Mine safety situation cooperative sensing method based on sensor network

Abstract

The invention provides a mine safety situation cooperative sensing method based on a sensor network, which relates to the technical field of intelligent sensing, and comprises the following steps of: constructing a shared active tuple space, defining a situation clue tuple, a pheromone tuple and a heartbeat tuple format, deploying a rule engine in a sensor node, and establishing a situation clue tuple space; semantic packaging and cooperative processing of the sensing data are realized; the active tuple space executes matching and cooperative calculation according to the association keys, and a high-order situation tuple is generated and written back to the space; sniffing and response among nodes are supported through a periodic diffusion and attenuation mechanism of pheromone tuples; the nodes periodically write heartbeat tuples and carry out information entropy detection, and when abnormal or significant variation is detected, priority matching and resource scheduling are triggered; based on priority arbitration and a comprehensive scoring mechanism, node responsibilities of coordinators are dynamically adjusted, and cooperative sensing among multiple nodes and intelligent judgment and self-adaptive response of mine safety situations are achieved.

Description

Technical Field

[0001] This invention relates to the field of intelligent sensing technology, specifically to a collaborative sensing method for mine safety situation based on sensor networks. Background Technology

[0002] The mine environment has extremely high requirements for safety monitoring; mines use a variety of sensors to collect gas, airflow, pressure and temperature parameters in real time, and wireless sensor networks and wired monitoring systems are widely deployed to achieve continuous observation. The industry trend is towards the integration of edge computing and intelligence, which puts forward higher requirements for low-power reliable communication and timely analysis. Existing solutions typically use centralized data platforms and message buses to report and store sensor data. The acquisition end usually performs filtering and threshold triggering to reduce the uplink load. The upper layer uses time-series databases and stream processing for aggregation analysis and alarms. Some studies introduce probabilistic fusion to improve the accuracy of multi-source information fusion, which facilitates unified management and auditing. Existing systems suffer from trade-offs between real-time performance and scalability, leading to latency and processing bottlenecks. Centralized processing generates single-point loads and limits local rapid response capabilities. Multi-source data fusion lacks evidence chains and priority awareness, resulting in ambiguous decision-making. In resource-constrained environments, dynamic scheduling and energy balancing support for abnormal situations are insufficient. Summary of the Invention

[0003] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a collaborative perception method for mine safety situation based on sensor networks. This method solves the problems mentioned in the background, such as delays and processing bottlenecks caused by trade-offs, single-point loads generated by centralized processing and limitations on local rapid response capabilities, ambiguity in decision-making due to the lack of evidence chains and priority perception in multi-source data fusion, and insufficient support for dynamic scheduling and energy balance in abnormal situations under resource-constrained environments.

[0004] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: a collaborative perception method for mine safety situation based on sensor networks, comprising the following steps: S1. Construct and initialize a shared active tuple space, and define standardized situational cue tuple, pheromone tuple, and heartbeat tuple formats in the active tuple space. The situational cue tuple format contains association keys, and a rule engine is deployed in each sensor node to perform local perception and semantic encapsulation. A coordinator node is also deployed in the active tuple space to manage matching tasks and resource scheduling within the active tuple space. S2, after any sensor node completes local perception, the rule engine encapsulates the raw readings according to the situational cue tuple format and performs a write operation to write the encapsulated situational cue tuple into the active tuple space, while triggering the matching task triggered by the write operation inside the active tuple space. S3 is a matching task triggered by the active tuple space response write operation. It searches for other tuples in the space that match the association key of the just-written situational clue tuple, and performs judgment according to the collaborative computing rules to generate a judgment result. The collaborative computing rules include threshold comparison, change rate criterion and tuple consistency judgment. S4, the active tuple space writes the judgment result back to the active tuple space in the form of a new higher-order situation tuple, and records the source tuple reference in the higher-order situation tuple. The source tuple reference points to the tuple that generated the judgment result. S5. After the high-order situation tuple is written back, the active tuple space is written with the pheromone tuple corresponding to the high-order situation tuple, and the pheromone tuple enters a periodic diffusion and decay cycle for other sensor nodes to sniff and respond. S6, during periodic operation, each sensor node periodically writes a heartbeat tuple to the active tuple space and periodically detects the information entropy of the local data stream; when the information entropy detects a significant change or the heartbeat tuple detects an anomaly, the active tuple space increases the priority of the corresponding tuple and drives priority matching and resource scheduling within the active tuple space.

