens and closes by switching between the inside air inlet 26 and the outside air inlet 27. Then, the air taken in from the inside air suction port 26 or the outside air suction port 27 is sent to the evaporator 18 by the blower fan 21 via the inside / outside air switching box 23. In addition, the air volume sent from the vehicle air conditioner 1 can be adjusted by adjusting the rotational speed of the blower fan 21.

  On the downstream side of the evaporator 18, an air mix door 28 and a heater core 29 are arranged in this order from the evaporator 18 side. The heater core 29 is circulated and supplied with cooling water used for cooling the in-vehicle engine 13 in order to warm the air passing through the heater core 29. The air conditioning case 20 has a bypass passage 30 that bypasses the heater core 29. The air mix door 28 is rotated by the temperature control servo motor 31 and passes through the bypass passage 30 and the warm air from the passage 32 passing through the heater core 29 in order to bring the air sent from each outlet to a predetermined temperature. Adjust the air volume ratio with the cool air.

  Further, on the downstream side of the air mixing section 33 where the cool air passing through the bypass passage 30 and the warm air from the passage 32 passing through the heater core 29 are mixed, a foot outlet 34 for sending conditioned air into the vehicle, a face A blowout port 35 and a defroster blowout port 36 are provided. Each outlet is provided with a foot door 37, a face door 38, and a defroster door 39 for opening and closing each outlet. The foot outlet 34 sends conditioned air to the feet of the driver seat or passenger seat, and the face outlet 35 sends conditioned air from the front panel toward the driver seat or passenger seat. Further, the defroster outlet 36 sends conditioned air toward the windshield. Each door 37, 38 and 39 is driven by a mode servo motor 40.

Next, various sensors that function as an information acquisition unit included in the vehicle air conditioner 1 will be described. The inside air temperature sensor 51 is installed together with an aspirator on an instrument panel or the like in the vicinity of the steering wheel in order to measure the temperature _{Tr in the} vehicle interior. The outside air temperature sensor 52 for measuring the temperature T _am of the vehicle exterior, is located outside the front of the vehicle front radiator grille of the condenser 15. Further, in order to measure the intensity (intensity of solar radiation) S of the sunlight shining into the vehicle interior, a solar radiation sensor 53 is attached in the vicinity of the windshield in the vehicle interior. The solar radiation sensor 53 is composed of a photodiode or the like. The inside air temperature T _r , the outside air temperature _Tam and the solar radiation amount S acquired by these sensors are used as air conditioning information, and are used by the control unit 60 to perform temperature control and air volume control. Details of temperature control and air volume control will be described later.

  Furthermore, an evaporator outlet temperature sensor for measuring the temperature of the air blown out from the evaporator 18 (evaporator outlet temperature), a heater inlet water temperature sensor for measuring the coolant temperature of the engine coolant to the heater core 29, and refrigeration A pressure sensor for measuring the pressure of the refrigerant circulating in the cycle R is provided. In addition, in the passenger compartment, one or more in-vehicle cameras 54 that function as a passenger information acquisition unit and photograph the faces of passengers on the driver's seat and other seats are installed. In addition, an out-of-vehicle camera 55 that captures the appearance outside the vehicle is also installed.

The vehicle air conditioner 1 acquires position information such as the current position of the vehicle, the traveling direction, the surrounding area information, and Gbook information from the navigation system 56 in addition to the sensing information from each sensor described above. Also, various operation information such as accelerator opening, steering wheel, brake, power window opening, wiper, turn lever or car audio ON / OFF, vehicle speed, vehicle behavior information, and the like are acquired from the vehicle operating device 57. Further, time information such as the day of the week and the current time is acquired from the in-vehicle clock 58. In addition, the vehicle air conditioner 1 may acquire an occupant's biological information by installing an electrocardiogram detection sensor, a heartbeat / respiration sensor, a body temperature sensor, a skin temperature sensor, or the like in the driver's seat.
Thus, the navigation system 56, the vehicle operating device 57, and the in-vehicle clock 58 also function as an information acquisition unit.

FIG. 2 is a functional block diagram of the control unit 60 of the vehicle air conditioner 1.
The control unit 60 includes one or a plurality of microcomputers (not shown) including a CPU, ROM, RAM, etc. (not shown), peripheral circuits thereof, a storage unit 61 consisting of an electrically rewritable nonvolatile memory, The communication unit 62 communicates with various sensors, the navigation system 56, the vehicle operation device 57, and the like according to an in-vehicle communication standard such as a control area network (CAN).

  Furthermore, the control unit 60 includes a collation unit 63, a control information correction unit 64, an air conditioning control unit 65, and a learning unit 66 as functional modules realized by the microcomputer and a computer program executed on the microcomputer.

  When the control unit 60 acquires state information such as the sensing information, position information, and vehicle behavior information from various sensors, navigator systems, vehicle operating devices, and the like, the control unit 60 temporarily stores them in the RAM. Similarly, the operation signal acquired from the A / C operation panel 59 as the operation unit is also temporarily stored in the RAM. Then, the control unit 60 controls the air conditioning unit 10 based on the state information and the operation signal. For example, the control unit 60 controls the electromagnetic clutch 14 to switch the compressor 11 on and off, and controls the drive motor 22 for adjusting the rotational speed of the blower fan 21. The controller 60 controls the inside / outside air servo motor 24, the temperature control servo motor 31, and the mode servo motor 40 to adjust the opening of each door. By performing these controls, the air volume ratio of the conditioned air sent from each outlet, the overall air volume, and the temperature are adjusted so that the temperature inside the vehicle approaches the temperature set by the passenger. Here, the control unit 60 inputs predetermined state information into an available probability model in order to determine the temperature of the conditioned air, the air volume, etc., and the passenger performs a predetermined operation (for example, lowers the set temperature, air volume, , Etc., and the probability of performing the setting in the inside air circulation mode, etc. If the probability is equal to or higher than a predetermined threshold, the predetermined operation is automatically performed.

  Furthermore, when the passenger operates the vehicle air conditioner 1, the control unit 60 accumulates the operation content and various information at the time of the operation. When a predetermined number of such information is accumulated, a statistical learning process is performed to generate a probability model. Or the control formula used for temperature control is corrected. Hereinafter, each functional module that performs these operations will be described.

  When the engine switch is turned on, the verification unit 63 performs verification and authentication of the passenger based on the image captured by the in-vehicle camera 54 and the verification information regarding the registered user registered in advance in the vehicle air conditioner 1. , It is determined which registered user the passenger is. Then, the identification information (ID) of the registered user determined to be the passenger and the personal information related to the registered user are read from the storage unit 61.

