A home scene device linkage control method based on an internet of things

Abstract

The application discloses a kind of family scene device linkage control method based on Internet of Things, it is related to Internet of Things control system technical field;Its technical points are: obtaining the multiple-source linkage operation data corresponding to target home automation network in current analysis window, executes time base correction, message folding, state edge extraction and action window merging, constructs linkage action instance identification, generates linkage event fragment, establishes device expected effective return interval baseline and extracts current queued action set;Obtain device description data and rule description data, construct device capability resource graph, generate current effective action set and conflict candidate relationship set;Combined with device expected effective return interval baseline, device capability resource graph, current effective action set, conflict candidate relationship set and multiple-source feedback evidence, calculate state inaccessibility credibility and basic conflict intensity value, generate conflict disposal result.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) control system technology, specifically to an IoT-based method for the coordinated control of devices in a home environment. Background Technology

[0002] As the types of home IoT devices continue to increase, the linkage control in residential scenarios has gradually evolved from the remote switching of a single device to multi-condition collaborative control of lighting, air conditioning, fresh air, curtains, door locks, water valves, security, and environmental regulation devices. In actual deployment, home automation networks typically connect to human presence sensors, door and window magnetic sensors, temperature and humidity sensors, water leakage sensors, power metering units, as well as voice terminals, mobile terminals, and gateway controllers. Multiple sets of automation rules are configured according to different scenarios such as leaving home, returning home, night, sleep, ventilation, energy saving, and security. In real use, the same execution object is often affected by timing policies, environmental threshold policies, alarm policies, and manual control policies simultaneously. For example, night linkage requires turning off lights and reducing air conditioning power, water leakage linkage requires closing the main water valve and stopping irrigation, and door and window opening linkage may require pausing cooling and starting ventilation. In addition, there are short-term link jitter, device reporting delays, inconsistent status subscription refreshes, and the duration of device actions in home networks. As a result, linkage control is no longer a simple triggering of a single rule on a single device, but a continuous coordination process of multiple source states, multiple human access points, and multiple device actions in the same home automation network.

[0003] Existing technologies typically employ rule engines, automated scripts, delayed execution, queuing, priority overriding, and online device status marking to achieve interconnected control in home scenarios. These technologies possess relatively mature capabilities for triggering and executing individual rules and can also provide interface-level alerts for unreachable devices. However, in scenarios where multiple rules concurrently operate around the same resource, most existing solutions still rely primarily on the validity of a single trigger, lacking unified modeling of device resource occupancy status, residual effects of actions, and replay boundaries during the recovery phase. Consequently, it is difficult to accurately distinguish between real device state changes and state gaps created by communication gaps, and it is also difficult to address persistent prohibition conditions formed after the execution of emergency rules. For example, in handling household water leaks, the system can immediately shut off the main water valve and stop irrigation after detecting a leak. However, when the network recovers briefly, previously queued irrigation tasks may be reassigned due to a simple re-release mechanism, causing the main water valve to be accidentally opened. Similarly, in a nighttime rest scenario, the lighting has already been turned off, but the human body sensor resumes reporting after a short period of unreachability. The old triggering conditions are reinterpreted as new valid events, which may still cause lights to turn on accidentally. It is evident that existing technologies still struggle to solve the problems of residual constraints remaining in effect after multiple concurrent rule conflicts and the erroneous replay control during the recovery phase in complex home automation networks. This has become a key technical gap affecting the stability, security, and user trust of linkage control. Summary of the Invention

[0004] To achieve the above objectives, the present invention provides the following technical solution: A method for controlling interconnected devices in a home environment based on the Internet of Things (IoT) includes the following steps: Acquire the multi-source linkage operation data of the target home automation network within the current analysis window, perform time base correction, message folding, state edge extraction and action window merging on the multi-source linkage operation data, construct linkage action instance identifiers, generate linkage event fragments, establish the baseline of the expected effective return interval of the device, and extract the current queued action set. Obtain device description data and rule description data, identify the execution resource unit mapping and effective action window for linkage action instance identification, construct device capability resource graph, and generate current effective action set and conflict candidate relationship set; By combining the baseline of expected effective return interval of equipment, equipment capability resource map, current effective action set, conflict candidate relationship set and multi-source feedback evidence, the credibility of state unreachability and basic conflict intensity value are calculated, and conflict resolution results are generated. Based on the conflict resolution results, residual constraint state variables are generated. The continued validity of the scenario is determined according to the rule condition expression and the current real-time state set. A replay qualification score is constructed, and the current queuing action set is replayed, suspended, or discarded. Queuing action replay screening results and linkage control results are formed. Samples that meet the preset sample quality gating conditions are written into the stable running sample set in the historical running samples.

[0005] Furthermore, the multi-source coordinated operation data includes automated rule triggering data, manual control data, control command issuance data, equipment response data, equipment status reporting data, equipment link quality data, queuing action data, and equipment execution feedback data; the coordinated action instance identifier is generated by combining the trigger source identifier, target equipment identifier, target action identifier, rule identifier, and window sequence number, and is used to characterize the unique execution instance of the same coordinated action within the current analysis window.

[0006] Furthermore, generating linkage event fragments includes: performing unified time base correction, duplicate message folding, state edge extraction, and action window merging on the multi-source linkage operation data; converting control command edge response quantities, device response consistency quantities, state switching quantities, and device execution feedback change quantities into dimensionless quantities; combining the dimensionless quantities according to preset weights to form event anchoring strengths; and merging candidate event anchor points with event anchoring strengths exceeding preset event anchoring thresholds and having the same linkage action instance identifier into linkage event fragments.

[0007] Furthermore, a baseline for the expected effective return interval of the equipment is established, including: for periodically reporting equipment, the median of the effective reporting intervals within the stable operating sample set in the historical operating samples is used to establish the baseline for the expected effective return interval of the equipment; for event-driven equipment, the median of the response interval between the issuance of control commands and the effective response or effective status report is used to establish the baseline for the expected effective return interval of the equipment; for execution confirmation equipment, the median of the confirmation interval between the issuance of control commands and the equipment execution feedback entering a stable state is used to establish the baseline for the expected effective return interval of the equipment; when the number of samples in the stable operating sample set is lower than the preset minimum sample threshold, the initial baseline for the expected effective return interval of the equipment is established using the equipment configuration default period, the protocol recommended response window, or the reference confirmation window of similar equipment.

[0008] Furthermore, the construction of the device capability resource map includes: mapping linkage action instances to corresponding resource units based on device capabilities, device object relationships, and a preset action semantic mapping table in the device description data, and generating an action resource occupancy vector; determining the mutual exclusion coefficient, concurrency coefficient, and priority coverage coefficient of each resource unit based on a preset resource constraint table; and setting the occupancy value of linkage actions of non-competitive limited power supply objects to zero in the limited power supply resource unit.

[0009] Furthermore, the current set of valid actions and the set of candidate conflict relationships are generated, including: determining the effective action window of each linked action instance based on the start and end times of the linked event fragments and the historical action duration statistics obtained from the stable operation sample set in the historical operation samples; when there is a stable point for device execution feedback, the moment when the device execution feedback enters a stable state is taken as the end time; when there is no stable point for device execution feedback, the estimated time corresponding to the historical median duration of the same action of the same device is taken as the end time; and then generating the current set of valid actions and the set of candidate conflict relationships based on the time overlap relationship between the effective action windows, the rule priority relationship, the recovery strategy determined by the recovery strategy in the rule description data, the recovery strategy relationship determined by the prohibition recovery flag, and the security blocking relationship.

[0010] Furthermore, the reliability of the unreachable state is calculated, including: extracting the silence duration of the target device since the most recent valid response, valid state report or valid execution feedback, current link quality, response consistency and device execution feedback consistency; Link quality is normalized from received signal strength, link quality indicator value, and packet loss rate; response consistency is normalized from the ratio of valid responses to expected responses within the observation window; and device execution feedback consistency is normalized from the degree of consistency between device execution feedback records and the expected direction of target actions. The unreachability confidence level is determined by the degree of deviation of the silence duration from the baseline of the expected effective response interval of the device, the degree of link quality degradation, the degree of decrease in response consistency, and the degree of decrease in device execution feedback consistency. For air conditioning and fresh air systems, when the current observation window length is less than the preset minimum thermal response time threshold, temperature data, humidity data, and air quality data are only used to determine the validity of the scenario and do not participate in the formation of consistent feedback from the equipment.

[0011] Furthermore, the calculation of the basic conflict intensity value and the generation of conflict resolution results include: for each action to be resolved, matching it one by one with each currently effective action in the current effective action set; extracting the resource conflict coupling amount and time overlap rate from the equipment capacity resource map and the effective action window of each linkage action instance; and obtaining the priority level deficit normalization amount by normalizing the difference between the rule priority level of the action to be resolved and the rule priority level of the currently effective action. The resource conflict coupling amount, time overlap rate, and priority level deficit normalization amount are combined to form the basic conflict intensity value of the action to be resolved. The basic conflict intensity value is then compared with the preset conflict threshold determined by the preset equipment risk level mapping table, and combined with the status unreachability confidence, water supply safety resource unit occupancy status, and safety blockade resource unit occupancy status, a conflict resolution result is generated from one of the following: immediate execution result, delayed observation result, blocked execution result, or suspended waiting result.