[0005] Preferably, the coordinator node identifier is registered in the active tuple space, and persistent storage and indexing are allocated; the format of situational clue tuples, pheromone tuples, and heartbeat tuples are defined within the space, and the field set and field semantics are specified; the field set includes type, location, value, confidence, timestamp, source identifier, and association key; the type field uses integer encoding, ranging from 1 to 255, to identify the sensor category and observation category; the location field uses mine coordinate strings or roadway number strings to accurately locate the observation point; the value field uses floating-point numbers or enumeration encoding to represent the measurement result; the confidence field ranges from 0 to 1, with an initial value of 0.8, for subsequent confidence aggregation calculation; the timestamp field uses millisecond-level Unix timestamps for event sorting and timeline reconstruction; the source identifier field uses a unique node identifier, with a length of 16 bytes, for tracing; the association key field uses a 64-bit hash value, and 1024 hash index buckets are established for the association key field to support selective scanning and fast matching; priority labels, emergency identifier writing rules, and lifecycle parameters are specified for the situational clue tuples; the priority label uses... Integers from 0 to 10, with a default value of 5, are used for visibility control and processing scheduling. The emergency flag uses a Boolean value and is jointly determined with the priority label to determine the visibility level and retrieval permissions. The lifecycle parameter represents the survival time: 600 seconds for situational cue tuples, 300 seconds for pheromone tuples, and 120 seconds for heartbeat tuples. Upon expiration, it is automatically cleaned up by the active tuple space and an immutable audit entry is written for traceability retrieval. Access levels are defined for visibility control: 0 for public, 1 for region-restricted, and 2 for node-restricted. An access control list containing a sequence of allowed-to-read node identifiers is written into the tuple during writing for retrieval authorization verification. Before the write operation, tuple fields are serialized using fixed bytes. Visibility attributes and access control attributes are written simultaneously during the write operation, maintaining a strictly monotonically increasing timestamp order with the heartbeat tuple write sequence to confirm the write node's activity and the write timeline through the heartbeat sequence. After initialization, the tuple template, field semantics, priority rules, lifecycle parameters, and index metadata are written to the metadata directory to support subsequent matching tasks, selective scanning, and resource scheduling decisions for coordinator nodes.

[0006] Preferably, within each sensor node, when the rule engine starts, it first loads the situational cue tuple template and binds the field set definition and field semantics in the template. After the sensor node collects raw readings, the rule engine immediately performs type identification, unit conversion, and validity verification on the raw readings. Based on the sensor type identifier, it matches and verifies the byte length and data format of the input data, and after confirming that there are no errors, it enters the value normalization stage. According to the corresponding sensor category, it calls the conversion table to uniformly convert the reading values ​​to international standard units. In the validity verification stage, the rule engine performs a judgment based on a preset threshold range. If the reading value exceeds the allowed range, an anomaly label is generated and written to the emergency identifier field of the situational cue tuple. After the data passes through the above steps, the rule engine encapsulates the type, location, value, confidence level, timestamp, source identifier, and association key fields sequentially according to the field set. The location field is determined by the sensor node's... Fixed coordinates and lane numbers are generated together; the initial value of the confidence field is derived from the status code, with the conversion rule being a linear mapping from status code 0 to 10 to confidence level 0 to 1; the timestamp field is generated by the local clock; the source identifier field contains a unique identifier for the sensor node, used to track the data source; the association key field uses a 64-bit value obtained by hashing the concatenated string of the location, timestamp, and type fields, used for indexing and locating in subsequent matching tasks; after the situational clue tuple is encapsulated, the rule engine adds a priority label and lifecycle parameter to the tuple and temporarily stores it in the local cache for 0.5 seconds to wait for the write window signal; when the active tuple space opens the write window, the rule engine performs the write operation, writing the structured tuple to the active tuple space and generating a write event record; after the write is completed, the rule engine records the summary information of the encapsulated tuple and the write time to the node log for subsequent auditing and tracing.

