Intelligent home ecological multi-device linkage control system based on big data

Abstract

This invention belongs to the field of smart home control technology, specifically a multi-device linkage control system for a smart home ecosystem based on big data. It includes a linkage control platform that communicates with multiple smart home devices. The platform includes a data acquisition module that collects operational status data, environmental perception data, and user operation behavior data from the multiple smart home devices. This invention uses a scene recognition module to mine frequent operation sequences from historical user operation behavior data using a sliding time window. Operation sequences that meet support and confidence thresholds are written into a scene pattern library, while scene patterns with confidence levels below the elimination threshold are removed. This achieves an automatic update and elimination mechanism for scene patterns based on real-time behavior data, overcoming the technical shortcomings of existing technologies where linkage rules rely entirely on manual presets and cannot be dynamically maintained as lifestyle habits evolve.

Description

Technical Field

[0001] This invention relates to the field of smart home control technology, and in particular to a multi-device linkage control system for a smart home ecosystem based on big data. Background Technology

[0002] With the deep penetration of IoT communication technology and smart terminal devices into home spaces, lighting systems, environmental control systems, security monitoring systems, and various household appliances constitute an increasingly complex home IoT ecosystem. Users' demands for home automation have evolved from independent remote control of single devices to collaborative linkage between multiple devices. In existing technologies, to improve user experience, a common solution is to preset scene linkage rules through mobile applications or smart voice assistants to achieve fixed logic such as "automatically turning on living room lighting and turning off security mode when the door lock is opened." However, this preset solution based on static thresholds cannot adapt to the dynamic living habits of family members.

[0003] Relevant technical documents, such as patent application CN120178696A, propose a smart home multi-device control system and method based on the Internet of Things. This method acquires multi-dimensional data and calculates multi-dimensional feature vectors and multi-dimensional matching indices, combines them with historical behavior continuity to form a comprehensive correlation score, executes user behaviors with scores greater than a threshold when device status changes, and generates execution control sequences according to priority when there are redundant control conflicts.

[0004] Another patent application, CN120315310A, discloses a method and system for big data analysis of living habits and behaviors in smart homes. It collects user behavior data, environmental data, and device status data, generates behavioral preference tags through multi-dimensional data processing, and outputs living behavior patterns from a behavioral analysis model. Combined with real-time data, it generates control strategies for smart home devices. While the aforementioned existing technical solutions achieve automation based on user behavior to a certain extent, their technical shortcomings are objectively manifested in the following ways: First, the generation or identification of linkage rules in existing solutions relies on statically set scoring thresholds. The system lacks a closed-loop mechanism for dynamically capturing and automatically updating the evolution of user habits over time. When users' schedules shift seasonally or their lifestyles change, the system cannot automatically adjust the linkage trigger conditions, resulting in a significant decrease in the hit rate of linkage rules over time. Secondly, existing solutions only use a single priority factor for sorting when handling concurrent conflicts of multiple commands, without classifying the underlying physical scenarios of command conflicts. For example, they do not distinguish between command conflicts caused by inconsistent control parameters of the same device and resource mutual exclusion conflicts caused by multiple device commands vying for the same power supply circuit or communication bandwidth. Different types of conflicts have significantly different physical consequences, and the single sorting strategy results in coarse arbitration granularity, making it impossible to achieve a precise trade-off between safety and comfort. Thirdly, the linkage control process of existing solutions terminates after the command is issued. Data on the actual execution results of the device and the user's on-site correction or cancellation of the linkage behavior are not collected and incorporated into the system training. As a result, the system cannot learn from user interaction feedback and correct its own linkage behavior. Inappropriate linkage commands may be repeatedly triggered in subsequent operations. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a big data-based smart home ecosystem multi-device linkage control system, which overcomes the deficiencies of existing technologies and effectively solves the problems mentioned in the background.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: A big data-based smart home ecosystem multi-device linkage control system includes a linkage control platform that communicates with multiple smart home devices. The linkage control platform includes: The data acquisition module collects operating status data, environmental perception data, and user operation behavior data from multiple smart home devices. The scene recognition module performs frequent operation sequence mining based on historical user operation behavior data according to a sliding time window, defines the mined frequent operation sequences as scene patterns and builds a scene pattern library. The scene recognition module also matches the real-time collected data with the scene pattern library to determine the current home scene. The linkage decision module distinguishes between device instruction conflicts and resource mutual exclusion conflicts in the conflict detection of instruction execution. Device instruction conflicts are conflicts that point to the same device but have inconsistent control parameters. Resource mutual exclusion conflicts are conflicts that multiple instructions contend for the same restricted resource. The module calculates an arbitration score based on static priority and scenario matching confidence, and generates an execution instruction sequence based on the arbitration score. The execution control module sends the execution instruction sequence to the corresponding smart home device, collects the execution results of the smart home device, and collects the user's takeover behavior in response to the execution results; The feedback correction module categorizes user takeover actions into three types: confirmation, modification, and cancellation. Based on the cancellation label, it lowers the confidence level of the corresponding scenario mode and removes the scenario mode from the scenario mode library when it falls below the elimination threshold. Based on the confirmation label, it increases the instruction weight. Based on the modification label, it updates the default value of the instruction parameters when the cumulative number of modifications of the same type reaches the update threshold.