[0012] Furthermore, residual constraint state quantities are generated based on the conflict resolution results, including: extracting historical resolution events that were delayed observation, blocked execution, suspended waiting, or covered by emergency actions from the conflict resolution results to form a set of residual constraint events; and classifying the residual constraint events into hard blocking residual constraint events or soft blocking residual constraint events based on the source rule type, equipment risk level, and occupied resource unit type of the historical resolution events. Specifically, when a historical incident originates from a water leakage protection rule, gas protection rule, security blockade rule, electrical overload protection rule, or water supply safety blockade rule, and the equipment risk level is not lower than the preset equipment risk level threshold, the corresponding residual constraint event is identified as a hard-blocking residual constraint event. Then, based on the severity coefficient of the historical incident, the coupling coefficient with the target resource unit, the effective time, the residual constraint time constant, and the release evidence factor formed by combining explicit release trigger data, reverse control command execution confirmation data, continuous stable operation data, and emergency alarm disappearance data according to preset weights, the residual constraint state quantity of the equipment on the corresponding resource unit is generated.

[0013] Furthermore, the queuing action replay filtering includes: determining the validity of the scenario based on the rule condition expression corresponding to each queuing action and the current real-time state set formed by summarizing multi-source linkage operation data at the current analysis time. Specifically, for threshold-type conditions, the current real-time state value is compared with the corresponding threshold interval to obtain the condition validity result; for event-type conditions, whether the current event is within the rule validity duration window is taken as the condition validity result; for combined conditions, the scenario validity result is obtained by evaluating each item according to AND, OR, and sequential relationships. The queuing action timeliness is determined based on the time decay relationship between the queuing action generation time and the current analysis time. A replay eligibility score is then formed by combining the queuing action timeliness, the normalized result corresponding to the basic conflict intensity value, the normalized result corresponding to the residual constraint state quantity, and the unreachability confidence of the target device state. When the replay eligibility score is not lower than the preset replay threshold and the resource unit corresponding to the queuing action does not have hard blocking residual constraints, the queuing action is replayed. When the replay eligibility score is lower than the preset replay threshold, the scenario is still valid, and the resource unit corresponding to the queuing action does not have hard blocking residual constraints, the queuing action is suspended. When the scenario is still invalid, the queuing action timeliness is lower than the preset failure threshold, or the resource unit corresponding to the queuing action has hard blocking residual constraints, the queuing action is discarded. Sample quality gating is applied to the actual effective return interval, historical action duration, actual recovery time, and device execution feedback consistency. The sample quality gating includes at least no gateway-level abnormal interruption, link quality higher than the preset link quality threshold, and device execution feedback consistency higher than the preset consistency threshold.

[0014] This invention provides a method for the coordinated control of devices in a home environment based on the Internet of Things (IoT), which has the following beneficial effects: 1. This invention employs techniques such as time base correction, message folding, state edge extraction, and action window merging to perform multi-source linkage operation data, and constructs linkage action instance identifiers, generates linkage event fragments, and establishes a baseline for the expected effective reporting interval of devices. This solves the problems of inconsistent recording times from multiple sources, strong interference from repeated reporting, and easy commingling of actions in the same direction in home automation networks. It achieves the effect of unifying action attribution and unifying time base, and also provides a stable event-level input basis for subsequent conflict determination and recovery control.

[0015] 2. This invention employs a technical approach that maps linkage action instances to resource units, constructs action resource occupancy vectors and device capability resource graphs, and combines action effective action windows to generate a set of currently effective actions and a set of conflict candidate relationships. This solves the problem in existing linkage control systems where conflicts are judged solely by device number or a single rule, making it difficult to identify target adversarial and security coverage relationships. It accurately depicts resource competition relationships and control target adversarial relationships, while also improving the completeness and consistency of conflict identification in complex home scenarios.

[0016] 3. This invention employs a technical means that combines the baseline of the expected effective return interval of the equipment, multi-source feedback evidence, equipment capability resource map and conflict candidate relationship set to form the state unreachability credibility and basic conflict intensity value, and generates conflict handling results. This solves the problem in the prior art of being unable to distinguish between short-term communication gaps, real equipment anomalies and actual resource conflicts, and improves the accuracy of risk judgment before the execution of linkage actions, while also reducing false blocking, false release and abnormal misjudgment.

[0017] 4. This invention employs a technique of generating residual constraint state variables based on conflict resolution results, and combining the still-validity of the scenario, the timeliness of queuing actions, and the replay eligibility score to perform replay, suspension, or discard processing on the current queuing action set. This solves the problems of accidental restoration, expired restoration, and failure of security constraints of historical queuing actions after network recovery. It effectively suppresses accidental replay and extends security constraints, while also improving the continuity, stability, and reliability of subsequent sample updates in linkage control. Attached Figure Description

[0018] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a schematic diagram of the S4 process of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figure 1 and Figure 2 This embodiment provides a method for the coordinated control of devices in a home environment based on the Internet of Things (IoT), including the following specific steps: In this embodiment, S1 is used to convert the multi-source linkage operation data formed by the target home automation network within the current analysis window into a set of linkage event fragments, a set of linkage action instances, a set of device expected effective return interval baselines, and a set of current queued actions that can be directly accepted by subsequent steps. The unified time base correction, duplicate message folding, state edge extraction, and action window merging in S1 provide a unified timing and data foundation for subsequent processing. The construction of linkage action instance identifiers, the generation of linkage event fragments, and the establishment of device expected effective return interval baselines are the core of this step. After adopting this processing method, S2 can directly construct a device capability resource map around the already assigned and delimited linkage action instances, S3 can calculate the state unreachability credibility around the device expected effective return interval baseline, and S4 can perform queued action replay filtering around the current queued action set.

[0021] In this embodiment, the multi-source coordinated operation data includes at least automated rule triggering data, manual control data, control command issuance data, device response data, device status reporting data, device link quality data, queuing action data, and device execution feedback data. Automated rule triggering data is used to characterize the triggering source and triggering time of the coordinated action by the rule engine within the current analysis window; manual control data is used to characterize the manual control actions of the mobile terminal, voice terminal, or local panel on the target device; control command issuance data is used to characterize the time, target action, and target device when the controller issues the action command to the target device; device response data is used to characterize the target device's confirmation of the control command; device status reporting data is used to characterize the state change process of the target device; device link quality data is used to characterize the communication quality of the current control link; queuing action data is used to characterize pending actions that have not yet been processed; and device execution feedback data is used to characterize whether the action has already had an actual effect on the device itself.

[0022] For lighting equipment, the preferred methods for feedback data execution are relay readback, current change recording, or power change recording; for water valve equipment, the preferred methods are valve position readback, flow rate change recording, or pipeline pressure change recording; for curtain equipment, the preferred methods are motor current change recording, travel feedback recording, or end position signal; for air conditioning and fresh air equipment, temperature, humidity, and air quality data are used in S1 to characterize scene conditions and subsequent release evidence sources, and do not participate in the formation of consistent equipment execution feedback during the immediate execution confirmation stage. The reason for this distinction is that there is a direct causal relationship between the feedback quantities of lighting, water valve, and curtain equipment and the target action, which can be used for immediate confirmation; the environmental quantities corresponding to air conditioning and fresh air equipment have thermal inertia and diffusion delay, and cannot accurately characterize the immediate execution status of control commands within a short observation window.

[0023] When performing time base correction on multi-source coordinated operation data, the gateway reception time is used as a unified reference time. Combined with the device's local time deviation record, all types of records are converted to the same time axis. The device's local time deviation record is preferably formed by the median of the historical pairing difference between the device's local timestamp and the gateway reception time. When the device does not provide a local timestamp, the gateway reception time is directly used as the valid time of the record. The duplicate message folding process adopts a two-level rule: when the record has a message identifier, records with the same message identifier and an arrival interval not greater than the preset duplicate suppression window are folded into a valid record. When the record does not have a message identifier, duplicate records with the same source, target device, target action, and status code are folded into a valid record when the arrival interval is not greater than the preset duplicate suppression window.

[0024] The preset repetition suppression window is preferably determined according to the protocol retransmission period, preferably 1 to 2 times the protocol retransmission period. When the protocol retransmission period cannot be directly read, 0.3s is preferred for wired bus devices and 1.0s is preferred for wireless devices. This value is based on common retransmission rhythms and gateway reception jitter distribution in home automation networks, and can cover normal retransmissions without amplifying the time interval between independent actions. State edge extraction is completed by comparing adjacent state values. For discrete state quantities, when the current state value is different from the previous state value, it is recorded as a state edge. For continuous feedback quantities, when the absolute value of the difference between the current value and the previous value is not less than a preset change threshold, it is recorded as a state edge. The preset change threshold is preferably equal to the sensor resolution. The action window merging should be performed using the larger of the step size and the standard deviation of the stable operating noise. Constraints of same device, same action, and same direction should be applied. When adjacent candidate records have the same target device identifier, target action identifier, and action direction, and the time interval is not greater than the preset merging time threshold, they are merged into the same candidate action window. The preset merging time threshold is preferably determined based on the 95th percentile confirmation delay formed from command issuance to execution confirmation in the historical stable operating samples of similar devices. When samples are insufficient, a value of 1 second is preferred for lighting equipment, 3 seconds for water valve equipment, and 5 seconds for air conditioning and fresh air equipment. This setting ensures that fast-response devices are not over-expanded, and slow-response devices are not split into multiple independent candidate action windows due to feedback delays.

[0025] After completing the above processing, the linkage action instance identifier is first constructed. Specifically, the trigger source identifier, target device identifier, target action identifier, rule identifier, and window sequence number are combined to form the linkage action instance identifier. The window sequence number is generated by incrementing the order of the start times of the candidate action windows in the current analysis window. When the start times of two candidate action windows are the same, they are further numbered according to the arrival order of the control command data. With this method, the same-direction actions triggered by different rules, different manual control entry points, or different command batches within the same analysis window for the same target device can be distinguished into different linkage action instance identifiers. They will not be mistakenly merged due to the same action direction. Compared with the method of directly describing a single action with the target device and target action, this processing can improve the stability of action attribution. It is particularly suitable for home automation scenarios where night mode, human presence rules, and manual control continuously act on the same device in a short period of time.