[0007] Preferably, when the rule engine initiates a write request, it performs final serialization of the tuple according to the situational cue tuple format and calculates the visibility and access control attributes; it creates a write transaction locally and generates a write event record, which includes the writer node identifier, write timestamp, write transaction identifier, and write tuple summary; and it submits the write event to the receiving interface of the live tuple space, simultaneously writing the visibility and access control attributes to the tuple metadata area to limit the retrieval scope; to ensure consistency of the write order, the write sequence number is bound to the local heartbeat tuple write sequence before executing the write transaction. The binding rule is that the write sequence number must be equal to the last value of the heartbeat sequence plus 1. After the live tuple space receives the write event, the coordinator node verifies the consistency between the write transaction and the heartbeat sequence. If the verification is successful, the write tuple is persisted and a matching task is triggered. Otherwise, the write transaction is marked as conflicting, the write transaction is rolled back, and the reason for the conflict is returned to the writer. After the write is completed, the complete information of the write event is recorded in the node log and an immutable audit record is written in the audit log of the live tuple space. After the write is completed, the write event triggers the matching task scheduler inside the live tuple space to start selectively scanning the associated key index to locate the matching tuple.

[0008] Preferably, when the active tuple space receives a write event, the associated key value of the written tuple is read and the index bucket number is calculated, with the calculation rule being the hash value of the associated key modulo 1024; tuple identifiers are read in the corresponding hash bucket in linked list order, and tuple header metadata is loaded sequentially for each candidate tuple to verify the timestamp and access level; during loading, a preliminary screening rule is executed on the candidate tuples, which includes a time window limit of 60 seconds, access level allowing matching identifiers, and a confidence threshold of 0.5; candidate tuples that pass the preliminary screening are collected into a matching tuple set and submitted to the collaborative computing processing queue; the collaborative computing processor evaluates the matching tuple set item by item according to the threshold comparison rule, the rate of change criterion, and the tuple consistency judgment; threshold The comparison rule is set to compare the target observation value with a threshold A, where the threshold A is 100 units; the rate of change criterion is set to trigger the rate of change condition when the change amplitude is greater than 10 units in the past minute; the tuple consistency judgment is set to have at least 3 matching tuples within a 60-second time window, and the range of the observation values ​​of these tuples does not exceed 5 units and the aggregate confidence value is not less than 0.7; when the collaborative computing processor generates multiple judgment results in parallel, it sorts them in descending order of confidence and places the judgments marked as urgent as true at the top, with the sorting rule being a confidence weight of 0.6, an emergency weight of 0.3, and a new / old weight of 0.1; after sorting, the final judgment is generated by weighted average and submitted to the high-order situation tuple generation function to complete the write-back process.

[0009] Preferably, after the collaborative computing processor outputs the final judgment result, a new tuple instance is created based on the judgment result, and a source tuple identifier set, timeline record, and confidence aggregation value are added to it during the generation phase. The source tuple identifier set consists of unique identifiers of all matching tuples participating in the collaborative computing, used to ensure traceability integrity. The timeline record field includes three items: start time, end time, and generation time, in milliseconds, used to reconstruct the situation evolution process. The confidence aggregation value is obtained by averaging the confidence scores of participating tuples according to weights, with the weight allocation rule being that the closer the time distance, the higher the weight, and the calculation range is limited to the past 120 seconds. A trigger identifier field is added before the high-order situation tuple is written back, and the trigger identifier is... A Boolean value of true indicates that the higher-order situation tuple will drive the pheromone release operation. The higher-order situation tuple is serialized and written to the persistent region of the active tuple space. During writing, an immutable audit entry is generated, containing the higher-order situation tuple identifier, the set of source tuple identifiers, the generation timestamp, the confidence aggregate value, and the writing node identifier. The audit entry is stored using a hash signature and recorded in the audit index table to ensure consistency for subsequent retrieval and verification. After writing back, the higher-order situation tuple is marked as a high-priority tuple in the active tuple space. The entire generation and write-back process ensures the integrity and traceability of the evidence chain for the higher-order situation tuple and provides a unified data foundation for subsequent situation propagation and collaborative perception.