[0007] Preferably, the scene mode includes triggering conditions, a set of linkage instructions, and confidence level. The triggering conditions include time window conditions, environmental conditions, and device status conditions. The scene recognition module counts operation sequences in historical user operation behavior data according to a sliding time window, calculates the support and confidence of the operation sequences, writes operation sequences with support not lower than the support threshold and confidence not lower than the confidence threshold as scene patterns into the scene pattern library, and removes scene patterns with confidence lower than the elimination threshold from the scene pattern library.

[0008] Preferably, the conflict detection performed by the linkage decision module includes conflict types such as device command conflicts that point to the same smart home device but have inconsistent control parameters, and resource mutual exclusion conflicts where multiple commands contend for the same restricted resource. The linkage decision-making module calculates arbitration scores for conflicting instructions based on static priority and scenario matching confidence. The static priority is set in a hierarchy where safety instructions are higher than health instructions, health instructions are higher than comfort instructions, and comfort instructions are higher than energy-saving instructions. Instructions with low arbitration scores are suspended or discarded, and the remaining instructions are sorted according to their arbitration scores to form the execution instruction sequence.

[0009] Preferably, the data acquisition module is deployed on an edge gateway, and the data acquisition module preprocesses the acquired data, including data deduplication, missing value imputation, time alignment and normalization. The data acquisition module performs hierarchical reporting according to data type, with event-type data being reported in real time and status-type data being reported in batches according to a preset period.

[0010] Preferably, the execution control module listens for the status feedback returned by the corresponding smart home device after the instruction is issued. If the feedback times out, the instruction is resent. If the number of resentments reaches the resentment limit, the preset degradation strategy is executed. The execution control module stores the instruction execution log in a structured database and marks user takeover actions. The types of marking include confirmation, modification, and revocation.

[0011] Preferably, the linkage control platform further comprises a protocol adaptation unit, which converts data of smart home devices with different communication protocols into a unified device model. The unified device model describes devices with device types, capability attributes and state parameters, and newly accessed smart home devices register according to the device type template to complete access to the linkage control platform.

[0012] Preferably, the linkage control platform stores a parameter configuration table, which records the correspondence between house type, family role and linkage parameters, and the linkage parameters include a confidence threshold, a degradation strategy and an initial instruction weight value; The edge gateway locally caches the scenario pattern library and the generation rule of the execution instruction sequence. When the network is interrupted, the edge gateway executes linkage control according to the local cache, and synchronizes the execution logs during the interruption after the network is restored.

[0013] Preferably, the feedback correction module reduces the confidence of the corresponding scenario pattern and the weight of the corresponding instruction according to the user takeover behavior marked as revocation, and removes the scenario pattern whose confidence falls below the elimination threshold from the scenario pattern library; The feedback correction module increases the weight of the corresponding instruction according to the user takeover behavior marked as confirmation, and the feedback correction module records the parameters adjusted by the user according to the user takeover behavior marked as modification. When the number of repeated occurrences of the parameters adjusted by the user in similar scenarios reaches the update threshold, the default value of the instruction parameter in the corresponding scenario pattern is updated.

[0014] A big data-based multi-device linkage control method for smart home ecology, the method comprising: Collecting operating state data, environment sensing data and user operation behavior data of a plurality of smart home devices; Performing frequent operation sequence mining according to a sliding time window based on historical user operation behavior data, constructing a scenario pattern library, and matching the real-time collected data with the scenario pattern library to determine the current home scenario; Generating a multi-device linkage instruction set according to the scenario pattern corresponding to the current home scenario, performing conflict detection on the instructions in the multi-device linkage instruction set, arbitrating the instructions according to the conflict detection result, and outputting an execution instruction sequence; Issuing the execution instruction sequence to the corresponding smart home devices, collecting the execution results of the smart home devices, and collecting the user's takeover behavior for the execution results; Correcting the confidence of the scenario patterns in the scenario pattern library and the weight of each instruction in the multi-device linkage instruction set according to the execution results and the user takeover behavior.

[0015] A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the steps of a method for multi-device linkage control of a smart home ecosystem based on big data are disclosed.