  Here, the collation part 63 performs a collation and authentication of a passenger, for example with the following method. The collating unit 63 binarizes an image photographed by the in-vehicle camera 54 or performs edge detection to identify an area corresponding to the passenger's face. Then, characteristic parts such as eyes, nose and lips are detected from the identified face area by means such as edge detection, and the size and relative positional relationship of the characteristic parts are used as a set of feature values. Extract. Next, the collation unit 63 compares the extracted feature quantity set with the feature quantity set obtained for each registered user stored in the storage unit 61 in advance, and uses a correlation calculation or the like. The degree of coincidence is calculated. And when the highest matching degree becomes more than a predetermined threshold value, the collation part 63 authenticates a passenger as the registered user who became the highest matching degree. In addition, said collation method is only an example and the collation part 63 can perform a collation and authentication of a passenger using another known collation method. For example, the collation unit 63 can use a face authentication system for a vehicle described in JP-A-2005-202786. It is also possible to use a method other than image authentication. For example, a passenger may be verified and authenticated using a smart key system. Further, as in the vehicle antitheft device described in Japanese Patent Application Laid-Open No. 2005-67353, the smart key system and image authentication may be combined for verification and authentication.

Based on the probability model, the control information correction unit 64 determines whether or not to automatically adjust the control parameters of the air conditioner 1, which are setting information that can be directly set by the passenger, such as the set temperature T _set and the air volume W.
In this embodiment, a Bayesian network is used as the probability model. A Bayesian network models a probabilistic causal relationship between a plurality of events, and is a network represented by an acyclic directed graph in which propagation between nodes is obtained with a conditional probability. For details on the Bayesian network, see Kazuo Shigeru et al., “Introduction to the Bayesian Network”, first edition, Baifukan, July 2006, or supervised by Morio Onoe, “Pattern Identification”, first edition, New Technology Communications, July 2001. And the like.

In the present embodiment, the probability model is generated for each user registered in the vehicle air conditioner 1. Further, the probability model is generated for each setting operation (for example, lowering or raising the _set temperature Tset, adjusting the air volume W, or setting the inside air circulation mode). The storage unit 61 stores probability model structure information in association with each user information and setting operation. Specifically, the user structure is identified along with a graph structure representing the connection relationship between each node constituting the probability model, the type of input information given to the input node, and a conditional probability table (hereinafter referred to as CPT) of each node. The number (ID), the setting operation number k uniquely corresponding to the content of the setting operation, the control parameter corrected by the setting operation and the correction value (for example, when the set temperature T _set is lowered by 3 ° C., (T _set , -3) and (W, W _max ) etc.) are defined for each probability model when the air volume W is _set to the maximum value W _max and stored in the storage unit 61.

  The control information correction unit 64 reads out from the storage unit 61 a probability model associated with a registered user identified as a passenger by the verification unit 63. The control information correction unit 64 inputs predetermined state information to each of the read one or more probability models, and obtains the probability that the passenger performs the setting operation associated with each probability model. That is, the probability of performing the setting operation that is uniquely defined for each probability model and is represented by the setting operation number k stored in the storage unit 61 together with each probability model is obtained. The probability can be calculated using, for example, a belief propagation method. Then, if the calculated probability is equal to or higher than a first threshold Th1 (for example, Th1 = 0.9) that is considered to be almost certain that the passenger will perform the setting operation, the control information correcting unit 64 performs the setting operation. Is automatically executed. Specifically, the control parameter value associated with the setting operation is associated with the probability model, that is, uniquely defined for the probability model, and stored in the storage unit 61 together with each probability model. Correct using the correction value of.

  In addition, when the calculated probability is less than the first threshold Th1, but is greater than or equal to a second threshold Th2 (for example, Th2 = 0.6) that is considered to be highly likely that the passenger will perform the setting operation, The control information correction unit 64 displays the contents of the setting operation through a display unit such as the A / C operation panel 59 or the navigation system 56 to notify the passenger. And it confirms whether a passenger performs the setting operation. When an operation for authorizing the passenger to perform the setting operation (for example, pressing a predetermined operation button) is performed through the A / C operation panel 59 or the like, the control information correction unit 64 performs the setting operation. . The passenger may be notified of the setting operation content by voice through the A / C operation panel 59 or the navigation system 56. Further, by connecting a microphone to the vehicle air conditioner 1 and installing a voice recognition program in the control unit 60, it may be confirmed whether or not the setting operation is performed in response to the voice of the passenger.

Hereinafter, an example in which the set temperature T _set is lowered by 3 ° C. will be described. Here, the first threshold Th1 is set to 0.9, and the second threshold Th2 is set to 0.6.
FIG. 3 shows an example of such a specific situation. The situation shown here is that the passenger (Mr. A) always plays tennis in an athletic park on Saturday afternoon, and then gets into a private car at around 4 o'clock to lower the set temperature of the vehicle air conditioner than usual. Like that. On the other hand, in other cases, for example, a case where such a setting operation is not performed when going home from work is considered.
FIG. 4 shows a graph structure of an example of a probability model used for automatically adjusting the control parameters of the vehicle air conditioner 1. In the probability model 101 shown in FIG. 4, three input nodes 102, 103, and 104 are connected to the output node 105, respectively. Further, the day of the week (x ₁ ), the time zone (x ₂ ), and the current position (x ₃ ) are given to the input nodes 102, 103, and 104 as state information that is input. The output node 105 outputs the probability of lowering the set temperature T _set by 3 ° C.

  5A to 5D show CPTs 106 to 109 for the nodes of the probability model 101 shown in FIG. The CPTs 106 to 108 correspond to the input nodes 102 to 104, respectively, and define prior probabilities for input state information. The CPT 109 corresponds to the output node 105 and defines a conditional probability distribution assigned to each input node information value.

Here, if the day of the week is Saturday (x ₁ = 1), the time zone is noon (x ₂ = 1), the current position is park (x ₃ = 1), and all the information given to each input node is known, the set temperature The probability P (x ₄ = 1 | x ₁ = 1, x ₂ = 1, x ₃ = 1) of lowering T _set by 3 ° C. is 0.95 from FIG. Therefore, since the obtained probability is equal to or higher than the first threshold Th1, the control information correcting unit 64 corrects the control parameter so as to lower the set temperature T _set by 3 ° C.

If the day of the week is Saturday (x ₁ = 1) and the time zone is noon (x ₂ = 1), for example, when the navigation system 56 is not turned on and the current position cannot be known, FIG. P (x ₄ = 1 | x ₁ = 1, x ₂ = 1, x ₃ ) is calculated using the prior probability P (x ₃ ) when the current position shown in (c) is a park. in this case,
P (x ₄ = 1 | x ₁ = 1, x ₂ = 1, x ₃ )
= P (x ₄ = 1 | x ₁ = 1, x ₂ = 1, x ₃ = 1) ・ P (x ₃ = 1)
+ P (x ₄ = 1 | x ₁ = 1, x ₂ = 1, x ₃ = 0) ・ P (x ₃ = 0)
= 0.95 ・ 0.15 + 0.55 ・ 0.85 = 0.61
It becomes. Therefore, since the obtained probability is smaller than the first threshold value Th1 but is equal to or greater than the second threshold value Th2, the control information correction unit 64 determines whether or not to lower the set temperature _Tset by 3 ° C. Confirm with the passenger through the operation panel 59 or the like.

