Air conditioner and control method and device thereof, storage medium and computer program product

Abstract

The invention discloses an air conditioner control method, device and system, a storage medium and a computer program product. The method comprises the steps that a broadcast packet of Bluetooth equipment and the signal strength of the broadcast packet are obtained; extracting multiple types of feature information in the broadcast packet to construct a device fingerprint; performing similarity matching based on the device fingerprint to associate the random MAC address in the broadcast packet with the same Bluetooth device to obtain an association result; filtering the signal intensity, and judging the existence state of the Bluetooth equipment according to the filtered signal intensity and the association result; and generating and executing a control strategy of the air conditioner according to the existence state. According to the scheme, the equipment existence recognition reliability of the air conditioner is improved, the automatic control precision of the air conditioner is improved, and the energy-saving effect and the user comfort experience are enhanced.

Description

Technical Field

[0001] This invention belongs to the field of air conditioning technology, specifically relating to an air conditioning control method, device, system, storage medium, and computer program product, and particularly to an air conditioning control method, device, system, storage medium, and computer program product based on the fusion of multi-feature fingerprints and RSSI. Background Technology

[0002] The automatic control of smart air conditioners largely relies on human presence detection technology to trigger adjustments in operating strategies in order to achieve the core requirement of "operation by humans, energy saving by humans". Existing detection solutions mainly include infrared sensors, millimeter-wave radar, cameras and Bluetooth sensing.

[0003] Among them, infrared sensors are inexpensive but easily affected by ambient temperature and airflow, and cannot detect stationary human bodies, limiting their detection range and reliability; millimeter-wave radar can identify stationary human bodies, but it has problems such as high cost, challenges in multi-target recognition, and detection blind spots; cameras have high detection accuracy, but the collection of image information poses a serious risk of privacy leakage, and is significantly affected by lighting and occlusion, limiting their application scenarios.

[0004] Bluetooth sensing technology, with its advantages of low cost, easy deployment, and no privacy leakage risk, has become the preferred direction for human presence detection in smart air conditioners. However, it faces key technical bottlenecks in practical applications: Bluetooth devices generally use MAC address randomization mechanisms to protect user privacy, causing traditional MAC address-based device identification methods to fail and making it impossible to stably associate Bluetooth devices of the same user; at the same time, Bluetooth signal strength (RSSI) is easily affected by environmental obstruction and multipath effects, resulting in large fluctuations, making it difficult to accurately determine whether a device is indoors; in addition, single feature recognition lacks the ability to adapt to user behavior habits, resulting in insufficient accuracy and robustness in device presence determination, which in turn leads to problems such as accidental start-up or shutdown of the air conditioner or delayed response of energy-saving strategies, failing to take into account both user comfort and energy-saving needs.

[0005] In summary, existing Bluetooth sensing solutions for smart air conditioners fail to effectively address issues such as MAC randomization, RSSI fluctuations, and insufficient behavior adaptation. Consequently, they cannot reliably identify device presence, making it difficult for the automatic control accuracy and energy-saving performance of air conditioners to meet practical application requirements.

[0006] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention

[0007] The purpose of this invention is to provide a control method, device, system, storage medium, and computer program product for an air conditioner, in order to solve the problem that the Bluetooth sensing scheme for air conditioners in related solutions cannot reliably identify the presence of devices due to MAC randomization, RSSI fluctuations, and insufficient behavioral adaptation, resulting in the air conditioner's automatic control accuracy and energy-saving effect failing to meet actual needs. This invention improves the reliability of intelligent air conditioners in identifying the presence of devices, increases the accuracy of automatic air conditioner control, and enhances energy-saving effect and user comfort experience.

[0008] This invention provides a method for controlling an air conditioner, comprising: acquiring a broadcast packet from a Bluetooth device and the signal strength of the broadcast packet; extracting multiple feature information from the broadcast packet to construct a device fingerprint; performing similarity matching based on the device fingerprint to associate a random MAC address in the broadcast packet with the same Bluetooth device, thereby obtaining an association result; filtering the signal strength; determining the presence state of the Bluetooth device based on the filtered signal strength and the association result; and generating and executing a control strategy for the air conditioner based on the presence state.

[0009] In some implementations, similarity matching is performed based on the device fingerprint to associate the random MAC address in the broadcast packet with the same Bluetooth device to obtain an association result. This includes: extracting historical device fingerprints corresponding to broadcast packets obtained from historical scans and constructing a historical fingerprint database; the historical fingerprint database stores the correspondence between historical random MAC addresses and historical device fingerprints; calculating the multi-dimensional similarity between the device fingerprint corresponding to the current broadcast packet and each historical device fingerprint in the historical fingerprint database; the multi-dimensional similarity includes numerical vector similarity and set similarity; weighted fusion of the multi-dimensional similarity to obtain a total similarity score; if the total similarity score reaches a preset association threshold, then the random MAC address in the current broadcast packet is associated with the Bluetooth device corresponding to the historical device fingerprint, and an association result is generated. The association result includes the association relationship between the current random MAC address and the corresponding Bluetooth device, the association confidence, and the unique identifier corresponding to the Bluetooth device.

[0010] In some implementations, filtering the signal strength includes: performing a first-level filter on the signal strength to eliminate short-term random fluctuations in the signal strength; and performing a second-level filter on the first-level filtered signal strength within a preset time window to extract stable features of the signal strength and characterize the signal propagation state of the device.

[0011] In some implementations, determining the presence status of the Bluetooth device based on the filtered signal strength and the association result includes: determining the unique identifier of the Bluetooth device corresponding to the current broadcast packet based on the association result; filtering target scan records within a preset time window based on the unique identifier, counting the number of scans by the Bluetooth device, and obtaining the current occurrence frequency; determining whether the filtered signal strength is within a preset effective strength range and whether the current occurrence frequency reaches a preset occurrence threshold; if the signal strength is within the preset effective strength range and the current occurrence frequency reaches the preset occurrence threshold, then determining the initial state of the Bluetooth device as a candidate for presence; if the signal strength is not within the preset effective strength range and the current occurrence frequency does not reach the preset occurrence threshold, then determining the initial state of the Bluetooth device as a candidate for absence; and correcting the initial state based on a preset machine learning model to obtain the presence status of the Bluetooth device.

[0012] In some implementations, the initial state is corrected based on a preset machine learning model to obtain the presence state of the Bluetooth device, including: using the initial state as an input constraint, assigning a basic weight of presence probability to the present candidate and a basic weight of departure probability to the departure candidate; inputting preset features into the preset machine learning model to calculate a presence probability value; if the presence probability value is greater than or equal to a first preset threshold, then determining the presence state of the Bluetooth device as present; if the presence probability value is less than or equal to a second preset threshold, then determining the presence state of the Bluetooth device as departing; if the presence probability value is greater than the second preset threshold and less than the first preset threshold, then maintaining the previous presence state.

[0013] In conjunction with the above method, another aspect of the present invention provides an air conditioner control device, comprising: an acquisition unit configured to acquire a broadcast packet of a Bluetooth device and the signal strength of the broadcast packet; a fingerprint construction unit configured to extract multiple types of feature information from the broadcast packet to construct a device fingerprint; a device association unit configured to perform similarity matching based on the device fingerprint to associate a random MAC address in the broadcast packet with the same Bluetooth device to obtain an association result; a state determination unit configured to filter the signal strength and determine the presence state of the Bluetooth device based on the filtered signal strength and the association result; and a control execution unit configured to generate and execute a control strategy for the air conditioner based on the presence state.

[0014] In some implementations, the device association unit performs similarity matching based on the device fingerprint to associate the random MAC address in the broadcast packet with the same Bluetooth device, obtaining an association result. This includes: extracting historical device fingerprints corresponding to broadcast packets obtained from historical scans and constructing a historical fingerprint database; the historical fingerprint database stores the correspondence between historical random MAC addresses and historical device fingerprints; calculating the multi-dimensional similarity between the device fingerprint corresponding to the current broadcast packet and each historical device fingerprint in the historical fingerprint database; the multi-dimensional similarity includes numerical vector similarity and set similarity; weighted fusion of the multi-dimensional similarity to obtain a total similarity score; if the total similarity score reaches a preset association threshold, then the random MAC address in the current broadcast packet is associated with the Bluetooth device corresponding to the historical device fingerprint, and an association result is generated. The association result includes the association relationship between the current random MAC address and the corresponding Bluetooth device, the association confidence, and the unique identifier corresponding to the Bluetooth device.

[0015] In some implementations, the state determination unit performs filtering processing on the signal strength, including: performing a first-level filtering on the signal strength to eliminate short-term random fluctuations in the signal strength; and performing a second-level filtering on the first-level filtered signal strength within a preset time window to extract stable features of the signal strength and characterize the signal propagation state of the device.

[0016] In some implementations, the state determination unit determines the presence state of the Bluetooth device based on the filtered signal strength and the association result, including: determining the unique identifier of the Bluetooth device corresponding to the current broadcast packet based on the association result; filtering target scan records within a preset time window based on the unique identifier, counting the number of scans by the Bluetooth device, and obtaining the current occurrence frequency; determining whether the filtered signal strength is within a preset effective strength range and whether the current occurrence frequency reaches a preset occurrence threshold; if the signal strength is within the preset effective strength range and the current occurrence frequency reaches the preset occurrence threshold, then determining the initial state of the Bluetooth device as a candidate for presence; if the signal strength is not within the preset effective strength range and the current occurrence frequency does not reach the preset occurrence threshold, then determining the initial state of the Bluetooth device as a candidate for absence; and correcting the initial state based on a preset machine learning model to obtain the presence state of the Bluetooth device.

[0017] In some implementations, the state determination unit corrects the preliminary state based on a preset machine learning model to obtain the presence state of the Bluetooth device, including: using the preliminary state as an input constraint, assigning a basic weight of presence probability to the presence candidate and a basic weight of departure probability to the departure candidate; inputting preset features into the preset machine learning model to calculate a presence probability value; if the presence probability value is greater than or equal to a first preset threshold, determining the presence state of the Bluetooth device as present; if the presence probability value is less than or equal to a second preset threshold, determining the presence state of the Bluetooth device as departing; if the presence probability value is greater than the second preset threshold and less than the first preset threshold, maintaining the previous presence state.

[0018] In conjunction with the above-described device, the present invention further provides a system comprising: the control device for the air conditioner described above.

[0019] In conjunction with the above method, the present invention further provides a storage medium comprising a stored program, wherein, when the program is executed, the device on which the storage medium is located controls the air conditioner control method described above to be performed.

[0020] In conjunction with the above method, the present invention further provides a computer program product comprising a computer program that, when processed and executed, implements the steps of the above-described air conditioner control method.

