Occupancy prediction using historical occupancy patterns

• US 9,176,485 B2
• Filed: 06/02/2011
• Issued: 11/03/2015
• Est. Priority Date: 06/02/2011
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Inventors: Krumm, John Charles, Scott, James W., Brush, Alice Jane Bernheim, Meyers, Brian R., Hodges, Stephen Edward
• Primary Examiner(s): Chaki, Kakali
• Assistant Examiner(s): Zidanic, Michael
• Application Number:
  US13/151,849
• Publication Number:
  US 20120310376A1
• Time in Patent Office: 1,615 Days
• Field of Search: 706/14, 706/45
• US Class Current: 1/1
• CPC Class Codes: G05B 15/02 electric
• CPC Class Codes: G05B 2219/2642 Domotique, domestic, home control, automation, smart house
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks

Abstract

Methods and systems for occupancy prediction using historical occupancy patterns are described. In an embodiment, an occupancy probability is computed by comparing a recent occupancy pattern to historic occupancy patterns. Sensor data for a room, or other space, is used to generate a table of past occupancy which comprises these historic occupancy patterns. The comparison which is performed identifies a number of similar historic occupancy patterns and data from these similar historic occupancy patterns is combined to generate an occupancy probability for a time in the future. In an example, time may be divided into discrete slots and binary values may be used to indicate occupancy or non-occupancy in each slot. An occupancy probability for a defined future time slot then comprises a combination of the binary values for corresponding time slots from each of the identified similar occupancy patterns.

Claims

1. A method comprising:

receiving sensor data;

creating a table of past occupancy based on the sensor data;

identifying a sequence of occupancy data for a space for an elapsed portion of a current calendar day and a portion of a day immediately prior to the current calendar day divided into discrete time slots;

comparing the identified sequence to portions of sequences of occupancy data extracted from the table of past occupancy for the space to select one or more similar sequences, the portions of sequences of occupancy data extracted from the table of past occupancy for the space being selected using a progressive algorithm, each sequence of occupancy data extracted from the table of past occupancy relating to a calendar day and a portion of a previous calendar day divided into discrete time slots, use of the progressive algorithm comprising conducting a first comparison at a first level of granularity to identify a set of similar sequences and then conducting at least a second comparison against the set of similar sequences at a second level of granularity, the second level of granularity being a shorter time increment than the first level of granularity;

computing a future occupancy probability for the space based on the selected one or more similar sequences; and

sending a control message to operate a system based on the future occupancy probability for the space.

2. A method according to claim 1, further comprising:

triggering an automatic action based on the computed occupancy probability for the space.

3. A method according to claim 2, wherein triggering an automatic action comprises:

sending a control message to cause a system or device to change modes based on the computed occupancy probability for the space.

4. A method according to claim 1, wherein creating a table of past occupancy based on the sensor data comprises:

recording a value indicating if the space was occupied for each of a plurality of discrete time slots.

5. A method according to claim 4, wherein comparing the identified sequence to portions of sequences of occupancy data extracted from the table of past occupancy for the space to select one or more similar sequences comprises:

excluding anomalous days from the portions of sequences of occupancy data extracted from the table of past occupancy for the space; and

comparing a value in each time slot in the identified sequence with values in corresponding time slots in portions of sequences of occupancy data extracted from the table of past occupancy for the space to select one or more similar sequences.

6. A method according to claim 5, wherein the one or more similar sequences are selected based on a distance metric between the identified sequence and each of the portions of sequences of occupancy data extracted from the table of past occupancy.

7. A method according to claim 5, wherein comparing the identified sequence to portions of sequences of occupancy data extracted from the table of past occupancy for the space to select one or more similar sequences further comprises:

creating a time shifted version of the identified sequence; and

comparing a bit value in each time slot in the time shifted version of the identified sequence with bit values in time slots in portions of sequences of occupancy data extracted from the table of past occupancy for the space to select one or more similar sequences.

8. A method according to claim 5, wherein computing a future occupancy probability for the space based on the selected one or more similar sequences:

for each future occupancy probability, computing an average of values in corresponding time slots in each of the selected one or more similar sequences.

9. A method according to claim 1, wherein the method further comprises:

filtering the sequences of occupancy data extracted from the table of past occupancy used in the comparing step using a filter based on at least one of;

a characteristic of the future time and a characteristic of the space at the future time.