[0010] Preferably, when a higher-order situation tuple is written back to the active tuple space with a true trigger flag, a pheromone tuple is immediately created and written to the diffusion queue. The pheromone tuple contains a pheromone identifier, intensity value, diffusion radius, attenuation coefficient, source tuple identifier, and timestamp. The initial pheromone intensity is set to 1.0, the diffusion radius to 50 spatial units, and the attenuation coefficient to 0.05. The pheromone tuple enters a periodic processing loop with a processing cycle of 5 seconds. At the beginning of each cycle, the delivery intensity contribution is calculated based on the current intensity value and diffusion radius and diffused to surrounding spatial units. The diffusion operation is performed according to spatial index partitions, receiving an intensity contribution value once per spatial unit within the diffusion radius and recording the contribution source identifier locally. At the end of each cycle, the pheromone intensity is decreased according to the attenuation coefficient, and when the intensity drops to 0... The system automatically clears pheromone tuples and writes cleanup records to the audit log table. Each sensor node distributed in the active tuple space periodically reads the pheromone intensity in its local spatial unit. When the read intensity exceeds the node response threshold of 0.6, the sensor node executes a response behavior. The response behavior includes adjusting the sampling frequency to twice the original sampling frequency, performing a local cross-validation task, and writing support tuples into the active tuple space. When the local cross-validation passes, the sensor node releases the same type of pheromone according to the positive feedback rule. The initial intensity of the release is 0.5, and the diffusion radius is 30 spatial units, so as to achieve the pheromone superposition effect. The pheromone periodic processing cycle continues until all active pheromone tuples are decayed to zero and cleared, so as to ensure that spatial resources are effectively utilized and that the collaborative sensing behavior remains continuous and efficient.

[0011] Preferably, during the periodic operation of the active tuple space, each sensor node writes a heartbeat tuple to the active tuple space at a fixed interval of 30 seconds. The heartbeat tuple contains a node identifier, timestamp, status code, and consecutive sequence number to reflect the node's activity status. The coordinator node continuously monitors the heartbeat tuple sequence and calculates the node activity index based on the sequence interval. A node with an activity index below 0.5 is identified as an abnormally silent node. Simultaneously, each sensor node performs information entropy detection locally, analyzing the data stream feature distribution over the past 120 seconds using a sliding window approach. If the calculated information entropy value increases by more than 0.3 compared to the previous period, it is determined to be a significant mutation event. When a significant mutation event occurs, the node immediately generates a situational cue tuple and writes a priority label into the tuple. A tag value of 10 is used to ensure that tuples are prioritized for processing within the active tuple space. After receiving a high-priority tuple triggered by entropy change, the coordinator node immediately triggers matching and resource scheduling operations based on the priority tag and records the trigger source in the scheduling table. When heartbeat monitoring identifies that no heartbeat tuples have been received for two consecutive cycles, the coordinator node marks the corresponding node as a high-interest node and raises the processing priority of its related tuples to the highest level. The active tuple space reorders the tuple processing queue according to the priority parameter, and high-priority tuples are given priority in allocating computing and matching resources in the next scheduling cycle. The joint operation of heartbeat and information entropy detection ensures a dynamic balance between sensor node activity, data stability, and spatial task scheduling order, thereby maintaining the continuity and real-time nature of the mine safety situation awareness process.

[0012] Preferably, when a regional silence pattern is detected or a parallel high-entropy event is identified by information entropy detection, the coordinator node initiates a resource scheduling process. The resource scheduling process first calculates a composite score, which is a weighted sum of three parameters: entropy magnitude, confidence level, and influence radius. The entropy magnitude has a weight of 0.5, the confidence level has a weight of 0.3, and the influence radius has a weight of 0.2. When the composite score is higher than 0.7, a key resource scheduling instruction is triggered. The coordinator node, according to the scheduling instruction, designates a moving node or allocates additional computing resources. The moving node adjusts the center point offset of the sampling area in the spatial coordinate system by 20 units to approach the abnormal area. When the composite score is lower than 0.4, the coordinator node only performs an observation frequency increase operation, shortening the sensor sampling period in the affected area from 30 seconds to 15 seconds. During scheduling execution, the coordinator node adjusts the communication bandwidth according to the score result. The allocation ratio is as follows: high-scoring areas receive 60% of the total bandwidth, and the remaining bandwidth is equally distributed among the remaining nodes according to their weights. Simultaneously, the coordinator node dynamically adjusts the allocation of computation time slices, prioritizing resources for matching tasks in high-scoring areas to ensure that critical events can be analyzed first. After resource scheduling is completed, the coordinator node generates a scheduling result tuple in the active tuple space. This tuple contains a list of target nodes, the type of allocated resources, the execution start time, and the expected duration. Once the scheduling result tuple is written, a scheduling record entry is created in the audit log table, along with the identifier of the scheduling initiating node and the execution signature, to ensure the traceability of the resource allocation process. The coordinator node periodically evaluates the scheduling effect and generates a responsibility change tuple when necessary to update the node's responsibility scope and effective time, thereby maintaining resource balance and computational stability during the mine safety situation collaborative perception process.