[0021] The present invention involves an air conditioner scanning the broadcast packets and signal strength of Bluetooth devices, extracting broadcast packet features to construct a device fingerprint, associating a random MAC address with the device through fingerprint matching, filtering the signal strength, and combining the association results to determine the device's presence status. Finally, based on this status, an air conditioner control strategy is generated and executed. This improves the accuracy of the air conditioner in recognizing the user's presence status, enhances the intelligence of air conditioner control, and balances user comfort with energy-saving efficiency.

[0022] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention.

[0023] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0024] Figure 1 This is a flowchart illustrating an embodiment of the air conditioner control method of the present invention;

[0025] Figure 2 This is a schematic diagram of the structure of an embodiment of the air conditioner control device of the present invention;

[0026] Figure 3 This is a flowchart illustrating an air conditioning control method based on the fusion of multi-feature fingerprints and RSSI.

[0027] Referring to the accompanying drawings, the reference numerals in the embodiments of the present invention are as follows:

[0028] 101-Acquisition unit; 102-Fingerprint construction unit; 103-Device association unit; 104-Status determination unit; 105-Control execution unit. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0030] According to embodiments of the present invention, an air conditioning control method is provided, such as... Figure 1 The flowchart of an embodiment of the method of the present invention is shown. The air conditioner control method may include steps S110 to S150.

[0031] In step S110, the broadcast packet of the Bluetooth device and the signal strength of the broadcast packet are obtained.

[0032] Bluetooth devices refer to electronic devices that support Bluetooth wireless communication technology, including but not limited to smartphones, smartwatches, Bluetooth headsets, and smart home terminals. Broadcast packets are data packets containing device information that Bluetooth devices periodically send out. They are the primary way Bluetooth devices transmit information without pairing and include core information such as device identification, service information, and manufacturer information. Signal strength refers to the signal reception power of an air conditioner's Bluetooth scanning module when it receives a broadcast packet from a Bluetooth device. It is usually measured in dBm. The signal strength value is negatively correlated with the distance between the Bluetooth device and the air conditioner; the closer the distance, the higher the signal strength value (closer to 0).

[0033] Bluetooth devices continuously send out broadcast packets, which the air conditioner receives via its built-in Bluetooth scanning module, recording the signal strength of each packet. Signal strength reflects the distance between the Bluetooth device and the air conditioner, serving as a key criterion for determining user presence; the characteristic information within the broadcast packets forms the basis for constructing device fingerprints and establishing device association.

[0034] Specifically, the air conditioner's Bluetooth scanning module starts scanning at a preset interval (e.g., once every second) to search for broadcast packets sent by all Bluetooth devices within the coverage area; for each broadcast packet received, the content of the broadcast packet and the corresponding signal strength value are recorded simultaneously.

[0035] In step S120, multiple feature information from the broadcast packet is extracted to construct a device fingerprint.

[0036] Device fingerprinting is a set of features built upon multiple types of characteristic information from Bluetooth device broadcast packets. It uniquely identifies a Bluetooth device and can be used to distinguish different Bluetooth devices, unaffected by random changes in the device's MAC address. These multiple types of characteristic information include: the Bluetooth device's broadcast type, service UUID, vendor-specific data, instantaneous signal strength values, and trends in signal strength.

[0037] Bluetooth devices' random MAC addresses change periodically and cannot serve as stable device identifiers. However, certain characteristic information in broadcast packets possesses uniqueness and stability (such as vendor-specific data and service UUIDs). Device fingerprints constructed based on this characteristic information can uniquely identify Bluetooth devices, solving the problem of device identification difficulties caused by random MAC addresses.

[0038] Specifically, the air conditioner's control module parses the received broadcast packets and extracts multiple types of feature information with stable identification characteristics. The extracted feature information is then structured and combined into a complete feature set according to preset algorithm rules. This feature set is the device fingerprint of the corresponding Bluetooth device.

[0039] In step S130, similarity matching is performed based on the device fingerprint to associate the random MAC address in the broadcast packet with the same Bluetooth device, thereby obtaining the association result.

[0040] A random MAC address is a media access control address that Bluetooth devices periodically generate to protect privacy. The same Bluetooth device may use different random MAC addresses at different times and in different scenarios. The association result refers to the set of information generated after establishing an association between the random MAC address in the current broadcast packet and the corresponding Bluetooth device through device fingerprint matching. This includes the correspondence between the random MAC address and the Bluetooth device, the trustworthiness of the association, and the unique identifier of the Bluetooth device.

[0041] Broadcast packets sent by the same Bluetooth device at different times may have different random MAC addresses, but their core feature information (i.e., device fingerprint) remains consistent. By matching the current device fingerprint with historical device fingerprints, the Bluetooth device to which the current broadcast packet belongs can be identified, achieving a stable association between the random MAC address and the Bluetooth device, and providing an accurate device identifier for subsequent status determination.

[0042] Specifically, the air conditioner compares the device fingerprint corresponding to the current broadcast packet with each historical device fingerprint in the pre-stored historical device fingerprint database; it calculates the similarity between the current device fingerprint and each historical device fingerprint using a preset algorithm, and selects the historical device fingerprint with the highest similarity; if the similarity reaches a preset association threshold, it determines that the current broadcast packet and the Bluetooth device corresponding to the historical device fingerprint are the same device, and establishes an association between the random MAC address in the current broadcast packet and the Bluetooth device, generating an association result that includes the association relationship, association confidence, and device unique identifier.

[0043] In some implementations, step S130 involves performing similarity matching based on the device fingerprint to associate the random MAC address in the broadcast packet with the same Bluetooth device, thereby obtaining an association result, including steps S210 to S240.

[0044] Step S210: Extract the historical device fingerprints corresponding to the broadcast packets obtained from historical scans and construct a historical fingerprint database; the historical fingerprint database stores the correspondence between historical random MAC addresses and historical device fingerprints.

[0045] Historical device fingerprints refer to the device fingerprints constructed by parsing Bluetooth device broadcast packets acquired during past scans of the air conditioner and extracting feature information. These fingerprints are historical data used to match the current device fingerprint. The historical fingerprint database refers to a collection of historical device fingerprints stored internally by the air conditioner or in a cloud-connected system. It also stores the mapping relationship between historical random MAC addresses and their corresponding historical device fingerprints.

[0046] Building a historical fingerprint database enables centralized storage and management of device fingerprints, facilitating rapid data retrieval for similarity matching. Simultaneously, storing the correspondence between historical random MAC addresses and historical device fingerprints allows for tracing device address change records, improving the stability of associations. Specifically, the air conditioner extracts all Bluetooth device broadcast packets stored during past scanning processes, performs feature analysis and device fingerprint construction on each broadcast packet to obtain the corresponding historical device fingerprint. Each historical device fingerprint is then bound to its corresponding historical random MAC address and device association information, organized and stored according to a preset data structure to form a complete historical fingerprint database. Furthermore, the system periodically updates the historical fingerprint database, deleting expired and invalid historical data to ensure data timeliness.

[0047] Step S220: Calculate the multi-dimensional similarity between the device fingerprint corresponding to the current broadcast packet and the historical device fingerprints in the historical fingerprint database; the multi-dimensional similarity includes numerical vector similarity and set similarity.

[0048] Multi-dimensional similarity calculates the similarity between the current device fingerprint and historical device fingerprints from different feature dimensions. It includes two categories: numerical vector similarity and set similarity, comprehensively reflecting the matching degree between two fingerprints. Numerical vector similarity targets quantifiable feature information in broadcast packets (such as instantaneous signal strength values, data field lengths, etc.), converting them into numerical vectors and calculating the vector similarity through algorithms, accurately reflecting the consistency of numerical features. Set similarity targets discrete feature information in broadcast packets (such as service UUID, broadcast type, etc.), treating them as set elements and calculating the set overlap through algorithms, effectively reflecting the matching degree of discrete features.

[0049] Single-dimensional similarity calculations cannot fully reflect the matching situation of device fingerprints and may lead to matching errors due to interference from some features. Calculating similarity using two dimensions—numerical vector similarity and set similarity—can cover the matching needs of both quantized and discrete features, improving the comprehensiveness and accuracy of similarity calculations. Specifically, quantized features in the current device fingerprint (such as signal strength values, data field lengths, etc.) are transformed into standardized numerical vectors, while discrete features (such as service UUID, broadcast type, etc.) are transformed into feature sets. For each historical device fingerprint in the historical fingerprint database, the cosine similarity algorithm is used to calculate the numerical vector similarity, and the Jaccard similarity algorithm is used to calculate the set similarity, obtaining the two-dimensional similarity results between the current device fingerprint and each historical device fingerprint.

[0050] The process of calculating the similarity of numerical vectors using the cosine similarity algorithm is as follows: First, calculate the weighted Euclidean distance. Then, use radial basis function (RBF) to convert it into similarity. .in, The weights corresponding to the i-th feature dimension are: Let x be the eigenvalue of the numerical vector x in the i-th dimension. Let be the eigenvalue of the numerical vector y in the i-th dimension. This represents the bandwidth parameter of the radial basis core.

[0051] The formula for calculating set similarity using the Jaccard similarity algorithm is as follows: Where A is the discrete feature set corresponding to the current device fingerprint, and B is the discrete feature set corresponding to the historical device fingerprint.

[0052] Step S230: Perform weighted fusion on the multi-dimensional similarity to obtain the total similarity score.

[0053] Weighted fusion is a process that assigns different weights to the similarity of different dimensions based on their importance, and then calculates the comprehensive similarity score through weighted summation. This can highlight the influence of core features and improve matching accuracy.

[0054] Numerical vector similarity and set similarity have different levels of importance in device identification. For example, the numerical vector similarity corresponding to manufacturer-specific data has a greater impact on device uniqueness. Weighted fusion can highlight the weight of core features, avoid matching deviations caused by interference from secondary features, and improve the reliability of the overall similarity score. Specifically, according to a preset weight allocation rule, weights are assigned to numerical vector similarity and set similarity respectively (e.g., numerical vector similarity accounts for 65% and set similarity accounts for 35%). The similarity results of the two dimensions are multiplied by their corresponding weights, and then the products are added together to obtain the overall similarity score between the current device fingerprint and the corresponding historical device fingerprint. For example, if the calculated numerical vector similarity is 92% and the set similarity is 88%, the overall similarity score, calculated using weighted fusion of 65% and 35%, is 92% × 65% + 88% × 35% = 90.6%.

[0055] Step S240: If the total similarity score reaches the preset association threshold, then the random MAC address in the current broadcast packet is associated with the Bluetooth device corresponding to the historical device fingerprint, and an association result is generated. The association result includes the association relationship between the current random MAC address and the corresponding Bluetooth device, the association confidence, and the unique identifier corresponding to the Bluetooth device.