Furthermore, if the day of the week is Monday (x ₁ = 0), the time zone is night (x ₂ = 0), and the current location is the workplace (x ₃ = 0), the probability P (x ₄ = 1 | x ₁ = 0, x ₂ = 0, x ₃ = 0) is 0.1 from FIG. Therefore, the probability obtained because less than the first threshold value Th1 and the second threshold value Th2, the control information correcting unit 64 does not change the set temperature T _{The set,} for changing the set temperature T _{The set,} We do not check with the passenger.

  In the above example, a two-layer network configuration is used for simplification, but a three-layer or more network configuration including an intermediate layer may be used. Moreover, although the day of the week as input information is divided into Saturday and other than that, it may be divided into other divisions, for example, every day of the week. Similarly, the current position may not be classified into a park and others, but may be classified for each place where a passenger frequently visits. Further, the time zone may be further subdivided or divided into morning, afternoon and the like.

In addition, when there are multiple probability models related to the same operation group (such as correction of set temperature, change of air volume, switching of inside / outside air or setting of air volume ratio), that is, the probability of correcting a specific control parameter When there are a plurality of output probability models, the control information correction unit 64 calculates the probability for each of the plurality of probability models. The specific control parameters include air volume, switching between inside and outside air, air volume ratio, and the like. Then, among the obtained probabilities, the largest one is selected and the above processing is performed. For example, consider a case where there are a probability model M1 (maximizing the air volume W) and M2 (making the air volume W medium) regarding the air volume setting. In this case, the control information correction unit 64 obtains the probability P _M1 that maximizes the air volume W based on the probability model M1, and similarly calculates the probability P _M2 that makes the air volume W moderate based on the probability model M2. . Then, if P _M1 > P _M2 , the control information correction unit 64 compares P _M1 with the threshold values Th1 and Th2, and determines whether or not to maximize the air volume W. On the other hand, if P _M2 > P _M1 , P _M2 is compared with the above-mentioned threshold values Th1 and Th2, and it is determined whether or not the air volume W is moderate.

In the above, in order to facilitate understanding, it is defined that the probability models M1 and M2 are associated with different setting operations. However, the probability models M1 and M2 may be associated with the same setting operation (for example, both maximize the air volume W). This corresponds to, for example, that the same operation may be performed in two or more situations where one passenger is different (one is sunny in the day and the other is on the way back to the gym). . If probability models corresponding to the respective situations are generated, these probability models are associated with setting operations belonging to the same operation group.
When the control information correction unit 64 corrects the control parameters such as the _set temperature T _set and the air volume W according to the above processing as necessary, the control information correction unit 64 controls the control parameters so that the control parameters can be used by each unit of the control unit 60. Temporarily stored in the RAM of the unit 60.

  The air conditioning control unit 65 reads out the values of the control parameters and the sensing information acquired from the sensors from the RAM, and controls the air conditioning unit 10 based on those values. For this purpose, the air conditioning control unit 65 includes a temperature adjustment unit 651, a compressor control unit 652, an outlet control unit 653, an inlet control unit 654, and an air flow rate setting unit 655. In addition, when the control parameter corrected by the control information correction unit 64 is stored in the RAM, the air conditioning control unit 65 reads out and uses the corrected parameter.

上式において、Doは、エアミックスドア２８の開度を表す。また、係数k_set、k_r、k_am、k_s、C、a、bは定数であり、T_set、T_r、T_am、Sは、それぞれ、設定温度、内気温、外気温及び日射量を表す。ここで、制御情報修正部６４が設定温度T_setを修正している場合、その修正された設定温度T_setを使用する。また、エアミックスドア２８の開度Doは、ヒータコア２９を経由する通路３２を閉じた状態（すなわち、冷房のみが動作する状態）を０％、バイパス通路３０を閉じた状態（すなわち、暖房のみが動作する状態）を１００％として設定される。温調制御式の各係数k_set、k_r、k_am、k_s、C及びエアミックスドアの開度を求める関係式の係数a、bは温調制御パラメータとして、登録済利用者ごとに設定され、登録済利用者の個人設定情報に含まれる。
なお、温度調節部６５１は、空調温度T_ao及びエアミックスドア２８の開度を、ニューラルネットワークを用いた制御やファジイ制御など、他の周知の制御方法を用いて決定してもよい。算出された空調温度T_aoは、制御部６０の他の部で参照できるように、記憶部６１に記憶される。 Based on the set temperature T _set and the measurement signals of the temperature sensors and the solar radiation sensor 53, the temperature adjustment unit 651 determines the necessary blow-out temperature (air-conditioning temperature T _ao ) of the conditioned air sent from each blow-out port. And the opening degree of the air mix door 28 is determined so that the temperature of the conditioned air becomes the air conditioning temperature _Tao, and the opening degree of the air mix door 28 is set to the temperature control servo motor 31. A control signal is transmitted to. For example, the opening degree of the air mix door 28 is obtained by inputting a value obtained by correcting the difference between the internal temperature _Tr and the set temperature T _set by the outside air temperature T _am , the solar radiation amount S, etc. It is determined based on the control expression. Here, the opening degree of the air mix door 28 is determined at regular time intervals (for example, every 5 seconds). A temperature control expression for _{obtaining the} air conditioning temperature _Tao from each measurement value for performing such control and a relational expression of the opening degree of the air mix door 28 are shown below.

In the above formula, Do represents the opening degree of the air mix door 28. The coefficients k _set , k _r , k _am , k _s , C, a, and b are constants, and T _set , T _r , T _am , and S are the set temperature, internal temperature, external temperature, and solar radiation, respectively. Represents. Here, when the control information correcting unit 64 corrects the set temperature T _set , the corrected set temperature T _set is used. The opening degree Do of the air mix door 28 is 0% when the passage 32 passing through the heater core 29 is closed (that is, only the cooling operation is performed), and when the bypass passage 30 is closed (that is, only heating is performed). The operating state) is set as 100%. Each coefficient k _set , k _r , k _am , k _s , C of the temperature control expression and the coefficients a and b of the relational expression for determining the opening of the air mix door are set for each registered user as temperature control parameters And is included in the personal setting information of registered users.
Note that the temperature adjustment unit 651 may determine the air conditioning temperature T _ao and the opening degree of the air mix door 28 using other known control methods such as control using a neural network and fuzzy control. The calculated air conditioning temperature _Tao is stored in the storage unit 61 so that it can be referred to by other units of the control unit 60.