[0026] When generating linkage event fragments, the control command edge response quantity, device response consistency quantity, state switching quantity, and device execution feedback change quantity are first extracted from the unified preprocessed multi-source linkage operation data. The above four types of quantities are then converted into dimensionless quantities in the range of 0 to 1. The control command edge response quantity is used to characterize the clarity of the starting edge of the control command within the candidate action window, and is preferably calculated by converting the command existence and the proximity of the command arrival time to the starting point of the candidate action window. The device response consistency quantity is used to characterize the matching degree between the device response received in the observation window and the response that should be sent, and is preferably calculated by the ratio of the number of valid responses to the number of responses that should be sent, after performing interval pruning. The state switching quantity is used to characterize whether the device state changes along the expected direction of the target action, and is preferably calculated by converting the state switching amplitude relative to the preset switching threshold corresponding to the device type. The device execution feedback change quantity is used to characterize whether the device body feedback changes along the direction of the target action, and is preferably calculated by converting the absolute value of the difference between the device execution feedback data before and after the action execution relative to the standard change amplitude of similar devices.

[0027] In this embodiment, the event anchoring strength is formed by weighting and summing the control command edge response, device response consistency, state transition, and device execution feedback change according to preset weights corresponding to the device type. For lighting devices, curtain devices, and water valve devices with direct device execution feedback data, the preset weight corresponding to the device execution feedback change is preferably higher than the preset weight corresponding to the control command edge response, because the causal relationship between the device feedback and the actual execution state is more direct. For devices lacking direct device execution feedback data, the preset weight corresponding to the device execution feedback change is proportionally allocated to the other three types of quantities. The preset event anchoring threshold is preferably the intersection of the event anchoring strength distributions corresponding to the actual linkage action records and repetitive noise records in the historical stable operation samples. Fork point determination; when historical samples are insufficient, the preset event anchoring threshold is preferably set to 0.65. Candidate event anchor points with event anchoring strength exceeding the preset event anchoring threshold and having the same linkage action instance identifier are grouped into linkage event segments in chronological order. Each linkage event segment includes at least a linkage action instance identifier, segment start time, segment end time, and a set of valid evidence types within the segment. Compared with the processing method in the prior art that directly represents the action event by triggering a single rule record or a single device status record, this implementation method can significantly reduce the impact of repeated reporting, late response, and link retransmission on action event identification by fusing control command evidence, device response evidence, status switching evidence, and device execution feedback evidence under the constraint of the same linkage action instance identifier.

[0028] When establishing the baseline for the expected effective return interval of equipment, time interval samples under the same equipment, communication method, and acknowledgment method are first extracted from the stable operation sample set in the historical operation samples. The baseline for the expected effective return interval of equipment is a dimensional time quantity, with the unit being seconds. It is used in subsequent S3 to convert the silence duration into a comparable time deviation. For periodically reporting equipment, the time interval between adjacent valid reports is statistically analyzed, and the median is taken as the baseline for the expected effective return interval of equipment. For event-driven equipment, the response interval between the issuance of control commands and valid responses or valid status reports is statistically analyzed, and the median is taken as the baseline for the expected effective return interval of equipment. For execution acknowledgment equipment, the acknowledgment interval between the issuance of control commands and the equipment execution feedback entering a stable state is statistically analyzed, and the median is taken as the baseline for the expected effective return interval of equipment. The median is used because occasional link jitter, repeated manual triggering, and short-term offline events in home automation networks can create outliers. The median can suppress the impact of a small number of outliers on the time baseline. The preset minimum sample threshold for the stable operating sample set is preferably 30. When there are fewer than 30 samples for the same device, it is preferable to call reference samples from devices of the same device type and communication method to supplement them. When there are still fewer than 30 samples, the initial expected effective return interval baseline for the device is established by using the device configuration default period, the protocol recommended response window, or the reference confirmation window of similar devices. This process makes the formation of subsequent unreachable credibility not dependent on a single nominal period, but on the historical behavior characteristics of the device in the real home network. Therefore, it is more suitable for mixed deployment scenarios of periodically reporting devices, event-driven devices, and execution confirmation devices.

[0029] When extracting the current queued action set, the system filters out linkage action instances whose action status is pending execution, delayed observation, or suspended waiting and which have not yet formed a linkage control result from the queued action data, and includes them in the current queued action set. To ensure the singleness of the objects processed in subsequent S4, linkage action instances that have formed a clear linkage control result are no longer included in the current queued action set. Thus, this step finally outputs a set of linkage event fragments, a set of linkage action instances, a set of baselines for expected effective return intervals of equipment, and a set of current queued actions. Among them, the set of linkage event fragments and the set of linkage action instances serve as inputs for S2 to construct the equipment capability resource map and generate the current effective action set and the conflict candidate relationship set. The set of baselines for expected effective return intervals of equipment serves as the time base for S3 to calculate the unreachability confidence of the status. The set of current queued actions serves as the object to be handled by S4 to perform queued action replay filtering.

[0030] In this embodiment, S2 is used to further convert the set of linked event fragments and the set of linked action instances output by S1 from event objects with completed time and action attribution into a device capability resource map that can represent resource competition relationships, control target confrontation relationships, and security coverage relationships. On this device capability resource map, combined with action effective action windows, rule priority level relationships, recovery strategy relationships, and security blocking relationships, a current effective action set and a conflict candidate relationship set are generated for direct use in subsequent S3 calculations of basic conflict intensity values ​​and generation of conflict resolution results. In this step, device description data and rule description data are obtained, a preset action semantic mapping table is called, and a scene target mapping table is called. The mapping table and the invocation of the pre-set resource constraint table are supporting existing technologies that provide rule carriers and mapping carriers for subsequent innovative processing. Based on the linkage action instances, the action resource occupancy vector is established, the device capability resource map is constructed, the resource conflict coupling quantity is formed, and the current effective action set and conflict candidate relationship set are generated under the constraints of the resource map and the effective action window. After adopting this processing method, the subsequent S3 conflict analysis is based on the unified judgment in the same resource space and the same time space. It can simultaneously identify physically mutually exclusive actions within the same device, resource conflict actions of different devices competing for the same controlled object, and antagonistic actions of different devices pointing to the same environmental target but in opposite directions.

[0031] In this embodiment, the equipment description data includes at least the equipment type, equipment capacity, restricted power supply object, valve-controlled object, airflow regulation object, shading control object, lighting control object, temperature control target object, humidity control target object, ventilation status object, water supply safety object, safety blockage object, equipment risk level, and historical action duration statistics. The equipment type is used to select the action mapping branch and time parameter branch; the equipment capacity is used to characterize the types and directions of actions that the equipment can execute; the restricted power supply object is used to characterize whether the actions compete for the same restricted power supply branch, the same energy storage branch, or the same current limiting module; the valve-controlled object, airflow regulation object, shading control object, and lighting control object are used to characterize the direct relationship between the actions and the physical execution objects; the temperature control target object, humidity control target object, and ventilation status object are used to characterize the actions. The data includes: the relationship between the action and the environmental control objectives; water supply safety objects and safety blockade objects to characterize whether an action enters the water supply protection and safety blockade control range; equipment risk levels to be used in subsequent S3 and S4 calls to the preset equipment risk level mapping table; historical action duration statistics to determine the termination time of the effective action window; rule description data including at least rule priority level, effective time period, recovery strategy, prohibition of recovery flag, and conditional expression; rule priority level to characterize the order of handling when multiple actions act simultaneously; effective time period to limit the effective interval of the rule within the current analysis window; recovery strategy to characterize the recovery method after conflict resolution; prohibition of recovery flag to characterize whether the corresponding action is allowed to be replayed before the release condition is met; and conditional expression to determine the validity of the scenario in subsequent S4.

[0032] When mapping linked action instances to resource units, a pre-defined action semantic mapping table and a scene target mapping table are first invoked. The pre-defined action semantic mapping table represents the correspondence between various actions and physical execution objects, while the scene target mapping table represents the correspondence between various actions and environmental control targets. These tables are preferably generated during the system deployment phase, and their formation is based on at least equipment description parameters, equipment installation and debugging records, equipment object relationships, control range configuration results, and rule configuration results. For example, the curtain opening action is mapped to the occlusion control object via the pre-defined action semantic mapping table and to the ventilation status object via the scene target mapping table; the air conditioning cooling action is mapped to the airflow adjustment object via the pre-defined action semantic mapping table and to the temperature control target object via the scene target mapping table; the dehumidification action is mapped to... The action is mapped to the humidity control target object; the main water valve closing action triggered by the leakage protection rule is mapped to the water supply safety object; the door lock blocking action triggered by the security blockade rule is mapped to the security blockade object; based on the equipment capabilities, equipment object relationships in the equipment description data and the above two mapping tables, each linkage action instance is mapped to the corresponding resource unit to form an action resource occupancy vector. Each dimension of the action resource occupancy vector corresponds to a uniformly defined resource unit, which includes at least the limited power supply resource unit, valve control resource unit, airflow regulation resource unit, shielding control resource unit, lighting control resource unit, temperature control target resource unit, humidity control target resource unit, ventilation status resource unit, water supply safety resource unit, and security blockade resource unit. The action resource occupancy vector is a dimensionless vector, and the occupancy value of each dimension is uniformly mapped to a dimensionless occupancy amount of 0 to 1.

