A generalized world model execution result hierarchical backflow updating method

Abstract

The application provides a generalized world model execution result hierarchical backflow updating method, which comprises the following steps: obtaining original field files recording the execution actions and completion states of a natural social system, extracting fragment summaries and collecting device fingerprints, reading transfer chain marks allowing missing, misrecording or clock drift, and forming field evidence spectrum records; tracing the reproduction positions of the fragment summaries and the device fingerprints along the chain, collecting isogenic fission files and constructing a transmission node topology map, locating the original collection source, and obtaining a relay transmission path with a collection starting point; tracing the isogenic fission file copies along the path, verifying the isogenic relationship of the copies, stripping repeated evidence, and collecting the original field files directly obtained by each collection terminal into an independent evidence set; correlating the independent evidence to the execution result, backtracking the repeated evidence to the node connection position, forming an evidence backflow mapping, and performing hierarchical backflow according to the execution result layer and the link layer to complete hierarchical backflow updating control of the execution result and the relay transmission path record.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a hierarchical backflow update method for the execution results of a generalized world model. Background Technology

[0002] The hierarchical feedback and update control of the execution results of natural and social systems is a crucial link in ensuring the efficient handling of major emergencies. In emergency response processes under complex environments, the accurate feedback of on-site evidence and status assessment directly affect the reliability of command decisions. Existing emergency command systems can use world models to uniformly represent the disaster environment, the response entities, the actions taken, the completion status, and communication nodes, and continuously update the execution result status and transmission link status in the world model based on on-site observations. Ideally, independent feedback evidence from different original acquisition terminals can reflect the on-site facts from different locations. The accumulation of independent evidence can quickly confirm the strength of the on-site facts and provide mutually independent observational basis for updating the world model's status. However, in actual operation, it has been found that this judgment method based on the accumulation of feedback quantity is easily misled by superficial phenomena, leading to misjudgments of the actual on-site situation. When the world model directly treats each feedback file as an independent on-site observation, the increase in the number of feedback files is interpreted as an increase in independent observation sources, causing the same execution results in the model to be repeatedly updated. Due to limited communication networks, a single piece of original evidence often needs to be transferred and relayed through multiple terminal devices before being transmitted back to the command center, thus generating a large number of copies in the network. Although each copy has a different terminal transfer timestamp and transfer chain marker, the recorded execution action, completion status, and original collection source remain unchanged. Therefore, they cannot be repeatedly used as independent on-site evidence in the execution result evaluation. To improve evaluation efficiency, conventional evaluation mechanisms generally equate the number of evidence collected with the strength of confirmation of on-site facts. While this quantity-based judgment method can quickly process massive amounts of returned information and improve evaluation speed, it ignores the physical propagation process of evidence in complex communication networks, leading to the incorrect identification of an increase in the number of returned copies as an increase in independent evidence. For systems that use world models to extrapolate event states, multiple copies of the same original evidence are continuously written to the execution result layer, repeatedly reinforcing the same execution action and completion status. However, the actual changes in transmission nodes and relay order reflected by each copy are not accurately preserved at the transmission link layer. This is because the same piece of evidence generates a large number of copies after being transferred and relayed by multiple terminal devices. The increase in the number of returned copies only represents an increase in transmission nodes, not an increase in on-site observation perspectives. For example, during mountain fire fighting, videos taken by rescuers in communication blind spots are relayed through multiple team members' terminals, resulting in multiple copies of the transmitted files for the command center. If these files are input into a world model as independent observations, the model will update the fire situation and response results for the same area multiple times, mistakenly believing there are multiple independent pieces of evidence from different perspectives, thus artificially inflating the credibility of the fire situation in that area. In this case, the newly added copies actually carry the link information of the evidence passing through new transmission nodes, rather than new on-site execution results.Therefore, in complex communication environments, accurately identifying the original source of evidence collection and transmission relay process, distinguishing between independent evidence used to update the execution result layer of the world model and duplicate evidence used to update the transmission link layer, stripping away the false superposition of the weight of the same source duplicate files on the on-site facts, and performing hierarchical backflow updates on the execution result records and relay transmission path records have become urgent technical challenges to be solved. Summary of the Invention

[0003] This invention provides a hierarchical backflow update method for the execution results of a generalized world model, comprising: Original field files recording the actions and completion status of natural and social systems are obtained from restricted networks in mountain fire communication blind spots. Fragment summaries and acquisition device fingerprints are extracted, and transfer chain markers are read. The transfer chain markers are allowed to have missing, incorrect, and terminal clock drift, forming a field evidence genealogy record. The fragment summary is traced back to the location of its recurrence and the fingerprint of the acquisition device along the chain to identify homologous fission traces, collect homologous fission files, and when the transfer chain is unreliable due to missing transfer chain markers, incorrect recording, or terminal clock drift, the homologous relationship between the field file entries is determined by comparing the consistency between the fragment summary and the fingerprint of the acquisition device. Read the terminal transfer timestamp and transfer chain marker of the same source fission file, trace the order of each terminal relay, and construct a transmission node topology map; Analyze the transmission node topology, identify the relay transmission path, locate the original data acquisition source, and obtain the relay transmission path with the data acquisition starting point; Tracing the same source fission document copies along the relay transmission path from the collection starting point, verifying the same source relationship of the copies, separating duplicate evidence, and classifying the original field documents directly obtained by each collection terminal into an independent evidence set; Link independent evidence to the execution results of the natural and social system, trace duplicate evidence back to the transmission node topology, determine the node and the order of succession, and link it to the node connection position to form an evidence backflow mapping; Independent evidence sets are fed back to the execution result layer according to the execution result layer, and duplicate evidence is fed back to the corresponding transmission node according to the link layer. Missing items and inconsistent items are added and corrected according to the continuity of execution status and relay path, thus completing the hierarchical backflow update control of execution result and relay transmission path records.

[0004] Preferably, the process of acquiring original on-site files recording the actions and completion status of natural and social systems from a restricted network in a mountain fire communication blind spot, extracting fragment summaries and acquisition device fingerprints, reading transfer chain markers, and forming an on-site evidence genealogy record includes: Retrieve the original field files recorded by the acquisition terminal, verify the integrity of the load, and remove any missing parts; A fixed-length sliding window is used to extract fragments from the file payload byte stream to generate fragment digest values; Read the hardware serial number, sensor model code and firmware version number, and concatenate them into a plaintext string by separating them with vertical lines in a preset order. Then, generate a fingerprint of the acquisition device as a 64-bit hexadecimal string by performing SHA-256 digest operation. Expand the relay terminal identifier string and the upstream node retrieval field, and archive them to form a field evidence genealogy record with the collection starting point as the root and the relay nodes as branches.

[0005] Preferably, the recurrence location of the trace fragment summary and the chain retention of the fingerprint of the acquisition device are used to identify homologous fission traces and collect homologous fission files, including: Extract fragment summaries and acquisition device fingerprints from each field document entry in the genealogy record. Compare the fragment summaries one by one between each relay node, retrieve the recurring locations, and register the corresponding relay terminal identifier strings. Those fingerprints from the same acquisition device and the same fragment summary value that are synchronously reproduced at multiple relay nodes are registered as homologous fission traces; Tracing back along the relay sequence to the collection starting point where the index field is empty, file entries pointing to the same collection starting point are grouped into a homogeneous fission file set.

[0006] Preferably, the step of reading the terminal transfer timestamp and transfer chain marker of the same-source fission file, tracing the relay order of each terminal, and constructing a transmission node topology map includes: Read the terminal transfer timestamp and transfer chain marker bound to each field file entry from the same source fission file set, and expand the relay terminal identifier string and upstream node back pointer field along the chain; The downstream terminals are paired with the upstream terminals based on the upstream node's back pointer field. The pairing is verified by checking the timing of the terminal transfer timestamps. The relay sequence of each terminal and its connection position with the upstream and downstream are determined, resulting in a terminal relay sequence with directional indication.

[0007] Preferably, the step of analyzing the transmission node topology, identifying the relay transmission path, locating the original data acquisition source, and obtaining the relay transmission path with the data acquisition starting point includes: Read the in-degree and out-degree values ​​of each node from the transmission node topology graph, determine the node with no in-edge and out-edge as the source node candidate, and determine the source node corresponding to the original collection source if the relay terminal identifier string corresponding to the source node candidate matches the collection terminal identifier string of the confirmed collection device fingerprint. Traverse forward along the directed edges from the source node, recording the relay terminal identifier strings and the order of the directed edges, until a node with no outgoing edges is reached, thus obtaining the candidate relay path chain. The field file entry numbers mounted on each node in the path chain are concatenated along the path. The upstream and downstream connection positions of the directed edge records between adjacent nodes are verified to match the upstream node's back pointer field of the corresponding file entry. The connection lines between adjacent nodes that do not match are identified as breakpoint branches and removed, thus obtaining a relay transmission path with the acquisition start point.

[0008] Preferably, the process of tracing the duplicate files along the relay transmission path from the data acquisition starting point, verifying the commonality of the duplicates, separating duplicate evidence, and grouping the original field files directly obtained by each acquisition terminal into an independent evidence set includes: Tracing the copies within the same fission file set along the relay transmission path, attaching the field file entries to each copy to the corresponding nodes in the path chain according to the relay node they belong to, and defining the field file entry at the starting node of the path chain as the starting file; Read the fragment summaries of the remaining copies and compare them with the fingerprints of the acquisition devices and the starting file for verification. Mark the ones that match as the transfer copies. Remove the transferred copies after the starting node from the candidate file sequence and add the starting file to the independent evidence set.

