Automotive air conditioner and method and apparatus for controlling automotive air conditioner

• US 20080195564A1
• Filed: 02/11/2008
• Published: 08/14/2008
• Est. Priority Date: 02/13/2007
• Status: Active Grant

• Current Assignee: DENSO Corporation
• Original Assignee: DENSO Corporation
• Inventors: Kojima, Yasufumi, Hara, Kousuke, Takeda, Hiroshi
• Granted Patent: US 7,966,280 B2
• Time in Patent Office: Days
• US Class Current: 706/12
• CPC Class Codes: B60H 1/00735 Control systems or circuits characterised by their input, i.e. by the detection, measurement or calculation of particular conditions, e.g. signal treatment, dynamic models
• CPC Class Codes: B60H 1/00771 the input being a vehicle position or surrounding, e.g. GPS-based position or tunnel
• CPC Class Codes: B60H 1/00971 Control systems or circuits characterised by including features for locking or memorising of control modes

Abstract

An automotive air conditioner includes a storage unit which stores a plurality of pieces of state information as respective learned data; a learning unit which constructs a probabilistic model; a control information correcting unit which corrects setting information, etc., related to a vehicle occupant'"'"'s setting operation in accordance with a calculated probability so as to achieve a specific setting operation; and an air-conditioning control unit which controls an air-conditioning unit in accordance with the corrected setting information, etc. The learning unit includes a clustering subunit for classifying the plurality of learned data into at least first and second clusters and for determining first and second ranges for the value of the state information from the learned data included in the respective clusters, and a probabilistic model constructing subunit for constructing the probabilistic model by determining the probabilities for the state information contained in the first and second ranges, respectively.

Claims

1. An automotive air conditioner comprising:

an air-conditioning unit for supplying conditioned air into a vehicle;

an information acquiring unit for acquiring state information indicating a state related to said vehicle;

a storage unit for storing a plurality of pieces of said state information as respective learned data;

a learning unit, by using said learned data, for constructing a probabilistic model into which said state information is entered in order to calculate the probability of a vehicle occupant performing a specific setting operation;

a control information correcting unit for calculating said probability by entering said state information into the probabilistic model constructed by said learning unit, and for correcting setting information or control information related to the setting operation of said occupant in accordance with said calculated probability so as to achieve said specific setting operation; and

an air-conditioning control unit for controlling said air-conditioning unit in accordance with said corrected setting information or control information, wherein

said learning unit comprises;

a clustering subunit for classifying said plurality of learned data stored in said storage unit into at least a first cluster and a second cluster, and for determining a first range for a value of said state information from the learned data included in said first cluster and a second range for the value of said state information from the learned data included in said second cluster; and

a probabilistic model constructing subunit for constructing said probabilistic model associated with said specific setting operation by determining the probability of occurrence of the value of said state information contained in said first range and the probability of occurrence of the value of said state information contained in said second range.

2. The automotive air conditioner according to claim 1, wherein said clustering subunit has a first clustering condition and a second clustering condition that define different ranges for the value of said state information, and generates said clusters after determining the ranges of the clusters to which said respective learned data belong by using said plurality of learned data and said first and second clustering conditions.

3. The automotive air conditioner according to claim 2, wherein the ranges for the value of said state information are each determined based on a Euclidean distance,

said first clustering condition is that the Euclidean distance between each pair of learned data in said plurality of learned data is not greater than a first Euclidean distance, and

said second clustering condition is that the Euclidean distance between each pair of learned data in said plurality of learned data is greater than said first Euclidean distance but not greater than a second Euclidean distance.

4. The automotive air conditioner according to claim 1, wherein said clustering subunit determines said first range so as to contain all of said learned data included in said first cluster but not to contain any learned data included in said second cluster, and determines said second range so as to contain all of said learned data included in said second cluster but not to contain any learned data included in said first cluster.

5. The automotive air conditioner according to claim 1, wherein said clustering subunit determines said first range for the value of said state information so as to be defined by an area whose center is located at the centroid of said learned data included in said first cluster and whose radius is defined by the Euclidean distance from said centroid to the learned data located farthest away therefrom among said learned data included in said first cluster.

6. The automotive air conditioner according to claim 1, wherein said probabilistic model has a node that takes said state information as an input and that outputs the probability of said occupant performing said specific setting operation, and said node has a conditional probability table that indicates said probability for the case where the value of said state information is contained in said first range as well as for the case where the value of said state information is contained in said second range, and

said probabilistic model constructing subunit obtains from said plurality of learned data the number of times that said specific setting operation has been performed for the case where the value of said state information is contained in said first range as well as for the case where the value of said state information is contained in said second range, creates said conditional probability table by dividing said number of times by the total number of said plurality of learned data and thereby obtaining said recommended probability for the case where the value of said state information is contained in said first range as well as for the case where the value of said state information is contained in said second range, and stores said conditional probability table in said storage unit by associating said conditional probability table with said node.

7. The automotive air conditioner according to claim 1, wherein said probabilistic model has a group of nodes consisting of a node that takes said state information as an input and that outputs a conditional probability of a specific event and at least one other node that takes the output of said node as an input and that outputs the probability of said occupant performing said specific setting operation, and said node has a conditional probability table that indicates said conditional probability for the case where the value of said state information is contained in said first range as well as for the case where the value of said state information is contained in said second range, and

said probabilistic model constructing subunit obtains from said plurality of learned data the number of times that said specific even has occurred for the case where the value of said state information is contained in said first range as well as for the case where the value of said state information is contained in said second range, creates said conditional probability table by dividing said number of times by the total number of said plurality of learned data and thereby obtaining said conditional probability for the case where the value of said state information is contained in said first range as well as for the case where the value of said state information is contained in said second range, and stores said conditional probability table in said storage unit by associating said conditional probability table with said node.