[0033] For switching actions, when the action completely occupies a resource unit, the corresponding dimension occupancy value is set to 1; when the action does not occupy the resource unit, the corresponding dimension occupancy value is set to 0. For regulating actions, the ratio of the action setting amplitude to the device's full-scale regulating amplitude is used for conventional normalization to obtain a dimensionless occupancy value in the range of 0 to 1. When the normalization result is less than 0, it is set to 0; when it is greater than 1, it is set to 1. For linkage actions that do not compete for limited power supply objects, their occupancy value on the limited power supply resource unit is set to 0. After adopting this method, the action resource occupancy vector can express both the degree of occupancy of the action on the physical execution object and the direction and intensity of the action's effect on the environmental control target, and can be directly used to form resource conflict coupling quantities in the future.

[0034] When constructing the equipment capability resource map, a pre-set resource constraint table is invoked. This table includes at least the mutual exclusion coefficient, concurrency coefficient, and priority coverage coefficient for each resource unit. All three coefficients are dimensionless, ranging from 0 to 1. The mutual exclusion coefficient characterizes the strength of the conflict when two actions occur simultaneously on the same resource unit; the concurrency coefficient characterizes the degree to which two actions are allowed to coexist on the same resource unit; and the priority coverage coefficient characterizes the suppression strength of a higher-priority action on a lower-priority action on the same resource unit. To ensure the values ​​have a practical basis, this implementation preferably uses a discrete-level assignment method to generate these coefficients. For reverse actions on the same physical execution object, protection actions and water supply actions on the same water supply safety object, and blocking actions and normal actions on the same safety blocking object, the mutual exclusion coefficient is 1, the concurrency coefficient is 0, and the priority coverage coefficient is 1 according to the relationship where a higher-priority action covers a lower-priority action. For actions acting on… For actions targeting the same environmental object but with opposite control directions, such as opening a window for ventilation and air conditioning for cooling, which act simultaneously in the same environmental area, the mutual exclusion coefficient is preferably 0.8, and the concurrency coefficient is preferably 0.2. For actions targeting the same environmental object with the same control direction, the mutual exclusion coefficient is preferably 0.2, and the concurrency coefficient is preferably 0.8. For actions that do not act on the same object or target, the mutual exclusion coefficient is 0, the concurrency coefficient is 1, and the priority coverage coefficient is 0. The reason for using a discrete level assignment method is that resource competition and target confrontation in the home scenario have clear physical and rule meanings. Using a discrete level can ensure stable values ​​and facilitate unified deployment across different device types. Based on the action resource occupancy vector and the above constraint coefficients, a device capability resource graph is constructed. Nodes in the device capability resource graph correspond to linked action instances, and edges correspond to two linked action instances having an occupancy relationship on at least one resource unit. The edge weight is represented by the resource conflict coupling quantity described later.

[0035] When generating resource conflict coupling, for any two linked action instances, the degree of overlap in resource unit usage for each action is calculated. For switch-type actions, the degree of overlap is determined by the smaller of the two actions' usage values ​​in that resource unit; for adjustment-type actions, the degree of overlap is determined by the product of the two actions' usage values ​​in that resource unit. Then, the degree of overlap is multiplied by the mutual exclusion coefficient corresponding to that resource unit to obtain the local conflict level in that resource unit. The local conflict level is dimensionless, ranging from 0 to 1. The higher the local conflict level, the less suitable the two actions are for parallel existence in that resource unit. Finally, a supplementary aggregation is performed on the local conflict levels of all resource units to obtain the resource conflict coupling between the two actions. The supplementary aggregation is represented by the following core relationship: ; in, This indicates the resource conflict coupling between two linked action instances. Indicates the first The degree of local conflict on each resource unit The formula represents the total number of resource units involved in the comparison. The calculation logic is as follows: first, calculate the degree of non-conflict on each resource unit, then multiply the degree of non-conflict on all resource units, and finally subtract the product from 1 to obtain the total degree of conflict between the two actions in the entire resource space. All terms in the formula are dimensionless, so the dimensions are consistent. The premise for the formula to hold is that the occupancy values ​​of all resource units involved in the calculation have been uniformly mapped to dimensionless quantities from 0 to 1. The role of the resource conflict coupling quantity in this step is to serve as the direct input for the subsequent S3 calculation of the basic conflict intensity value. The reason for using the supplementary aggregation is that weak conflicts on multiple resource units can be superimposed into significant conflicts in a home scenario. For example, when directional opposition on a temperature control target object and directional opposition on a ventilation status object occur simultaneously, it should have a higher conflict weight than a weak conflict on a single target. At the same time, strong conflicts on a single key resource unit, such as conflicts on water supply security objects or security blockade objects, can also directly increase the resource conflict coupling quantity.

[0036] When determining the effective action window for each linkage action instance, the linkage event fragment generated in S1 is first read. The start time of the linkage event fragment is directly used as the start point of the effective action window for the corresponding linkage action instance. The end time is determined in two levels. In the first level, when there is a stable point for the device execution feedback, the moment when the device execution feedback enters a stable state is taken as the end time. The stable point for the device execution feedback is preferably determined by continuous stability detection. That is, when the change amplitude of the device execution feedback does not exceed the upper limit of the corresponding feedback noise in three consecutive sampling periods, it is determined to have entered a stable state. The upper limit of the feedback noise is preferably taken as the standard deviation of the corresponding feedback quantity in the historical stable operation samples. The first level is twice the standard level. The second level is when there is no stable point for device execution feedback. The stable operation sample set in the historical operation sample is called to count the historical action duration of the same device and the same action. The median historical duration is preferred and superimposed on the start time of the linkage event segment to obtain the estimated termination time. For actions with rule effective time period restrictions, the intersection of the natural action window and the effective time period in the rule description data is taken to obtain the final action effective action window. After adopting this processing method, the action effective action window can reflect both the continuous occupancy characteristics of the device action body and the restriction of the rule effective time period on the action effective boundary.

[0037] When generating the current effective action set, the current analysis time is used as the criterion. Instances of linked actions that fall within the effective action window at the current analysis time, have not received an explicit termination record, and are not completely covered by higher priority safety blocking actions are included in the current effective action set. Explicit termination records include at least action completion confirmation records, equipment shutdown confirmation records, and manual cancellation records. Actions that have been covered by higher priority actions on water supply safety objects or safety blocking objects are no longer included in the current effective action set, even if their effective action window has not yet naturally ended. Instead, they participate in conflict analysis as pending actions in subsequent S3. The reason for this is that the current effective action set is used to characterize the action objects that are still actually occupying resource units or acting on control targets at the current time, so both the time validity and control failure conditions must be met simultaneously.

[0038] When generating the set of candidate conflict relationships, for each action to be handled, it is matched one by one with each currently effective action in the current set of effective actions. The resource conflict coupling, the time overlap between the effective action windows, the rule priority level, the recovery strategy, and the security blockade are judged sequentially. First, when the resource conflict coupling between two actions is not lower than the preset candidate conflict threshold, and the effective action windows of the two actions have a non-zero time overlap, the action pair is retained as a candidate conflict pair. The preset candidate conflict threshold is preferably determined by the intersection of the resource conflict coupling distribution of real conflict pairs and non-conflict pairs in historical operation samples; when the sample is insufficient, it is preferably set to 0.2. Second, when two actions have an overlay relationship on water supply security objects or security blockade objects, even if the resource conflict coupling is not lower than the preset candidate conflict threshold, the action pair is retained as a candidate conflict pair. If the total number of conflict pairs is lower than the preset candidate conflict threshold, the action pair will be directly included in the conflict candidate relationship set. Then, based on the rule priority relationship, recovery strategy relationship, and security block relationship, the candidate conflict pair will be further screened. The recovery strategy relationship is determined by the recovery strategy and the prohibition of recovery flag in the rule description data. The recovery strategy includes at least four types: immediate recovery, conditional recovery, delayed recovery, and prohibition of recovery. When the recovery strategy of the action to be dealt with is prohibition of recovery, or when the rule corresponding to the current effective action sets the prohibition of recovery flag for the action to be dealt with, it is considered that there is a recovery strategy constraint relationship between the two. When the current effective action occupies a security block object and its rule priority is higher than that of the action to be dealt with, it is considered that there is a security block relationship between the two. After the above processing, the conflict candidate relationship set is obtained.

[0039] This step ultimately outputs a device capability resource map, a set of currently valid actions, and a set of conflict candidate relationships. The device capability resource map serves as the structured carrier for extracting resource conflict coupling in S3. The effective action window and rule priority level relationships serve as inputs for S3 to calculate the basic conflict intensity value and generate conflict resolution results. The set of conflict candidate relationships serves as the object to be analyzed in S3. The continuous occupancy state and security blocking relationship in the set of currently valid actions serve as the basis for generating residual constraint state quantities in S4. Compared to existing technologies that directly determine conflicts by matching device numbers or configuring single rules, this implementation method maps physical execution conflicts, target adversarial conflicts, and security coverage constraints to the device capability resource map in S2 using a pre-set action semantic mapping table, a scene target mapping table, and a pre-set resource constraint table. Then, it combines the effective action window and rule constraint relationships to generate the set of currently valid actions and the set of conflict candidate relationships. Subsequent conflict determination and replay filtering are all based on a unified, calculable, and verifiable input.

[0040] In this embodiment, S3, based on the equipment capability resource map, current effective action set, and conflict candidate relationship set already formed in S2, further combines the equipment expected effective return interval baseline established in S1 and multi-source feedback evidence in multi-source linkage operation data to distinguish the possibility that the target equipment is currently in a normal reachable state, a short-term communication gap state, or a stable execution confirmation state. On this basis, the degree of conflict between the action to be handled and the current effective action is uniformly quantified, and the conflict handling result is output as the direct basis for S4 to generate residual constraint state quantities and perform queued action replay screening. After adopting this processing method, S4 can generate residual constraint state quantities and perform queued action replay screening under the premise that the current equipment communication reliability, current action resource conflict degree, and current security resource occupancy degree are clear, thereby reducing the probability of false replay and false suspension.