[0009] Preferably, the step of associating independent evidence with the execution results of the natural and social system, checking duplicate evidence back to the transmission node topology graph, determining the node and the order of successive relays, associating it with the node connection position, and forming an evidence backflow mapping includes: Extract each starting point file from the independent evidence set, read the on-site handling action and completion status fields, match the fingerprint of the collection device corresponding to the starting point file with the execution action entries registered in the execution result layer of the natural and social system according to the handling action category and collection coordinate position, and attach the completion status field to the matched execution result entries as independent observation values ​​to obtain the independent evidence attachment mapping of the execution result layer; The terminal transfer timestamp and transfer chain mark corresponding to the duplicate evidence are retrieved back to the transmission node topology map. The node is located according to the relay terminal identifier string. Adjacent nodes are linked according to the upstream node back pointer field and verified by the order of terminal transfer timestamps to determine the node and the relay order. The on-site document entry numbers of duplicate evidence are attached to the corresponding node connection positions marked by the directed edge, and merged and archived with the attachment mapping of independent evidence to form an evidence backflow mapping covering the execution result layer and the link layer.

[0010] Preferably, the step of backflowing independent evidence sets to the execution result layer, backflowing duplicate evidence to the corresponding transmission node at the link layer, and supplementing missing items and revising inconsistencies based on the continuity of execution status and relay path, thereby completing the hierarchical backflow update control of the execution result and relay transmission path records, includes: The execution result layer observation values ​​of each starting point file are backfilled into the execution result entries of the natural and social system according to the fingerprint of the acquisition device. The node connection position of duplicate evidence is backfilled into the corresponding transmission node record of the link layer according to the relay terminal identifier string. Read the completion status field of the execution result entries arranged along the collection time, mark the gap period where the progress jump exceeds the set increase range as the record missing item, and only fill in the independent evidence value that the collection time falls within the gap period and originates from other collection starting points. If there is no independent evidence that meets the conditions, mark the record missing item as unfillable. Read the upstream and downstream connection positions of the transmission node records arranged in the relay order, mark the upstream and downstream breakpoint positions as missing items in the path and fill them into the corresponding node connection positions to obtain the filling results of missing record items and missing path items.

[0011] Preferably, the directional terminal relay sequence includes: Each relay terminal identifier string is abstracted as a node, and the upstream and downstream connection positions are abstracted as directed edges pointing from the upstream terminal to the downstream terminal. The corresponding terminal transfer timestamp and on-site file entry number are attached to the node, and the pointing record of the upstream node back pointer field is attached to the directed edge. A transmission node topology diagram with the collection starting point as the source node and the aggregation node as the sink node is drawn.

[0012] Preferably, the results of filling in the missing record items and missing path items include: Each field in the world model is compared with the observed values ​​after reflow to determine if they are consistent. Fields with inconsistent values ​​are defined as inconsistencies. For inconsistencies at the execution result layer, the independent evidence value corresponding to the later acquisition time is used to overwrite the original field value corresponding to the earlier acquisition time, without using the terminal transfer timestamp as the basis for overwriting at the execution result layer. For inconsistencies at the link layer, the node connection position corresponding to the later terminal transfer timestamp is used to revise the earlier record, thus completing the hierarchical reflow update control of execution result records and relay transmission path records.

[0013] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: This invention discloses a hierarchical backflow update method for the execution results of a generalized world model. Addressing the challenges of ambiguous sources, redundancy, and missing execution result records in the relay transfer of on-site evidence during mountain fires in communication-blind, limited-network environments, the method obtains original on-site files recording the actions and completion states of the natural and social system from the limited network. It extracts fragment summaries from the original files, collects device fingerprints, and reads the transfer chain markers. These transfer chain markers allow for missing, incorrect, and terminal clock drift records, forming a phylogenetic record of on-site evidence. Furthermore, it traces the recurrence locations of fragment summaries and the retention of device fingerprints along the chain, aggregating homogeneous fission. The system constructs a transmission node topology map and locates the original collection starting point based on the terminal transfer timestamp and transfer chain marker. It verifies the same-source relationship of the copies along the relay path, removes duplicate evidence to form an independent evidence set, associates the independent evidence with the execution result, and traces duplicate evidence back to the node connection position to form an evidence backflow mapping. Based on the continuity of the execution state and the continuity of the relay path, it fills gaps and corrects errors, realizing hierarchical backflow updates at the execution result layer and the link layer. This allows the world model state to evolve round by round according to the state update equation driven by the evidence backflow mapping, which is different from static data storage. This effectively improves the accuracy of evidence tracing and the completeness and consistency of the world model record in blind spot environments. Attached Figure Description

[0014] Figure 1 This is a flowchart of a hierarchical backflow update method for the execution results of a generalized world model according to the present invention.

[0015] Figure 2 This is a schematic diagram of a hierarchical backflow update method for the execution results of a generalized world model according to the present invention.

[0016] Figure 3 This is another schematic diagram of a hierarchical backflow update method for the execution results of a generalized world model according to the present invention. Detailed Implementation

[0017] To further understand the content of this invention, a detailed description of the invention is provided in conjunction with the accompanying drawings and embodiments. The specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention. It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0018] like Figures 1-3 This embodiment of a hierarchical backflow update method for the execution results of a generalized world model may specifically include: Step S101: Obtain original field files recording the actions and completion status of natural and social systems from the restricted network in the communication blind zone of the mountain fire, extract the original file fragment summary and the fingerprint of the acquisition device, read the transfer chain mark, and form a field evidence genealogy record for use by the world model.

[0019] Raw field files collected by rescue terminals are retrieved from restricted networks in mountain fire communication blind spots. These raw field files record the actions and completion status of firefighters. For each raw field file, the file payload and accompanying metadata are read. The integrity of the payload is determined by verifying the file header fields, encapsulation container structure, and payload byte length item by item. Damaged payloads and missing fields are removed to obtain field file entries with complete payloads and metadata. For each field file entry to be parsed, a fixed-length sliding window is used to sequentially truncate the file payload byte stream at fixed steps, and a fragment digest value is generated for each fragment. The hardware serial number, sensor model code, and firmware version number of the acquisition terminal are read from the accompanying metadata, and concatenated into a plaintext string separated by vertical lines in a preset order. The plaintext string is then generated by SHA-256 digest operation to produce a 64-bit hexadecimal string fingerprint of the acquisition device, resulting in a field file entry with a fragment digest and acquisition device fingerprint. For the field document entries containing the attached fragment summary and the fingerprint of the acquisition device, the relay terminal identifier string, the terminal transfer timestamp, and the upstream node retracement field in the transfer chain marker are read. The relay terminal identifier string is expanded segment by segment according to the order of the terminal transfer timestamps and the pointing relationship of the upstream node retracement field. The expanded relay sequence is bound and archived with the fragment summary and the fingerprint of the acquisition device on a per-field document basis, forming a field evidence genealogy record with the acquisition starting point as the root and the relay nodes as branches, for use by the world model.

[0020] In mountain fire fighting operations, firefighters often venture deep into valleys, back slopes, and forest gaps where communication infrastructure is weak. These areas constitute typical communication blind spots. Videos, fire line photos, smoke concentration sampling footage, and recordings of firefighting actions collected by rescue terminals cannot be directly transmitted back to the command center. Instead, they must be relayed and saved through a chain of portable terminals carried by team members along the route. The aforementioned original field files refer to the original payload records generated directly by the rescue terminals in these blind spots, documenting the actions and completion status of firefighters.

[0021] In one possible implementation, the process of retrieving raw field files from a restricted network includes listening for backhaul channel requests sent by any relay terminal along the chain to the aggregation node. The aggregation node is the relay terminal at the end of the transfer chain responsible for external data retrieval, and is responsible for uniformly collecting the data to be backhauled generated by each upstream node on the chain. A temporary buffer is deployed in the local non-volatile storage area of ​​each relay terminal. Its capacity is determined by the product of the maximum dead zone dwell time per terminal and the average daily field file output, plus a margin, and is planned to be 512 megabytes to 2 gigabytes. This buffer is used to initially write the data to disk before the link is established. Each file packet generated by a node or transferred from an upstream node is appended with a transfer chain marker when written. The transfer chain marker is embedded in the extended attribute area of ​​the file in the form of a linked list. Each write operation adds a node corresponding to the current terminal to the end of the linked list. The node consists of three items: the relay terminal identifier string, the terminal transfer timestamp, and the upstream node back pointer field. After receiving the packet, the aggregation node merges the file packets in the same chain according to the back pointer relationship between the nodes and restores the path order. After receiving the file packets to be transmitted with the transfer chain marker in the buffer, the aggregation node performs deduplication and reordering according to the marker before sending them to the external processing flow.

[0022] For example, a video of a fire spreading halfway up a mountain may be relayed from a rescuer's handheld terminal to three team members descending the slope using wrist-mounted terminals, and finally uploaded to the command center via a relay terminal at the foot of the mountain.