[0056] The preset association threshold is a pre-defined similarity judgment threshold. When the total similarity score reaches or exceeds this threshold, the current device fingerprint is determined to match the historical device fingerprint, and a device association can be established. The association confidence refers to the degree of credibility in establishing an association between the current random MAC address and the corresponding Bluetooth device. It is usually positively correlated with the total similarity score and is used to measure the reliability of the association results.

[0057] A preset association threshold is the core criterion for determining device matching, ensuring that only highly similar device fingerprints can be associated, thus avoiding false associations. Generating association results containing multiple types of information provides comprehensive data support for subsequent existence status determination and device tracking. Specifically, the calculated total similarity score is compared with a preset association threshold (e.g., 80%). If the total similarity score reaches or exceeds this threshold, the Bluetooth device corresponding to the current broadcast packet is determined to be the same device as the Bluetooth device corresponding to the historical device fingerprint. The current random MAC address is then bound to the unique identifier of the Bluetooth device to establish an association. Simultaneously, an association confidence score is calculated based on the total similarity score (e.g., a score of 90% corresponds to a confidence score of 0.9). Finally, the association relationship, association confidence score, and unique device identifier are integrated to generate and store a complete association result.

[0058] In step S140, the signal strength is filtered, and the presence status of the Bluetooth device is determined based on the filtered signal strength and the correlation result.

[0059] Filtering refers to the process of processing the raw signal strength acquired through scanning to eliminate interference and fluctuations in the signal, and extracting stable signal characteristics that truly reflect the status of the Bluetooth device. The presence status of the Bluetooth device refers to whether the user corresponding to the Bluetooth device is within the effective control range of the air conditioner, including the two core states of presence and absence, as well as the pending confirmation state in between.

[0060] The raw signal strength is easily affected by environmental interference (such as wall obstruction or interference from other electronic devices), causing fluctuations that cannot directly reflect the true state of the Bluetooth device. Filtering can eliminate random interference in the signal and extract stable signal characteristics; combined with the unique device identifier in the association results, signal changes of a specific Bluetooth device can be accurately tracked, thereby accurately determining whether the user corresponding to that device is within the air conditioner's control range.

[0061] Specifically, the air conditioner filters the raw signal strength obtained by scanning, first eliminating short-term random fluctuations, and then extracting stable signal features within a preset time window; based on the unique identifier of the Bluetooth device in the association results, it matches the historical existence record of the device; combining the filtered stable signal strength, the historical existence record of the device, and the frequency of signal occurrence, it comprehensively determines the existence status of the Bluetooth device (present, absent, or pending confirmation) through preset judgment rules and machine learning models.

[0062] In some implementations, step S140, filtering the signal strength, includes: performing a first-level filter on the signal strength to eliminate short-term random fluctuations in the signal strength; and performing a second-level filter on the first-level filtered signal strength within a preset time window to extract stable features of the signal strength and characterize the signal propagation state of the device.

[0063] The first-level filtering is a preliminary data processing step targeting the raw signal strength. Its core function is to eliminate short-term random interference in the signal, providing relatively stable basic data for subsequent processing. The second-level filtering, building upon the first-level filtering, performs in-depth processing on the signal data within a preset time window. Its core function is to extract stable trends and characteristic values ​​of the signal, accurately characterizing the signal propagation state of the Bluetooth device. The stable characteristics of the signal strength refer to the signal characteristics that, after filtering, accurately reflect the distance and positional relationship between the Bluetooth device and the air conditioner, including the average signal strength, signal fluctuation amplitude, and signal duration. The signal propagation state refers to the transmission of the Bluetooth signal from the device to the air conditioner, directly related to the distance between the device and the air conditioner, intermediate obstacles, and the intensity of environmental interference, and can indirectly reflect the device's presence.

[0064] The original signal strength is highly susceptible to external environmental interference. For example, slight shaking of the Bluetooth device by the user or movement of people nearby obstructing the signal can cause instantaneous changes in signal strength. These short-term fluctuations are unrelated to the actual presence of the device. The first-level filter smooths out data fluctuations, eliminates abnormal instantaneous values, and retains the core trend of the signal, avoiding random interference from affecting subsequent judgment results. However, the first-level filter only eliminates instantaneous interference; the signal strength may still fluctuate slowly with device movement and environmental changes. By performing a second-level filter within a preset time window, the overall characteristics of the signal can be extracted from the time dimension, such as the average signal strength, fluctuation amplitude, and sustained stability. These stable characteristics can accurately reflect the signal propagation state of the Bluetooth device, thus providing a reliable basis for determining its presence.

[0065] Specifically, the air conditioner retrieves the original signal strength sequence acquired through scanning and processes it using algorithms suitable for eliminating short-term interference, such as moving average filtering and median filtering. For example, when using moving average filtering, the original signal strength over 3-5 consecutive scanning cycles is used as a window, and the average value within the window is calculated as the signal strength value for that time period, replacing the original instantaneous value. This eliminates random fluctuations in a single cycle, resulting in relatively stable first-stage filtered data. A preset time window of 10-30 seconds is set, and all signal strength data after the first-stage filtering is collected within this time period. Deep processing is then performed using algorithms such as Kalman filtering and exponentially weighted average filtering. By predicting signal change trends and weightedly integrating historical and current data, stable features such as the mean, maximum, minimum, and variance of the signal strength within the time window are extracted. These features collectively constitute the signal propagation state characterization of the Bluetooth device, used for subsequent existence state determination.

[0066] In some implementations, step S140, determining the presence status of the Bluetooth device based on the filtered signal strength and the correlation result, includes steps S310 to S360.

[0067] Step S310: Determine the unique identifier of the Bluetooth device corresponding to the current broadcast packet based on the association result.

[0068] The unique identifier of a Bluetooth device is a special identification information used to accurately distinguish different Bluetooth devices. This identification information does not change with the random MAC address of the Bluetooth device, enabling continuous tracking of the same Bluetooth device.

[0069] The association results include the mapping between the current random MAC address and the Bluetooth device, as well as the device's unique identifier. This unique identifier allows for precise location of the target Bluetooth device, avoiding errors caused by device confusion. Historical records contain past state information of the target device, providing a reference for current state determination and improving the consistency and accuracy of the results. Specifically, the air conditioner parses the association results and extracts the unique identifier of the Bluetooth device corresponding to the current broadcast packet.

[0070] Step S320: Based on the unique identifier, filter the target scan records within the preset time window, count the number of scans by the Bluetooth device, and obtain the current occurrence frequency.

[0071] The preset time window is the time interval used to count the number of scans by Bluetooth devices. Its length needs to be determined based on the broadcast cycle of the Bluetooth devices and the actual application scenario requirements to ensure that the statistical results reflect the actual occurrence of the devices. The current occurrence frequency refers to the ratio of the number of times the air conditioner scanning module successfully obtained the broadcast packet of the target Bluetooth device within the preset time window to the total number of scans within that time window, or simply the number of successful scans within the preset time window, used to characterize the frequency of Bluetooth device occurrence in the current scenario.

[0072] Scan results at a single point in time may be subject to chance (such as false scans or missed scans caused by a single interference). By statistically analyzing the number of scans within a preset time window, the impact of chance factors can be reduced, providing a more objective reflection of whether Bluetooth devices are consistently present in the current scenario. The current frequency of occurrence is one of the important bases for determining the presence of a device; a higher frequency indicates a greater likelihood of the device's presence. Specifically, a preset time window (e.g., 30 seconds) is set, and all scan records within that time window are retrieved. Valid scan records containing the unique identifier of the target Bluetooth device are selected from the scan records, and the number of valid scan records is counted. This number of valid scan records is used as the current frequency of occurrence (or the ratio of the number of valid scan records to the total number of scans within the time window is calculated as the current frequency of occurrence), and this frequency data is stored.

[0073] Step S330: Determine whether the signal strength after filtering is within a preset effective strength range, and whether the current occurrence frequency reaches a preset occurrence threshold.

[0074] Step S340: If the signal strength is within the preset effective strength range and the current occurrence frequency reaches the preset occurrence threshold, then the initial state of the Bluetooth device is determined to be a candidate for presence.

[0075] Step S350: If the signal strength is not within the preset effective strength range and the current occurrence frequency has not reached the preset occurrence threshold, then the initial state of the Bluetooth device is determined to be an off-field candidate.

[0076] The preset effective strength range is a range of signal strength values ​​that reflects the Bluetooth device's position within the effective control range of the air conditioner. This range is determined based on factors such as the performance of the air conditioner's Bluetooth scanning module and the intensity of environmental interference in the application scenario. The preset occurrence threshold is the critical number (or frequency value) used to determine whether the Bluetooth device's occurrence is valid. When the current occurrence frequency reaches or exceeds this threshold, it indicates that the Bluetooth device's occurrence is relatively stable in the current scenario. The preliminary state refers to the preliminary result of the Bluetooth device's presence state obtained based on hard rules determined by signal strength and occurrence frequency, including two categories: present candidates and absent candidates.

[0077] The signal strength being within the effective range indicates that the device is relatively close to the air conditioner and is likely within its effective control range; the frequency of occurrence reaching the threshold indicates that the device is appearing stably, ruling out the possibility of accidental detection. When both conditions are met simultaneously, it can be preliminarily determined that the user corresponding to the device is likely present, thus the preliminary state is identified as a candidate for presence.

[0078] Specifically, the signal strength value after filtering is compared with a preset effective strength range (e.g., -90dBm to -70dBm) to determine if the signal strength is within the effective range. Simultaneously, the statistically obtained current occurrence frequency is compared with a preset occurrence threshold (e.g., 5 times / 30 seconds) to determine if the occurrence frequency meets the standard. If both conditions are met, a candidate status identifier for being present is generated, and this identifier is used as the initial status of the current Bluetooth device. If neither condition is met, a candidate status identifier for being absent is generated, and this identifier is used as the initial status of the current Bluetooth device. If the signal strength is within the preset effective strength range and the current occurrence frequency does not reach the preset occurrence threshold, or if the signal strength is not within the preset effective strength range and the current occurrence frequency reaches the preset occurrence threshold, then the initial status of the Bluetooth device is determined to be a candidate for correction.

[0079] Step S360: Correct the initial state based on a preset machine learning model to obtain the existence state of the Bluetooth device.

[0080] Preliminary states obtained solely through hard rules may be biased. For example, in scenarios where signal strength is at a critical value or frequency is close to a threshold, hard rules cannot accurately determine the state. Pre-defined machine learning models can analyze multi-dimensional features comprehensively, correcting the preliminary state and improving the accuracy of the judgment results.

[0081] Specifically, the initial state (present candidate or absent candidate) and the corresponding feature data (such as filtered signal strength, current frequency of occurrence, relevant data in historical existence records, etc.) are input into a preset machine learning model; the model analyzes and processes the input data, and corrects the initial state by combining the judgment rules learned during the training process; the corrected Bluetooth device presence state (present or absent) is output to complete the determination of the presence state.