[0016] The beneficial effects of this invention are as follows: The present invention provides a smart home ecosystem multi-device linkage control system based on big data. Through a scene recognition module, frequent operation sequence mining is performed on historical user operation behavior data by sliding time windows. Operation sequences that meet the support threshold and confidence threshold are written into the scene pattern library, and scene patterns with confidence below the elimination threshold are removed. This realizes an automatic update and elimination mechanism for scene patterns based on real-time behavior data. It overcomes the technical defects of existing technologies where linkage rules rely entirely on manual presets and cannot be dynamically maintained with the evolution of living habits. This allows the linkage triggering conditions of the system to evolve synchronously with changes in user living habits, significantly reducing the frequency of manual intervention in rule maintenance. This invention presents a big data-based smart home ecosystem multi-device linkage control system. Through a linkage decision module, conflicts in the multi-device linkage command set are categorized into device command conflicts and resource mutual exclusion conflicts. An arbitration score is calculated based on static priority and scene matching confidence, and then an execution command sequence is generated by sorting the arbitration scores. This ensures that the conflict arbitration dimensions encompass the differences in the physical type of the conflict and the urgency of the current scene. Compared to the coarse-grained arbitration method of single priority sorting in existing technologies, this significantly improves the accuracy of command arbitration and scene adaptability. Simultaneously, the execution control module collects execution results and labels user takeover behavior. The feedback correction module performs differentiated increases and decreases in scene confidence and command weights based on three types of labeling results: confirmation, modification, and revocation. This constructs a complete closed loop from execution feedback to rule correction, ensuring that system linkage behavior continuously converges to the user's true preferences and that the trigger rate of inappropriate linkages gradually decreases as the system's operating cycle extends. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall process of a smart home ecosystem multi-device linkage control system based on big data proposed in this invention. Detailed Implementation

[0018] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0019] Reference Figure 1 A smart home ecosystem multi-device linkage control system based on big data. I. System Overall Architecture and Operating Principles The system described in this invention comprises a device layer, an edge layer, and a platform layer in its physical deployment. The device layer includes lighting equipment, environmental sensing equipment, security alarm equipment, home appliance actuators, and user interaction terminals. The edge layer deploys edge gateways, which carry data acquisition modules and establish data links with various smart home devices via wireless personal area networks (WLANs), wireless local area networks (WLANs), or infrared communication channels. The platform layer is deployed on a home private cloud or public cloud server, carrying scene recognition modules, linkage decision-making modules, execution control modules, feedback correction modules, and storage and parameter configuration units. During system operation, the edge layer data acquisition module obtains raw data from the device layer according to a preset sampling period, performs cleaning and alignment, and then reports it to the platform layer in a hierarchical manner. The platform layer scene recognition module mines a scene pattern library based on historical behavior data and matches it with real-time data streams to determine the current home scene. The linkage decision module expands the linkage instruction set based on the current home scene, and generates an execution instruction sequence after conflict detection and arbitration. The execution control module sends the instruction sequence to the target device through the edge gateway and continuously listens for device feedback to confirm the execution status. The feedback correction module takes the execution log and user takeover behavior as input, dynamically adjusts the confidence level in the scene pattern library and the instruction weight in the instruction set, thereby forming a complete closed-loop evolution link.

[0020] II. Detailed Workflow of the Data Acquisition Module The data acquisition module is deployed on the edge gateway. This module performs preprocessing on the received raw data, specifically including data deduplication, missing value imputation, time alignment, and normalization. The data deduplication step removes duplicate data records caused by network retransmissions based on device identifiers and timestamps. The missing value imputation step uses a sliding linear interpolation method for environmental sensor data. For data segments with consecutive missing time windows not exceeding 120 seconds, linear interpolation is performed based on the valid sample values ​​before and after the missing values ​​to fill in the missing values. The time alignment step uses the edge gateway's system clock as a unified reference, uniformly correcting the time stamp of data reported by each sensor to the gateway's local time stamp, eliminating timing deviations caused by asynchronous reporting from different devices. The normalization step performs maximum-minimum normalization transformation for data with different physical dimensions.

[0021] After preprocessing, the data acquisition module implements a tiered reporting strategy based on the urgency and volume of the data type: event-related data, including door lock opening, human movement triggering, and smoke alarm triggering, are immediately reported to the platform layer via a message queue after preprocessing; status-related data, including ambient temperature and humidity values, device power consumption, and device on / off status, are batch-packaged and reported according to a preset period, which is a fixed duration between 30 and 120 seconds. This tiered reporting mechanism ensures that the transmission delay of high-priority event data is limited to the underlying network transmission latency, while low-priority status data is compressed to reduce uplink bandwidth usage. The data streams output by the data acquisition module are used to trigger real-time matching calculations by the scene recognition module and historical behavior accumulation analysis by the feedback correction module.