[0013] Preferably, when processing tuples in the active tuple space, each tuple has a priority score calculated. This priority score is derived from a weighted combination of four parameters: information entropy trigger flag, confidence aggregation value, pheromone intensity, and diffusion radius. The information entropy trigger flag has a weight of 0.4, the confidence aggregation value has a weight of 0.3, the pheromone intensity has a weight of 0.2, and the diffusion radius has a weight of 0.1. After the priority score is calculated, the coordinator node arbitrates the parallel decision tasks based on the score. The arbitration rule is that tasks with higher scores have priority in occupying computing resources and communication bandwidth. In resource-constrained situations, the coordinator node dynamically determines the communication bandwidth allocation ratio, computing time slice allocation, and energy allocation based on the priority score. High-priority tuples receive 70% of the total bandwidth, 60% of the total computing time slices, and 50% of the total energy. When a new coordinator node is assigned, the active tuple space dynamically adjusts the coordinator node's responsibilities based on the priority scores of the tuples to be processed. The new responsibilities and effective time are written into the active tuple space in tuple form. The writing action includes generating tuple identifiers, recording the original coordinator node identifier, the new coordinator node identifier, and updating the start timestamp and validity period. During the arbitration process, if multiple tuples have equal priority scores, they are arranged in order of their generation time, with the earlier ones being processed first. The priority arbitration mechanism and the comprehensive scoring mechanism, combined with heartbeat monitoring, information entropy detection, and pheromone intensity, enable dynamic adjustment of resource allocation, task scheduling, and coordinator node management within the active tuple space. This ensures that mine safety situation collaborative perception tasks are executed in the order of urgency and importance, and guarantees the traceability and continuity of resource use.

[0014] Preferably, within the active tuple space, the list of tuples to be processed is periodically scanned, and a priority score is calculated for each tuple. The priority score is obtained by a weighted combination of four parameters: information entropy trigger flag, confidence aggregation value, pheromone intensity, and diffusion radius. The information entropy trigger flag has a weight of 0.4, the confidence aggregation value has a weight of 0.3, the pheromone intensity has a weight of 0.2, and the diffusion radius has a weight of 0.1. After calculation, the coordinator node arbitrates the parallel decision-making tasks based on the priority scores. The arbitration rule is that tuples with higher scores have priority in occupying computing resources, communication bandwidth, and energy allocation. Under resource constraints, the coordinator node dynamically adjusts the communication bandwidth ratio, computing time slice allocation, and energy allocation based on the priority scores. Tuples with a priority score of 0.8 or higher receive 70% of the total bandwidth and a certain percentage of the total computing time slices. 60% of the total energy allocation; when it is determined that a new coordinator node needs to be assigned, the active tuple space dynamically updates the scope of responsibility of the coordinator node according to the priority score, and generates a responsibility change tuple record of the new coordinator node identifier, the original coordinator node identifier, the scope of responsibility, the start timestamp and the validity period, and writes it into the active tuple space to ensure the traceability of responsibility transfer; during the arbitration process, if multiple tuples have the same priority score, they are arranged in order of tuple generation time, and the earlier one is executed first; the heartbeat writing and information entropy detection, pheromone release and decay mechanism work together to realize the internal resource allocation, parallel judgment execution order scheduling and dynamic adjustment of coordinator node responsibilities in the active tuple space, thereby ensuring that the mine safety situation collaborative perception task is executed in order of urgency and importance, and maintaining the continuity and traceability of processing resources.

[0015] (III) Beneficial Effects This invention provides a collaborative perception method for mine safety status based on sensor networks. It has the following beneficial effects: 1. This invention enables collaborative perception among multiple sensor nodes by constructing a shared active tuple space and introducing a standardized situational cue tuple format; each node encapsulates and writes semantic information through a rule engine to achieve unified data expression and dynamic matching; the active tuple space can perform rapid retrieval and collaborative computation based on association keys; thereby significantly improving the real-time performance and consistency of mine safety situational awareness and reducing the risk of information silos.