[0082] Precise tracking of Bluetooth devices is achieved through unique identifiers, and historical presence records provide a reference for status determination, reducing errors caused by device confusion and random factors. By statistically analyzing the frequency of occurrence within a preset time window, the stability of device presence can be objectively reflected. Combined with filtered signal strength, hard rule determination is performed to quickly obtain the initial status, ensuring efficiency. Furthermore, a preset machine learning model refines the initial status, compensating for the limitations of hard rule determination and improving the accuracy of presence status determination.

[0083] For example, the air conditioner analysis results determine that the unique identifier of the smartphone corresponding to the current broadcast packet is "Device-001". The historical presence records corresponding to this identifier are retrieved, revealing that the user is typically at home (within the effective control range of the air conditioner) between 18:00 and 22:00 daily. A preset time window of 30 seconds is set, and the scan records within this window are counted. There are 6 valid scan records for "Device-001", with a current occurrence frequency of 6 times / 30 seconds. The preset effective signal strength range is -90dBm to -70dBm, and the current filtered signal strength is -68dBm, within the effective range. The preset occurrence threshold is 5 times / 30 seconds, and the current occurrence frequency of 6 times / 30 seconds meets the threshold. Since the signal strength is effective and the occurrence frequency meets the threshold, the smartphone is preliminarily identified as a candidate for presence. The system inputs data such as the presence candidate, signal strength of -68dBm, frequency of occurrence (6 times / 30 seconds), and historical records showing "frequently present from 18:00-22:00" into a pre-defined machine learning model. After analysis, the model determines that the current conditions meet the presence criteria, confirms the initial state (without correction), and finally outputs the presence state as "present." If the user gradually moves away from the air conditioner with their smartphone, and there are 2 valid scan records within the 30-second time window (not reaching the 5-scan threshold), and the filtered signal strength is -95dBm (outside the valid range), then the initial state is determined as an absence candidate. After inputting this data into the model, the model, considering the continuously weakening signal strength, corrects the output state to "absent." If a user carries a smartphone and moves around at the edge of the effective control range of the air conditioner, and the effective scan record is 3 times within 30 seconds (not reaching the threshold), but the signal strength is -75dBm (within the effective range), then it belongs to the intermediate case of "effective signal strength but frequency not meeting the standard", and the preliminary state is determined to be a candidate to be corrected; after inputting into the model, the model combines the historical activity trajectory to determine that the user is still present, and the output status is present.

[0084] In some implementations, step S360 involves correcting the initial state based on a preset machine learning model to obtain the presence state of the Bluetooth device, including steps S410 to S450.

[0085] Step S410: Using the preliminary state as an input constraint, assign basic weights to the presence probability of the candidates present and to the departure probability of the candidates leaving the field.

[0086] Input constraints use the initial state as a pre-defined input condition for the machine learning model. By assigning basic weights to different initial states, they define the range of probability calculations for the model, provide initial decision biases, avoid indiscriminate calculations across the entire range, and improve computational efficiency and decision accuracy. The presence probability basic weight refers to the pre-assigned basic probability weight value for the initial state when it is a candidate for presence, used to clarify the model's initial decision bias for that state (default: high probability of presence). The departure probability basic weight refers to the pre-assigned basic probability weight value for the initial state when it is a candidate for departure, used to clarify the model's initial decision bias for that state (default: high probability of departure).

[0087] The initial state is a rapid determination result based on hard rules of signal strength and frequency of occurrence, which can reflect the general tendency of the device's state. Using it as an input constraint and assigning corresponding basic weights allows the pre-set machine learning model to focus on the probability range that matches the initial state during calculation, avoiding indiscriminate calculation across the entire range. This not only adapts to the limited computing resources on the air conditioning end, but also reduces the interference of abnormal features on the determination results, improving computational efficiency and accuracy.

[0088] Specifically, if the initial state is a candidate for presence, it is assigned a preset basic weight for presence probability (e.g., 0.7), explicitly indicating that the model assumes the device is likely present. If the initial state is a candidate for departure, it is assigned a preset basic weight for departure probability (e.g., 0.3), explicitly indicating that the model assumes the device is likely to leave. The initial state with the basic weight is used as an input constraint and passed to the preset machine learning model. If the initial state is a candidate to be corrected, it is assigned a neutral basic weight (e.g., 0.5), meaning the model has no initial tendency to be present or depart, and only performs full-range probability calculations based on preset features, ultimately determining the state based on the threshold range into which the presence probability value falls.

[0089] Step S420: Input preset features into the preset machine learning model and calculate the presence probability value.

[0090] Preset features refer to multi-dimensional feature data input into a pre-defined machine learning model to assist in accurately calculating the presence probability value. These features can reflect information related to the presence status of Bluetooth devices from different perspectives. The presence probability value is a numerical value that characterizes the probability that the user corresponding to the Bluetooth device is within the effective control range of the air conditioner. The value range is usually 0-1 (or 0%-100%), and the higher the value, the greater the probability of presence.

[0091] A single initial state and basic weights cannot accurately cover the state determination needs in complex scenarios (such as critical signal fluctuations, changes in environmental interference, etc.). Preset features contain multi-dimensional data reflecting the device state, which can provide the model with a more comprehensive basis for judgment; the model combines input constraints and preset features for comprehensive calculation, which can achieve accurate correction of the initial state and output more reliable presence probability values.

[0092] Specifically, the air conditioner's control module retrieves the raw data corresponding to preset features from the storage unit, including the filtered mean signal strength, signal strength variance, current frequency of occurrence, and time period characteristics corresponding to the Bluetooth device; it standardizes this raw data (e.g., normalizes it to the same numerical range) to form a preset feature vector that meets the model's input requirements; it inputs the input constraints with basic weights and the preset feature vector together into a preset machine learning model; the model performs weighted analysis and feature fusion on the input data using a preset algorithm to calculate the presence probability value representing the likelihood of the device being present.

[0093] Step S430: If the presence probability value is greater than or equal to the first preset threshold, then the presence status of the Bluetooth device is determined to be present.

[0094] Step S440: If the presence probability value is less than or equal to the second preset threshold, then the presence state of the Bluetooth device is determined to be absent.

[0095] Step S450: If the presence probability value is greater than the second preset threshold and less than the first preset threshold, then the previous presence state is maintained.

[0096] The first preset threshold is a critical probability value used to determine the presence status as "present". When the presence probability value reaches or exceeds this threshold, the presence status can be clearly determined. The second preset threshold is a critical probability value used to determine the absence status as "absent". When the presence probability value is lower than or equal to this threshold, the absence status can be clearly determined. The first preset threshold is greater than the second preset threshold.

[0097] Specifically, the presence probability value output by the model is compared with a first preset threshold (e.g., 0.8) and a second preset threshold (e.g., 0.3). If the presence probability value is greater than or equal to the first preset threshold, the preliminary state "present candidate" is verified as valid, and the presence state of the Bluetooth device is determined to be present. If the presence probability value is less than or equal to the second preset threshold, the preliminary state "exit candidate" is verified as valid, and the presence state of the Bluetooth device is determined to be absent. If the second preset threshold is less than the presence probability value and less than the first preset threshold, the presence state (present or absent) obtained from the previous determination is retrieved. This state is maintained unchanged, and the adjustment of the air conditioning control strategy is not triggered. At the same time, the current fuzzy state is recorded to provide a reference for the determination of subsequent scanning cycles.

[0098] In step S150, a control strategy for the air conditioner is generated and executed based on the existing state.

[0099] The control strategy of an air conditioner refers to the operation adjustment plan formulated for the air conditioner based on the presence status of Bluetooth devices, including specific operation commands such as temperature adjustment, fan speed adjustment, operation mode switching, and on / off control.

[0100] The presence of Bluetooth devices directly reflects whether the user is in the air conditioning usage scenario. By formulating corresponding control strategies for different states, the air conditioner can be operated intelligently, reducing energy consumption while meeting user needs.

[0101] Specifically, based on the determined presence status of the Bluetooth device, the air conditioner invokes pre-stored strategy mapping rules: if the presence status is "present," a control strategy to maintain comfortable operation is generated (such as maintaining the preset temperature and automatic fan speed); if the presence status is "absent," an energy-saving control strategy is generated (such as increasing the temperature, decreasing the fan speed, or timed shutdown); if the presence status is "pending confirmation," the current operating state is maintained. After the control strategy is generated, the air conditioner adjusts its operating parameters according to the strategy instructions to complete the control operation.

[0102] This solution achieves accurate determination of user presence status through Bluetooth device broadcast packet identification and signal analysis, thereby generating and executing corresponding air conditioning control strategies. It not only solves the device identification problem caused by random MAC addresses of Bluetooth devices, ensuring the stability and accuracy of device association and providing a reliable foundation for presence status determination; but also improves the accuracy of the determination results by eliminating signal interference through filtering and combining multi-dimensional information to comprehensively determine presence status, avoiding misjudgments caused by signal fluctuations. Ultimately, it achieves intelligent operation of the air conditioner, providing a comfortable user experience when the user is present and automatically switching to energy-saving mode when the user leaves, effectively reducing energy consumption and meeting the dual needs of user experience and energy conservation.

[0103] For example, when a user brings their smartphone into the home, the air conditioner's Bluetooth scanning module scans every second, receiving broadcast packets from the smartphone and recording a signal strength of -65dBm. The air conditioner parses the broadcast packet, extracting feature information such as broadcast type, service UUID, and manufacturer-specific data to construct a device fingerprint. It compares the current device fingerprint with data in the historical fingerprint database, finding a 95% similarity (exceeding the preset 80% association threshold) with the user's smartphone's historical device fingerprint. Therefore, it associates the random MAC address in the current broadcast packet with the user's smartphone, generating an association result. The -65dBm signal strength is filtered to eliminate fluctuations caused by environmental interference, confirming signal stability. Combined with the smartphone's historical presence records, the air conditioner determines the user's presence. Based on this presence, the air conditioner generates and executes a comfort control strategy, maintaining the indoor temperature at 26℃ and automatically adjusting the fan speed to meet the user's needs. When a user leaves home with their smartphone, the air conditioner's Bluetooth scanning module cannot receive the phone's broadcast packets for an extended period. The filtered signal strength is below the effective range, indicating that the user is away. The system then generates an energy-saving strategy, raising the temperature to 28°C and automatically shutting off the unit after 30 minutes to reduce energy consumption.

[0104] Figure 3 The flowchart of the air conditioning control method based on the fusion of multi-feature fingerprint and RSSI is shown, including steps 1 to 7.