[0041] In this embodiment, the inputs of S3 include at least the baseline set of expected effective return intervals for equipment, the equipment capability resource map, the current effective action set, the conflict candidate relationship set, and multi-source feedback evidence. The baseline set of expected effective return intervals for equipment is output by S1 and is a dimensional time quantity set in seconds, used to characterize the expected return rhythm of each device under stable operating conditions. The equipment capability resource map, the current effective action set, and the conflict candidate relationship set are output by S2. The equipment capability resource map is used to provide action resource occupancy relationships and resource conflict coupling relationships. The current effective action set is used to characterize action objects that are still in an effective occupancy state at the current analysis time. The conflict candidate relationship set is used to limit the range of action pairs that need to further calculate the basic conflict intensity value. The multi-source feedback evidence is obtained from multi-source linkage operation data and includes at least equipment response data, equipment status reporting data, equipment link quality data, and equipment execution feedback data. The equipment response data is used to form response consistency, the equipment status reporting data is used to form silent duration and return timing support, the equipment link quality data is used to form the current link quality, and the equipment execution feedback data is used to form equipment execution feedback consistency.

[0042] In this implementation, the basic quantities required for the unreachable state credibility are first established. The silence duration is used to characterize the duration of no feedback from the target device since the most recent valid report. Valid reports include three types: valid response, valid status report, and valid execution feedback. As long as any type of report meets the valid confirmation conditions of the device under the current action type, it is considered a valid report. The silence duration is obtained by subtracting the unified time base time corresponding to the most recent valid report from the current analysis time. It is a dimensional time quantity with the unit being seconds. Valid response, valid status report, and valid execution feedback are all included in the endpoint determination of the silence duration because there are objective differences in the main confirmation channels of different devices. Periodically reporting devices mainly rely on status reports, event-driven devices mainly rely on valid responses, and execution confirmation devices mainly rely on device execution feedback. After the three types of valid reports participate in the determination, the silence duration can more accurately reflect the actual reporting behavior of the device.

[0043] The current link quality characterizes the stability of the target device's current communication link. It is composed of received signal strength, link quality indicator (LQI), and packet loss rate. Received signal strength is a dimensional quantity measured in dBm. Preferably, it is calculated using a linear range conversion based on the upper and lower limits of the received signal strength supported by the corresponding communication protocol. For Zigbee devices, the received signal strength is preferably mapped from -95dBm to -45dBm to 0 to 1; for Wi-Fi devices, it is preferably mapped from -85dBm to -40dBm to 0 to 1; and for Bluetooth Low Energy devices, it is preferably mapped from -100dBm to -50dBm to 0 to 1. The LQI is preferably calculated using a ratio conversion based on the full-scale value defined by the protocol to obtain the link quality within the range of 0 to 1. The packet loss rate is obtained by comparing the number of lost packets to the number of packets expected within the current observation window. Subtracting the packet loss rate from 1 yields the packet loss quality. Then, a weighted average is used to fuse the signal quality, link quality, and packet loss quality to obtain the current link quality. The current link quality is a dimensionless quantity ranging from 0 to 1; a higher value indicates a more stable link. The current observation window is preferably 2 to 4 times the baseline of the device's expected effective reporting interval. When the baseline of the device's expected effective reporting interval is less than 1 second, the observation window is preferably 2 to 4 seconds. When the baseline of the device's expected effective reporting interval is not less than 1 second, the observation window is preferably 3 times that baseline. This value is based on the time coverage requirements of a normal reporting process and a retransmission process in a home automation network, accommodating both short-cycle and long-cycle devices.

[0044] Response consistency is used to characterize the completeness of the target device's response to the issued control commands within the observation window. Response consistency is preferably formed by the ratio of the number of valid responses to the number of responses that should be sent within the observation window. The number of valid responses is obtained by comparing the target device identifier, target action identifier, and linkage action instance identifier in the response data. The number of responses that should be sent is obtained by counting the number of control commands sent to the device in the current analysis window that respond according to the protocol requirements. For devices that do not require explicit responses in the protocol itself, the number of responses that should be sent is set to 1, and the valid status report is used instead of response data for statistics. For batch control commands under the same linkage action instance, the batch command is regarded as a single response event that should be sent. Response consistency is a dimensionless quantity in the range of 0 to 1. The larger the value, the more complete the response chain.

[0045] Equipment execution feedback consistency is used to characterize whether the equipment's feedback forms an effective change along the expected direction of the target action. First, the baseline value of the equipment execution feedback before the execution of the linked action instance is taken. Then, the equipment execution feedback value at the end of the current observation window is taken. The difference between the two constitutes the change in equipment execution feedback. Next, it is determined whether this change in equipment execution feedback is consistent with the expected direction of the target action. If consistent, the change in equipment execution feedback is normalized relative to the standard feedback change amplitude of the same equipment and the same action, resulting in equipment execution feedback consistency within the range of 0 to 1. If inconsistent, the equipment execution feedback consistency is directly set to 0. The standard feedback change amplitude is preferably taken from the stable operating sample set in the historical operating samples, for the same equipment and the same action. The median of the absolute value of the feedback change is a dimensionless quantity. For air conditioning and fresh air systems, when the current observation window length is less than the preset minimum thermal response time threshold, temperature, humidity, and air quality data are only used for the validity determination of the scenario in subsequent S4 and do not participate in the formation of the device's feedback consistency. The preset minimum thermal response time threshold is preferably determined by the median of the time from the issuance of the control command to the first sustained unidirectional change of the environmental quantity within the stable operating sample set in the historical operating samples. When the sample is insufficient, the preferred value is 60s to 180s for air conditioning and 30s to 120s for fresh air systems. The above values ​​conform to the actual laws of indoor thermal inertia and air volume diffusion in residential scenarios.

[0046] After establishing the aforementioned basic quantities, the unreachability credibility is further determined. Specifically, the deviation of the silence duration from the baseline of the device's expected effective response interval is first converted into a dimensionless time deviation. The conversion method is as follows: divide the silence duration by the baseline of the corresponding device's expected effective response interval to obtain the silence deviation multiple, and then subtract 1 from the silence deviation multiple to obtain the time deviation. The larger the time deviation, the more the current silence level exceeds the device's normal response rhythm. Subsequently, the degradation of the current link quality, the decrease in response consistency, and the decrease in device execution feedback consistency are respectively determined. The degradation of the current link quality equals 1 minus the current link quality, the decrease in response consistency equals 1 minus response consistency, and the decrease in device execution feedback consistency equals 1 minus device execution feedback consistency. Since the current link quality, response consistency, and device execution feedback consistency are all dimensionless quantities within the range of 0 to 1, the above three decreases are also dimensionless quantities within the range of 0 to 1. After measuring the dimensional quantities, the time deviation, link quality degradation, response consistency decrease, and device execution feedback consistency decrease are weighted and summed according to the preset weights corresponding to the device type to obtain the original comprehensive quantity of the target device's current unreachability risk. For periodically reporting devices, the weights of time deviation and link quality degradation are preferably higher than those of response consistency decrease and device execution feedback consistency decrease; for event-driven devices, the weight of response consistency decrease is preferably increased; for execution confirmation devices, the weight of device execution feedback consistency decrease is preferably increased. The preset weights preferably satisfy the condition that the sum is 1, and the individual weights are preferably between 0.15 and 0.45. Finally, the original comprehensive quantity is monotonically compressed using the Logistic compression method to obtain the state unreachability confidence level in the range of 0 to 1. The closer the state unreachability confidence level is to 1, the more likely the target device is to be in an unreachable state; the closer it is to 0, the more likely the target device is still in a normal reachable state.

[0047] Before forming the basic conflict intensity value, the resource conflict coupling quantity and time overlap rate are extracted from the equipment capability resource map and action effective action window output by S2. These are then combined with rule description data to form a priority level deficit normalization quantity. The resource conflict coupling quantity is directly output by S2 and is a dimensionless quantity within the range of 0 to 1, representing the degree of conflict between two actions in a unified resource space. The time overlap rate characterizes the overlap intensity of two actions in the time dimension. The time overlap rate is formed by dividing the intersection length of the action effective action windows of the action to be handled and the currently effective action by the shorter of the two action effective action window lengths. The result is then cropped to the range of 0 to 1. Priority level deficit normalization is then performed. The quantity is calculated from the difference in rule priority levels between the action to be processed and the currently effective action. Specifically, the difference between the highest and lowest configurable rule priority levels in the system is taken as the maximum priority level difference. When the rule priority level of the action to be processed is lower than or equal to the rule priority level of the currently effective action, the absolute value of the difference between the two is divided by the maximum priority level difference to obtain the priority level deficit normalization quantity. When the rule priority level of the action to be processed is higher than the rule priority level of the currently effective action, the priority level deficit normalization quantity is 0. This processing can highlight the actual control logic that when the action to be processed is at a disadvantage in terms of priority level, its conflict risk should be further amplified.

[0048] After the aforementioned basic quantities are formed, a basic conflict intensity value is further formed. Specifically, for each action to be handled, it is matched one by one with each currently effective action in the current effective action set. First, a single-pair conflict contribution value is formed for each pair of actions. The single-pair conflict contribution value is obtained by multiplying the resource conflict coupling amount by the time overlap rate, and then multiplying it by the priority amplification factor formed by the priority level deficit normalization amount. The priority amplification factor is formed by adding 1 to the priority level deficit normalization amount. Subsequently, the single-pair conflict contribution values ​​formed by the action to be handled and all currently effective actions in the current effective action set are accumulated to obtain the basic conflict intensity value of the action to be handled. Since the resource conflict coupling amount, time overlap rate and priority level deficit normalization amount are all dimensionless, the single-pair conflict contribution value and the final basic conflict intensity value are also dimensionless. The larger the basic conflict intensity value, the less suitable the action to be handled is to be executed directly at the current moment.