[0023] Specifically, payload integrity verification refers to a preliminary screening of audio and video files before they enter the parsing process. The purpose is to eliminate incomplete files caused by transmission interruptions, storage corruption, or abnormal acquisition, preventing empty or incorrect reads in subsequent parsing stages. This verification targets three aspects: file header fields, encapsulation container structure, and payload byte length. File header field verification includes checking the existence of the file magic number identifier and encoding format declaration, and adding specific declaration fields according to file type: for video and action recordings, verifying the duration field; for photos, verifying the image format declaration and resolution field; and for smoke concentration sampling segments, verifying the sampling rate declaration and range field. Encapsulation container structure verification involves reading the index structure within the container according to file type: for audio and video files, reading the audio and video track index table and keyframe position table; for photo files, reading the image block index table; and for sensor data files, reading the sampling sequence index table and sampling point count field to determine if there are any breaks or missing items. Payload byte length verification compares the declared file tail length with the actual number of bytes read, with an allowable deviation of no more than 16 bytes. This deviation is determined by the maximum number of bytes padded at the end of the encapsulation container. If all three criteria are met, the file is included in the list of files to be parsed; if any one criterion fails, such as a missing video keyframe index or a byte length mismatch, the file is removed as an incomplete file.

[0024] It should be noted that accompanying metadata refers to auxiliary information fields generated and recorded along with the file body, recording the acquisition environment and equipment attributes. These include the latitude and longitude, orientation, hardware serial number of the acquisition terminal, sensor model code, and firmware version number at the time of acquisition. In the fragment summary generation stage, a fixed-length sliding window is used to sequentially extract fragments along the file payload byte stream.

[0025] Preferably, the window length can be 8 kilobytes, and the step size is equal to the window length to achieve non-overlapping cutting; for each segment, the SHA-256 algorithm is used to generate a 256-bit digest value, which is used as the fragment digest value of that segment.

[0026] For example, a 120-second video of a fire spreading is cut into several segments, and the segment digest value corresponding to each segment is written into the digest field of the file entry. The fingerprint of the acquisition device is constructed as follows: the hardware serial number, sensor model code, and firmware version number are concatenated into a plaintext string by separating them with vertical lines in a preset order. The plaintext string is then subjected to a SHA-256 digest operation to obtain a fixed-length fingerprint of 256 bits, registered as a 64-bit hexadecimal string. This fingerprint serves as the verification credential for determining which rescue terminal the file originated from, and uses the same digest algorithm as the segment digest value.

[0027] In one embodiment, the transfer chain marker is embedded in the extended attribute area of ​​the file in the form of a linked list. Each node in the linked list consists of three items: a relay terminal identifier string, a terminal transfer timestamp, and an upstream node retrace field. The relay terminal identifier string is used to record the number of the terminal involved in the transfer, the terminal transfer timestamp is used to record the time when the terminal received and transferred the file, and the upstream node retrace field is used to point to the node number of the previous hop transfer terminal. Furthermore, the unfolding process is divided into two stages executed sequentially: The first stage is node index construction, which reads all nodes in the extended attribute area, sorts them in ascending order by terminal transfer timestamp, and takes the node with the largest timestamp as the current convergence node. For example, if a file contains 4 nodes with timestamps of 1724200000, 1724200035, 1724200072, and 1724200118, then the node with sequence number 4 is the convergence node. The second stage is backtracking and positioning. Starting from the convergence node, it reads the node sequence number pointed to by the upstream node's backreference field, jumps to the node corresponding to that sequence number, and repeats this jump operation until it reaches the node where the upstream node's backreference field is empty. This node corresponds to the initial data collection rescuer terminal. Reversing the node sequence traversed in the second stage yields an ordered relay sequence from the data collection starting point to the convergence node.

[0028] It is understandable that the expanded ordered relay sequence, along with the fragment summary value and the fingerprint of the acquisition device, is bound and archived on a per-entry basis as on-site documents, resulting in a tree-like on-site evidence genealogy record with the acquisition starting point as the root and each relay node as a branch. The world model mentioned here refers to a structured data model used by the command center to uniformly represent on-site spatiotemporal entities, personnel and equipment, event flow, and evidence relationships. It includes three types of primary key-associated data sets: entity table E, relationship table R, and evidence table V. Each row in V corresponds to an on-site document entry and carries the genealogy root node identifier `rid` and the node path `path`. The world model interface layer refers to the service layer that exposes the read, write, and query capabilities of this data model. It interfaces downwards with the archive storage and upwards with unified interfaces `query(rid)`, `query(path)`, and `query(fileId)` for upper-layer business calls. The on-site evidence genealogy records are serialized and written into the corresponding fields of table V. Foreign keys pointing to the collection points and relay nodes are established in E and R. After writing, the interface layer returns the primary key fileId and the genealogy root rid. Subsequent calls by file entry retrieve each entry individually using fileId as the input parameter, while calls by collection starting point retrieve the entire genealogy subtree under that starting point at once using rid as the input parameter. The response time for a single query is controlled within 200 milliseconds, and the number of fields per entry is approximately 20 to 30, meeting the command center's needs for tracing, comparing, and linking on-site evidence. The state of the world model is not a static data set, but a dynamic state that evolves with each round of evidence feedback: let M be the state of the world model before the k-th round of feedback update. k =(E k ,R k V k ,L k ), where L k For the link layer transmission node record set; the k-th round of flow update satisfies the state update equation M {k+1} =F(M k ,Φ k ), where Φ k For this round of evidence backflow mapping, F is the hierarchical backflow update operator, and Φ is mapped by the primary key. kThe incremental changes are merged into entity table E, relation table R, evidence table V, and link layer L, respectively. The details of the state transition of the completion status field of the execution result item are given in S107. Accordingly, the technical differences between the world model and ordinary databases are at least as follows: First, the updates of ordinary databases are based on transactions, and the update content and writing order are determined by the external caller. In contrast, the updates of the world model in this application are based solely on evidence backflow mapping. Independent evidence and duplicate evidence are separated according to evidence attributes. Observation values ​​only enter the execution result layer, and relay path information only enters the link layer. The update path is determined by the evidence homology judgment result rather than specified by the caller. Second, the completion status field of the execution result item evolves according to the coverage rule with historical revision stack based on the collection time, constituting a state transition semantic with backtracking capability, rather than the stateless field overwriting of the database. Third, the update trigger depends on the two types of model-internal judgment logic: execution state continuity and relay path continuity. Missing entries are constrained by the starting point of independent evidence collection. Ordinary databases do not have a mechanism to select the update layer and update order based on the source attributes of the observed evidence.

[0029] Step S102: Trace the recurrence location of the original document fragment summary in the on-site evidence genealogy record and the fingerprint retention along the chain of the acquisition device, identify homologous fission traces, and collect homologous fission documents.

[0030] From the aforementioned on-site evidence genealogy record, the original document fragment summary and the acquisition device fingerprint of each on-site document entry are extracted sequentially. For each relay node within the genealogy record, the original document fragment summary is compared line by line, and recurrence locations where the summary value appears more than once are retrieved. The relay terminal identifier string corresponding to each recurrence location and the on-site document entry number to which it belongs are registered, resulting in a summary recurrence location list. For this summary recurrence location list, the traces left along the relay sequence by the acquisition device fingerprint are investigated to determine whether the same acquisition device fingerprint appears repeatedly in document entries with multiple relay terminal identifier strings. If the same acquisition device fingerprint and the same fragment summary value are synchronously reproduced at multiple relay nodes, they are registered as homologous fission traces, resulting in a homologous fission trace annotation table. Based on the aforementioned homogeneous fission trace annotation table, the fragment summary values ​​are precisely matched and compared with the fingerprints of the acquisition devices within the on-site evidence genealogy record. All on-site document entries carrying the same fragment summary value and the same acquisition device fingerprint are detected. The process is then reversed along the relay sequence of each document entry back to the acquisition starting point where the reference field is empty. Document entries pointing to the same acquisition starting point are grouped into a homogeneous fission file set. The determination of the aforementioned homogeneous relationship is based on the content comparison between the fragment summary and the acquisition device fingerprint, and does not depend on the completeness of the transfer chain markers. Even when the transfer chain markers are missing, incorrectly recorded, or the terminal clock drift causes the link to be unreliable, the identification of homogeneous fission traces and the grouping of homogeneous fission file sets still hold, providing a content-side benchmark for the construction of the transmission node topology and the tracing of the relay order.

[0031] To eliminate the false weighting of on-site facts caused by the original on-site files from the same rescue terminal being copied into duplicate evidence by various transfer terminals along the way in mountain fire communication blind spots, it is necessary to trace and collect the fission traces of these duplicate evidences based on the on-site evidence spectrum record obtained in the previous stage, and complete the deduplication process before updating the world model state.

[0032] In one possible implementation, the original document fragment summary and the acquisition device fingerprint bound to each on-site document entry are sequentially read from the on-site evidence genealogy record. The original document fragment summary is a sequence of summary values ​​generated segment by segment after being truncated by a fixed-length sliding window. The acquisition device fingerprint is a fixed-length fingerprint registered as a 64-bit hexadecimal string, obtained by concatenating the hardware serial number, sensor model code, and firmware version number of the acquisition terminal in a preset order separated by vertical lines into a plaintext string and performing SHA-256 digest operation.

[0033] Specifically, the retrieval of abstract recurrence locations is carried out between the branches of each relay node within the genealogical record. For the same segment abstract value, the number of times it appears in different field document entries is compared one by one. If a certain abstract value appears simultaneously in two or more field document entries, the abstract value is determined to have recurred.