[0022] III. Scene Recognition Module's Mining Rules and Matching Mechanism The scene recognition module is deployed at the platform layer, and its core functions include scene pattern mining and construction, and real-time scene matching. In the scene pattern mining and construction stage, this module mines frequent operation sequences based on historical user operation behavior data and follows a sliding time window.

[0023] Set the time window width to W, and the historical behavior data set to... The scene recognition module from Extract the set T of operation sequences consisting of a triplet of device identifier, action type, and relative time interval. For each candidate operation sequence... The scene recognition module calculates the support of the candidate operation sequence. and confidence level .

[0024] The formula for calculating support is:

[0025] The formula for calculating confidence level is:

[0026] If the calculated support is not lower than a preset support threshold and the confidence is not lower than a preset confidence threshold, the scene recognition module defines the frequent operation sequence as a scene pattern and writes the scene pattern into the scene pattern library. The scene pattern includes triggering conditions, a set of linked instructions, and confidence. The triggering conditions are specifically divided into time window conditions, environmental conditions, and device status conditions; among them, the time window condition matches the time value falling within the range from the start time to the end time of the frequent operation sequence occurrence period, the environmental condition matches the environmental perception data value falling within the environmental perception feature range when the frequent operation sequence occurs, and the device status condition matches the device operating state being consistent with the device state before the frequent operation sequence occurred.

[0027] At the same time, the scene recognition module performs regular elimination inspection on all existing scene modes in the scene mode library. If the current confidence of a certain scene mode is lower than the preset elimination threshold, the scene mode is removed from the scene mode library. The entry threshold is higher than the retention threshold, forming hysteresis to prevent frequent entry and exit of modes from the library.

[0028] In the real-time scene matching link, the scene recognition module receives the real-time data stream uploaded by the data acquisition module, and compares each data item in the real-time data stream with the trigger condition of each scene mode in the scene mode library one by one. For the matched scene mode, the scene matching confidence is calculated , and the confidence is equal to the ratio of the number of data items that match and hit the scene trigger condition to the total number of scene trigger conditions. Among all the fully matched scene modes, the scene corresponding to the scene mode with the highest scene matching confidence that is not lower than the confidence threshold is selected and output as the current home scene. If the scene matching confidence of all scene modes is lower than the confidence threshold, the system determines that there is no home scene that needs to be responded to currently, and continues to monitor subsequent real-time data.

[0029] IV. Instruction set generation and conflict arbitration logic of the linkage decision module The linkage decision module receives the current home scene output by the scene recognition module and the corresponding scene matching confidence. This module extracts the linkage instruction set in the scene mode corresponding to the current home scene from the scene mode library, and expands the linkage instruction set into a specific device control instruction list.

[0030] Subsequently, the linkage decision module performs conflict detection, which covers two pre-defined conflict types. The first conflict type is device instruction conflict. When more than two instructions point to the same smart home device and the set values of the control parameters of the device by these instructions are inconsistent, it is determined that a device instruction conflict occurs. The second conflict type is resource mutual exclusion conflict. When multiple instructions require simultaneous occupation of the same limited physical resource, it is determined that a resource mutual exclusion conflict occurs; the limited physical resources include the total power capacity of a single power supply loop, the available bandwidth of a single wireless local area network, and the exclusive right to use a single audio and video output channel.

[0031] For the detected conflicting instruction group, the linkage decision module calculates the arbitration score of each conflicting instruction. The calculation formula of the arbitration score is:

[0032] wherein, represents the arbitration score of the i-th instruction; represents the static priority score corresponding to the instruction; represents the scene matching confidence of the current home scene; and These are the weighted coefficients for the static priority score and the scenario matching confidence level, respectively. The static priority is set in a hierarchy where safety-related instructions are higher than health-related instructions, health-related instructions are higher than comfort-related instructions, and comfort-related instructions are higher than energy-saving instructions.

[0033] After calculating the arbitration score of each instruction in the conflicting instruction group, the linkage decision module determines that the instruction with the lowest arbitration score should be suspended or discarded. The remaining instructions are then reordered according to their arbitration scores from highest to lowest, forming the final output sequence of execution instructions. When the conflicting instruction group contains only two instructions, one of which is a safety instruction and the other is an energy-saving instruction, the arbitration score of the safety instruction will be significantly higher than that of the energy-saving instruction because the static priority score of the safety instruction is much higher than that of the energy-saving instruction. Even if the scenario matching confidence of the energy-saving instruction is high, the arbitration score ranking result still prioritizes the execution of the safety instruction, thereby ensuring the system's compliance in the safety priority dimension.

[0034] V. Command Issuance and Execution Receipt Collection of the Execution Control Module The execution control module receives the execution instruction sequence output by the linkage decision module. This module converts each instruction in the execution instruction sequence into a communication protocol message recognizable by the target device via the protocol adaptation unit of the edge gateway, according to the sorting order, and sends it to the smart home device.