[0049] When generating conflict resolution results, the basic conflict intensity value is compared with the preset conflict threshold determined by the preset equipment risk level mapping table. This is combined with the unreachability confidence level, the water supply security resource unit occupancy status, and the security blockade resource unit occupancy status to generate one of the following results: immediate execution, delayed observation, blocked execution, or suspended waiting. The preset equipment risk level mapping table is used to characterize the conflict threshold corresponding to different equipment risk levels. The higher the equipment risk level, the lower the preset conflict threshold is preferred. Preferably, the equipment risk levels are divided into three levels: the preset conflict threshold for high-risk equipment is preferably 0.20 to 0.35, the preset conflict threshold for medium-risk equipment is preferably 0.35 to 0.55, and the preset conflict threshold for low-risk equipment is preferably 0.55 to 0.75. The values ​​are based on the differences in the potential risk consequences of different equipment in the event of erroneous execution. For high-risk equipment, a smaller basic conflict intensity value is sufficient to trigger conservative handling; for low-risk equipment, a higher conflict tolerance is allowed.

[0050] The specific handling logic is as follows: When the water supply safety resource unit occupancy status or safety blockage resource unit occupancy status shows that the current effective action has already occupied the corresponding safety resource unit, and the action to be handled has a resource conflict coupling relationship with the safety resource unit, a suspension waiting result is directly generated. This logic is applicable to scenarios that require priority to maintain the protection status, such as water leakage protection, gas protection, and security blockage. In addition to the above-mentioned forced suspension situation, when the basic conflict intensity value is lower than the preset conflict threshold of the corresponding equipment risk level, and the status unreachability confidence is lower than the preset unreachability confidence threshold, an immediate execution result is generated. The preset unreachability confidence threshold is preferably taken as 0.4 to 0.6, and the value is based on the intersection interval of the status unreachability confidence distribution corresponding to the actual unreachable status and the recoverable short-term jitter status in the historical operation sample. When the basic conflict intensity value is lower than the preset conflict threshold of the corresponding equipment risk level, but the status unreachability confidence is not lower than the preset unreachability confidence threshold, an immediate execution result is generated. The delayed observation result indicates that the current conflict intensity has not yet reached the direct blocking condition, but the current feedback evidence from the device is still insufficient. Execution should be suspended and the system should continue to wait for new multi-source feedback evidence. When the basic conflict intensity value is not lower than the preset conflict threshold of the corresponding device risk level, and the rule priority level of the action to be handled is lower than the rule priority level corresponding to the current effective action, a blocking execution result is generated. When the basic conflict intensity value is not lower than the preset conflict threshold of the corresponding device risk level, and the rule priority level of the action to be handled is higher than the rule priority level corresponding to the current effective action, an immediate execution result is generated for the action to be handled, and the current effective action is marked as a historical handling event that forms a residual constraint state in subsequent S4. With this handling logic, high-priority safety actions and high-priority control actions can penetrate the occupancy state formed by low-priority comfort actions, which meets the actual control requirements of safety protection priority and high-priority rule priority in home automation networks.

[0051] Through the above processing, this step finally outputs the unreachability confidence level, the basic conflict intensity value, and the conflict resolution result. Among them, the unreachability confidence level serves as one of the suppression quantities when S4 constructs the replay qualification score. The basic conflict intensity value serves as an important input for S4 to construct the replay qualification score and evaluate whether the action is recoverable. The conflict resolution result serves as the direct source for S4 to extract historical resolution events and generate residual constraint state quantities. Compared with the existing technology that disperses the processing of equipment online status, rule priority level, and conflict determination, this implementation introduces the equipment expected effective return interval baseline, multi-source feedback evidence, equipment capability resource map, action effective action window, and preset equipment risk level mapping table in S3 to jointly quantify whether the current unreachability is current and whether the current conflict is current. Then, the two are uniformly mapped to the conflict resolution result. After adopting this method, it is possible to identify pseudo unreachability caused by short-term communication jitter, as well as real resource conflicts and target confrontation conflicts. It is also possible to directly incorporate the water supply security resource unit occupancy status and the security blockade resource unit occupancy status into the resolution logic.

[0052] In this implementation, S4, based on the conflict resolution results, unreachability confidence level, and basic conflict intensity value already formed in S3, further transforms the subsequent constraints caused by delayed observation, blocked execution, suspension waiting, and coverage by emergency actions into residual constraint state quantities that can be continuously calculated. At the current analysis time, it determines whether the scenario is still valid based on the rule condition expression and the current real-time state set. Combining the queuing action timeliness, the normalized result corresponding to the basic conflict intensity value, the normalized result corresponding to the residual constraint state quantity, the unreachability confidence level of the target device state, and the continuous occupation status of the target resource unit by the current valid action set, a replay qualification score is formed. Then, the current queuing action set is replayed, suspended, or discarded, and the queuing action replay screening result and linkage control result are output. At the same time, the sample quality gating is performed on the samples generated in this execution process and written into the stable running sample set in the historical running samples.

[0053] In this step, the current real-time state set aggregation, rule condition expression evaluation, time decay conversion, and sample quality gating are supporting existing technologies that provide input and update foundations for innovative processing. The extraction of residual constraint event sets, the division of hard-blocking and soft-blocking residual constraint events, the formation of residual constraint state quantities, the formation of replay qualification scores, and the screening of queuing action replays are the core innovative processing in this step. After adopting this processing method, the system's subsequent control of queuing actions is based on the combined effect of the current conflict handling results, the current resource occupancy status, the current scenario conditions, and the current device reachability status, which can effectively suppress the erroneous replay of outdated and high-risk actions after network recovery.

[0054] In this embodiment, the inputs to S4 include at least the conflict resolution result, the basic conflict intensity value, the state unreachability confidence level, the current queued action set, the current valid action set, the rule condition expression in the rule description data, and the current real-time state set formed by summarizing multi-source linkage operation data at the current analysis time. The conflict resolution result is output by S3 and includes at least the immediate execution result, the delayed observation result, the blocking execution result, and the suspension waiting result. The basic conflict intensity value is output by S3 and is a dimensionless quantity used to characterize the current conflict level of the action to be resolved. The state unreachability confidence level is output by S3 and is a dimensionless quantity in the range of 0 to 1, used to characterize the probability that the target device is currently communicating or reporting an anomaly. The current queued action set is output by S1 and used to characterize the current state at the current time. Actions that are pending execution, delayed observation, or suspended waiting and have not yet formed a clear linkage control result; the current effective action set is output by S2 and is used to characterize the action objects that are still in an effective occupancy state at the current time; the rule condition expression is obtained from the rule description data and is used to determine whether the triggering condition on which the corresponding queuing action depends is still valid at the current time; the current real-time state set is obtained by summarizing the multi-source linkage operation data at the current analysis time according to the target device identifier, target area identifier, and rule condition source identifier, which includes at least the current device state, current environment state, current safety alarm state, and current manual control state. The above input objects jointly participate in the formation of residual constraint state quantities, the determination of the still validity of the scene, the formation of replay qualification scores, and the writing back of samples in S4.

[0055] In this implementation, a set of residual constraint events is first extracted based on the conflict resolution results. Specifically, historical events that were delayed observation, blocked execution, suspended waiting, or covered by emergency actions are extracted from the conflict resolution results in the current analysis window. These events are then categorized according to the target equipment identifier and the resource unit occupied, forming a set of residual constraint events. Historical events corresponding to delayed observation results represent actions where the current conflict level has not yet reached the blocking threshold but the equipment feedback evidence is insufficient. Historical events corresponding to blocked execution results represent actions where the current conflict level has reached the blocking condition. Historical events corresponding to suspended waiting results represent actions where the current safety resource unit has been occupied by a high-priority action. Historical events covered by emergency actions represent actions where the original action was suppressed by a high-priority action on the water supply safety object or safety blockade object. After all the above historical events are uniformly included in the set of residual constraint events, all subsequent constraints can be continuously calculated on a unified event carrier.

[0056] After forming the set of residual constraint events, the residual constraint events are further divided into hard-blocking residual constraint events and soft-blocking residual constraint events based on the source rule type, equipment risk level, and resource unit type of the historical handling events. When the historical handling events originate from water leakage protection rules, gas protection rules, security blockade rules, electrical overload protection rules, or water supply safety blockade rules, and the equipment risk level is not lower than the preset equipment risk level threshold, the corresponding residual constraint event is determined as a hard-blocking residual constraint event. The preset equipment risk level threshold is preferably determined through a preset equipment risk level mapping table. When the equipment risk level is divided into 3 levels, the preset equipment risk level threshold is preferably set to 2 levels. When the historical handling events originate from comfort linkage rules, energy-saving linkage rules, ordinary equipment coordination rules, or general delay observation rules, the corresponding residual constraint events are determined as soft-blocking residual constraint events.