[0105] Step 1: Start scanning. The air conditioner automatically starts the Bluetooth Low Energy (BLE) scanning function according to a preset cycle (e.g., 1 second / time), or starts scanning under specific conditions (e.g., the air conditioner is turned on or the environmental sensor detects a change), and continuously receives broadcast packets sent out by all Bluetooth devices within the coverage area.

[0106] Step 2, raw data preprocessing: The air conditioner filters and denoises the received signal strength (RSSI) corresponding to the broadcast packets obtained by scanning, for example by using moving average, Kalman filtering, etc., to eliminate short-term random fluctuations and environmental interference in the signal and obtain relatively stable RSSI data.

[0107] Step 3: Device fingerprint association. The air conditioner extracts feature information (such as service UUID, manufacturer data, etc.) from the current broadcast packet to construct a device fingerprint. The fingerprint is then matched with the data in the pre-stored historical fingerprint database for similarity. If the match is successful, the random MAC address of the current broadcast packet is associated with the corresponding device entry. If the match fails, a new device entry is created and its fingerprint information is stored.

[0108] Step 4, Existence determination: The air conditioner counts the number of times the target device appears based on a preset short time window (e.g., 30 seconds), combines the pre-processed RSSI to determine whether it is within the effective threshold range, the rate of change of RSSI (reflecting whether the device has moved), and then makes a comprehensive determination through priority rules, finally outputting the user status corresponding to the device: "present", "absent", or "uncertain".

[0109] Step 5, Behavioral Learning and Strategy Decision-Making: The air conditioner learns users' daily departure times and stay patterns in different areas through local storage or cloud synchronization. Based on this data, it generates adaptive timeouts (such as turning off the air conditioner 30 minutes after the user usually leaves home) and formulates corresponding energy-saving actions (such as raising the temperature and lowering the fan speed).

[0110] Step 6, Execution and Feedback: Based on the presence determination result and strategy decision, the air conditioner triggers the corresponding operation mode switch (e.g., switch to energy-saving mode and shut down after timeout when the user is "not present"; maintain comfort mode when the user is "present"); at the same time, it pushes reminders to the user's APP or records operation logs and reports them to the cloud when necessary (e.g., long-term inactivity or abnormal status).

[0111] Step 7, Manual / APP Coverage: Users can manually adjust the air conditioner's operating mode (covering the automatic strategy) through the air conditioner's control panel or the corresponding mobile APP. They can also configure a whitelist (only recognize specific devices), a blacklist (ignore specific devices), and set sensitive time periods (such as not triggering state switching during rest periods).

[0112] In some embodiments, the air conditioner includes a BLE scanning module, a data preprocessing module, a device fingerprint association module, an existence determination module, a behavior learning and strategy decision-making module, and an execution and feedback module.

[0113] The BLE scanning module is installed inside the air conditioner and is used to initiate Bluetooth Low Energy (BLE) scanning operations according to a preset cycle or triggering conditions, continuously receiving broadcast packets sent by Bluetooth devices within its communication coverage area.

[0114] The data preprocessing module is connected to the BLE scanning module and receives the received signal strength (RSSI) data corresponding to the broadcast packets transmitted by the module. It processes the raw RSSI through filtering and denoising algorithms (such as moving average filtering and Kalman filtering) to eliminate short-term random fluctuations and environmental interference, and outputs stable RSSI characteristic data.

[0115] The device fingerprint association module communicates with the data preprocessing module to extract feature information (such as service UUID, manufacturer-specific data, etc.) from the broadcast packet to construct the current device fingerprint. The fingerprint is then matched with the pre-stored historical fingerprint database for similarity. If the match is successful, the association between the random MAC address of the current broadcast packet and the corresponding device entry is established. If the match fails, a new device entry is created and its fingerprint information is stored to achieve unique identification and stable association of Bluetooth devices.

[0116] The existence determination module communicates with the device fingerprint association module. Based on a preset short time window, it counts the number of times the target device appears. Combined with the preprocessed RSSI threshold range and RSSI change rate (characterizing the device's motion state), it performs a comprehensive analysis through priority determination rules and outputs three user status results: "present", "absent", and "uncertain".

[0117] The behavior learning and strategy decision-making module communicates with the existence determination module. It learns and stores users' daily behavior data (such as time away from home and stay patterns) through local storage or cloud synchronization. Based on this data, it generates adaptive timeout time and corresponding energy-saving control strategies (such as temperature adjustment and fan speed adjustment) to adapt the control strategies to users' personalized usage habits.

[0118] The execution and feedback module communicates with the behavior learning and strategy decision-making module. Based on the existence determination result and decision-making strategy, it triggers the switching operation of the air conditioner operation mode. At the same time, under preset conditions (such as abnormal status or timeout trigger), it pushes reminder information to the user terminal APP, or records operation logs and reports them to the cloud, so as to realize the execution of control actions and interactive feedback.

[0119] Furthermore, the Bluetooth device is equipped with a broadcast packet sending module, which periodically sends BLE broadcast packets containing its own characteristic information as the data source for various modules on the air conditioner side; at the same time, the user terminal APP corresponding to the Bluetooth device is equipped with a manual interaction module, which supports users to manually override automatic control policies, configure device whitelists / blacklists and sensitive time periods, and realize personalized intervention and configuration of automatic processes.

[0120] The technical solution of this embodiment scans the broadcast packets and signal strength of Bluetooth devices, extracts broadcast packet features to construct a device fingerprint, associates a random MAC address with the device through fingerprint matching, filters the signal strength, and combines the association results to determine the device's presence status. Finally, based on this status, an air conditioning control strategy is generated and executed. This improves the accuracy of the air conditioner in recognizing the user's presence status, enhances the intelligence of air conditioning control, and balances user comfort with the air conditioner's energy efficiency.

[0121] According to an embodiment of the present invention, an air conditioner control device corresponding to the air conditioner control method is also provided. See also Figure 2 The schematic diagram shown is a structural diagram of an embodiment of the device of the present invention. The control device of the air conditioner may include: an acquisition unit 101, a fingerprint construction unit 102, a device association unit 103, a status determination unit 104, and a control execution unit 105.

[0122] Acquisition unit 101 is configured to acquire broadcast packets from a Bluetooth device and the signal strength of the broadcast packets. For the specific functions and processing of this unit, please refer to step S110.

[0123] Bluetooth devices refer to electronic devices that support Bluetooth wireless communication technology, including but not limited to smartphones, smartwatches, Bluetooth headsets, and smart home terminals. Broadcast packets are data packets containing device information that Bluetooth devices periodically send out. They are the primary way Bluetooth devices transmit information without pairing and include core information such as device identification, service information, and manufacturer information. Signal strength refers to the signal reception power of an air conditioner's Bluetooth scanning module when it receives a broadcast packet from a Bluetooth device. It is usually measured in dBm. The signal strength value is negatively correlated with the distance between the Bluetooth device and the air conditioner; the closer the distance, the higher the signal strength value (closer to 0).

[0124] Bluetooth devices continuously send out broadcast packets, which the air conditioner receives via its built-in Bluetooth scanning module, recording the signal strength of each packet. Signal strength reflects the distance between the Bluetooth device and the air conditioner, serving as a key criterion for determining user presence; the characteristic information within the broadcast packets forms the basis for constructing device fingerprints and establishing device association.

[0125] Specifically, the air conditioner's Bluetooth scanning module starts scanning at a preset interval (e.g., once every second) to search for broadcast packets sent by all Bluetooth devices within the coverage area; for each broadcast packet received, the content of the broadcast packet and the corresponding signal strength value are recorded simultaneously.

[0126] The fingerprint construction unit 102 is configured to extract multiple types of feature information from the broadcast packet to construct a device fingerprint. For the specific functions and processing of this unit, please refer to step S120.

[0127] Device fingerprinting is a set of features built upon multiple types of characteristic information from Bluetooth device broadcast packets. It uniquely identifies a Bluetooth device and can be used to distinguish different Bluetooth devices, unaffected by random changes in the device's MAC address. These multiple types of characteristic information include: the Bluetooth device's broadcast type, service UUID, vendor-specific data, instantaneous signal strength values, and trends in signal strength.

[0128] Bluetooth devices' random MAC addresses change periodically and cannot serve as stable device identifiers. However, certain characteristic information in broadcast packets possesses uniqueness and stability (such as vendor-specific data and service UUIDs). Device fingerprints constructed based on this characteristic information can uniquely identify Bluetooth devices, solving the problem of device identification difficulties caused by random MAC addresses.

[0129] Specifically, the air conditioner's control module parses the received broadcast packets and extracts multiple types of feature information with stable identification characteristics. The extracted feature information is then structured and combined into a complete feature set according to preset algorithm rules. This feature set is the device fingerprint of the corresponding Bluetooth device.

[0130] The device association unit 103 is configured to perform similarity matching based on the device fingerprint to associate the random MAC address in the broadcast packet with the same Bluetooth device, thereby obtaining an association result. For the specific functions and processing of this unit, please refer to step S130.

[0131] A random MAC address is a media access control address that Bluetooth devices periodically generate to protect privacy. The same Bluetooth device may use different random MAC addresses at different times and in different scenarios. The association result refers to the set of information generated after establishing an association between the random MAC address in the current broadcast packet and the corresponding Bluetooth device through device fingerprint matching. This includes the correspondence between the random MAC address and the Bluetooth device, the trustworthiness of the association, and the unique identifier of the Bluetooth device.

[0132] Broadcast packets sent by the same Bluetooth device at different times may have different random MAC addresses, but their core feature information (i.e., device fingerprint) remains consistent. By matching the current device fingerprint with historical device fingerprints, the Bluetooth device to which the current broadcast packet belongs can be identified, achieving a stable association between the random MAC address and the Bluetooth device, and providing an accurate device identifier for subsequent status determination.

[0133] Specifically, the air conditioner compares the device fingerprint corresponding to the current broadcast packet with each historical device fingerprint in the pre-stored historical device fingerprint database; it calculates the similarity between the current device fingerprint and each historical device fingerprint using a preset algorithm, and selects the historical device fingerprint with the highest similarity; if the similarity reaches a preset association threshold, it determines that the current broadcast packet and the Bluetooth device corresponding to the historical device fingerprint are the same device, and establishes an association between the random MAC address in the current broadcast packet and the Bluetooth device, generating an association result that includes the association relationship, association confidence, and device unique identifier.

[0134] In some implementations, the device association unit 103 performs similarity matching based on the device fingerprint to associate the random MAC address in the broadcast packet with the same Bluetooth device, obtaining an association result, including:

[0135] The device association unit 103 is further configured to extract historical device fingerprints corresponding to broadcast packets acquired through historical scanning, and construct a historical fingerprint database; the historical fingerprint database stores the correspondence between historical random MAC addresses and historical device fingerprints. For the specific functions and processing of this unit, please refer to step S210.