[0034] For example, a video of a fire spreading was relayed from the wrist-worn terminals of three retreating team members to a relay terminal at the foot of the mountain. In the on-site evidence genealogy record, it was presented as the entry for the collection starting point and the copy entries for each relay node. The corresponding segment summary values ​​on the three relay copies were completely identical.

[0035] In one embodiment, the list of summary reproduction locations is stored as fielded entries in the comparison cache of the command center. Each entry in the list consists of three fields: a reproduction summary value, the field file entry number to which it belongs, and the relay terminal identifier string where the reproduction occurred. The reproduction summary value is used to indicate the segment involved in the reproduction; the field file entry number is used to connect to the specific file in the genealogy record; and the relay terminal identifier string is used to point to the relay node position of the file entry.

[0036] It should be noted that identical fragment digest values ​​alone are insufficient to determine whether two file entries are homologous copies. The fragment digest value is obtained by the SHA-256 algorithm on the fragmented binary data, outputting a fixed-length digest of 256 bits, recorded as a 64-bit hexadecimal string, used for content consistency comparison at the fragment level. The probability of the digest algorithm producing the same digest value for different payloads is extremely low, and digest values ​​of similar payloads are uncorrelated and lack similarity. In mountain fire scenes, rescuers may take multiple independent photographs of the same fire-prone area, resulting in different fragment digest values ​​from each photograph, which can be distinguished by the digest values ​​themselves. However, the fragment digest value only reflects the fragment content and does not carry source terminal information. To determine whether entries with identical digest values ​​originate from the same original file from the same acquisition terminal, it is necessary to further pinpoint the original terminal source of the file using the fingerprint of the acquisition device. Both factors are then used to jointly determine the homologous copy relationship. The fingerprint of the acquisition device is obtained by concatenating fixed fields read from the local machine by the acquisition terminal during file generation and then performing SHA-256 calculation. The fields involved in the concatenation are consistent with those described in S101, namely, hardware serial number, sensor model code, and firmware version number. Each field is separated by a vertical bar to form a plaintext string. Then, a digest is taken from this plaintext string, and finally, it is registered along with the file entry in the form of a 64-bit hexadecimal string. The acquisition device fingerprints generated by the same terminal remain stable throughout its life cycle. Therefore, the retention of the acquisition device fingerprint along the relay sequence is further investigated based on the digest reproduction location list. Further, the process of investigating the retention traces of the acquisition device fingerprint along the chain includes: for all field file entries corresponding to the same acquisition device fingerprint, reading the relay terminal identifier string of each entry, and determining whether the acquisition device fingerprint appears repeatedly in file entries corresponding to multiple different relay terminal identifier strings. If the fingerprint of the same acquisition device and the same fragment summary value are reproduced synchronously at multiple relay node locations, they are registered as homologous fission traces; if the fragment summary values ​​are consistent but the acquisition device fingerprints are inconsistent, the document entries involved are determined to originate from different acquisition terminals and are considered as independent observations; if the acquisition device fingerprints are consistent but the fragment summary values ​​are inconsistent, the document entries involved are determined to be multiple independent acquisitions by the same terminal at different times and are also considered as independent observations.

[0037] For example, a video recording of a rescue action generated by rescuer A's handheld terminal is successively transferred to the wrist terminals of team members B, C, and D along the chain. A's collection device fingerprint and the corresponding segment summary value will appear simultaneously on the file entries at the three relay nodes of B, C, and D, and will be registered as homologous fission traces.

[0038] Preferably, the homologous fission trace annotation table uses the fingerprint of the acquisition device as the group index. Under the fingerprint of the same acquisition device, several homologous fission trace entries are attached according to the reproduction summary value. Each entry records the corresponding field document entry number and the relay terminal identifier string.

[0039] It is understandable that the homologous fission trace annotation table only marks the trace indexes that are determined to be homologous. A complete collection of document entries is still needed within the aforementioned field evidence genealogy record. The purpose of the collection step is to expand the trace index into a set of document entries that can be directly accessed for subsequent independent evidence stripping, and simultaneously supplement document entries not covered by the trace annotation table that are located at the end of the same relay sequence and have not yet participated in the reproduction comparison. In the collection operation, starting with each collection device fingerprint group in the homologous fission trace annotation table, all field document entries belonging to that group are extracted, and their fragment summary value sequences are compared one by one with the collection device fingerprint string. When both are consistent segment by segment and character by character, it is considered a precise match. In the collection embodiment, for each precise match set of field document entries, the process traces backward along the relay sequence bound to each entry, that is, tracing upwards level by level according to the upstream node's backreference field until a node with an empty backreference field is reached. This node corresponds to the initial collection rescuer terminal and is marked as the collection starting point for that document entry. If multiple file entries that are precisely matched point to the same collection starting point after tracing back along their respective relay sequences, they are determined to constitute a homogeneous fission cluster, and the file entries under this cluster are all included in the homogeneous fission file set. In one embodiment, the homogeneous fission file set is stored in independent groups with the collection starting point as the identifier. Each file entry in the group retains the original record of its relay terminal identifier string, terminal transfer timestamp, and upstream node back-pointing field, so as to facilitate the subsequent construction of the transmission node topology map and the separation of independent evidence along the relay path.

[0040] Step S103: Read the terminal transfer timestamp and transfer chain marker of the same source fission file, trace the relay order of each terminal, and construct a transmission node topology map.

[0041] From the homogeneous fission file set, each field file entry is read sequentially, along with its associated terminal transfer timestamp and transfer chain marker. The relay terminal identifier string and upstream node retrace field in the transfer chain marker are expanded along the chain to obtain the timing information and upstream / downstream pointing relationships for each terminal relay entry. Based on this timing information and upstream / downstream pointing relationships, all relay terminal identifier strings under the same acquisition device fingerprint are paired, with downstream terminals matched to their upstream terminals using the upstream node retrace field. The pairing rationality is verified using the order of the terminal transfer timestamps to determine the relay sequence and upstream / downstream connection position of each terminal, resulting in a terminal relay sequence with directional pointing. Based on the terminal relay sequence with direction pointing, each relay terminal identifier string is abstracted as a node, and the paired upstream and downstream connection positions are abstracted as directed edges pointing from the upstream terminal to the downstream terminal. The corresponding terminal transfer timestamp and on-site file entry number are attached to the node, and the pointing record of the upstream node back pointer field is attached to the directed edge. A transmission node topology diagram with the collection starting point as the source node and the aggregation node as the sink node is drawn.

[0042] In mountain fire fighting scenarios, although the various field file entries within the same source fission file set have been determined to originate from the same original acquisition terminal, their relay paths in the blind zone network have not yet been explicitly depicted. In order to restore this relay information from scattered timestamps and backreference fields into a structured topology that can be called by the world model link layer, it is necessary to trace the relay order and draw the transmission node topology map based on the aforementioned source fission file set.

[0043] In one possible implementation, the terminal transfer timestamp and transfer chain marker of each field file entry are read sequentially from the homogeneous fission file set. The terminal transfer timestamp records the absolute moment when the entry was written on the corresponding relay terminal, expressed in a standard time format; the transfer chain marker is embedded in the file's extended attribute area in the form of a linked list, containing a relay terminal identifier string and an upstream node backreference field. After unrolling along the chain, the timing information and upstream / downstream pointing relationships of each terminal relay entry are obtained.

[0044] It should be noted that during mountain fire fighting, the relay coordination among team members is often affected by the terrain's undulations. A video of a fire spreading from a collection point halfway up the mountain may be relayed from Team A's handheld terminal to Team B's wrist-mounted terminal, then from Team B to Team C's backpack terminal, and finally from Team C to the relay terminal at the foot of the mountain. Each relay generates a new field file entry on the next hop terminal, with its upstream node backreference field pointing to the node sequence number of the previous hop terminal.

[0045] Specifically, the relay order restoration process uses the upstream node's back-pointing field as the main thread and the terminal's saved timestamp as the verification method. For all relay terminal identifier strings under the same acquisition device fingerprint, the downstream terminal is paired with its direct upstream terminal along the upstream node's back-pointing field, forming an upstream-to-downstream pairing relationship.

[0046] For example, if the upstream node pointer field of the file entry corresponding to Team B's wrist-mounted terminal points to Team A's handheld terminal, then Team A and Team B form a pair, with Team A being the upstream and Team B the downstream. Further, after pairing, the order of the terminal transfer timestamps is used for validity verification. The verification rule is set so that the transfer timestamp of the upstream terminal should be earlier than that of the downstream terminal, and the time difference should fall within a preset reasonable relay duration range. The reasonable relay duration range is determined based on the quantile statistics of historical relay task handover times, and is set between 2 and 60 seconds. This range corresponds to the normal time required for adjacent terminals to complete the handover of weapons, confirm passwords, and transfer materials or information in mountain fire fighting. Less than 2 seconds usually indicates that the same terminal is repeatedly registering while moving, while more than 60 seconds exceeds the normal relay rhythm and may involve pauses, detours, or non-relay actions. If the upstream transfer timestamp is later than the downstream transfer timestamp in a pairing, or the time difference Δt does not fall within the aforementioned interval, the pairing is marked as a suspicious pairing and enters the cross-validation stage. During validation, three independent pieces of information are retrieved simultaneously: the upstream terminal identifier U1 pointed to by the upstream node's retracement field in the downstream node entry, the relay terminal identifier string U2 in the upstream terminal's local field file entry, and the positioning coordinate distance d between the two ends at the pairing time. When U1 and U2 match and d is less than 30 meters, it indicates that the relay relationship between the two ends has been mutually verified by the topology field and spatial evidence. The timestamp deviation is attributed to the terminal's local clock drift. The pairing is retained based on the topology relationship pointed to by the retracement field, and the timestamp is corrected using the downstream clock reference. When U1 and U2 are inconsistent, or although consistent, d is greater than 30 meters, it indicates that the upstream and downstream relationship declared by the retracement field cannot be supported by independent fields and field spatial evidence. This is determined to be a retracement field error, and the incorrect pairing is discarded. Pairings that pass the above validation are sequentially connected along the direction of the acquisition starting point to form a terminal relay sequence with directional pointing.