[0057] Before generating residual constraint state quantities, various intermediate quantities related to them are first formed. The severity coefficient of historical handling events is used to characterize the degree of inhibition of subsequent action recovery by different historical handling events. The severity coefficient is a dimensionless quantity in the range of 0 to 1. It is preferably determined by statistical analysis of the risk of false replay and the frequency of manual intervention before and after recovery of similar events in the stable operation sample set of historical operation samples. When the sample is insufficient, the severity coefficient of hard blocking residual constraint events is preferably 0.85 to 1.00, the severity coefficient of soft blocking residual constraint events corresponding to ordinary blocking execution is preferably 0.55 to 0.80, and the severity coefficient of soft blocking residual constraint events corresponding to delayed observation is preferably 0.30 to 0.55. The coupling coefficient between historical handling events and target resource units is used to characterize the degree of direct correlation between the historical handling event and the target resource unit. It is also a dimensionless quantity in the range of 0 to 1. When the historical handling event and the target resource unit correspond completely, the coupling coefficient is preferably 1. When the source units belong to the same type of control target but are not the same object, the coupling coefficient is preferably 0.6 to 0.8; when there is an indirect linkage between the historical handling event and the target resource unit, the coupling coefficient is preferably 0.2 to 0.5. The residual constraint time constant of the historical handling event is used to characterize the natural decay scale of the constraint strength of the historical handling event. It is a dimensional time quantity with the unit being seconds. Preferably, the residual constraint time constant is determined by the median of the actual recovery time between the generation and the allowed recovery of the same type of historical handling event in the stable operation sample set in the historical operation sample. When the sample is insufficient, the residual constraint time constant is preferably 120s to 600s for water leakage protection events, preferably 300s to 1800s for gas protection events, preferably 300s to 3600s for security blockage events, and preferably 30s to 300s for ordinary comfort blocking events. The above values ​​are all derived from the difference in the recovery rhythm of protective actions and comfort actions in the actual automation system.

[0058] The release evidence factor is used to characterize whether the historical handling event has an evidentiary basis for releasing the constraint at the current analysis time. The release evidence factor is a dimensionless quantity ranging from 0 to 1. The inputs to the release evidence factor include at least explicit release trigger data, reverse control command execution confirmation data, continuous stable operation data, and emergency alarm disappearance data. Explicit release trigger data characterizes that the user or high-priority rule has explicitly released the historical handling event; reverse control command execution confirmation data characterizes that the control command in the opposite direction to the historical handling event has been successfully issued and confirmed; continuous stable operation data characterizes that the relevant equipment or resource objects have remained in a stable state for a period of time; and emergency alarm disappearance data characterizes that the original alarm that triggered the safety-related historical handling event has been released. The release evidence factor is formed using a weighted combination method. For hard-blocking residual constraint events, the release evidence factor is allowed to exceed 0.5 only if either explicit release trigger data or emergency alarm disappearance data is satisfied. When only continuous stable operation data is available but there is no explicit release trigger data or emergency alarm disappearance data, the upper limit of the release evidence factor is preferably limited to 0.3. For soft-blocking residual constraint events, the release evidence factor can be formed by combining continuous stable operation data, reverse control command execution confirmation data, and explicit release trigger data according to preset weights. Preferably, the weight corresponding to explicit release trigger data is 0.4 to 0.6, the weight corresponding to reverse control command execution confirmation data is 0.2 to 0.3, the weight corresponding to continuous stable operation data is 0.1 to 0.3, and the weight corresponding to emergency alarm disappearance data is 0.3 to 0.5. The above weight range is derived from the differences in the directness and credibility of different release evidence in releasing historical constraints.

[0059] After the aforementioned intermediate quantities are generated, residual constraint state quantities are further generated. Specifically, for each device's set of residual constraint events on each resource unit, the residual constraint contribution of each historical handling event at the current analysis time is calculated one by one. The residual constraint contribution of a single historical handling event is jointly determined by the severity coefficient, the coupling coefficient with the target resource unit, the degree of remaining time decay, and the complement value of the release evidence factor. Preferably, the interval between the current analysis time and the effective time of the historical handling event is first calculated, and then the interval is converted into a time decay ratio relative to the residual constraint time constant of the historical handling event. The time decay ratio preferably adopts a piecewise linear decay method, that is, from 0 to 1 after the historical handling event takes effect. Within a single residual constraint time constant, the residual constraint decays linearly from 1 to 0, and then returns to 0 after exceeding one residual constraint time constant. Next, the severity coefficient, coupling coefficient, and time decay ratio are multiplied together, and then multiplied by the complement of the release evidence factor to obtain the residual constraint contribution of the historical event to the target resource unit at the current analysis time. Finally, on the same device and the same resource unit, the maximum residual constraint contribution of all historical events is taken to obtain the residual constraint state quantity of the device on that resource unit. The residual constraint state quantity is a dimensionless quantity within the range of 0 to 1. After aggregating the maximum values, the strongest constraint event on the same resource unit can determine whether recovery is currently allowed. The numerical logic is consistent with the actual control logic.

[0060] After generating the residual constraint state variables, the validity of the scenario is further determined based on the rule condition expressions corresponding to each queuing action and the current real-time state set. The current real-time state set is formed by summarizing multi-source linkage operation data at the current analysis time, and includes at least the current device state, current environment state, current safety alarm state, and current manual control state. For threshold-type conditions, the current real-time state value is compared with the corresponding threshold interval. If it falls within the threshold interval, the condition is valid; if it exceeds the threshold interval, the condition is invalid. For event-type conditions, whether the current event is still within the valid duration window of the rule is taken as the result of the condition being valid. For combined conditions, each item is evaluated according to the AND, OR, and sequential relationships in the rule condition expression. When the expression as a whole is valid, the scenario is still valid; otherwise, the scenario is still invalid. When the unreachability confidence of the current state source device on which the rule condition expression depends is not lower than the preset unreachability confidence threshold, the corresponding condition is directly determined to be invalid. After this processing, the automatic recovery action is always based on the premise that the current scenario conditions are truly valid.

[0061] After determining that the scenario is still valid, a queuing action timeliness is further formed. The queuing action timeliness is used to characterize the degree to which the queuing action still has recovery value at the current analysis time. It is a dimensionless quantity in the range of 0 to 1. First, the waiting time between the queuing action generation time and the current analysis time is calculated. Then, the allowable retention time corresponding to the queuing action is determined. The allowable retention time is preferably determined by the smaller value of the remaining time of the rule's effective duration window and the statistical value of the historical action duration plus twice the baseline of the expected effective return interval of the device. Finally, the ratio of the waiting time to the allowable retention time is converted into a time decay ratio, and the queuing action timeliness is formed by subtracting the time decay ratio from 1. When the waiting time exceeds the allowable retention time, the queuing action timeliness is set to 0. This process can ensure that the queuing action gradually loses its recovery value as the waiting time increases.

[0062] When generating replay eligibility scores, the basic conflict intensity value and residual constraint state quantity are first normalized. The normalization result for the basic conflict intensity value is preferably obtained by dividing the basic conflict intensity value by a preset conflict threshold corresponding to the equipment risk level, and the result is cropped to the range of 0 to 1. Since the residual constraint state quantity is already in the range of 0 to 1, it is directly used as the normalized result corresponding to the residual constraint state quantity. Then, the scenario's still-validity is used as the scenario release factor, set to 1 if valid and 0 if invalid. The queuing action timeliness is used as the timeliness release factor. Finally, the normalized result corresponding to the basic conflict intensity value, the normalized result corresponding to the residual constraint state quantity, the target equipment state unreachability confidence level, and the current valid action set are considered together. The continuous occupation results of the resource units corresponding to the queued actions are used as a suppression factor. The continuous occupation results of the resource units corresponding to the queued actions in the current set of effective actions are preferably formed in a binary manner: when there are still actions in the current set of effective actions occupying the same target resource unit and the rule priority level is not lower than that queued action, the value is 1; otherwise, the value is 0. The replay qualification score is formed by multiplying the scene release factor and the timeliness release factor, and then multiplying by the product of the supplementary values ​​of each suppression factor. After this processing, the replay qualification score decreases monotonically as the scene is still effective, the timeliness of the queued actions decreases, the basic conflict intensity value increases, the residual constraint state quantity increases, the state unreachability confidence increases, and the continuous occupation degree of the current effective actions increases.

[0063] When performing replay, suspension, or discard on the current set of queued actions, the system first determines whether there are hard blocking residual constraints on the resource unit corresponding to the queued action. If hard blocking residual constraints exist, the queued action is discarded directly. If no hard blocking residual constraints exist, the replay eligibility score is compared with a preset replay threshold. The preset replay threshold is preferably determined by the intersection of the distribution of replay eligibility scores corresponding to actually successfully recovered actions and erroneously recovered actions in the stable operation sample set of historical operation samples. When the sample is insufficient, the preset replay threshold is preferably between 0.55 and 0.75. When the replay eligibility score is not lower than the preset replay threshold and there are no hard blocking residual constraints on the resource unit corresponding to the queued action, the system will proceed accordingly. When the queueing action ends, the queuing action is replayed. If the replay eligibility score is lower than the preset replay threshold, the scenario is still valid, and there are no hard blocking residual constraints on the resource unit corresponding to the queuing action, the queuing action is suspended. If the scenario is still invalid, the queuing action timeliness is lower than the preset failure threshold, or there are hard blocking residual constraints on the resource unit corresponding to the queuing action, the queuing action is discarded. The preset failure threshold is preferably between 0.10 and 0.20. The value is based on the statistical results that the recovery value decreases significantly when the action waiting time is close to the end of the allowed retention time. Through the above processing logic, S4 finally forms the queuing action replay screening result and linkage control result.

[0064] After obtaining the queuing action replay screening results and linkage control results, sample quality gating is further implemented, and samples that meet the gating conditions are written into the stable operation sample set in the historical operation samples. The sample quality gating includes at least no gateway-level abnormal interruption, link quality higher than the preset link quality threshold, device execution feedback consistency higher than the preset consistency threshold, and no manual forced intervention. No gateway-level abnormal interruption is determined by continuous recording of network door-beats. The preset link quality threshold is preferably determined by the intersection of the stable operation sample set and abnormal samples in the link quality distribution in the historical operation samples. When the sample is insufficient, a value of 0.65 is preferred. The preset consistency threshold is preferably... The quartiles are determined by the distribution of device execution feedback consistency in the stable operation samples. When the sample is insufficient, 0.75 is preferred. When the sample quality gating condition is met, the actual effective return interval, historical action duration, actual recovery time and device execution feedback consistency formed during this execution are written into the stable operation sample set in the historical operation samples. This is used for subsequent updates such as S1 updating the baseline of the expected effective return interval of the device, S2 updating the statistical value of the historical action duration, and S4 updating the residual constraint time constant. Samples that do not meet the sample quality gating condition are not written into the stable operation sample set, but are only kept in the original operation log for fault diagnosis and manual review.