[0136] Historical device fingerprints refer to the device fingerprints constructed by parsing Bluetooth device broadcast packets acquired during past scans of the air conditioner and extracting feature information. These fingerprints are historical data used to match the current device fingerprint. The historical fingerprint database refers to a collection of historical device fingerprints stored internally by the air conditioner or in a cloud-connected system. It also stores the mapping relationship between historical random MAC addresses and their corresponding historical device fingerprints.

[0137] Building a historical fingerprint database enables centralized storage and management of device fingerprints, facilitating rapid data retrieval for similarity matching. Simultaneously, storing the correspondence between historical random MAC addresses and historical device fingerprints allows for tracing device address change records, improving the stability of associations. Specifically, the air conditioner extracts all Bluetooth device broadcast packets stored during past scanning processes, performs feature analysis and device fingerprint construction on each broadcast packet to obtain the corresponding historical device fingerprint. Each historical device fingerprint is then bound to its corresponding historical random MAC address and device association information, organized and stored according to a preset data structure to form a complete historical fingerprint database. Furthermore, the system periodically updates the historical fingerprint database, deleting expired and invalid historical data to ensure data timeliness.

[0138] The device association unit 103 is further configured to calculate the multi-dimensional similarity between the device fingerprint corresponding to the current broadcast packet and the historical device fingerprints in the historical fingerprint database; the multi-dimensional similarity includes numerical vector similarity and set similarity. For the specific functions and processing of this unit, please refer to step S220.

[0139] Multi-dimensional similarity calculates the similarity between the current device fingerprint and historical device fingerprints from different feature dimensions. It includes two categories: numerical vector similarity and set similarity, comprehensively reflecting the matching degree between two fingerprints. Numerical vector similarity targets quantifiable feature information in broadcast packets (such as instantaneous signal strength values, data field lengths, etc.), converting them into numerical vectors and calculating the vector similarity through algorithms, accurately reflecting the consistency of numerical features. Set similarity targets discrete feature information in broadcast packets (such as service UUID, broadcast type, etc.), treating them as set elements and calculating the set overlap through algorithms, effectively reflecting the matching degree of discrete features.

[0140] Single-dimensional similarity calculations cannot fully reflect the matching situation of device fingerprints and may lead to matching errors due to interference from some features. Calculating similarity using two dimensions—numerical vector similarity and set similarity—can cover the matching needs of both quantized and discrete features, improving the comprehensiveness and accuracy of similarity calculations. Specifically, quantized features in the current device fingerprint (such as signal strength values, data field lengths, etc.) are transformed into standardized numerical vectors, while discrete features (such as service UUID, broadcast type, etc.) are transformed into feature sets. For each historical device fingerprint in the historical fingerprint database, the cosine similarity algorithm is used to calculate the numerical vector similarity, and the Jaccard similarity algorithm is used to calculate the set similarity, obtaining the two-dimensional similarity results between the current device fingerprint and each historical device fingerprint.

[0141] The process of calculating the similarity of numerical vectors using the cosine similarity algorithm is as follows: First, calculate the weighted Euclidean distance. Then, use radial basis function (RBF) to convert it into similarity. .in, The weights corresponding to the i-th feature dimension are: Let x be the eigenvalue of the numerical vector x in the i-th dimension. Let be the eigenvalue of the numerical vector y in the i-th dimension. This represents the bandwidth parameter of the radial basis core.

[0142] The formula for calculating set similarity using the Jaccard similarity algorithm is as follows: Where A is the discrete feature set corresponding to the current device fingerprint, and B is the discrete feature set corresponding to the historical device fingerprint.

[0143] The device association unit 103 is further configured to perform weighted fusion on the multi-dimensional similarity to obtain a total similarity score. The specific functions and processing of this unit are described in step S230.

[0144] Weighted fusion is a process that assigns different weights to the similarity of different dimensions based on their importance, and then calculates the comprehensive similarity score through weighted summation. This can highlight the influence of core features and improve matching accuracy.

[0145] Numerical vector similarity and set similarity have different levels of importance in device identification. For example, the numerical vector similarity corresponding to manufacturer-specific data has a greater impact on device uniqueness. Weighted fusion can highlight the weight of core features, avoid matching deviations caused by interference from secondary features, and improve the reliability of the overall similarity score. Specifically, according to a preset weight allocation rule, weights are assigned to numerical vector similarity and set similarity respectively (e.g., numerical vector similarity accounts for 65% and set similarity accounts for 35%). The similarity results of the two dimensions are multiplied by their corresponding weights, and then the products are added together to obtain the overall similarity score between the current device fingerprint and the corresponding historical device fingerprint. For example, if the calculated numerical vector similarity is 92% and the set similarity is 88%, the overall similarity score, calculated using weighted fusion of 65% and 35%, is 92% × 65% + 88% × 35% = 90.6%.

[0146] The device association unit 103 is further configured to, if the total similarity score reaches a preset association threshold, establish an association between the random MAC address in the current broadcast packet and the Bluetooth device corresponding to the historical device fingerprint, and generate an association result. The association result includes the association relationship between the current random MAC address and the corresponding Bluetooth device, the association confidence level, and the unique identifier corresponding to the Bluetooth device. For the specific functions and processing of this unit, please refer to step S240.

[0147] The preset association threshold is a pre-defined similarity judgment threshold. When the total similarity score reaches or exceeds this threshold, the current device fingerprint is determined to match the historical device fingerprint, and a device association can be established. The association confidence refers to the degree of credibility in establishing an association between the current random MAC address and the corresponding Bluetooth device. It is usually positively correlated with the total similarity score and is used to measure the reliability of the association results.

[0148] A preset association threshold is the core criterion for determining device matching, ensuring that only highly similar device fingerprints can be associated, thus avoiding false associations. Generating association results containing multiple types of information provides comprehensive data support for subsequent existence status determination and device tracking. Specifically, the calculated total similarity score is compared with a preset association threshold (e.g., 80%). If the total similarity score reaches or exceeds this threshold, the Bluetooth device corresponding to the current broadcast packet is determined to be the same device as the Bluetooth device corresponding to the historical device fingerprint. The current random MAC address is then bound to the unique identifier of the Bluetooth device to establish an association. Simultaneously, an association confidence score is calculated based on the total similarity score (e.g., a score of 90% corresponds to a confidence score of 0.9). Finally, the association relationship, association confidence score, and unique device identifier are integrated to generate and store a complete association result.

[0149] The state determination unit 104 is configured to filter the signal strength and determine the presence state of the Bluetooth device based on the filtered signal strength and the correlation result. For the specific function and processing of this unit, please refer to step S140.

[0150] Filtering refers to the process of processing the raw signal strength acquired through scanning to eliminate interference and fluctuations in the signal, and extracting stable signal characteristics that truly reflect the status of the Bluetooth device. The presence status of the Bluetooth device refers to whether the user corresponding to the Bluetooth device is within the effective control range of the air conditioner, including the two core states of presence and absence, as well as the pending confirmation state in between.

[0151] The raw signal strength is easily affected by environmental interference (such as wall obstruction or interference from other electronic devices), causing fluctuations that cannot directly reflect the true state of the Bluetooth device. Filtering can eliminate random interference in the signal and extract stable signal characteristics; combined with the unique device identifier in the association results, signal changes of a specific Bluetooth device can be accurately tracked, thereby accurately determining whether the user corresponding to that device is within the air conditioner's control range.

[0152] Specifically, the air conditioner filters the raw signal strength obtained by scanning, first eliminating short-term random fluctuations, and then extracting stable signal features within a preset time window; based on the unique identifier of the Bluetooth device in the association results, it matches the historical existence record of the device; combining the filtered stable signal strength, the historical existence record of the device, and the frequency of signal occurrence, it comprehensively determines the existence status of the Bluetooth device (present, absent, or pending confirmation) through preset judgment rules and machine learning models.

[0153] In some implementations, the state determination unit 104 performs filtering processing on the signal strength, including: performing a first-level filtering on the signal strength to eliminate short-term random fluctuations in the signal strength; and performing a second-level filtering on the first-level filtered signal strength within a preset time window to extract stable features of the signal strength and characterize the signal propagation state of the device.

[0154] The first-level filtering is a preliminary data processing step targeting the raw signal strength. Its core function is to eliminate short-term random interference in the signal, providing relatively stable basic data for subsequent processing. The second-level filtering, building upon the first-level filtering, performs in-depth processing on the signal data within a preset time window. Its core function is to extract stable trends and characteristic values ​​of the signal, accurately characterizing the signal propagation state of the Bluetooth device. The stable characteristics of the signal strength refer to the signal characteristics that, after filtering, accurately reflect the distance and positional relationship between the Bluetooth device and the air conditioner, including the average signal strength, signal fluctuation amplitude, and signal duration. The signal propagation state refers to the transmission of the Bluetooth signal from the device to the air conditioner, directly related to the distance between the device and the air conditioner, intermediate obstacles, and the intensity of environmental interference, and can indirectly reflect the device's presence.

[0155] The original signal strength is highly susceptible to external environmental interference. For example, slight shaking of the Bluetooth device by the user or movement of people nearby obstructing the signal can cause instantaneous changes in signal strength. These short-term fluctuations are unrelated to the actual presence of the device. The first-level filter smooths out data fluctuations, eliminates abnormal instantaneous values, and retains the core trend of the signal, avoiding random interference from affecting subsequent judgment results. However, the first-level filter only eliminates instantaneous interference; the signal strength may still fluctuate slowly with device movement and environmental changes. By performing a second-level filter within a preset time window, the overall characteristics of the signal can be extracted from the time dimension, such as the average signal strength, fluctuation amplitude, and sustained stability. These stable characteristics can accurately reflect the signal propagation state of the Bluetooth device, thus providing a reliable basis for determining its presence.

[0156] Specifically, the air conditioner retrieves the original signal strength sequence acquired through scanning and processes it using algorithms suitable for eliminating short-term interference, such as moving average filtering and median filtering. For example, when using moving average filtering, the original signal strength over 3-5 consecutive scanning cycles is used as a window, and the average value within the window is calculated as the signal strength value for that time period, replacing the original instantaneous value. This eliminates random fluctuations in a single cycle, resulting in relatively stable first-stage filtered data. A preset time window of 10-30 seconds is set, and all signal strength data after the first-stage filtering is collected within this time period. Deep processing is then performed using algorithms such as Kalman filtering and exponentially weighted average filtering. By predicting signal change trends and weightedly integrating historical and current data, stable features such as the mean, maximum, minimum, and variance of the signal strength within the time window are extracted. These features collectively constitute the signal propagation state characterization of the Bluetooth device, used for subsequent existence state determination.

[0157] In some implementations, the state determination unit 104 determines the presence state of the Bluetooth device based on the filtered signal strength and the correlation result, including:

[0158] The status determination unit 104 is further configured to determine the unique identifier of the Bluetooth device corresponding to the current broadcast packet based on the association result. See step S310 for the specific functions and processing of this unit.