[0035] After each instruction is issued, the execution control module starts a receipt monitoring timer. If a status receipt is received from the target device within the set receipt timeout threshold, and the receipt content shows that the instruction was executed successfully, the instruction is marked as completed. If no status receipt is received within the receipt timeout threshold, the execution control module starts a retransmission mechanism to retransmit the instruction, restarting the receipt monitoring timer each time it is retransmitted. When the number of retransmissions reaches the preset retransmission limit and no valid receipt arrives, the module performs a degradation operation according to a preset degradation strategy. The degradation strategy includes migrating the instruction to an alternative device in the same area for execution, or marking the instruction as failed and skipping other linked instructions that depend on the result of the instruction's execution.

[0036] The execution control module stores the execution log of each instruction in a structured database, categorized by device identifier, instruction content, issuance time, receipt time, execution result, and execution status. Simultaneously, the module collects user actions on the linkage results within a preset observation window after the linkage execution, including adjustments, shutdowns, or rollbacks performed by the user via mobile application, voice speaker, or physical panel. This module labels the user's intervention based on the specific operation type, categorizing these labels into three types: confirmation, modification, and cancellation, based on the differentiated effects of the operation consequences. When the user does not intervene in the observation window, it is labeled as confirmation; when the user manually adjusts the target parameters of the linkage device in the observation window, it is labeled as modification; and when the user actively shuts down or rolls back the execution result of the linkage device in the observation window, it is labeled as cancellation.

[0037] VI. Confidence and Weight Adjustment Mechanism of Feedback Correction Module The feedback correction module receives structured execution logs and user takeover annotation data output by the execution control module. This module uses scenario modes as the basic statistical unit and attributes each event in the execution log to the scenario mode that triggered the event.

[0038] For linked events marked as confirmed, the feedback correction module increases the weight of the instruction in the corresponding scenario mode. The increase in weight is a preset fixed increment or a cumulative index based on the confirmation frequency, so that the instruction can obtain a higher arbitration score in the subsequent similar scenario instruction arbitration process.

[0039] For linked events marked as canceled, the feedback correction module reduces the weight of the instruction in the corresponding scene mode and also reduces the confidence of the scene mode itself. When the confidence falls below the preset elimination threshold, the scene recognition module is triggered to remove the scene mode from the scene mode library to prevent the linked rule from being triggered repeatedly in subsequent periods.

[0040] For linked events marked as modified, the feedback correction module records the user's modified target parameters and writes them into the candidate parameter pool. When the cumulative occurrence of the same instruction modification operation in the candidate parameter pool reaches a preset update threshold under the same scenario mode, the feedback correction module updates the default parameters of that instruction in the current scenario mode to the user's modified target parameters. When the scenario mode is triggered again subsequently, the linkage decision module will generate a linkage instruction set with the updated parameter values.

[0041] The aforementioned feedback mechanism based on confirmation, modification, and cancellation ensures that the user's subjective attitude towards the automatic linkage results can be quantitatively embedded into the core decision-making logic of the system, making the system's linkage behavior increasingly closer to the user's actual operating preferences and suppressing the recurrence of inappropriate linkage rules.

[0042] VII. Protocol Adaptation Unit and Unified Device Model The linkage control platform and edge gateway deploy protocol adaptation units. These units address the heterogeneous access issues of smart home devices from different brands and using different communication protocols.

[0043] This unit maintains a unified device model, which describes each device in three dimensions: device type, capability attributes, and status parameters. The device type identifies the physical function category to which the device belongs, the capability attributes list the types of instructions and actions that the device can execute, and the status parameters list the data fields that the device can report during operation.

[0044] For newly connected smart home devices, the protocol adaptation unit matches a preset device type template based on the reported physical layer device information. This template pre-declares the set of capability attributes and status parameters for that type of device, as well as the parsing rules for physical communication protocol messages corresponding to each attribute parameter. The protocol adaptation unit then completes the registration and configuration of the new device based on the device type template, enabling the newly connected device to be recognized and invoked by other modules of the linkage control platform as a unified device model instance.

[0045] For processes requiring platform-layer commands to be sent to a target device, the protocol adaptation unit converts command operations based on the unified device model into the physical message format of the communication protocol of the target device. For status reporting data received from the target device, the protocol adaptation unit parses the physical messages and maps them to status parameter values ​​of the unified device model, thereby providing a clean and standardized data source for the data acquisition module. The existence of the protocol adaptation unit eliminates the burden of developing specific device drivers for each device, thus reducing the engineering cost of connecting heterogeneous devices to the system.