The compressor control unit 652 controls ON / OFF of the compressor 11 based on the air conditioning temperature (necessary outlet temperature) T _ao , the set temperature T _{set, the} evaporator outlet temperature, and the like obtained by the temperature adjustment unit 651. The compressor control unit 652 operates the compressor 11 and the refrigeration cycle R in principle when cooling the interior of the vehicle or operating the defroster, for example. However, in order to avoid the evaporator 18 from being frosted, the compressor 11 is stopped when the evaporator outlet temperature falls close to the temperature at which the evaporator 18 is frosted. Then, when the evaporator outlet temperature rises to some extent, the compressor 11 is operated again. Since the compressor 11 can be controlled using a known method such as variable displacement control, detailed description thereof is omitted here.

The air outlet control unit 653 is configured to output each air outlet based on the set value of the air flow ratio set by the passenger through the A / C operation panel 59, the air conditioning temperature T _ao obtained by the temperature adjusting unit 651, the set temperature T _{set, and the} like. The air volume ratio of the conditioned air sent from the air is determined, and the opening degree of the foot door 37, the face door 38, and the defroster door 39 is determined so as to correspond to the air volume ratio. The air outlet control unit 653 adjusts the opening degree of each door 37 to 39 according to a control expression representing the relationship between the set value of the air flow ratio, the air conditioning temperature T _ao , the setting temperature T _set and the opening degree of each door 37 to 39. decide. Such a control formula is defined in advance and is incorporated in a computer program executed in the control unit 60. In addition, the blower outlet control part 653 can also determine the opening degree of each door 37-39 using another known method. And the mode servomotor 40 is controlled so that each door 37-39 becomes the determined opening degree.
Further, when the control information correction unit 64 corrects the set value or the set temperature T _set of the air flow ratio, each of the outlet control units 653 uses the corrected set value or the set temperature T _set. The opening degree of the doors 37 to 39 is determined.

The air inlet control unit 654 is configured to allow the vehicle air conditioner 1 to move from the air inlet 26 based on the air inlet setting, the set temperature T _set , the air conditioning temperature T _ao , the internal air temperature _{Tr, and the} like acquired from the A / C operation panel 59. The ratio of the air to be sucked and the air to be sucked from the outside air inlet 27 is set. Inlet control unit 654 determines the opening degree of outside air switching door 25 according to the control equation that represents the relationship between the outside temperature T _am, like the difference between the inner temperature T _r and the set temperature T _set and the intake ratio. Such a control formula is set in advance and is incorporated in a computer program executed in the control unit 60. In addition, the suction inlet control part 654 can also determine the opening degree of the inside / outside air switching door 25 using another known method. The inlet control unit 654 controls the inside / outside air servomotor 24 to rotate the inside / outside air switching door 25 so as to obtain the obtained intake ratio. In addition, when the control information correction unit 64 corrects the intake set value or set temperature T _set , the intake port control unit 654 uses the corrected intake set value or set temperature T _set to change the inside / outside air The opening degree of the switching door 25 is determined.

Air volume setting unit 655, based on air volume W acquired from A / C control panel 59, the set temperature T _{The set,} air-conditioning temperature T _ao, inside temperature T _r, and the like outside temperature T _am and the amount of solar radiation S, the blower fan 21 Determine the rotation speed. Then, a control signal is transmitted to the drive motor 22 so that the rotational speed of the blower fan 21 becomes a set value. For example, when the air volume setting is a manual setting, the air volume setting unit 655 determines the rotation speed of the blower fan 21 so that the air volume W obtained from the A / C operation panel 59 is obtained. In addition, when the air volume setting is automatically set, the air volume setting unit 655 rotates the blower fan 21 according to the air volume control expression that represents the relationship between the internal air temperature T _r , the air conditioning temperature T _ao , and the like and the air volume W. Determine the speed. Alternatively, the air volume control expression may directly represent the relationship between the _set air temperature T _set, air conditioning information (inside air temperature T _r , outside air temperature _Tam, and solar radiation amount S) and the air volume W. As such an airflow control formula, various well-known ones can be used. Such a control formula is set in advance and incorporated in a computer program executed in the control unit 60. Alternatively, the air flow rate setting unit 655 prepares a map that defines the relationship between the air conditioning information and the air flow W in advance, and performs map control that determines the air flow W corresponding to the air conditioning information measured with reference to the map. The rotation speed of the blower fan 21 can also be determined using other known methods. In addition, when the control information correction unit 64 corrects the air volume W or the set temperature T _set , the air volume setting unit 655 uses the corrected air volume W or the set temperature T _set of the blower fan 21. Determine the rotation speed.

  The learning unit 66 determines whether or not to generate a new probability model or to update an existing probability model when a passenger operates the air conditioner 1 for a vehicle. In this case, a probability model is generated or updated. In addition, the learning unit 66 corrects a control expression such as the above-described temperature control control expression or air flow control expression as necessary.

ここで、d_ijkは、各状態情報の値である。iは、上記の操作回数i_Akを示す。また、jは、状態情報の各値に対して便宜的に指定される状態項目番号であり、本実施形態では、j=1に対して内気温T_r、j=2に対して外気温T_am、j=3に対して日射量Sが割り当てられる。そして、j=4以降に、位置情報、車両挙動情報、生体情報などが割り当てられる。また、kは設定操作番号である。
これら学習情報D_Ak及び操作回数i_Akは、登録済み利用者及び設定操作ごとに別個に記憶される。 Generally, the passenger performs a setting operation of the vehicle air conditioner 1 when the interior of the vehicle is not in an air conditioning state appropriate for the passenger. Therefore, when the passenger frequently performs the setting operation of the vehicular device 1, it is considered necessary to construct a probability model for estimating the passenger's setting operation. However, in order to construct an appropriate probabilistic model, enough data is needed to make a statistically correct estimate. Therefore, each time the setting operation of the vehicle air conditioner 1 is performed, the learning unit 66 obtains each state information (air condition information such as the outside temperature Tam, position information such as the current position of the vehicle, vehicle speed, etc.) acquired at the time of the operation. Vehicle behavior information, biological information such as heart rate) is stored as learning information in the storage unit 61 in association with the setting operation number k and the passenger ID described above. Further, the number of operations i _Ak that the passenger A has performed the setting operation α corresponding to the setting operation number k (for example, lowering the set temperature by 3 ° C. or maximizing the air volume W) is also stored in the storage unit 61. The learning information D _Ak is expressed by the following equation, for example.

Here, _dijk is a value of each state information. i indicates the number of operations i _Ak described above. Further, j is a state item number designated for convenience with respect to each value of the state information. In this embodiment, the internal temperature T _r for j = 1 and the external temperature T for j = 2. _The solar radiation amount S is assigned to _am and j = 3. Then, after j = 4, position information, vehicle behavior information, biological information, and the like are assigned. K is a setting operation number.
The learning information D _Ak and the number of operations i _Ak are stored separately for each registered user and setting operation.