[0047] In one embodiment, the node is attached with the corresponding terminal transfer timestamp and on-site file entry number to record when the node received and transferred which file entry; the directed edge is attached with the pointer record of the upstream node's back pointer field to trace back the original source of the relay.

[0048] Understandably, the drawn transmission node topology uses the collection starting point as the source node and the convergence node as the sink node. The intermediate layer nodes are arranged sequentially according to the relay order, and the directed edges between nodes reflect the actual relay path of evidence in the blind zone network. The transmission node topology is stored independently in the command center's link layer buffer area, using the fingerprint of the collection device as the group identifier, for subsequent relay transmission path identification and independent evidence separation stages to be called by group.

[0049] Step S104: By analyzing the transmission node topology, identify the relay transmission path of the same source fission files, locate the original acquisition source, and obtain the relay transmission path with the acquisition starting point.

[0050] The in-degree and out-degree values ​​of each node are read from the transmission node topology graph. Nodes with an in-degree of zero and a non-zero out-degree are identified as candidate source nodes. The relay terminal identifier string corresponding to the candidate source node is matched with the identifier string of the rescuer terminal to which the fingerprint of the acquisition device confirmed in the previous step belongs. If they match, the candidate source node is determined to be the source node corresponding to the original acquisition source. Based on the source node, a forward traversal is performed along the directed edges in the transmission node topology graph, sequentially passing through the downstream nodes pointed to by each directed edge. The relay terminal identifier strings and the order of the directed edges are recorded until a node with an out-degree of zero is reached, thus obtaining a candidate relay path chain from the source node to the node with an out-degree of zero. Based on the candidate relay path chain, the field file entry numbers attached to each node on the path chain are sequentially connected along the path. It is verified whether the upstream and downstream connection positions recorded by the directed edges between adjacent nodes on the path chain match the upstream node back pointer field of the corresponding file entry in the same source fission file set. The connection between adjacent nodes where there is no match is determined as a breakpoint branch and is removed to obtain a relay transmission path with the acquisition starting point.

[0051] In the topology analysis phase, candidate source nodes are first identified based on the condition that the in-degree is 0 and the out-degree is not 0. The original acquisition node is then confirmed by fingerprint binding verification from the acquisition device. A forward traversal is then performed along the directed edges to obtain several candidate relay path chains originating from the source node. At this stage, the candidate chains exist only as intermediate products and are not the final deliverables. Subsequently, each candidate chain is checked hop-by-hop to verify whether the upstream and downstream connection positions recorded on the directed edges between adjacent nodes match the upstream node backreference field of the corresponding file entry in the same source fission file set. Edges that do not match are identified as breakpoints and removed, and branches that cannot be traced back to the source node due to breakpoints are discarded. After removal, if multiple branches extending from the source node still exist, the branch with the most valid pairings is converged as the unique link. If there is a tie, the branch with the most field file entries attached to it is selected.

[0052] In one possible implementation, the in-degree and out-degree values ​​of each node are read sequentially from the transmission node topology graph. The in-degree value indicates the number of directed edges of that node, and the out-degree value indicates the number of directed edges emitted by that node. Nodes with an in-degree of zero and a non-zero out-degree are determined as source node candidates. In the example, the node set of the transmission node topology graph only covers relay terminals within the blind zone; the feedback from the foothill relay terminal to the command center is not considered a directed edge within the topology graph. The foothill relay terminal receives relay files pushed by the wrist-worn terminal of rescuer B and the backpack terminal of rescuer C upstream along the chain, with an in-degree of 2; however, it has no downstream terminals within the topology graph, resulting in an out-degree of 0. Therefore, it is a sink node with a non-zero in-degree and is not included in the source node candidate list. Rescuer A, located halfway up the mountain, holds a handheld terminal. No terminal upstream pushes relay files to him, so his in-degree is 0. Downstream, he points to Rescuer B's wrist-mounted terminal and completes one forwarding, so his out-degree is 1. This satisfies the condition that the in-degree is 0 and the out-degree is not 0, so he is judged as a candidate source node.

[0053] It should be noted that the candidate source nodes obtained solely based on in-degree and out-degree determination may result in multiple isolated nodes with zero in-degree in the event of a broken link in some relay paths. Therefore, it is still necessary to use the fingerprint of the data collection device for binding and verification in order to truly confirm the original data collection source.

[0054] Specifically, the binding and verification step of the source node candidate compares the relay terminal identifier string corresponding to the source node candidate with the acquisition rescuer terminal identifier string to which the acquisition device fingerprint confirmed in the previous step belongs. The acquisition device fingerprint is a fixed-length hexadecimal string obtained by concatenating the hardware serial number, sensor model code, and firmware version number in a preset order separated by vertical lines and then performing SHA-256 digest operation. The acquisition rescuer terminal identifier string to which it belongs has been bound and archived as a group identifier when grouping the same source fission file set. If the relay terminal identifier strings of the two are identical character by character, the source node candidate is confirmed to be the source node corresponding to the original acquisition source; if they are inconsistent, the source node candidate is added to the entry list to be reviewed and its participation in path construction is temporarily suspended until new field file entries or genealogy records are added under the acquisition device fingerprint group and then matched again.

[0055] For example, the identifier string of Team Member A's handheld terminal appears as a candidate source node at a position with an in-degree of zero in the topology graph, and is also bound to the fingerprint group of the data collection device as the identifier string of the rescuer's terminal. Since the two are consistent, Team Member A's handheld terminal node is identified as the source node. Further, based on the source node, a forward traversal is performed along the directed edges in the transmission node topology graph. The forward traversal starts from the source node and goes deeper layer by layer along each outgoing directed edge, passing through the downstream nodes pointed to by the directed edges in sequence. Along the way, the order of the relay terminal identifier strings and the pointing of the directed edges is recorded until the sink node with an out-degree of zero is reached, thus obtaining the candidate relay path chain from the source node to the sink node.

[0056] For example, starting from the handheld terminal of team member A, passing through the wrist terminal of team member B and the backpack terminal of team member C, and arriving at the aggregation node corresponding to the relay terminal at the foot of the mountain, a candidate relay path chain is formed.

[0057] Preferably, if there are branching situations in the topology graph, the forward traversal process extends independently downward at each branching node to obtain multiple parallel candidate relay path chains, each path chain starting from the source node and ending at the convergence node.

[0058] In one embodiment, the path chain verification step sequentially concatenates the field document entry numbers attached to each node on the candidate relay path chain along the path, reads the upstream and downstream connection positions recorded by the directed edges between adjacent nodes, and compares them with the upstream node retrace field carried by the corresponding document entry in the same source fission document set. If the pointing relationships of the two match item by item, the adjacent node connection segment is retained; if the directed edge corresponding to a certain adjacent node connection segment does not match the upstream node retrace field of the document entry, the connection segment is determined as a breakpoint branch and is removed.

[0059] Understandably, after removing breakpoint branches, the remaining continuous nodes are sequentially connected along the source node direction, covering the previously generated candidate relay path chain to form a uniquely delivered relay transmission path, no longer retaining candidate chain forms. The path starting point is marked as the original data collection source terminal confirmed through binding and verification, intermediate nodes are arranged sequentially according to the relay order, and the end points to the aggregation node. The entire chain is independently stored in the database using the fingerprint of the data collection device as the group identifier. The link layer cache carrying this result is deployed between the command center access server and the source tracing analysis server, logically located above the session aggregation layer and below the business persistence layer. The capacity is planned to be no more than 2MB per fingerprint group and no less than 4GB in total capacity, organized using a key-value structure, where the key is the fingerprint of the data collection device and the value is an ordered node sequence and a timestamp. The cache exposes two types of interfaces: one for the source tracing analysis module to pull the complete path by fingerprint for backtracking and positioning, and the other for the scheduling module to look up the upstream relay node by the end aggregation node for task assignment. Path entries are rolled out by the cleanup thread after the next round of data collection is triggered or the retention time exceeds 30 minutes.

[0060] Step S105: Tracing the homologous fission file copies along the relay transmission path with the collection starting point, and based on the homologous relationship between the original file fragment summary and the fingerprint verification copy of the collection device, the duplicate evidence formed by multiple transfers is separated, and the original field files directly obtained by each original collection terminal are classified into an independent evidence set.