[0016] 2. This invention introduces pheromone diffusion and heartbeat monitoring mechanisms into the active tuple space to achieve adaptive scheduling of node state and situation changes; when a sudden change in information entropy or node anomaly is detected, it can automatically increase the priority of relevant tuples and trigger resource reallocation; combined with a comprehensive scoring mechanism, it completes priority arbitration and dynamic adjustment of coordinator nodes; thereby ensuring the continuity of perception and the efficiency of emergency response in complex mining environments. Detailed Implementation

[0017] The technical solutions in the embodiments of the present invention will be clearly and completely described below. 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 skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1: This invention provides a collaborative perception method for mine safety situation based on sensor networks. The method includes deploying sensor nodes in the main and branch roadways of the mine. Each sensor node includes a temperature sensor, a gas concentration sensor, a wind speed sensor, and a vibration sensor. A rule engine and a local clock are deployed in each sensor node. First, a coordinator node identifier is registered in the active tuple space, and a persistent storage area and associated index structure are allocated. Then, the active tuple space defines the formats for situational clue tuples, pheromone tuples, and heartbeat tuples, and specifies the field set, field semantics, priority label, emergency identifier, lifecycle parameters, and access control rules for each type of tuple. The field set includes a type field and a location field. The system includes a value field, a confidence field, a timestamp field, a source identifier field, and a correlation key field. The type field uses an integer encoding range of 1 to 255 to identify the sensor category and observation category. The location field is a string generated by combining the mine coordinates and the roadway number for precise location of the observation point. The value field uses a floating-point number to represent the measurement result. The confidence field ranges from 0 to 1 with an initial value of 0.8. The timestamp field is generated by the local clock and synchronized with the coordinator node. The source identifier field records the unique identifier of the sensor node with a length of 16 bytes. The correlation key field uses a hash operation on the location field, timestamp field, and type field to obtain a 64-bit value and creates 1024 index buckets to support selective scanning and fast matching.

[0019] After the sensor nodes collect raw readings, the rules engine immediately calls the data parsing function to complete type identification, unit conversion, and validity verification. Type identification strictly matches the byte length and format of the sensor input data, uniformly converting temperature to degrees Celsius, gas concentration to percentage, wind speed to meters per second, and vibration amplitude to millimeters per second squared. Validity verification judges based on preset thresholds: temperature range 0 to 60 degrees Celsius, gas concentration 0 to 5% percentage, wind speed 0 to 15 meters per second, and vibration amplitude 0 to 50 millimeters per second squared. If the range is exceeded, the emergency identification field is set to true and the abnormal type is recorded. The sealing process is then completed. After installation, the rule engine sequentially writes the type, location, value, confidence level, timestamp, source identifier, and association key fields into the situational clue tuple, and assigns a priority tag of 5. Lifecycle parameters are also assigned: situational clue tuple 600 seconds, pheromone tuple 300 seconds, heartbeat tuple 120 seconds. The serialized tuple is temporarily stored for 0.5 seconds to await the write window. When the write window arrives, the rule engine performs the write operation, generating a write event record containing node identifier, write timestamp, transaction identifier, and tuple summary. Visibility and access control attributes are written to the tuple metadata area. The write operation maintains strict monotonically increasing consistency with the local heartbeat sequence.

[0020] After a write event occurs in the active tuple space, the index is located by modulo 1024 based on the hash value of the association key and the corresponding index bucket is scanned. The selection criteria are: within a 60-second time window, read access allowed, and confidence level not lower than 0.5. Matching tuple sets are submitted to the collaborative computing processor. The processor performs a step-by-step evaluation based on threshold comparison, rate of change criterion, and tuple consistency judgment. The threshold comparison criteria include a temperature of 100 degrees Celsius, a gas concentration of 4% (presumably referring to a percentage), a wind speed of 12 meters per second, and a vibration amplitude of 40 millimeters per second squared. The rate of change criterion is the change in observed values ​​within the past minute. For triggers exceeding 10 units, the triplet consistency judgment requires at least 3 matching tuples within 60 seconds, with an observation range not exceeding 5 units and a confidence aggregation value not lower than 0.7. The parallel judgment results are sorted by confidence weight 0.6, emergency identifier weight 0.3, and new / old time weight 0.1, and then merged to generate the final judgment result. The final judgment result is encapsulated into a high-order situation tuple and appended with the source tuple identifier set, timeline record, confidence aggregation value, and trigger identifier field. It is then serialized and written back to the live tuple space and an immutable audit record entry is generated.