When the operation number i _Ak becomes equal to a predetermined number n1 (for example, 10 times), the learning unit 66 uses the learning information D _Ak stored in the storage unit 61 to construct a probability model M _Aqk related to the setting operation. To do. Note that q (= 1, 2,...) Represents the number of probability models constructed for the setting operation of the setting operation number k of the passenger A. Thereafter, when the occupant A further repeats the setting operation α, the learning unit 66 _{causes the} number of operations i _Ak after the previous construction of the probability model M _Aqk to reach n1 times (that is, the number of operations i _Ak = n1 · j (where j = 1, 2,...)), the probability information M _Aqk is updated using the learning information D _Ak stored in the storage unit 61.

Here, the learning unit 66 uses only the input parameters of the constructed probability model as input parameters of a control expression such as a temperature control control expression, that is, air conditioning information (inner temperature T _r , outside temperature _Tam, and solar radiation amount S). Judge whether to include. When the input parameter includes only the air conditioning information, it is considered that the temperature control is not optimized with respect to the sensation of the passenger, so the learning unit 66 corrects the control equation. Note that which control expression is to be corrected is determined in relation to the setting operation α. If the setting operation α is related to the setting of the air conditioning temperature, the learning unit 66 corrects the temperature control expression, and if the setting operation α is related to the setting of the air volume, the learning unit 66 corrects the air volume control expression.
On the other hand, when the input parameter includes information other than the air conditioning information, it is considered that the probability model for performing the air conditioning setting corresponding to the specific situation is constructed, and thus the control formula is not corrected.

When the number of operations i _Ak becomes equal to a predetermined number n2 (for example, 30 times), the learning unit 66 establishes the probability model M _Aqk stored in the storage unit 61 at that time, and thereafter the probability model M _Aqk is not updated. The learning unit 66 _attaches flag information indicating that the established probability model M _Aqk is not updated. For example, the update flag f is associated with the probability model and stored in the storage unit 61. When the update flag f is “1”, updating (that is, rewriting) is prohibited, and when the update flag f is “0”, updating is possible. As shown in FIG. Then, the learning unit 66 erases the learning information D _Ak stored in the storage unit 61, initializes the number of operations i _Ak , and resets the value to 0. Note that the predetermined number of times n2 is larger than n1 and corresponds to the number of data considered that a statistically sufficiently accurate probability model can be constructed. The predetermined times n1 and n2 can be optimized empirically and experimentally.

When the passenger A repeats the same setting operation α after the probability model M _Aqk is established, a new probability model M _{Aq + 1k} is constructed according to the same procedure as described above. In this way, by constructing a plurality of probability models as necessary, when there are a plurality of specific situations in which the same type of setting operation is performed (for example, as a situation in which an operation for setting the inside-air circulation mode is performed, In the case of being inside and in the case of being behind a large truck). In addition, for a specific situation with a high occurrence frequency, a lot of information corresponding to the situation is included in the learning information, so that a probability model corresponding to an early stage is constructed. For the specific situation in which the corresponding probability model is constructed, the control information correction unit 64 automatically performs the setting operation based on the probability inference based on the probability model. The setting operation of the device 1 is not performed. Therefore, the learning unit 66 constructs a probability model corresponding to a specific situation with a low occurrence frequency because the passenger performs a setting operation only when a specific situation with a low occurrence frequency occurs as learning progresses. You can also.

Next, the procedure for constructing the probability model will be described.
In order to construct a general-purpose probability model that can cope with various situations, it is necessary to construct a very large probability model including a large number of nodes. However, learning of such a probabilistic model requires a very long calculation time, and the hardware resources necessary for learning are enormous. Therefore, in this embodiment, two layers are selected as input parameters that are likely to be particularly relevant to the setting operation from the state information, and the probability of performing the setting operation based on a conditional probability for the combination of the input parameters is obtained. Fifteen types of configuration graph structures were prepared as standard models. However, the number of standard models is not limited to 15 types. The number of standard models can be optimized as appropriate depending on the number of state information obtained and the type of setting operation to be learned. Further, the standard model may have only one input parameter, or may have all the state information that can be acquired as input parameters. Further, the standard model is not limited to the graph structure having a two-layer structure, and a graph structure having three or more layers may be used as the standard model according to the ability of the CPU constituting the control unit 60.
Those standard models are stored in the storage unit 61. At the time of learning, for each standard model, a conditional probability between each node included in the standard model is determined to construct a temporary probability model. Thereafter, a temporary probability model having the most appropriate graph structure is selected using the information criterion. The selected model becomes the constructed probability model.

Hereinafter, it demonstrates in detail using figures.
FIGS. 6A to 6D show four of the 15 standard models as an example. Each of the standard models 501 to 504 shown in FIGS. 6A to 6D is a two-layer Bayesian network composed of an input node and an output node. Each standard model 501 to 504 is different in the parameters given to the input node.

In addition, for the input nodes of the standard models 501 to 504, a CPT that defines the prior probabilities for the input parameters assigned to the input nodes is set. The input information is classified using a technique such as clustering. For example, in the standard model 502 shown in FIG. 6B, clustering is performed on the input node having the current position as the input parameter (parameter y ₁₁ ) using a technique such as a k-average method based on the learning information. Perform parameter value classification. Alternatively, home when _{y 11 = 0, y 11 =} 1 when the workplace, may be determined in advance classification as y ₁₁ = 2 when the neighborhood park. Similarly, a CPT indicating a conditional probability distribution based on information given to the input node is set for the output node. In the initial state, the CPT is set to be the same value for all states.

The flowchart shown in FIG. 7 is a procedure for constructing a probability model.
When learning is started, the learning unit 66 first extracts a target input parameter from the learning information D _Ak for each standard model, _obtains a conditional probability of each node, creates a CPT, and creates a probability. A model is constructed (step S201).
Therefore, the learning unit 66 counts the number n corresponding to each parameter state for each node from the learning information D _Ak read from the storage unit 61. Then, a value obtained by dividing the number n by the total number of events N is set as the value of the prior probability and the conditional probability. For example, the standard model 502 in FIG. 6B will be described as an example. Here, there is learning information D _Ak including a set of 30 data, and when checking the current position assigned to one of the input nodes, the number of times at home (y ₁₁ = 0) is 15 times , If the number of times at work (y ₁₁ = 1) is 12 times and the number of times in the neighborhood park (y ₁₁ = 2) is 3, the prior probability P (y ₁₁ ) for the current position is P (y ₁₁ = 0) = 0.5, P (y ₁₁ = 1) = 0.4, and P (y ₁₁ = 2) = 0.1. Similarly, for the output node, for each combination of possible values of the current position (y ₁₁ ), day of the week (y ₁₂ ), and time zone (y ₁₃ ) of the input information given to each input node that is the parent node, The conditional probability can be obtained by calculating the number of occurrences in the learning information D _Ak and dividing it by 30 which is the total number of data. In this way, the CPT corresponding to each node is determined by obtaining the prior probability and the conditional probability.