[0061] The process involves sequentially tracing all copies within the homologous fission file set along the relay transmission path with the acquisition starting point. Each copy's associated field file entry is attached to the corresponding node on the path chain according to its relay node position. The field file entry number and relay terminal identifier string for each copy at each node position are recorded, resulting in an attachment list for each copy at each node position on the path chain. The field file entry located at the starting node of the path chain in the attachment list is designated as the starting file. Based on the attachment list, the original file fragment summary and acquisition device fingerprint of each copy on the path chain (excluding the starting file) are read one by one. These are compared and verified segment by segment and character by character with the original file fragment summary and acquisition device fingerprint carried by the starting file. If they match, the copy is determined to have a homologous relationship with the starting file and is marked as a transferred copy. If they differ, the copy is removed from the current path chain's tracing scope and returned to the homologous fission file set for further verification. Based on the calibration results of the transferred copies, all transferred copies located after the starting node on the path chain are stripped from the candidate file sequence participating in the world model execution result layer update, and only the starting file is retained as the original field file directly obtained by the original acquisition terminal, and the starting file is included in the independent evidence set.

[0062] In the scenario of evidence transmission in communication blind spots during mountain fire fighting, the relay transmission path obtained in the previous stage, with the starting point of the acquisition, has clearly identified the source node of the original acquisition terminal and the intermediate relay nodes along the chain. However, the copies in the same-source fission file set have not yet been verified and separated according to the path distribution, and there is still a risk that they may be mistakenly treated as independent observations and repeatedly updated to the world model execution result layer. To eliminate this risk, based on the starting file at the source node, the original file fragment summary and acquisition device fingerprint of each copy at the intermediate relay node are read one by one, and compared segment by segment and character by character with the original file fragment summary and acquisition device fingerprint carried by the starting file. Copyes that match both are determined to be same-source transfer copies and are separated from the candidate file sequence participating in the update of the world model execution result layer, with only the starting file retained for the update. Copyes newly added due to the update of the phylogenetic record or those with inconsistent comparisons are returned to the same-source fission file set for review. After the above path-level verification, the original field files directly obtained by each original acquisition terminal can be used as independent evidence items to enter the world model execution result layer.

[0063] In one possible implementation, a hash index is first established at the starting node of the path, using the relay terminal identifier string as the key and the node sequence number as the value. The index capacity is reserved according to the path length, and the load factor is controlled within 0.75. Then, starting from the starting point, the relay transmission path is advanced hop by hop. At each hop, the field file entry corresponding to the current node's relay terminal identifier string is read. A batch retrieval method is used, with a single batch limit of 200 entries. Each entry is compared with the original file fragment digest bound to the entry in the same source fission file set and the fingerprint of the acquisition device. If a match is found, the entry is written to the mounting slot of the corresponding node on the path chain according to its node sequence number. If a hop read times out (the timeout threshold is set to 3 seconds), the hop is skipped and the mounting slot is marked as empty, and the traversal continues to the next node to avoid the entire path being interrupted due to a single point of blockage. If there are multiple matching records for the same node, they are arranged in ascending order of the writing timestamp and mounted together. After all nodes have been traversed, the nodes corresponding to the empty slots are backfilled, and only the empty slot nodes are checked once, thus completing the full return of duplicate evidence while maintaining the relay order.

[0064] Specifically, the mount list is organized and stored using the path chain node number as the primary key. Several mount entries are registered under each node position. The content of each entry includes the on-site file entry number, the relay terminal identifier string, and the terminal transfer timestamp corresponding to the file entry.

[0065] For example, if the relay transmission path consists of four nodes connected in series: Team A's handheld terminal, Team B's wrist-mounted terminal, Team C's backpack terminal, and the relay terminal at the foot of the mountain, then the mounting list will register the field document item number received by each of these four nodes.

[0066] It should be noted that the "rescue worker terminal" mentioned here refers to the handheld terminal of Team Member A at the starting node of the path chain. This terminal is worn by the rescuer at the forefront of the fire line, directly observing the fire and recording raw data such as fire spread videos, smoke concentration sampling clips, handling action videos, and location coordinates. It is the only device in the entire relay link responsible for on-site data collection and initial data transfer, forming a clear division of labor with downstream relay nodes such as Team Member B's wrist-worn terminal, Team Member C's backpack terminal, and the foothill relay terminal, which only perform data transfer and verification functions. The on-site file entries at this starting node in the load list are all directly generated and written to local storage by Team Member A's handheld terminal the moment data collection is completed, without any intermediate relay transfer or format repackaging. These are marked as the starting file and serve as the comparison benchmark for subsequent source verification, used to compare the original file fragment summary with the fingerprint of the collection device against the identical copies at the three level nodes (B, C, and the foothill).

[0067] For example, the fire spread video file entry mounted under the path starting node where Team Member A's handheld terminal is located is the starting file. Further, the same-source verification step is performed on each copy in the mounting list except for the starting file. The verification process reads the original file fragment summary and acquisition device fingerprint carried by each copy, and compares them segment by segment and character by character with the original file fragment summary and acquisition device fingerprint carried by the starting file. The comparison of the original file fragment summary adopts a sequence-level strict matching method; the original file fragment summaries carried by the two file entries must be completely identical in three aspects: the total number of fragments, the order of fragments, and the value of each fragment summary. The comparison of the acquisition device fingerprint adopts a string-level strict matching method; every character of the two fingerprint strings must be completely identical. If both comparisons pass, the copy is determined to have a same-source relationship with the starting file and is marked as a transferred copy. If the acquisition device fingerprint or the original file fragment summary has missing fields due to overwriting of extended attribute areas during the transfer process, causing the comparison to fail, the copy is added to the supplementary list and temporarily suspended from participating in this round of stripping.

[0068] Preferably, the calibration results of the transferred copies are stored in the stripping and processing cache as field entries. Each entry records the entry number of the on-site file corresponding to the transferred copy, the location of the path node, and the entry number of the corresponding starting file, which is used to connect the starting file with all the transferred copies under its name.

[0069] In one embodiment, the stripping process is performed sequentially based on the calibration results of the transferred copies. The stripping process removes all transferred copies located after the starting node on the path chain from the candidate file sequence participating in the world model execution result layer update. The transferred copies themselves are not discarded, but are backfilled into the corresponding acquisition device fingerprint group under the homogeneous fission trace annotation table according to their corresponding relay terminal identifier string and terminal transfer timestamp, for subsequent world model link layer updates.

[0070] Understandably, after the stripping is complete, only the starting point file located at the path start node is retained as the original field file directly obtained by the original acquisition terminal. This starting point file is then categorized into the independent evidence set according to its corresponding acquisition device fingerprint. Each starting point file in the independent evidence set corresponds to one independent observation, and the number of independent observations is counted in the update count of the world model execution result layer based on the number of entries in the starting point file.

[0071] Step S106: Based on the independent evidence set, associate each independent piece of evidence with the execution result of the natural and social system in the world model, and look up the terminal transfer timestamp and transfer chain mark corresponding to the duplicate evidence to the transmission node topology map, determine the node and the order of successive relays, and associate it with the node connection position to form an evidence backflow mapping.

[0072] Each starting point file is extracted one by one from the independent evidence set. The on-site handling actions and completion status fields recorded in the starting point file are read. The fingerprint of the acquisition device corresponding to the starting point file is matched with the registered execution action entries in the execution result layer of the natural and social system in the world model according to the handling action category and acquisition coordinate position. The completion status field of the starting point file is attached to the matched execution result entry as the independent observation value of the execution result, thus obtaining the independent evidence attachment mapping of the execution result layer. The terminal transfer timestamp and transfer chain mark corresponding to each transfer copy are extracted sequentially from the calibration results of the transferred copies. The data is then retrieved back to the transmission node topology diagram. The node where the transferred copy is located is located according to the relay terminal identifier string. Then, the adjacent nodes are connected in series according to the pointer field of the upstream node. The rationality of the series connection result is verified by the order of the terminal transfer timestamps. The node and its relay order are determined to obtain the location result of the transferred copy in the transmission node topology diagram. Based on the node location results of the topology graph, the field file entry number of the transferred copy is attached to the node connection position marked by the directed edge between the corresponding node and its upstream and downstream nodes in the transmission node topology graph. The independent evidence attachment mapping of the execution result layer and the node connection attachment of the transferred copy on the topology graph are merged and archived to obtain the evidence backflow mapping covering the execution result layer and the link layer.

[0073] At the mountain fire fighting command center, the world model provides a unified representation of the disaster environment, response entities, actions taken, completion status, and communication nodes, divided into two update layers: the natural-social system execution result layer and the link layer. The starting point documents within the independent evidence set obtained in the previous stage have been stripped of duplicate evidence and can be considered independent observations directly participating in the update of the execution result layer; while the transferred copies remain under the homologous fission trace annotation table, only carrying relay path information, and should flow back to the link layer. Based on this, an evidence backflow mapping is formed, covering both the execution result layer and the link layer update paths.

[0074] In one possible implementation, each starting point file is extracted one by one from the independent evidence set, and the on-site handling actions and completion status fields recorded in the starting point files are read. The on-site handling action field represents the specific action category performed by the rescuer in mountain fire fighting using an enumerated encoding method. The action category values ​​include four categories: creating firebreaks, suppressing open flames, clearing remaining fires, and evacuating personnel. The completion status field consists of two items: progress percentage and end marker, used to characterize the progress of the handling action at the time of data collection.