10. A method according to claim 1, wherein the table of past occupancy comprises data for a plurality of spaces and wherein creating a table of past occupancy based on the sensor data comprises:

for each space, recording a value indicating occupancy for each of a plurality of discrete time slots.

11. A method according to claim 10, wherein the identifying, comparing and computing steps are performed for a composite space comprising a combination of at least two of a plurality of spaces.

12. A method according to claim 1, further comprising:

applying a threshold to the computed occupancy probabilities.

13. A method according to claim 1, wherein the table of past occupancy comprises per-person occupancy data for one or more people and wherein the identifying and comparing steps are performed using data for a particular person such that the occupancy probability for the space comprises an occupancy probability for the particular person in the space.

14. A method according to claim 1, wherein each sequence of occupancy data extracted from the table of past occupancy comprises occupancy data for a calendar day augmented with additional data from a portion of a previous day and wherein the sequence of occupancy data for a space for a period of time prior to a time t comprises occupancy data for a current calendar day prior to time t augmented with additional occupancy data from a portion of a day directly prior to the current calendar day.

15. A method according to claim 14, wherein each sequence of occupancy data extracted from the table of past occupancy comprises occupancy data for a calendar day augmented with additional data from both a portion of a subsequent day and a portion of a previous day.

16. A system comprising:

at least one sensor arranged to detect occupancy of a space;

at least one memory arranged to store data collected by the at least one sensor; and

at least one processor arranged to perform steps comprising;

creating a table of past occupancy based on the sensor data;

identifying a sequence of occupancy data for the space for an elapsed portion of a current calendar day and a portion of a day immediately prior to the current calendar day divided into discrete time slots;

comparing the identified sequence to portions of sequences of occupancy data extracted from the table of past occupancy for the space to select one or more similar sequences, the portions of sequences of occupancy data extracted from the table of past occupancy for the space being selected using a progressive algorithm, each sequence of occupancy data extracted from the table of past occupancy relating to a calendar day and a portion of a previous calendar day divided into discrete time slots, use of the progressive algorithm comprising conducting a first comparison at a first level of granularity to identify a set of similar sequences and then conducting at least a second comparison against the set of similar sequences at a second level of granularity, the second level of granularity being a shorter time increment than the first level of granularity;

computing a future occupancy probability for the space based on the selected one or more similar sequences; and

sending a control message to operate a system based on the future occupancy probability for the space.

17. A system according to claim 16, further comprising:

generating an occupancy prediction for the space and the future time interval by applying a threshold to the occupancy probability.

18. A system according to claim 16, further comprising:

sending a control message to a system associated with the space based on the generated occupancy probability.

19. One or more device-readable media, the device-readable media being hardware, with device-executable instructions that, when executed by a computing system, direct the computing system to perform steps comprising:

creating a table of past occupancy based on sensor data;

identifying a sequence of occupancy data for a space for an elapsed portion of a current calendar day and a portion of a day immediately prior to the current calendar day divided into discrete time slots;

comparing the identified sequence to portions of sequences of occupancy data extracted from the table of past occupancy for the space to select one or more similar sequences, the portions of sequences of occupancy data extracted from the table of past occupancy for the space being selected using a progressive algorithm, each sequence of occupancy data extracted from the table of past occupancy relating to a calendar day and a portion of a previous calendar day divided into discrete time slots, use of the progressive algorithm comprising conducting a first comparison at a first level of granularity to identify a set of similar sequences and then conducting at least a second comparison against the set of similar sequences at a second level of granularity, the second level of granularity being a shorter time increment than the first level of granularity;

computing a future occupancy probability for the space based on the selected one or more similar sequences; and

sending a control message to operate a system based on the future occupancy probability for the space.

20. One or more device-readable media according to claim 19, wherein:

the table of past occupancy comprises discrete occupancy data for one or more people,

the sequence of occupancy data for a space for an elapsed portion of a current calendar day and a portion of a day immediately prior to the current calendar day divided into discrete time slots comprises occupancy data for one of said people;

the comparison is performed using sequences of occupancy data for that person extracted from the table of past occupancy, such that the computed future occupancy probability for the space relates to that person; and

wherein the control message is also based on a personal preference of that person.