[0159] The unique identifier of a Bluetooth device is a special identification information used to accurately distinguish different Bluetooth devices. This identification information does not change with the random MAC address of the Bluetooth device, enabling continuous tracking of the same Bluetooth device.

[0160] The association results include the mapping between the current random MAC address and the Bluetooth device, as well as the device's unique identifier. This unique identifier allows for precise location of the target Bluetooth device, avoiding errors caused by device confusion. Historical records contain past state information of the target device, providing a reference for current state determination and improving the consistency and accuracy of the results. Specifically, the air conditioner parses the association results and extracts the unique identifier of the Bluetooth device corresponding to the current broadcast packet.

[0161] The status determination unit 104 is further configured to filter target scan records within a preset time window based on the unique identifier, count the number of scans by the Bluetooth device, and obtain the current occurrence frequency. For the specific functions and processing of this unit, please refer to step S320.

[0162] The preset time window is the time interval used to count the number of scans by Bluetooth devices. Its length needs to be determined based on the broadcast cycle of the Bluetooth devices and the actual application scenario requirements to ensure that the statistical results reflect the actual occurrence of the devices. The current occurrence frequency refers to the ratio of the number of times the air conditioner scanning module successfully obtained the broadcast packet of the target Bluetooth device within the preset time window to the total number of scans within that time window, or simply the number of successful scans within the preset time window, used to characterize the frequency of Bluetooth device occurrence in the current scenario.

[0163] Scan results at a single point in time may be subject to chance (such as false scans or missed scans caused by a single interference). By statistically analyzing the number of scans within a preset time window, the impact of chance factors can be reduced, providing a more objective reflection of whether Bluetooth devices are consistently present in the current scenario. The current frequency of occurrence is one of the important bases for determining the presence of a device; a higher frequency indicates a greater likelihood of the device's presence. Specifically, a preset time window (e.g., 30 seconds) is set, and all scan records within that time window are retrieved. Valid scan records containing the unique identifier of the target Bluetooth device are selected from the scan records, and the number of valid scan records is counted. This number of valid scan records is used as the current frequency of occurrence (or the ratio of the number of valid scan records to the total number of scans within the time window is calculated as the current frequency of occurrence), and this frequency data is stored.

[0164] The state determination unit 104 is further configured to determine whether the signal strength after filtering is within a preset effective strength range, and whether the current occurrence frequency reaches a preset occurrence threshold. The specific function and processing of this unit are described in step S330.

[0165] The state determination unit 104 is further configured to determine the initial state of the Bluetooth device as a candidate for presence if the signal strength is within the preset effective strength range and the current occurrence frequency reaches a preset occurrence threshold. The specific function and processing of this unit are described in step S340.

[0166] The state determination unit 104 is further configured to determine the initial state of the Bluetooth device as an off-field candidate if the signal strength is not within the preset effective strength range and the current occurrence frequency has not reached the preset occurrence threshold. The specific function and processing of this unit are described in step S350.

[0167] The preset effective strength range is a range of signal strength values ​​that reflects the Bluetooth device's position within the effective control range of the air conditioner. This range is determined based on factors such as the performance of the air conditioner's Bluetooth scanning module and the intensity of environmental interference in the application scenario. The preset occurrence threshold is the critical number (or frequency value) used to determine whether the Bluetooth device's occurrence is valid. When the current occurrence frequency reaches or exceeds this threshold, it indicates that the Bluetooth device's occurrence is relatively stable in the current scenario. The preliminary state refers to the preliminary result of the Bluetooth device's presence state obtained based on hard rules determined by signal strength and occurrence frequency, including two categories: present candidates and absent candidates.

[0168] The signal strength being within the effective range indicates that the device is relatively close to the air conditioner and is likely within its effective control range; the frequency of occurrence reaching the threshold indicates that the device is appearing stably, ruling out the possibility of accidental detection. When both conditions are met simultaneously, it can be preliminarily determined that the user corresponding to the device is highly likely to be present, therefore the preliminary state is identified as a candidate for presence.

[0169] Specifically, the signal strength value after filtering is compared with a preset effective strength range (e.g., -90dBm to -70dBm) to determine if the signal strength is within the effective range. Simultaneously, the statistically obtained current occurrence frequency is compared with a preset occurrence threshold (e.g., 5 times / 30 seconds) to determine if the occurrence frequency meets the standard. If both conditions are met, a candidate status identifier for being present is generated, and this identifier is used as the initial status of the current Bluetooth device. If neither condition is met, a candidate status identifier for being absent is generated, and this identifier is used as the initial status of the current Bluetooth device. If the signal strength is within the preset effective strength range and the current occurrence frequency does not reach the preset occurrence threshold, or if the signal strength is not within the preset effective strength range and the current occurrence frequency reaches the preset occurrence threshold, then the initial status of the Bluetooth device is determined to be a candidate for correction.

[0170] The state determination unit 104 is further configured to correct the initial state based on a preset machine learning model to obtain the presence state of the Bluetooth device. For the specific functions and processing of this unit, please refer to step S360.

[0171] Preliminary states obtained solely through hard rules may be biased. For example, in scenarios where signal strength is at a critical value or frequency is close to a threshold, hard rules cannot accurately determine the state. Pre-defined machine learning models can analyze multi-dimensional features comprehensively, correcting the preliminary state and improving the accuracy of the judgment results.

[0172] Specifically, the initial state (present candidate or absent candidate) and the corresponding feature data (such as filtered signal strength, current frequency of occurrence, relevant data in historical existence records, etc.) are input into a preset machine learning model; the model analyzes and processes the input data, and corrects the initial state by combining the judgment rules learned during the training process; the corrected Bluetooth device presence state (present or absent) is output to complete the determination of the presence state.

[0173] Precise tracking of Bluetooth devices is achieved through unique identifiers, and historical presence records provide a reference for status determination, reducing errors caused by device confusion and random factors. By statistically analyzing the frequency of occurrence within a preset time window, the stability of device presence can be objectively reflected. Combined with filtered signal strength, hard rule determination is performed to quickly obtain the initial status, ensuring efficiency. Furthermore, a preset machine learning model refines the initial status, compensating for the limitations of hard rule determination and improving the accuracy of presence status determination.

[0174] In some implementations, the state determination unit 104 corrects the preliminary state based on a preset machine learning model to obtain the presence state of the Bluetooth device, including:

[0175] The state determination unit 104 is further configured to use the preliminary state as an input constraint, assign basic weights to the presence probability of the candidates present, and assign basic weights to the departure probability of the candidates leaving. The specific functions and processing of this unit are described in step S410.

[0176] Input constraints use the initial state as a pre-defined input condition for the machine learning model. By assigning basic weights to different initial states, they define the range of probability calculations for the model, provide initial decision biases, avoid indiscriminate calculations across the entire range, and improve computational efficiency and decision accuracy. The presence probability basic weight refers to the pre-assigned basic probability weight value for the initial state when it is a candidate for presence, used to clarify the model's initial decision bias for that state (default: high probability of presence). The departure probability basic weight refers to the pre-assigned basic probability weight value for the initial state when it is a candidate for departure, used to clarify the model's initial decision bias for that state (default: high probability of departure).

[0177] The initial state is a rapid determination result based on hard rules of signal strength and frequency of occurrence, which can reflect the general tendency of the device's state. Using it as an input constraint and assigning corresponding basic weights allows the pre-set machine learning model to focus on the probability range that matches the initial state during calculation, avoiding indiscriminate calculation across the entire range. This not only adapts to the limited computing resources on the air conditioning end, but also reduces the interference of abnormal features on the determination results, improving computational efficiency and accuracy.

[0178] Specifically, if the initial state is a candidate for presence, it is assigned a preset basic weight for presence probability (e.g., 0.7), explicitly indicating that the model assumes the device is likely present. If the initial state is a candidate for departure, it is assigned a preset basic weight for departure probability (e.g., 0.3), explicitly indicating that the model assumes the device is likely to leave. The initial state with the basic weight is used as an input constraint and passed to the preset machine learning model. If the initial state is a candidate to be corrected, it is assigned a neutral basic weight (e.g., 0.5), meaning the model has no initial tendency to be present or depart, and only performs full-range probability calculations based on preset features, ultimately determining the state based on the threshold range into which the presence probability value falls.

[0179] The state determination unit 104 is further configured to input preset features into the preset machine learning model and calculate the presence probability value. For the specific function and processing of this unit, please refer to step S420.

[0180] Preset features refer to multi-dimensional feature data input into a pre-defined machine learning model to assist in accurately calculating the presence probability value. These features can reflect information related to the presence status of Bluetooth devices from different perspectives. The presence probability value is a numerical value that characterizes the probability that the user corresponding to the Bluetooth device is within the effective control range of the air conditioner. The value range is usually 0-1 (or 0%-100%), and the higher the value, the greater the probability of presence.

[0181] A single initial state and basic weights cannot accurately cover the state determination needs in complex scenarios (such as critical signal fluctuations, changes in environmental interference, etc.). Preset features contain multi-dimensional data reflecting the device state, which can provide the model with a more comprehensive basis for judgment; the model combines input constraints and preset features for comprehensive calculation, which can achieve accurate correction of the initial state and output more reliable presence probability values.

[0182] Specifically, the air conditioner's control module retrieves the raw data corresponding to preset features from the storage unit, including the filtered mean signal strength, signal strength variance, current frequency of occurrence, and time period characteristics corresponding to the Bluetooth device; it standardizes this raw data (e.g., normalizes it to the same numerical range) to form a preset feature vector that meets the model's input requirements; it inputs the input constraints with basic weights and the preset feature vector together into a preset machine learning model; the model performs weighted analysis and feature fusion on the input data using a preset algorithm to calculate the presence probability value representing the likelihood of the device being present.

[0183] The state determination unit 104 is further configured to determine that the Bluetooth device is present if the presence probability value is greater than or equal to a first preset threshold. The specific function and processing of this unit are described in step S430.

[0184] The state determination unit 104 is further configured to determine that the presence state of the Bluetooth device is "away" if the presence probability value is less than or equal to a second preset threshold. The specific function and processing of this unit are described in step S440.

[0185] The state determination unit 104 is further configured to maintain the previous presence state if the presence probability value is greater than a second preset threshold and less than a first preset threshold. The specific function and processing of this unit are described in step S450.

[0186] The first preset threshold is a critical probability value used to determine the presence status as "present". When the presence probability value reaches or exceeds this threshold, the presence status can be clearly determined. The second preset threshold is a critical probability value used to determine the absence status as "absent". When the presence probability value is lower than or equal to this threshold, the absence status can be clearly determined. The first preset threshold is greater than the second preset threshold.