Note that the learning unit 66 may estimate the probability distribution using the beta distribution when it is considered that the number of data used for learning is not sufficient. In addition, if there is no combination of values of some input information in the learning information D _Ak , that is, there is unobserved data, the probability distribution for the unobserved data is estimated, and the expected value is calculated based on the distribution. By calculating, the corresponding conditional probability is calculated. As for such conditional probability learning, for example, the method described in Kazuo Shigeru et al., “Outline of Bayesian Network”, First Edition, Baifukan, July 2006, p.35-38, p.85-87. Can be used.

ここで、AIC_mは、確率モデルＭに対するＡＩＣを表す。また、θ_mは、確率モデルＭのパラメータ集合を、l_m(θ_m|X)は、データXを所与としたときの確率モデルＭにおけるそのデータの最大対数尤度の値を、k_mは確率モデルＭのパラメータ数をそれぞれ表す。ここでl_m(θ_m|X)は、以下の手順で計算できる。まず、各ノードにおいて、親ノードの変数の各組み合わせについて、学習情報D_Akから出現頻度を求める。その出現頻度に条件付き確率の対数値を乗じた値を求める。最後にそれらの値を足し合わせることでl_m(θ_m|X)が算出される。また、k_mは、各ノードにおける、親ノード変数の組み合わせの数を足し合わせることで求められる。 When the CPT for each standard model is obtained, the learning unit 66 calculates an information criterion for each probability model in order to evaluate the constructed probability model (step S202).
In this embodiment, AIC (Akaike information criterion) is used as the information criterion. The AIC can be obtained based on the following equation based on the maximum log likelihood of the probability model and the number of parameters.

Here, AIC _m represents the AIC for the probability model M. Further, theta _m is the parameter set of probabilistic model _{M, l m (θ m |} X) is the value of the maximum log-likelihood of the data in the probability model M when the data X and a given, k _m Represents the number of parameters of the probability model M, respectively. Here, l _m (θ _m | X) can be calculated by the following procedure. First, in each node, the appearance frequency is obtained from the learning information D _Ak for each combination of the variables of the parent node. A value obtained by multiplying the appearance frequency by the logarithm of the conditional probability is obtained. Finally, l _m (θ _m | X) is calculated by adding these values. Also, k _m is at each node is calculated by adding the number of combinations of parent node variables.

When the learning unit 66 obtains the AIC for all the probability models, the learning unit 66 selects the model having the smallest AIC value as the probability model to be used, and stores it in the storage unit 61 (step S203). Then, other probability models are deleted (step S204).
Regarding the selection of the probability model using the information criterion (in other words, learning of the graph structure), other information such as Bayesian information criterion (BIC), Takeuchi information criterion (TIC), minimum description length (MDL) criterion, etc. An information criterion may be used. Furthermore, the information amount reference obtained by reversing the sign of the information amount reference calculation formula may be used. In this case, the probability model that maximizes the information criterion value is selected as the probability model to be used.

The learning unit 66 stores the constructed probability model in the storage unit 61. In addition, the passenger ID and the setting operation number k associated with the learning information D _Ak are acquired and stored in the storage unit 61 in association with the constructed probability model. Further, the control parameter and the correction value that are corrected based on the probability model are specified based on the setting operation number k, and are stored in the storage unit 61 in association with the probability model. The relationship between the setting operation number k, the control parameter to be corrected, and the correction value is defined in advance as a look-up table, for example, and held in the storage unit 61.

Next, correction of the control formula will be described.
As an example, when correcting the temperature control equation, the learning unit 66 uses the temperature control parameters K _set , K _r , K a, based on the learning information D _Ak , the set temperature T _set after the setting operation, and the air conditioning temperature T _ao . A corrected temperature control parameter is obtained by setting up simultaneous equations with K _am , K _S and C as variables and solving the simultaneous equations. Alternatively, as described in JP-A-5-147421, the learning unit 66 distributes the amount ΔT _set that the passenger changes the set temperature T _set and the amount of solar radiation S when the setting operation is performed. The set temperature change amount ΔT _set may be approximately expressed by a linear expression of the amount of solar radiation S, and the temperature control parameter K _S may be corrected based on the approximation result. Further, the learning unit 66 is provided with various other types as described in JP 2000-293204 A, JP 2000-071060 A, JP 5-208610 A, or JP 5-169963 A. You may make it correct a temperature control type | formula or an air volume control type | formula using a known method. When the air volume or the like is determined by map control, the learning unit 66 can correct the map using the learning information D _Ak based on a known method.

  Hereinafter, the air conditioning control operation of the vehicle air conditioner 1 to which the present invention is applied will be described with reference to the flowcharts shown in FIGS. 8 and 9. The air conditioning control operation is performed by the control unit 60 in accordance with a computer program incorporated in the control unit 60.

  As shown in FIG. 8, first, when the engine switch is turned on, the control unit 60 operates the vehicle air conditioner 1. And each state information is acquired from each sensor, the navigation system 56, the vehicle operation apparatus 57, etc. through the communication part 62 (step S101). Similarly, each setting information is acquired from the storage unit 61. Next, the verification unit 63 of the control unit 60 performs verification / authentication of the passenger (step S102). And the personal setting information of the registered user determined to be a passenger is read from the storage unit 61 (step S103).

Next, the control unit 60 determines whether or not the passenger has performed a setting operation of the vehicle air conditioner 1 (step S104). When an operation signal is received from the A / C operation panel 59, it is determined that a setting operation has been performed. When the passenger has not performed the setting operation, the control information correction unit 64 of the control unit 60 controls the control parameter (for example, the control parameter related to any operation group among the probability models M _Aqk associated with the passenger). The observed state information is input to the probabilistic model related to the correction of the set temperature _Tset ). Then, the probability of performing the setting operation associated with the probability model is calculated (step S105). And the highest probability is calculated | required as the appearance probability P among the probability calculated about the setting operation in the same operation group relevant to the control parameter.

  Next, the appearance probability P is compared with a first predetermined value Th1 (step S106). When the appearance probability P is equal to or greater than a first predetermined value Th1 (for example, 0.9), the control information correction unit 64 is based on correction information associated with a probability model (hereinafter referred to as a selection probability model) that has output the appearance probability P. Then, the control parameter of the corresponding vehicle air conditioner 1 is corrected (step S107). On the other hand, when the appearance probability P is less than the first predetermined value Th1, the control information correction unit 64 compares the appearance probability P with a second predetermined value Th2 (for example, 0.6) (step S108). If the appearance probability P is equal to or greater than the second predetermined value Th2, the control information correction unit 64 corresponds to the setting operation number k associated with the selection probability model through the display unit of the A / C operation panel 59 or the like. Whether or not to perform a setting operation is displayed and confirmed (step S109). If the passenger approves the setting operation, the control parameter is corrected based on the correction information associated with the selection probability model (step S107). On the other hand, if the passenger does not approve, the control parameter is not corrected. That is, the setting operation related to the control parameter associated with the selection probability model is not performed. In step S108, even when the appearance probability P is less than the second predetermined value Th2, the control parameter is not corrected.