[0075] Specifically, the matching process between the starting point file and the execution result layer of the natural and social system uses the fingerprint of the data acquisition device as the primary index and the category of the action and the data acquisition coordinates as the matching criteria. Each registered action entry in the execution result layer is associated with the fingerprint of the data acquisition device of the corresponding subject, the action category label, and the coordinate block number of the area it occupies. The matching process reads the fingerprint of the data acquisition device corresponding to the starting point file, searches for a subset of action entries with matching fingerprints in the execution result layer, further compares the action categories within this subset, and retains only entries with matching action categories; then compares the data acquisition coordinates with the coordinate block number associated with the action entry. If both fall into the same coordinate block, it is determined that the starting point file and the action entry are matched.

[0076] For example, the starting point file generated by rescuer A's handheld terminal has its device fingerprint identified as A's terminal fingerprint string, the action category as "opening a quarantine zone," and the collected coordinates fall within the fire zone numbered B3. An execution action entry under A's name, with the action category "opening a quarantine zone" and coordinate zone number B3, is detected in the execution result layer and matches the starting point file. Further, the completion status field of the starting point file is attached to the matched execution action entry as an independent observation value for that action. The attachment adopts an overlay registration method, that is, an independent observation slot is created under the execution action entry, and the field file entry number corresponding to the starting point file, along with the progress percentage and end marker of the completion status field, are sequentially written into the slot. Simultaneously, the collection time in the metadata accompanying the starting point file is registered as the observation time reference, for S107 to perform execution status continuity judgment and inconsistency correction according to the collection time sequence. The terminal transfer timestamp of the starting point file is only registered for attachment at the link layer connection position and is not used as the time basis for the execution result layer. If multiple starting files from fingerprints of different acquisition devices are received successively under the same execution action entry, independent observation slots are opened according to the fingerprints of each acquisition device, without overlapping each other, thus obtaining independent evidence attachment mapping at the execution result layer.

[0077] It should be noted that, in parallel with attaching the starting file to the execution result layer, the transferred copy only carries information on the changes in transmission nodes and the relay order, and its attachment object is the transmission node topology map within the link layer.

[0078] Specifically, the terminal transfer timestamp and transfer chain marker corresponding to each transfer copy are sequentially extracted from the calibration results of the transferred copies, and the data is then retrieved back to the transmission node topology graph. The retrieval process first searches for a node with the same name in the topology graph based on the relay terminal identifier string recorded by the node with the largest terminal transfer timestamp within the transfer chain marker, and attaches the transfer copy to that node. Then, based on the pointer of the upstream node's backreference field, the direct upstream node of that node is found, and the connection between the two is abstracted as a concatenation criterion. Finally, the rationality of the concatenation criterion is verified by the order of the terminal transfer timestamps, requiring that the transfer timestamps of other attached copies on the upstream node are earlier than the transfer timestamp of this particular transfer copy.

[0079] In one embodiment, if both the upstream node's pointer and the terminal's transfer timestamp verification criteria pass simultaneously, the node containing the transferred copy and its relay order are confirmed. If the upstream node's pointer is readable but the transfer timestamps are in reverse order, the pointer of the pointer is taken as the correct one. If the pointer itself is missing, the node whose timestamp is closest to and earlier than the transferred copy's timestamp is selected as its upstream node, and the node whose timestamp is closest to and later than the transferred copy's timestamp is selected as its downstream node, based on the order of the terminal's transfer timestamps. The nodes are ultimately used to form the node location result of the topology map of the transferred copy. This result is a four-tuple record, which includes the node identifier N to which the copy belongs, its upstream node identifier U, its downstream node identifier D, and the sequence number K of the copy in the entire relay link. For example, if a copy is identified as N3, its upstream node identifier N2, its downstream node identifier N4, and its relay sequence number K is 3 after the above determination, then the four-tuple N3, N2, N4, and 3 will be written into the location result table and synchronized to the corresponding node position in the topology map for subsequent tracing and link reconstruction.

[0080] Preferably, the field document entry number of the transferred copy is attached to the node connection position marked by the directed edge pointing from its upstream node to the node where the copy is located. When attaching, the directed edge is used as the carrying unit, and the field document entry numbers of the transferred copy are registered on the edge in sequence according to the transfer timestamp. The two ends of the edge indicate the upstream terminal and the downstream terminal, respectively.

[0081] For example, if team member A's handheld terminal is upstream and team member B's wrist-type terminal is downstream, then the transferred copy attached to node B is registered on the directed edge from A to B, reflecting that the copy is formed by a relay from A to B.

[0082] It is understandable that a blind zone network refers to adjacent transmission segments where stable communication links cannot be directly established between data collection nodes. It consists of data collection terminals, relay terminals, and a convergence node. These nodes are connected sequentially according to their spatial adjacency, and evidence must be relayed through these nodes to reach the convergence terminal. Understanding this concept, the independent evidence mapping at the execution result layer can be combined with the node mapping and archiving of copies on the topology map, resulting in an evidence backflow mapping covering both the execution result layer and the link layer. This mapping uses the fingerprint of the data collection device as a common primary key for cross-layer concatenation. The execution result layer records the independent observation values ​​of on-site handling actions and completion status, while the link layer records the relay path and the order of preceding and following nodes in the blind zone network. Both parts are stored side-by-side under the same data collection device fingerprint.

[0083] Step S107: Based on the evidence backflow mapping, the independent evidence set is backflowed to the execution result of the natural and social system in the world model according to the execution result layer, and the duplicate evidence is backflowed to the corresponding transmission node according to the link layer. Based on the execution state continuity and relay path continuity in the world model, missing records and inconsistent items are added and revised, thus completing the hierarchical backflow update control of the execution result record and the relay transmission path record.

[0084] Based on the evidence backflow mapping, the observation values ​​of the execution result layer attached to each starting file in the independent evidence set are backfilled into the execution result entries of the natural and social system in the world model according to the fingerprint of the acquisition device. The node connection positions of duplicate evidence attached to the transmission node topology map are backfilled into the corresponding transmission node records in the link layer of the world model according to the relay terminal identifier string, so as to obtain the initial backflow results of the execution result layer and the link layer. For the initial backflow results, the completion status field of the execution result entries arranged chronologically along the acquisition time is read to determine the continuity of its execution status in the world model. If the percentage jump in the progress of two adjacent observations exceeds the preset reasonable progress increase range, the intermediate gap period is marked as a missing record item. Only independent evidence values ​​that fall within the gap period and whose acquisition device fingerprints are different from the fingerprints of the two adjacent observations that caused the jump are retrieved from the evidence backflow mapping and added. If no independent evidence that meets the conditions is retrieved, the missing record item is marked as unfillable and the gap is retained. The two adjacent observation values ​​that caused the jump cannot be used as filler values. Then, the upstream and downstream connection positions of the transmission node records arranged chronologically along the relay order are read to determine the continuity of the relay path. The upstream and downstream breakpoint positions are marked as missing path items, and the corresponding node connection positions are retrieved from the link layer part of the evidence backflow mapping and added. Based on the results of filling in the missing records and missing paths, each field in the world model is compared with the observed values ​​after the backflow. Fields with inconsistent values ​​are defined as inconsistencies. For inconsistencies at the execution result layer, the independent evidence value corresponding to the later acquisition time is used to cover the original field value corresponding to the earlier acquisition time, based on the order of acquisition time in the evidence backflow mapping. The execution result layer does not use the terminal transfer timestamp as the sorting and overwriting basis. For inconsistencies at the link layer, the node connection position is revised based on the order of terminal transfer timestamps, thus completing the hierarchical backflow update control of execution result records and relay transmission path records.

[0085] At the mountain fire fighting command center, the previous stage had obtained an evidence backflow mapping covering the execution result layer and the link layer. Within this mapping, observation values ​​of independent evidence and node connections of duplicate evidence were stored side-by-side, using the fingerprints of the data acquisition devices as the common primary key. At this point, the original execution result entries and transmission node records of the natural and social systems in the world model had not yet received the backflow data injection, retaining the old values ​​from the previous update. Furthermore, there were record gaps due to communication interruptions in blind spots. Therefore, it was necessary to implement hierarchical backflow update control for the two layers of records based on the aforementioned evidence backflow mapping.

[0086] In one possible implementation, the initial backflow is divided into two parallel paths: the execution result layer and the link layer. The execution result layer path uses the fingerprint of the acquisition device as the backfill index, and writes the progress percentage, end identifier, and on-site file entry number of the completion status fields of each starting file in the independent evidence set into the corresponding natural and social system execution result entries in the world model according to the alignment and hit relationship registered in the evidence backflow mapping. The link layer path uses the relay terminal identifier string as the backfill index, and writes the connection position of the directed edge node on the transmission node topology graph of the duplicate evidence, along with the on-site file entry number, into the corresponding transmission node record in the link layer of the world model.

[0087] For example, the starting point file generated by rescuer A's handheld terminal is written into the independent observation slot under the execution action item with action category "opening a quarantine zone" and coordinate block B3 in the execution result layer, with a progress percentage of 70% and an end marker of "not finished"; duplicate evidence generated by the same original evidence on the terminals of relay team members B and C is written into the node connection positions of the two directed edges from A to B and from B to C in the link layer according to the relay terminal identifier string, to obtain the initial backflow result.