[0046] 8. Local caching and network outage self-governance mechanism of edge gateway The edge gateway has a local cache containing the scene pattern library and the generation rules for the execution instruction sequence. The generation rules specifically include scene matching logic parameters and instruction arbitration configuration parameters.

[0047] When the wide area network link between the linkage control platform and the edge gateway is interrupted, the edge gateway completes autonomous operation of device linkage control through locally cached data and rules. In autonomous operation mode, the edge gateway's data acquisition module collects and processes field device data in real time, and then performs scene matching according to the locally cached scene pattern library. If the matching is successful, the linkage decision module inside the edge gateway generates a sequence of instructions according to the locally cached generation rules, and sends it directly to the target device through the edge gateway's communication interface without the need for platform layer intervention.

[0048] During network outages, the edge gateway writes all autonomous operation command issuance records, device acknowledgment records, and user takeover operation records to a local transaction log file in timestamp order. When the WAN link is restored, the edge gateway uploads the local transaction log file to the platform layer. The platform layer's execution control module and feedback correction module then perform database entry and feedback correction processing on the execution logs from the outage period. This ensures that coordinated execution behaviors and user intervention behaviors that occurred during the network outage are also included in the system's big data statistical model, guaranteeing the continuity and integrity of the system's closed-loop evolution mechanism over long-term operation.

[0049] IX. Example 1: Whole-house linkage control in a multi-family home scenario This embodiment is set in a duplex family home, which is equipped with environmental sensors, smart lighting, smart air conditioning, smart curtains, security cameras, and smart door locks, totaling 28 connected devices. The edge gateway is a home central gateway with dual-band wireless LAN communication capabilities, and the platform layer is deployed on the home's own private cloud server.

[0050] During the first operating cycle after system initialization, the scene recognition module performs frequent operation sequence mining on the user's historical operation behavior data with a sliding time window width of 30 days. During data collection, a support threshold of 5%, a confidence threshold of 0.70, and a rejection threshold of 0.40 were set. On the 18th day, the system identified a high-frequency operation sequence of "returning home in the evening," characterized by the user simultaneously turning on the living room lights and air conditioner within 120 seconds after unlocking the door between 5:30 PM and 6:30 PM from Monday to Friday. This frequent operation sequence was written into the scene pattern library and assigned an initial confidence level.

[0051] On the evening of the 45th weekday of system operation, a door lock unlocking event was triggered. The scene recognition module matched the scene mode based on real-time data, and the scene matching confidence score reached 0.85, determining the current home scene as "returning home." The linkage decision module then deployed the linkage command set for this scene mode, which included turning on the main living room light, turning on the air conditioner's cooling mode, and closing the curtains. During conflict detection, the linkage decision module found that the sum of the power requested by the air conditioner cooling command and the water heater heating command exceeded the allowable power of a single branch in the main power supply circuit of the house, thus determining that this set of commands constituted a resource mutual exclusion conflict. The linkage decision module then applied the formula... Calculate the arbitration score for each command. Cooling commands belong to the comfort category. Their static priority score is lower than that of safety commands but higher than that of energy-saving commands. After weighting with a scenario matching confidence of 0.85, the arbitration score of the cooling command is higher than that of the heating command. Therefore, the heating command is suspended and the cooling command takes precedence.

[0052] After the execution control module issues a cooling instruction, it receives a device receipt within 2 seconds to confirm successful execution, then releases the restricted power resources, and then issues a heating instruction, thereby successfully completing the system linkage. This embodiment shows that the present system can realize safe and controllable sequential execution of multiple instructions according to physical constraint conditions in a resource-scarce scenario.

[0053] X. Embodiment 2: Multi-device cooperative linkage in a comprehensive office space This embodiment is set in an open office space, which is equipped with smart lighting devices, a smart fresh air system, projection display devices, and a smart conference control system, with a total of 35 access devices. The platform layer is deployed on the enterprise intranet server, and the edge gateway is deployed in the floor weak current well, aiming to optimize the energy consumption management and conference experience of the office space.

[0054] In the parameter configuration, the sliding time window is set to 14 days, the support threshold is set to 3%, and the confidence threshold is set to 0.65. In the initial stage of system operation, the scene recognition module excavates two high-frequency scenes of "daily office" and "conference mode". Under the "daily office" scene mode, the linkage instruction set includes adjusting the lighting brightness of the office area according to the value of the illumination sensor, and automatically turning on the fresh air system according to the indoor carbon dioxide concentration. At noon on a working day, the value of the illumination sensor suddenly increases, the scene recognition module matches the current environmental conditions to the "daily office" scene, and generates a brightness adjustment instruction and a fresh air turn-on instruction. The linkage decision module checks whether the two instructions conflict. Since the brightness adjustment instruction and the fresh air turn-on instruction do not point to the same device and do not contend for the same power supply branch, no conflict occurs, and the two instructions are issued in sequence according to the scene matching confidence.