  Thereafter, the control information correction unit 64 determines whether or not the adjustment of all the control parameters has been completed by checking whether or not the probabilities have been calculated for all the probability models (step S110). If there is a probability model for which the probability has not yet been calculated, that is, if there is an operation group for which the setting information has not been corrected, the control is returned to before step S105. On the other hand, when the probability calculation has been completed for all probability models, the air conditioning control unit 65 can obtain a desired air conditioning temperature, air volume, and the like based on the control parameters modified as necessary. Then, the number of rotations of the air mix door and the blower fan and the opening degree of each outlet are adjusted (step S111).

As shown in FIG. 9, when the passenger performs a setting operation of the vehicle air conditioner 1 in step S104, the setting signal is identified with reference to the setting signal (step S112). Then, the ID of the passenger, and the setting operation number k corresponding to the performed setting operation, the setting operation in association with an operation performed the number of times i _Ak, elements of the learning information D _Ak each state information acquired Is stored in the storage unit 61 (step S113).

Thereafter, the learning unit 66 of the control unit 60 determines whether or not the operation number i _Ak is equal to the predetermined number n1 * j (j = 1, 2, 3) (step S114). The predetermined number n1 is, for example, 10 times. Then, the learning unit 66, if it is determined that i _Ak = n1 * j, using the learned information D _Ak of the occupant and is associated with the setting operation number k stored in the storage unit 61, a probability model M _Aqk Is constructed (step S115). The probability model M _Aqk is constructed according to the procedure shown in the flowchart of FIG. Then, the probability model M _Aqk is stored in the storage unit 61 in association with the passenger ID and the like. On the other hand, if i _Ak is not equal to n1 * j in step S114, control is transferred to before step S118.

Then, the learning unit 66 determines the input parameters of the probabilistic model M _Aqk is, the air conditioning information (inside temperature T _r, the outside temperature T _am and solar radiation S) whether consisting only (step S116). If only the air conditioning information is used as an input parameter, the learning unit 66 determines that the temperature control is not optimized for the passenger's sense of temperature, and corrects the control expression related to the setting operation α. (For example, when the setting operation α is a change of the set temperature, the constants K _set , K _r , _Kam , K _S and C of the temperature control expression are adjusted) (step S117). Then, the probability model M _Aqk is discarded. On the other hand, in step S116, when information other than the air conditioning information is included in the input parameters of the probability model M _Aqk , the learning unit 66 determines that a probability model corresponding to a specific situation has been constructed. In this case, the learning unit 66 does not correct the control expression related to the setting operation α, and shifts the control to before step S118.

Next, the learning unit 66 determines whether or not the operation number i _Ak is equal to a predetermined number n2 (for example, n2 = 30) (step S118). If i _Ak is not equal to n2, i _Ak is incremented by 1 (step S119), and the control shifts to before step S111. On the other hand, if i _Ak = n2 in step S118, the learning unit 66 stores the learning information stored in the storage unit 61 in association with the passenger and the set operation number k stored in the storage unit 61. D _Ak is deleted (step S120). Also, i _Ak is initialized and i _Ak = 0. Thereafter, the control shifts before step S111.

  Thereafter, the vehicle air conditioner 1 repeats the control in steps S101 to S120 described above until the operation is stopped.

  As described above, the vehicle air conditioner to which the present invention is applied can learn separately according to the passenger's feeling of warmth and according to the specific situation, so that it is adapted to the passenger's feeling of warmth. Not only can the air conditioning be automatically set, but the optimum air conditioning can be automatically set according to a specific situation.

In addition, this invention is not limited to said embodiment. For example, the passenger is not limited to the driver. By determining who performed the setting operation of the vehicle air conditioner, it can be suitably used even when a passenger other than the driver operates. For example, when two A / C operation panels 59 for a vehicle air conditioner are prepared for a driver seat and a passenger seat, the control unit 60 determines which A / C operation panel 59 is operated, It may be determined whether the driver has operated or the passenger has operated. Further, as described in Japanese Patent Application Laid-Open No. 2002-29239, the control unit 60 is provided with an operation occupant detection sensor composed of an infrared temperature sensor or the like on the A / C operation panel 59, so It may be determined which of the persons has performed the operation.
When the passenger performs an operation, as in the case of the driver's verification and authentication, the passenger is also verified and authenticated based on the image data captured by the in-vehicle camera 54. Status information such as values is stored in association with the passenger, not the driver.

  Further, when the passenger is limited to a specific person, or when a probability model is constructed for a setting operation that is performed regardless of who is driving, the matching unit 63 may be omitted. In this case, the learning information used for learning the probability model and the probability model is used in common regardless of who the passenger is.

Further, the state information used for the construction of the probability model and the setting operation using the probability model may include setting information (setting temperature, air volume, etc.) of the vehicle air conditioner when the state information is acquired.
Furthermore, in the above embodiment, the control parameter corrected by the control information correction unit 64 is a parameter that can be directly set by the passenger through the A / C operation panel 59, such as the set temperature and the air volume. However, the control information correction unit 64 sets the control parameter to be corrected based on the probability model as the air conditioning temperature _Tao calculated using the temperature control control equation or the rotation speed of the blower fan 21 calculated using the air flow control equation. Control information directly related to the operation of each part of the air conditioning unit 10 such as the opening degree of the air mix door 28 may be used.
Further, the present invention can be widely applied to a case where an air conditioner is automatically controlled based on state information not directly related to air conditioning. For example, when the control unit 60 receives a signal for operating the wiper, the defroster is activated, the cigarette lighter is used, the outside air mode is set, or the car audio switch is turned on, Control such as lowering the air volume can be performed automatically. Furthermore, the control parameter to be automatically corrected may not be directly related to the control of the vehicle air conditioner. For example, when the air volume is set to 0, the power window may be automatically opened. In such a case, a control signal is transmitted from the controller 60 to the vehicle operating device.

Further, in the construction of the probability model, in the above-described embodiment, a standard model that defines a graph structure is prepared in advance. Instead of preparing such a standard model, a graph structure search is performed using a K2 algorithm or a genetic algorithm. May be performed. For example, in the case of using a genetic algorithm, a plurality of genes having each element as to whether or not each node is connected is prepared. Then, the fitness of each gene is calculated using the above information criterion. Select a gene with a fitness level greater than or equal to a predetermined level, and perform the operation such as crossover and mutation to create the next generation gene. Such an operation is repeated a plurality of times to select the gene having the highest fitness. The graph structure described by the selected gene is used to construct the probability model. Furthermore, these algorithms and construction of a probability model from a standard model may be used in combination.
Furthermore, in the above-described embodiment, the Bayesian network is used as the probability model, but other probability models such as a hidden Markov model may be used.

  The air conditioner to which the present invention is applied may be any type of front single, left and right independent, rear independent, four seat independent, and vertical independent. When the present invention is applied to any independent type air conditioner, a plurality of inside air temperature sensors, solar radiation sensors, and the like may be mounted.