[0021] After writing back, a pheromone tuple is created with an initial strength of 1.0, a diffusion radius of 50 spatial units, and a decay coefficient of 0.05. Every 5-second processing cycle, it contributes strength to surrounding spatial units according to spatial index partitions and records the source of the contribution. The strength decreases according to the decay coefficient, and when the strength drops to 0, it is automatically cleaned up and recorded in the audit log. Sensor nodes periodically read the pheromone strength. A threshold of 0.6 triggers an adjustment of the sampling frequency to twice the normal value, cross-validation, writing of a support tuple, and releasing the same type of pheromone with an initial strength of 0.5 and a diffusion radius of 30 units after successful validation. Each sensor node periodically writes a heartbeat tuple every 30 seconds. The heartbeat tuple contains the node identifier, timestamp, status code, and sequence number. The coordinator node monitors the continuous heartbeat sequence; nodes with an activity index below 0.5 are considered abnormally silent. Sensor nodes use a local entropy detection sliding window of 120 seconds. An increase in entropy greater than 0.3 triggers the generation of a situational cue tuple with priority. Tag 10: Upon receiving a high-priority tuple, the coordinator node immediately triggers matching and resource scheduling operations. Tuples with a score higher than 0.7 are assigned a mobile node, allocated additional computing resources, and have their sampling area center offset adjusted by 20 units. High-scoring areas receive 60% bandwidth, 60% computing time slice, and 50% energy allocation. The scheduling result is written into the active tuple space, recording the execution start time and duration, and generating immutable audit record entries. The coordinator node periodically scans the pending tuples, calculating the priority score, information entropy trigger flag weight (0.4), confidence aggregation value weight (0.3), pheromone intensity weight (0.2), and diffusion radius weight (0.1). The coordinator node arbitrates and determines tasks in parallel based on the scores. Tuples with high scores have priority in occupying computing resources and communication bandwidth. Tuples with equal scores are sorted by their generation time. After arbitration, tasks are executed sequentially to ensure that mine safety situation collaborative perception tasks are completed in order of urgency and importance, while maintaining resource continuity and traceability.

[0022] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A collaborative perception method for mine safety situation based on sensor networks, characterized in that, Includes the following steps: S1. Construct and initialize a shared active tuple space, and define standardized situational cue tuple, pheromone tuple, and heartbeat tuple formats in the active tuple space. The situational cue tuple format contains association keys, and a rule engine is deployed in each sensor node to perform local perception and semantic encapsulation. A coordinator node is also deployed in the active tuple space to manage matching tasks and resource scheduling within the active tuple space. S2, after any sensor node completes local perception, the rule engine encapsulates the raw readings according to the situational cue tuple format and performs a write operation to write the encapsulated situational cue tuple into the active tuple space, while triggering the matching task triggered by the write operation inside the active tuple space. S3 is a matching task triggered by the active tuple space response write operation. It searches for other tuples in the space that match the association key of the just-written situational clue tuple, and performs judgment according to the collaborative computing rules to generate a judgment result. The collaborative computing rules include threshold comparison, change rate criterion and tuple consistency judgment. S4, the active tuple space writes the judgment result back to the active tuple space in the form of a new higher-order situation tuple, and records the source tuple reference in the higher-order situation tuple. The source tuple reference points to the tuple that generated the judgment result. S5. After the high-order situation tuple is written back, the active tuple space is written with the pheromone tuple corresponding to the high-order situation tuple, and the pheromone tuple enters a periodic diffusion and decay cycle for other sensor nodes to sniff and respond. S6, during periodic operation, each sensor node periodically writes a heartbeat tuple into the active tuple space and periodically detects the information entropy of the local data stream. When the information entropy detects a significant mutation or the heartbeat tuple detects an anomaly, the active tuple space increases the priority of the corresponding tuple and drives priority matching and resource scheduling within the active tuple space.

2. The method for collaborative perception of mine safety status based on sensor networks according to claim 1, characterized in that, S1 includes: S11 specifies the set of fields in the situational clue tuple and defines the semantics of the fields, including type, position, value, confidence level, timestamp, source identifier and association key, so that each encapsulated action can provide semantic labels to the outside world; S12, Define the writing rules and lifecycle parameters for priority labels and emergency identifiers in the situational clue tuples so that the active tuple space can implement visibility control accordingly; S13, a rule engine deployed in each sensor node, after sensing the raw readings, converts the readings into structured tuples conforming to the situational cue tuple format according to the field set, and records the source of the tuples and the local confidence level during the conversion process for subsequent matching and auditing.