[0065] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0066] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0067] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for controlling the linkage of home scene devices based on the Internet of Things, characterized in that Includes the following steps: Acquire the multi-source linkage operation data of the target home automation network within the current analysis window, perform time base correction, message folding, state edge extraction and action window merging on the multi-source linkage operation data, construct linkage action instance identifiers, generate linkage event fragments, establish the baseline of the expected effective return interval of the device, and extract the current queued action set. Obtain device description data and rule description data, identify the execution resource unit mapping and effective action window for linkage action instance identification, construct device capability resource graph, and generate current effective action set and conflict candidate relationship set; By combining the baseline of expected effective return interval of equipment, equipment capability resource map, current effective action set, conflict candidate relationship set and multi-source feedback evidence, the credibility of state unreachability and basic conflict intensity value are calculated, and conflict resolution results are generated. Based on the conflict resolution results, residual constraint state variables are generated. The continued validity of the scenario is determined according to the rule condition expression and the current real-time state set. A replay qualification score is constructed, and the current queuing action set is replayed, suspended, or discarded. Queuing action replay screening results and linkage control results are formed. Samples that meet the preset sample quality gating conditions are written into the stable running sample set in the historical running samples.

2. The method for controlling the linkage of home scene devices based on the Internet of Things according to claim 1, wherein: Multi-source coordinated operation data includes automated rule triggering data, manual control data, control command issuance data, equipment response data, equipment status reporting data, equipment link quality data, queuing action data, and equipment execution feedback data; The linkage action instance identifier is generated by combining the trigger source identifier, target device identifier, target action identifier, rule identifier, and window sequence number, and is used to represent the unique execution instance of the same linkage action within the current analysis window.

3. The method for controlling the linkage of home scene devices based on the Internet of Things according to claim 2, wherein: Generating linkage event fragments includes: performing unified time base correction, duplicate message folding, state edge extraction, and action window merging on the multi-source linkage operation data; converting control command edge response quantities, device response consistency quantities, state switching quantities, and device execution feedback change quantities into dimensionless quantities; combining the dimensionless quantities according to preset weights to form event anchoring strengths; and merging candidate event anchor points with event anchoring strengths exceeding preset event anchoring thresholds and having the same linkage action instance identifier into linkage event fragments.

4. A method for controlling the linkage of home scene devices based on the Internet of Things according to claim 1, characterized in that: Establish the baseline for the expected effective return interval of the equipment, including: for periodically reporting equipment, using the median effective reporting interval within the stable operating sample set in the historical operating samples to establish the baseline for the expected effective return interval; for event-driven equipment, using the median response interval between the issuance of control commands and the effective response or effective status report to establish the baseline for the expected effective return interval; for execution confirmation equipment, using the median confirmation interval between the issuance of control commands and the equipment execution feedback entering a stable state to establish the baseline for the expected effective return interval; when the number of samples in the stable operating sample set is lower than the preset minimum sample threshold, using the equipment configuration default period, the protocol recommended response window, or the reference confirmation window of similar equipment to establish the initial baseline for the expected effective return interval.

5. The method for controlling the linkage of home scene devices based on the Internet of Things according to claim 1, wherein: The construction of the device capability resource map includes: mapping linkage action instances to corresponding resource units based on device capabilities, device object relationships, and a preset action semantic mapping table in the device description data, and generating an action resource occupancy vector; determining the mutual exclusion coefficient, concurrency coefficient, and priority coverage coefficient of each resource unit based on a preset resource constraint table; and setting the occupancy value of linkage actions of non-competitive limited power supply objects to zero in the limited power supply resource unit.

6. The method for controlling the linkage of home scene devices based on the Internet of Things according to claim 5, characterized in that: The process generates a set of currently valid actions and a set of candidate conflict relationships, including: determining the effective action window for each linked action instance based on the start and end times of the linked event fragments and the historical action duration statistics obtained from the stable operation sample set in the historical operation samples; when there is a stable point for device execution feedback, the moment when the device execution feedback enters a stable state is taken as the end time; when there is no stable point for device execution feedback, the estimated time corresponding to the historical median duration of the same action of the same device is taken as the end time; and then generating the set of currently valid actions and a set of candidate conflict relationships based on the time overlap relationship between the effective action windows, the rule priority relationship, the recovery strategy relationship determined by the recovery strategy in the rule description data and the recovery strategy relationship determined by the prohibition recovery flag, and the security blocking relationship.

7. A method for controlling the linkage of home scene devices based on the Internet of Things according to claim 1, characterized in that: The reliability of the unreachable status is calculated, including: extracting the silence duration of the target device since the most recent valid response, valid status report or valid execution feedback, current link quality, response consistency and device execution feedback consistency; Link quality is normalized from received signal strength, link quality indicator value, and packet loss rate; response consistency is normalized from the ratio of valid responses to expected responses within the observation window; and device execution feedback consistency is normalized from the degree of consistency between device execution feedback records and the expected direction of target actions. The unreachability confidence level is determined by the degree of deviation of the silence duration from the baseline of the expected effective response interval of the device, the degree of link quality degradation, the degree of decrease in response consistency, and the degree of decrease in device execution feedback consistency. For air conditioning and fresh air systems, when the current observation window length is less than the preset minimum thermal response time threshold, temperature data, humidity data, and air quality data are only used to determine the validity of the scenario and do not participate in the formation of consistent feedback from the equipment.

8. The method for controlling the linkage of home scene devices based on the Internet of Things according to claim 7, wherein: The calculation of the basic conflict intensity value and the generation of conflict resolution results include: for each action to be resolved, matching it one by one with each currently effective action in the current effective action set; extracting the resource conflict coupling amount and time overlap rate from the equipment capacity resource map and the effective action window of each linkage action instance; and obtaining the priority level deficit normalization amount by normalizing the difference between the rule priority level of the action to be resolved and the rule priority level of the currently effective action. The resource conflict coupling amount, time overlap rate, and priority level deficit normalization amount are combined to form the basic conflict intensity value of the action to be resolved. The basic conflict intensity value is then compared with the preset conflict threshold determined by the preset equipment risk level mapping table, and combined with the status unreachability confidence, water supply safety resource unit occupancy status, and safety blockade resource unit occupancy status, a conflict resolution result is generated from one of the following: immediate execution result, delayed observation result, blocked execution result, or suspended waiting result.

9. A method for controlling the linkage of home scene devices based on the Internet of Things according to claim 1, characterized in that: The residual constraint state is generated based on the conflict resolution results, including: extracting historical resolution events that were delayed observation, blocked execution, suspended waiting, or covered by emergency actions from the conflict resolution results to form a set of residual constraint events; and classifying the residual constraint events into hard blocking residual constraint events or soft blocking residual constraint events based on the source rule type, equipment risk level, and occupied resource unit type of the historical resolution events. Specifically, when a historical incident originates from a water leakage protection rule, gas protection rule, security blockade rule, electrical overload protection rule, or water supply safety blockade rule, and the equipment risk level is not lower than the preset equipment risk level threshold, the corresponding residual constraint event is identified as a hard-blocking residual constraint event. Then, based on the severity coefficient of the historical incident, the coupling coefficient with the target resource unit, the effective time, the residual constraint time constant, and the release evidence factor formed by combining explicit release trigger data, reverse control command execution confirmation data, continuous stable operation data, and emergency alarm disappearance data according to preset weights, the residual constraint state quantity of the equipment on the corresponding resource unit is generated.

10. The method for controlling the linkage of home scene devices based on the Internet of Things according to claim 9, wherein: The process of performing queue action replay filtering includes: determining the validity of a scenario based on the rule condition expressions corresponding to each queue action and the current real-time state set formed by summarizing multi-source linkage operation data at the current analysis time. Specifically, for threshold-type conditions, the current real-time state value is compared with the corresponding threshold interval to obtain the result of the condition being valid. For event-type conditions, whether the current event is within the rule validity duration window is taken as the result of the condition being valid. For combined conditions, the result of the scenario still being valid is obtained by evaluating each item according to the AND, OR, and sequential relationships. The queuing action timeliness is determined based on the time decay relationship between the queuing action generation time and the current analysis time. A replay eligibility score is then formed by combining the queuing action timeliness, the normalized result corresponding to the basic conflict intensity value, the normalized result corresponding to the residual constraint state quantity, and the unreachability confidence of the target device state. When the replay eligibility score is not lower than the preset replay threshold and the resource unit corresponding to the queuing action does not have hard blocking residual constraints, the queuing action is replayed. When the replay eligibility score is lower than the preset replay threshold, the scenario is still valid, and the resource unit corresponding to the queuing action does not have hard blocking residual constraints, the queuing action is suspended. When the scenario is still invalid, the queuing action timeliness is lower than the preset failure threshold, or the resource unit corresponding to the queuing action has hard blocking residual constraints, the queuing action is discarded. Sample quality gating is applied to the actual effective return interval, historical action duration, actual recovery time, and device execution feedback consistency. The sample quality gating includes at least no gateway-level abnormal interruption, link quality higher than the preset link quality threshold, and device execution feedback consistency higher than the preset consistency threshold.

Images