Furthermore, the present invention can be applied to other than air conditioners. Control information is calculated according to a predetermined control expression that receives a plurality of state information, and inputs a part of the state information and setting information related to a setting operation performed by the user, and a predetermined operation according to the control information. The present invention can be applied to any apparatus that performs the above. By applying the present invention, it is possible to determine whether a setting operation performed by the user is caused by state information used as an input of the predetermined control equation or other state information, thereby determining a predetermined control equation. It is possible to appropriately determine whether or not to correct.
As described above, various modifications can be made within the scope of the present invention.

It is a block diagram which shows the whole structure of the vehicle air conditioner to which this invention is applied. It is a functional block diagram of the control part of a vehicle air conditioner. It is a figure which shows an example of a specific situation. It is a figure which shows the graph structure of an example of the probability model used for the automatic adjustment of the setting value of a vehicle air conditioner. (A)-(d) is a figure which shows the conditional probability table | surface about each node of the probability model shown in FIG. 4, respectively. (A)-(d) is a figure which shows the standard model which has the graph structure used as the basis of a probability model, respectively. It is a flowchart which shows the stochastic model construction operation | movement of the vehicle air conditioner to which this invention is applied. It is a flowchart which shows the control operation | movement of the vehicle air conditioner to which this invention is applied. It is a flowchart which shows the control operation | movement of the vehicle air conditioner to which this invention is applied.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Vehicle air conditioner 10 Air conditioning part 11 Compressor 21 Blower fan 22 Driving motor 24 Inside / outside air servo motor 25 Inside / outside air switching door 28 Air mix door 31 Temperature control servo motor 37 Foot door 38 Face door 39 Defroster door 40 Mode servo motor 51 Inside Air temperature sensor 52 Outside air temperature sensor 53 Solar radiation sensor 54 In-vehicle camera 55 Out-of-vehicle camera 56 Navigation system 57 Vehicle operating device 58 In-vehicle clock 59 A / C operation panel 60 Control unit 61 Storage unit 62 Communication unit 63 Verification unit 64 Control information correction unit 65 Air conditioning Control unit 651 Temperature control unit 652 Compressor control unit 653 Air outlet control unit 654 Suction port control unit 655 Air flow rate setting unit 66 Learning unit 101 Probability model 102-105 Node 106-109 Conditional probability table (CP T)
501 to 504 Standard model

Claims (6)

A vehicle air conditioner,
An air conditioning unit (10) for supplying conditioned air into the vehicle;
An information acquisition unit (51, 52, 53) that acquires state information including at least one of air conditioning information inside and outside the vehicle, vehicle position information, vehicle behavior information, time information, or biological information of a passenger of the vehicle. 55, 56, 57, 58),
An operation unit (59) for outputting setting information corresponding to an air conditioning setting operation of the air conditioner;
A control equation for calculating control information based on the air conditioning information and setting information input through the operation unit (59) is provided, and the air conditioning unit (10) is controlled according to the control information calculated by the control equation. An air conditioning control unit (65);
Having at least one probability model related to a predetermined setting operation, and inputting the state information into the at least one probability model to calculate an appearance probability that the predetermined setting operation is performed; A control information correction unit (64) that corrects the control information or setting information using correction information associated with the at least one probability model when a predetermined condition is satisfied;
A storage unit (61) for storing the state information at the time of the predetermined setting operation each time the predetermined setting operation is performed through the operation unit (59);
When the number of operations for performing the predetermined setting operation is equal to or greater than a predetermined number, based on the state information stored in the storage unit (61), select state information related to the predetermined setting operation, When the selected status information does not include information other than the air conditioning information, the control formula is corrected. On the other hand, when the selected status information includes information other than the air conditioning information, the predetermined setting operation is performed. A learning unit (66) for constructing a probability model ;
A vehicle air conditioner comprising: The learning unit (66) determines a graph structure and a conditional probability of a node included in the graph structure using the state information stored in the storage unit (61) in association with the predetermined operation. , Build a temporary probability model for calculating the probability of occurrence of the predetermined setting operation,
Among the temporary probability models, select the most suitable temporary probability model according to a predetermined criterion,
The status information as an input of the temporary probabilistic models said selected, is selected as the state information related to the predetermined setting operation, air-conditioning system according to claim 1. The learning unit (66) includes a plurality of standard models having a predetermined graph structure,
When the number of operations is equal to or greater than the predetermined number of times, with respect to each of the plurality of standard models, with the condition information of nodes included in the predetermined graph structure using the state information stored in the storage unit (61) By determining the probability, construct a temporary probability model for calculating the appearance probability that the predetermined setting operation is performed,
Among the temporary probability models, select the most suitable temporary probability model according to a predetermined criterion,
The status information as an input of the temporary probabilistic models said selected, is selected as the state information related to the predetermined setting operation, air-conditioning system according to claim 1. The predetermined determination criterion is an information amount criterion, and the learning unit (66) selects a temporary probability model in which the value of the information amount criterion calculated for each of the temporary probability models is minimum or maximum. The air conditioner for vehicles according to claim 2 or 3 . The learning unit (66) according to any one of claims 2 to 4 , wherein, when the selected state information includes information other than the air conditioning information, the temporary probability model is set as a probability model related to the predetermined setting operation. The vehicle air conditioner according to one item. An air conditioning unit (10) for supplying conditioned air into the vehicle;
An information acquisition unit (51, 52, 53, which acquires state information including at least one of air conditioning information inside and outside the vehicle, position information of the vehicle, behavior information of the vehicle, time information, or biological information of a passenger of the vehicle) 55, 56, 57, 58),
An operation unit (59) for outputting setting information corresponding to an air conditioning setting operation of the air conditioner;
An air conditioning control unit having a control expression for calculating control information based on the air conditioning information and setting information input through the operation unit (59), and controlling the air conditioning unit according to the control information calculated by the control expression and (65),
Having at least one probability model related to a predetermined setting operation, and inputting the state information into the at least one probability model to calculate an appearance probability that the predetermined setting operation is performed; When the predetermined condition is satisfied, the vehicle air conditioner control method includes a control information correction unit (64) that corrects the control information or the setting information using correction information associated with the at least one probability model. And
Storing the state information at the time of the predetermined setting operation each time the predetermined setting operation is performed through the operation unit (59);
The step of selecting state information related to the predetermined setting operation based on the state information stored in the storage unit (61) when the number of operations for performing the predetermined setting operation is equal to or greater than a predetermined number of times. When,
Determining whether the selected status information includes information other than the air conditioning information;
In the determination step, when it is determined that the selected state information does not include information other than the air conditioning information, the control formula is corrected , while in the determination step, the selected state information is converted into the air conditioning information. When it is determined that information other than information is included, constructing a probability model related to the predetermined setting operation ;
A control method characterized by comprising:

Images