[0055] In the subsequent conference period, the smart conference system triggers the "conference mode" linkage instruction set, which includes turning on the projection device, closing the curtains, reducing the lighting brightness of the conference area, and increasing the air volume of the fresh air system. At this time, the fresh air volume increase instruction of the conference mode and the fresh air turn-on instruction of the daily office mode point to the same fresh air device, and the control parameters are inconsistent (one is to maintain the air volume, and the other is to increase the air volume). The linkage decision module determines the conflict as a device instruction conflict. The linkage decision module calculates the arbitration score, and the scene matching confidence of the conference mode is significantly higher than that of the daily office mode. Therefore, the conference mode instruction has a higher score, the daily office instruction is discarded, and finally the system executes the instruction sequence according to the conference mode scene, so that the office space smoothly enters the conference preparation state.

[0056] The above Embodiment 2 verifies the refined arbitration capability of the present system under the scenarios of multi-scene superposition and instruction conflict, avoids disordered overwriting of instructions in different modes, and ensures the priority execution of high-level scene requirements.

[0057] XI. Embodiment 3: Safety-first linkage in a barrier-free apartment This embodiment is set in an apartment management project, which includes 12 independent apartments. Each apartment is connected to 10 devices, including smart mattresses, floor lights, emergency call buttons, corridor lights, and human motion sensors. The platform layer is deployed on the apartment management center server, and the edge gateway is deployed in the floor's low-voltage room. The system configuration sets the family role attribute to elderly resident, and the initial value of the security command weight is increased by one level in the system parameter configuration table.

[0058] After the system operates within a 14-day sliding time window, it identifies a "nighttime bed-leaving" scenario mode based on frequent operation sequences. The trigger condition for this mode is set to a time window from 11 PM to 5 AM the following day, plus a bed-leaving signal output by the smart mattress pressure sensor. The associated command set for this scenario mode includes the gradual illumination of the floor lights, the constant illumination of the corridor lights, and the call button being on standby. At 3 AM one day, a certain apartment triggered a bed-leaving signal. During this period, the edge gateway was unable to communicate with the platform layer due to an interruption in external communication lines. The system automatically switched to local autonomous mode. The edge gateway used its locally cached scenario mode library to match the real-time data. The scenario matching was successful, and the command sequence was sent to the target device via the wireless personal area network according to the generation rules. The floor lights and corridor lights illuminated within one second, and the call button was put on standby.

[0059] Meanwhile, the system dynamically tracks the behavior of residents in the apartment when they leave bed. If, for three consecutive nights, the system detects that a user manually turns off the corridor light after the system has activated it, the data acquisition module marks this as a cancellation operation. The feedback correction module then reduces the weight of the corridor light activation command and correspondingly lowers the confidence level of the apartment resident in the current scenario. After these adjustments, when the apartment triggers the "leave bed" scenario, the linkage decision-making module generates a simplified command set that only includes turning on the floor lights, thus accurately matching the resident's actual lighting needs.

[0060] After a 70-minute network outage, the network was restored. The edge gateway synchronously uploaded 16 execution logs from the outage period to the platform layer. The feedback correction module performed closed-loop corrections on these 16 records, completing this round of rule evolution. This embodiment verifies the system's ability to finely adapt to specific user habits in complex security scenarios and its fault-tolerant survivability during network outages.

[0061] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A multi-device linkage control system for a smart home ecosystem based on big data, comprising a linkage control platform that communicates with multiple smart home devices, characterized in that, The linkage control platform includes: The data acquisition module collects operating status data, environmental perception data, and user operation behavior data from multiple smart home devices. The scene recognition module performs frequent operation sequence mining based on historical user operation behavior data according to a sliding time window, defines the mined frequent operation sequences as scene patterns and builds a scene pattern library, and also matches the real-time collected data with the scene pattern library to determine the current home scene; The linkage decision module distinguishes between device instruction conflicts and resource mutual exclusion conflicts in the conflict detection of instruction execution. Device instruction conflicts are conflicts that point to the same device but have inconsistent control parameters. Resource mutual exclusion conflicts are conflicts that multiple instructions contend for the same restricted resource. The module calculates an arbitration score based on static priority and scenario matching confidence, and generates an execution instruction sequence based on the arbitration score. The execution control module sends the execution instruction sequence to the corresponding smart home devices, collects the execution results of the smart home devices, and collects the user's takeover behavior in response to the execution results; The feedback correction module categorizes user takeover actions into three types: confirmation, modification, and cancellation. Based on the cancellation label, it lowers the confidence level of the corresponding scenario mode and removes the scenario mode from the scenario mode library when it falls below the elimination threshold. Based on the confirmation label, it increases the instruction weight. Based on the modification label, it updates the default value of the instruction parameters when the cumulative number of modifications of the same type reaches the update threshold.