[0187] Specifically, the presence probability value output by the model is compared with a first preset threshold (e.g., 0.8) and a second preset threshold (e.g., 0.3). If the presence probability value is greater than or equal to the first preset threshold, the preliminary state "present candidate" is verified as valid, and the presence state of the Bluetooth device is determined to be present. If the presence probability value is less than or equal to the second preset threshold, the preliminary state "exit candidate" is verified as valid, and the presence state of the Bluetooth device is determined to be absent. If the second preset threshold is less than the presence probability value and less than the first preset threshold, the presence state (present or absent) obtained from the previous determination is retrieved. This state is maintained unchanged, and the adjustment of the air conditioning control strategy is not triggered. At the same time, the current fuzzy state is recorded to provide a reference for the determination of subsequent scanning cycles.

[0188] The control execution unit 105 is configured to generate and execute a control strategy for the air conditioner based on the stated state. The specific functions and processing of this unit are described in step S150.

[0189] The control strategy of an air conditioner refers to the operation adjustment plan formulated for the air conditioner based on the presence status of Bluetooth devices, including specific operation commands such as temperature adjustment, fan speed adjustment, operation mode switching, and on / off control.

[0190] The presence of Bluetooth devices directly reflects whether the user is in the air conditioning usage scenario. By formulating corresponding control strategies for different states, the air conditioner can be operated intelligently, reducing energy consumption while meeting user needs.

[0191] Specifically, based on the determined presence status of the Bluetooth device, the air conditioner invokes pre-stored strategy mapping rules: if the presence status is "present," a control strategy to maintain comfortable operation is generated (such as maintaining the preset temperature and automatic fan speed); if the presence status is "absent," an energy-saving control strategy is generated (such as increasing the temperature, decreasing the fan speed, or timed shutdown); if the presence status is "pending confirmation," the current operating state is maintained. After the control strategy is generated, the air conditioner adjusts its operating parameters according to the strategy instructions to complete the control operation.

[0192] This solution achieves accurate determination of user presence status through Bluetooth device broadcast packet identification and signal analysis, thereby generating and executing corresponding air conditioning control strategies. It not only solves the device identification problem caused by random MAC addresses of Bluetooth devices, ensuring the stability and accuracy of device association and providing a reliable foundation for presence status determination; but also improves the accuracy of the determination results by eliminating signal interference through filtering and combining multi-dimensional information to comprehensively determine presence status, avoiding misjudgments caused by signal fluctuations. Ultimately, it achieves intelligent operation of the air conditioner, providing a comfortable user experience when the user is present and automatically switching to energy-saving mode when the user leaves, effectively reducing energy consumption and meeting the dual needs of user experience and energy conservation.

[0193] Since the processing and functions implemented by the device in this embodiment are basically the same as the embodiments, principles and examples of the aforementioned methods, any details not covered in the description of this embodiment can be found in the relevant descriptions in the aforementioned embodiments, and will not be repeated here.

[0194] The technical solution of this invention involves an air conditioner scanning the broadcast packets and signal strength of Bluetooth devices, extracting broadcast packet features to construct a device fingerprint, associating a random MAC address with the device through fingerprint matching, filtering the signal strength, and combining the association results to determine the device's presence status. Finally, based on this status, an air conditioner control strategy is generated and executed. This improves the accuracy of the air conditioner in recognizing the user's presence status, enhances the intelligence of air conditioner control, and balances user comfort with energy-saving efficiency.

[0195] According to an embodiment of the present invention, a system corresponding to an air conditioner control device is also provided. This system may include the air conditioner control device described above.

[0196] Since the processing and functions implemented by the system in this embodiment are basically the same as the embodiments, principles and examples of the aforementioned devices, any details not covered in the description of this embodiment can be found in the relevant descriptions in the aforementioned embodiments, and will not be repeated here.

[0197] The technical solution of this invention involves an air conditioner scanning the broadcast packets and signal strength of Bluetooth devices, extracting broadcast packet features to construct a device fingerprint, associating a random MAC address with the device through fingerprint matching, filtering the signal strength, and combining the association results to determine the device's presence status. Finally, based on this status, an air conditioner control strategy is generated and executed. This improves the accuracy of the air conditioner in recognizing the user's presence status, enhances the intelligence of air conditioner control, and balances user comfort with energy-saving efficiency.

[0198] According to an embodiment of the present invention, a storage medium corresponding to an air conditioner control method is also provided, the storage medium including a stored program, wherein the program controls the device where the storage medium is located to execute the air conditioner control method described above when it is executed.

[0199] Since the processing and functions implemented by the storage medium in this embodiment are basically the same as the embodiments, principles and examples of the aforementioned methods, any details not covered in this embodiment can be found in the relevant descriptions in the aforementioned embodiments, and will not be repeated here.

[0200] The technical solution of this invention involves an air conditioner scanning the broadcast packets and signal strength of Bluetooth devices, extracting broadcast packet features to construct a device fingerprint, associating a random MAC address with the device through fingerprint matching, filtering the signal strength, and combining the association results to determine the device's presence status. Finally, based on this status, an air conditioner control strategy is generated and executed. This improves the accuracy of the air conditioner in recognizing the user's presence status, enhances the intelligence of air conditioner control, and balances user comfort with energy-saving efficiency.

[0201] According to an embodiment of the present invention, a computer program product corresponding to the control method for an air conditioner is also provided. The computer program product includes a computer program that, when processed and executed, implements the steps of the control method for the air conditioner described above.

[0202] Since the processing and functions implemented by the computer program product in this embodiment are basically corresponding to the embodiments, principles and examples of the aforementioned methods, any details not covered in the description of this embodiment can be found in the relevant descriptions in the aforementioned embodiments, and will not be repeated here.

[0203] The technical solution of this invention involves an air conditioner scanning the broadcast packets and signal strength of Bluetooth devices, extracting broadcast packet features to construct a device fingerprint, associating a random MAC address with the device through fingerprint matching, filtering the signal strength, and combining the association results to determine the device's presence status. Finally, based on this status, an air conditioner control strategy is generated and executed. This improves the accuracy of the air conditioner in recognizing the user's presence status, enhances the intelligence of air conditioner control, and balances user comfort with energy-saving efficiency.

[0204] In summary, it is readily understood by those skilled in the art that, without conflict, the aforementioned advantageous methods can be freely combined and superimposed.

[0205] The above description is merely an embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of the claims of the present invention.

Claims

1. A method for controlling an air conditioner, characterized in that, include: Obtain the broadcast packets from the Bluetooth device and the signal strength of the broadcast packets; Extract multiple types of feature information from the broadcast packets to construct a device fingerprint; Similarity matching is performed based on the device fingerprint to associate the random MAC address in the broadcast packet with the same Bluetooth device, and the association result is obtained; The signal strength is filtered, and the presence status of the Bluetooth device is determined based on the filtered signal strength and the correlation result. The air conditioning control strategy is generated and executed based on the stated state.

2. The control method of the air conditioner according to claim 1, characterized by, Similarity matching is performed based on the device fingerprint to associate the random MAC address in the broadcast packet with the same Bluetooth device, and the association result is obtained, including: Extract the historical device fingerprints corresponding to the broadcast packets obtained from historical scans, and construct a historical fingerprint database; the historical fingerprint database stores the correspondence between historical random MAC addresses and historical device fingerprints; Calculate the multi-dimensional similarity between the device fingerprint corresponding to the current broadcast packet and each historical device fingerprint in the historical fingerprint database; the multi-dimensional similarity includes numerical vector similarity and set similarity; The multi-dimensional similarity is weighted and fused to obtain the total similarity score; If the total similarity score reaches the preset association threshold, the random MAC address in the current broadcast packet is associated with the Bluetooth device corresponding to the historical device fingerprint, and an association result is generated. The association result includes the association relationship between the current random MAC address and the corresponding Bluetooth device, the association confidence, and the unique identifier corresponding to the Bluetooth device.

3. The control method of the air conditioner according to claim 1, characterized by, Filtering the signal strength includes: The signal strength is subjected to a first-stage filter to eliminate short-term random fluctuations in the signal strength; Within a preset time window, the signal strength after the first-stage filtering is subjected to a second-stage filtering to extract stable characteristics of the signal strength and characterize the signal propagation state of the device.

4. The control method of the air conditioner according to any one of claims 1 to 3, characterized by, Determining the presence status of the Bluetooth device based on the filtered signal strength and the correlation result includes: Based on the association results, determine the unique identifier of the Bluetooth device corresponding to the current broadcast packet; Based on the unique identifier, target scan records within a preset time window are filtered, and the number of scans by the Bluetooth device is counted to obtain the current frequency of occurrence. Determine whether the signal strength after filtering is within a preset effective strength range, and whether the current occurrence frequency reaches a preset occurrence threshold; If the signal strength is within the preset effective strength range and the current occurrence frequency reaches the preset occurrence threshold, then the initial state of the Bluetooth device is determined to be a candidate for presence. If the signal strength is not within the preset effective strength range and the current occurrence frequency does not reach the preset occurrence threshold, then the initial state of the Bluetooth device is determined to be an off-field candidate. The initial state is corrected based on a preset machine learning model to obtain the existence state of the Bluetooth device.

5. The control method of the air conditioner according to claim 4, characterized by, The initial state is corrected based on a preset machine learning model to obtain the existence state of the Bluetooth device, including: Using the initial state as an input constraint, assign basic weights to the presence probability of the candidates in the field and basic weights to the departure probability of the candidates out of the field. Input preset features into the preset machine learning model and calculate the presence probability value; If the presence probability value is greater than or equal to the first preset threshold, then the presence status of the Bluetooth device is determined to be present. If the presence probability value is less than or equal to the second preset threshold, then the presence state of the Bluetooth device is determined to be absent. If the probability value of presence is greater than the second preset threshold and less than the first preset threshold, then the previous presence state is maintained.

6. A control device of an air conditioner, characterized by comprising: include: The acquisition unit is configured to acquire broadcast packets from a Bluetooth device and the signal strength of the broadcast packets. A fingerprint construction unit is configured to extract multiple types of feature information from the broadcast packet to construct a device fingerprint; The device association unit is configured to perform similarity matching based on the device fingerprint to associate the random MAC address in the broadcast packet with the same Bluetooth device and obtain an association result; The state determination unit is configured to filter the signal strength and determine the presence state of the Bluetooth device based on the filtered signal strength and the correlation result. The control execution unit is configured to generate and execute a control strategy for the air conditioner based on the stated state.

7. A system, characterized by include: The air conditioner control device as described in claim 6.

8. A storage medium, characterized by The storage medium includes a stored program, wherein, when the program is executed, the device containing the storage medium is controlled to perform the air conditioning control method according to any one of claims 1 to 5.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the air conditioning control method according to any one of claims 1 to 5.

Images