3. The method for collaborative perception of mine safety situation based on sensor networks according to claim 1, characterized in that, In S2, write operations include: S21, The write operation generates a write event, and the active tuple space generates a write event to trigger subsequent processing; S22, the write operation simultaneously marks the visibility attribute and access control attribute in the situational clue tuple, and sets the searchable range of the situational clue tuple in the space according to the visibility attribute to support reading and sniffing. S23, When the write operation is executed, the sensor node keeps the write transaction consistent with the local heartbeat tuple write sequence so that the active tuple space can verify the activity of the write node and the write timeline through the heartbeat tuple write sequence.

4. The method for collaborative perception of mine safety situation based on sensor networks according to claim 1, characterized in that, S3 includes: S31. After receiving a write event, the active tuple space selectively scans the spatial index according to the association key of the newly written situational cue tuple to locate the matching tuple and submits the matching result to the processing mechanism inside the active tuple space. S32, the active tuple space performs a judgment on the set of matching tuples according to the collaborative computing rules; the collaborative computing rules consist of threshold comparison, change rate criterion and tuple consistency judgment, and the output Boolean judgment or level judgment is used as the judgment result. S33, when multiple decision results are generated in parallel, the parallel decisions are sorted in the active tuple space according to the confidence or urgency of the decision results, and the sorted decision results are merged to generate the final high-order situation tuple.

5. The method for collaborative perception of mine safety situation based on sensor networks according to claim 1, characterized in that, S4 includes: S41, before the high-order situation tuple is written back, the active tuple space adds source tuple identifiers, timelines and confidence aggregation values ​​to it to ensure the integrity of the evidence chain. S42, when the higher-order situation tuple is written back, it carries a trigger flag to drive the subsequent pheromone release, so as to identify and respond to the higher-order situation tuple. S43, the high-order situation tuple write-back action simultaneously generates immutable audit records in the active tuple space for subsequent traceability retrieval and decision verification.

6. The method for collaborative perception of mine safety situation based on sensor networks according to claim 1, characterized in that, S5 includes: S51. After the pheromone tuple is written, it enters a periodic processing loop. In each processing loop, the intensity contribution is delivered to the surrounding spatial units according to the diffusion radius of the pheromone tuple, and the contribution source is accumulated and recorded at the target unit. S52, the pheromone tuple decreases its current intensity according to the decay coefficient in each processing cycle, and is automatically cleared from the active tuple space when the intensity drops to zero, so as to avoid expired information from continuously occupying resources. S53, the sensor nodes distributed in the active tuple space periodically read the pheromone intensity, and when the read intensity exceeds the preset response threshold of the sensor node, they perform response behavior; the response behavior includes adjusting the sampling frequency, performing local cross-validation or writing support tuples into the active tuple space, and releasing the same type of pheromone according to the positive feedback rule when the local validation is successful, so as to achieve superposition.

7. The method for collaborative perception of mine safety situation based on sensor networks according to claim 1, characterized in that, S6 includes: S61, each sensor node periodically writes a heartbeat tuple to the active tuple space so that the active tuple space or coordinator node can detect the active state of the sensor node or region through a continuous heartbeat tuple sequence and identify abnormal silence patterns. S62, perform information entropy detection periodically at sensor nodes or coordinator nodes to determine whether there are significant changes in the local data stream or tuple sequence, and write the resulting situational cue tuples into the active tuple space with priority labels when entropy change detection is triggered. S63, when heartbeat monitoring identifies a regional silence pattern or information entropy detection identifies a parallel high-entropy event, a resource scheduling process is triggered based on a composite score of entropy magnitude, confidence level, and influence radius to specify the allocation of computing and communication resources for mobile nodes or adjust the coordinator node for key reconnaissance or key synthesis.

8. The method for collaborative perception of mine safety status based on sensor networks according to claim 1, characterized in that: A comprehensive scoring mechanism of priority arbitration and resource allocation is added to the overall operation of S1 to S6. The comprehensive scoring mechanism calculates a priority score for each tuple to be processed in the active tuple space. The priority score is a weighted combination of information entropy trigger flag, confidence aggregation value, pheromone intensity and diffusion radius to determine the processing order. Priority scores are used at the coordinator node to arbitrate the execution order of parallel decisions, and to determine the decentralization strategy for communication bandwidth, computation time slices, and energy allocation when resources are limited. When the comprehensive scoring mechanism assigns a new coordinator node, the active tuple space dynamically adjusts the scope of responsibilities of the coordinator node according to the priority score, and writes the change of responsibilities and its effective time into the active tuple space in the form of tuples to ensure the traceability of the transfer of responsibilities.