2. The smart home ecosystem multi-device linkage control system based on big data according to claim 1, characterized in that, The scene mode includes triggering conditions, linkage instruction set and confidence level. The triggering conditions include time window conditions, environmental conditions and device status conditions. The scene recognition module counts operation sequences in historical user operation behavior data according to a sliding time window, calculates the support and confidence of the operation sequences, writes operation sequences with support not lower than the support threshold and confidence not lower than the confidence threshold as scene patterns into the scene pattern library, and removes scene patterns with confidence lower than the elimination threshold from the scene pattern library. The elimination threshold is lower than the confidence threshold.

3. The smart home ecosystem multi-device linkage control system based on big data according to claim 1, characterized in that, The conflict detection performed by the linkage decision module includes conflicting device commands that point to the same smart home device but have inconsistent control parameters, as well as resource mutual exclusion conflicts where multiple commands contend for the same restricted resource. The linkage decision-making module calculates arbitration scores for conflicting instructions based on static priority and scenario matching confidence. The static priority is set in a hierarchy where safety instructions are higher than health instructions, health instructions are higher than comfort instructions, and comfort instructions are higher than energy-saving instructions. Instructions with low arbitration scores are suspended or discarded, and the remaining instructions are sorted according to their arbitration scores to form the execution instruction sequence.

4. The smart home ecosystem multi-device linkage control system based on big data according to claim 1, characterized in that, The data acquisition module is deployed on the edge gateway. The data acquisition module preprocesses the acquired data, including data deduplication, missing value imputation, time alignment and normalization. The data acquisition module performs hierarchical reporting according to data type, with event-type data being reported in real time and status-type data being reported in batches according to a preset period.

5. The smart home ecosystem multi-device linkage control system based on big data according to claim 1, characterized in that, After the instruction is issued, the execution control module listens for the status feedback returned by the corresponding smart home device. If the feedback times out, the instruction is resent. If the number of resentments reaches the resentment limit, the preset degradation strategy is executed. The execution control module stores the instruction execution log in a structured database and marks user takeover actions. The types of marking include confirmation, modification, and revocation.

6. The smart home ecosystem multi-device linkage control system based on big data according to claim 1, characterized in that, The linkage control platform also includes a protocol adaptation unit, which converts data from smart home devices with different communication protocols into a unified device model. The unified device model describes the device with device type, capability attributes, and status parameters. Newly connected smart home devices register according to the device type template to complete the connection to the linkage control platform.

7. A multi-device linkage control system for a smart home ecosystem based on big data as described in claim 1, characterized in that, The linkage control platform stores a parameter configuration table, which records the correspondence between apartment type, family role and linkage parameters. The linkage parameters include confidence threshold, degradation strategy and initial value of instruction weight. The edge gateway locally caches the scene mode library and the generation rules of the execution instruction sequence. When the network is interrupted, the edge gateway performs linkage control based on the local cache. After the network is restored, the execution logs during the interruption are synchronized.

8. A multi-device linkage control system for a smart home ecosystem based on big data as described in claim 1, characterized in that, The feedback correction module adjusts the confidence level of the corresponding scenario mode and the weight of the corresponding instruction based on the user takeover behavior marked as revocation, and removes the scenario mode whose confidence level falls below the elimination threshold from the scenario mode library. The feedback correction module increases the weight of the corresponding instruction based on the user takeover behavior marked as confirmed. The feedback correction module records the adjusted parameters based on the user takeover behavior marked as modified. When the number of times the user-adjusted parameters appear repeatedly in the same scenario reaches the update threshold, the default value of the instruction parameters in the corresponding scenario mode is updated.

9. A smart home ecosystem multi-device linkage control method based on big data, characterized in that, The method applied to the big data-based smart home ecosystem multi-device linkage control system according to any one of claims 1 to 8, the method comprising: Collect operational status data, environmental perception data, and user operation behavior data from multiple smart home devices; Based on historical user behavior data, frequent operation sequences are mined according to sliding time windows to build a scene pattern library. The real-time collected data is then matched with the scene pattern library to determine the current home scene. Generate a multi-device linkage instruction set based on the scene mode corresponding to the current home scene, perform conflict detection on the instructions in the multi-device linkage instruction set, arbitrate the instructions based on the conflict detection results, and output the execution instruction sequence; The execution command sequence is sent to the corresponding smart home devices, the execution results of the smart home devices are collected, and the user's takeover behavior in response to the execution results is also collected. Based on the execution results and user takeover behavior, the confidence level of the scene modes in the scene mode library and the weight of each instruction in the multi-device linkage instruction set are adjusted.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the multi-device linkage control method for smart home ecosystem based on big data as described in claim 9.

Images