8. The automotive air conditioner according to claim 1, wherein said state information is said vehicle'"'"'s current location information.

9. The automotive air conditioner according to claim 1, wherein said state information is time information indicating current time.

10. The automotive air conditioner according to claim 1, wherein said state information is said vehicle'"'"'s current location information combined with time information indicating current time.

11. A method for controlling an automotive air conditioner, said air conditioner comprising an air-conditioning unit for supplying conditioned air into a vehicle, an information acquiring unit for acquiring state information indicating a state related to said vehicle, a storage unit, a control information correcting unit for having at least one probabilistic model associated with a specific setting operation, calculating the probability of said specific operation by entering said state information into said probabilistic model, and correcting setting information or control information related to the setting operation of a vehicle occupant in accordance with said calculated probability so as to achieve said specific setting operation, and an air-conditioning control unit for controlling said air-conditioning unit in accordance with said corrected setting information or control information, the method comprising:

storing said state information as learned information in said storage unit;

selecting a plurality of learned data from said learned data stored in said storage unit;

classifying said plurality of learned data stored in said storage unit into at least a first cluster and a second cluster, and determining a first range for a value of said state information from the learned data included in said first cluster and a second range for the value of said state information from the learned data included in said second cluster; and

constructing said probabilistic model associated with said specific setting operation by determining the probability of occurrence of said state information contained in said first range and the probability of occurrence of said state information contained in said second range.

12. The method according to claim 11, wherein each time said specific setting operation is performed, said storing said state information stores said state information in said storage unit as the learned data related to said specific setting operation.

13. The method according to claim 11, wherein when the number of times that said specific setting operation has been performed reaches a predetermined number of times, said selecting a plurality of learned data selects said plurality of learned data related to said specific setting operation.

14. The method according to claim 11, wherein when a predetermined time period has elapsed after the construction of said probabilistic model associated with said specific setting operation, said selecting a plurality of learned data selects said plurality of learned data related to said specific setting operation in order to construct another probabilistic model associated with said specific setting operation.

15. The method according to claim 11, wherein said classifying said plurality of learned data generates said clusters after determining the ranges of the clusters to which said respective learned data belong by using said plurality of learned data and also using a first clustering condition and a second clustering condition that define different ranges for the value of said state information.

16. The method according to claim 11, wherein the ranges for the value of said state information are each determined based on a Euclidean distance,

said first clustering condition is that the Euclidean distance between each pair of learned data in said plurality of learned data is not greater than a first Euclidean distance, and

said second clustering condition is that the Euclidean distance between each pair of learned data in said plurality of learned data is greater than said first Euclidean distance but not greater than a second Euclidean distance.

17. A controller comprising:

an information acquiring unit for acquiring state information of an apparatus to be controlled;

a storage unit for storing a plurality of pieces of said state information as respective learned data;

a learning unit, by using said learned data, for constructing a probabilistic model into which said state information is entered in order to calculate the probability of performing a specific setting operation;

a control information correcting unit for calculating said probability by entering said state information into the probabilistic model constructed by said learning unit, and for correcting setting information or control information related to said specific setting operation in accordance with said calculated probability so as to achieve said specific setting operation; and

a control unit for controlling said apparatus in accordance with said corrected setting information or control information, wherein

said learning unit comprises;

a clustering subunit for classifying said plurality of learned data stored in said storage unit into at least a first cluster and a second cluster, and for determining a first range for a value of said state information from the learned data included in said first cluster and a second range for the value of said state information from the learned data included in said second cluster; and

a probabilistic model constructing subunit for constructing said probabilistic model associated with said specific setting operation by determining said probability for the value of said state information contained in said first range and said probability for the value of said state information contained in said second range.

18. The controller according to claim 17, wherein said state information is location information indicating current location of said apparatus to be controlled, and

said clustering unit classifies said plurality of learned data in such a manner that the learned data whose distance relative to each other is not greater than a first distance are classified as belonging to said first cluster and the learned data whose distance relative to each other is greater than said first distance but not greater than a second distance are classified as belonging to said second cluster, and determines said first range for the value of said state information so as to be defined by a circular area whose center is located at the centroid of said learned data included in said first cluster and whose radius is defined by the distance from said first centroid to the learned data located farthest away therefrom among said learned data included in said first cluster, while determining said second range for the value of said state information so as to be defined by a circular area whose center is located at the centroid of said learned data included in said second cluster and whose radius is defined by the distance from said centroid to the learned data located farthest away therefrom among said learned data included in said second cluster.

19. The controller according to claim 17, wherein said state information is time information indicating time of day acquired by said information acquiring unit, and

said clustering unit classifies said plurality of learned data in such a manner that the learned data whose time difference relative to each other is not greater than a first period are classified as belonging to said first cluster and the learned data whose time difference relative to each other is greater than said first period but not greater than a second period are classified as belonging to said second cluster, and determines said first range for the value of said state information so as to be bounded at one end by the earliest occurring learned data and at the other end by the latest occurring learned data among said learned data included in said first cluster, while determining said second range for the value of said state information so as to be bounded at one end by the earliest occurring learned data and at the other end by the latest occurring learned data among said learned data included in said second cluster.