[0088] It should be noted that the initial backflow results have not yet addressed the missing time periods and inconsistent values ​​in the old values. Further additions and revisions are needed based on the continuity of execution status and relay path in the world model. For the initial backflow results, the progress percentage of two adjacent observations under the same execution result entry is read sequentially according to the acquisition time, and the continuity of execution status is determined pair by pair. The preset reasonable progress increase range is set to an upper limit of 5 per minute for the progress percentage increase, and negative increases are not allowed. For observations with an interval of T minutes between two adjacent acquisition times, the difference in their progress percentages is denoted as D. When 0 <= D <= 5 * T, it is considered continuous; if the interval is less than 1 minute, it is counted as 1 minute. If D is greater than 5*T, and the progress jumps from 20% to 80% within 2 minutes (D=60, far exceeding the upper limit of 10), the intermediate gap is marked as a missing record. Independent evidence values ​​are retrieved from the evidence backflow mapping that fall within this gap and whose acquisition device fingerprints differ from the acquisition starting point of the adjacent observation causing the jump. These values ​​are then inserted in the order of acquisition time. If no independent evidence meets the conditions, the missing record is marked as unrecoverable, and the gap is retained. If the observations causing the jump originate from the same acquisition starting point, the jump in its own observation sequence indicates a discontinuous acquisition rhythm at that starting point. Interpolating with its own values ​​can only smooth out a single source of loss and cannot cross-verify the on-site progress. Therefore, it is prohibited to use values ​​from the same starting point file or the same acquisition starting point as filler values. If D is less than 0, meaning the progress percentage of the later observation is lower than the previous observation, there is no intermediate gap. This is marked as an inconsistency and proceeds to the following inconsistency revision process, handled according to the late acquisition time coverage rule. Similarly, the upstream and downstream connection positions recorded by two adjacent transmission nodes under the same acquisition device fingerprint are read in relay order to determine the continuity of their relay path. If the downstream connection position and the upstream connection position of two adjacent nodes do not match, such as the downstream terminal identifier string recorded by the previous node being B while the upstream terminal identifier string recorded by the next node is D, then the breakpoint position is marked as a missing item in the path, and the node connection position with the same fingerprint is retrieved from the link layer part of the evidence backflow mapping to fill in the missing item.

[0089] In one embodiment, after the data is added, the original execution result fields in the world model are compared with the observed values ​​after the backflow. The comparison uses the fingerprint of the acquisition device plus the acquisition time as the joint primary key to locate field pairs under the same execution action entry. The original field values ​​in the world model and the observed values ​​in the evidence backflow mapping are checked for numerical consistency. Any entries with inconsistent values ​​are marked as inconsistent items.

[0090] Preferably, the revision of inconsistencies at the execution result layer is based on the fingerprints of the acquisition devices and the order of acquisition times in the evidence backflow mapping. The revision process adopts the rule of later acquisition times overriding earlier acquisition times; that is, the observation value of independent evidence with a later acquisition time is used to overwrite the original field value. Simultaneously, the overridden original field value, along with its acquisition time, is retained in the historical revision stack of that entry for easy backtracking. The terminal transfer timestamp is only used to determine the link layer connection order and is not used as the basis for overriding at the execution result layer. This is because the transfer time of the starting file transmitted back via the relay link depends on the link length and transmission delay. When different independent evidence is transferred via links of different lengths, the order of transfer times becomes disconnected from the order of the on-site facts. Sorting by transfer time would mistakenly interpret link delay as on-site time sequence. The revision of inconsistencies at the link layer is based on the node connection position corresponding to the later relay terminal transfer timestamp to update the directed edge attachment record.

[0091] Understandably, after the execution result layer completes the filling of missing items and the correction of inconsistencies, its execution result entries are sequentially concatenated with the progress percentages and end flag values ​​of each collection in chronological order. After the link layer completes the corresponding processing, its transmission node records are sequentially concatenated with the upstream and downstream connection positions of each segment in relay order, forming a two-layer record along the fingerprint group of the collection device and stored in the world model.

[0092] In one embodiment, the hierarchical backflow update control is expressed using a state update equation. For each execution action entry e within the execution result layer, its completion state field is denoted as the entry state vector s. k (e)=(p k ,f k H k (e)), where p k f represents the progress percentage. k H is the end marker. k (e) represents the history revision stack; the set of independent observations attached to e in the k-th cycle is denoted as O. k (e)={(p i ,f i ,t i )},t i The acquisition time for this observation is s; the state transition equation for the entry in the execution result layer is s. {k+1} (e)=σ(s k (e),O k (e)), where σ is the hierarchical backflow update operator: first merge O in ascending order according to the acquisition time. k (e) and s k(e) Registered observation values; for gaps in the progress jump exceeding the reasonable increase range, interpolation is performed only if there are independent observations originating from other acquisition points and their acquisition times fall within the gap, in the order of acquisition times; otherwise, the missing item is marked as unfillable; for field bits with inconsistent completion status values, the independent observation value with the later acquisition time is used to cover the earlier value, and the covered value along with its acquisition time is pushed into H. k (e) Backtracking is provided. The state update equation for the link layer is L {k+1} =L k ∪ΔL k (Φ k ), where ΔL k (Φ k This refers to the directed edge attachment increment driven by the link layer portion of the evidence return mapping in this round. It is added in relay order, and in case of inconsistency, the node connection position corresponding to the later terminal's transfer timestamp is used as the standard for revision. The two formulas combined are M. {k+1} =F(M k ,Φ k ).

[0093] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A hierarchical backflow update method for the execution results of a generalized world model, characterized in that, include: Original field files recording the actions and completion status of natural and social systems are obtained from restricted networks in mountain fire communication blind spots. Fragment summaries and acquisition device fingerprints are extracted, and transfer chain markers are read. The transfer chain markers are allowed to have missing, incorrect, and terminal clock drift, forming a field evidence genealogy record. The fragment summary is traced back to the location of its recurrence and the fingerprint of the acquisition device along the chain to identify homologous fission traces, collect homologous fission files, and when the transfer chain is unreliable due to missing transfer chain markers, incorrect recording, or terminal clock drift, the homologous relationship between the field file entries is determined by comparing the consistency between the fragment summary and the fingerprint of the acquisition device. Read the terminal transfer timestamp and transfer chain marker of the same source fission file, trace the order of each terminal relay, and construct a transmission node topology map; Analyze the transmission node topology, identify the relay transmission path, locate the original data acquisition source, and obtain the relay transmission path with the data acquisition starting point; Tracing the same source fission document copies along the relay transmission path from the collection starting point, verifying the same source relationship of the copies, separating duplicate evidence, and classifying the original field documents directly obtained by each collection terminal into an independent evidence set; Link independent evidence to the execution results of the natural and social system, trace duplicate evidence back to the transmission node topology, determine the node and the order of succession, and link it to the node connection position to form an evidence backflow mapping; Independent evidence sets are fed back to the execution result layer according to the execution result layer, and duplicate evidence is fed back to the corresponding transmission node according to the link layer. Missing items and inconsistent items are added and corrected according to the continuity of execution status and relay path, thus completing the hierarchical backflow update control of execution result and relay transmission path records.

2. The hierarchical backflow update method for the execution results of a generalized world model according to claim 1, characterized in that, The process involves acquiring original on-site documents recording the actions and completion status of natural and social systems from restricted networks in mountain fire communication blind spots, extracting fragment summaries and acquisition device fingerprints, reading transfer chain markers, and forming an on-site evidence genealogy record, including: Retrieve the original field files recorded by the acquisition terminal, verify the integrity of the load, and remove any missing parts; A fixed-length sliding window is used to extract fragments from the file payload byte stream to generate fragment digest values; Read the hardware serial number, sensor model code and firmware version number, and concatenate them into a plaintext string by separating them with vertical lines in a preset order. Then, generate a fingerprint of the acquisition device as a 64-bit hexadecimal string by performing SHA-256 digest operation. Expand the relay terminal identifier string and the upstream node retrieval field, and archive them to form a field evidence genealogy record with the collection starting point as the root and the relay nodes as branches.

3. The hierarchical backflow update method for the execution results of a generalized world model according to claim 1, characterized in that, The recurrence location of the trace fragment summary and the chain retention of the fingerprint of the acquisition device are used to identify homologous fission traces and collect homologous fission files, including: Extract fragment summaries and acquisition device fingerprints from each field document entry in the genealogy record. Compare the fragment summaries one by one between each relay node, retrieve the recurring locations, and register the corresponding relay terminal identifier strings. Those fingerprints from the same acquisition device and the same fragment summary value that are synchronously reproduced at multiple relay nodes are registered as homologous fission traces; Tracing back along the relay sequence to the collection starting point where the index field is empty, file entries pointing to the same collection starting point are grouped into a homogeneous fission file set.

4. The hierarchical backflow update method for the execution result of a generalized world model according to claim 1, characterized in that, The terminal transfer timestamp and transfer chain marker for reading the homogeneous fission file are used to trace the relay sequence of each terminal and construct a transmission node topology graph, including: Read the terminal transfer timestamp and transfer chain marker bound to each field file entry from the same source fission file set, and expand the relay terminal identifier string and upstream node back pointer field along the chain; The downstream terminals are paired with the upstream terminals based on the upstream node's back pointer field. The pairing is verified by checking the timing of the terminal transfer timestamps. The relay sequence of each terminal and its connection position with the upstream and downstream are determined, resulting in a terminal relay sequence with directional indication.

5. The hierarchical backflow update method for the execution results of a generalized world model according to claim 1, characterized in that, The analysis of the transmission node topology map identifies the relay transmission path, locates the original data acquisition source, and obtains the relay transmission path with the data acquisition starting point, including: Read the in-degree and out-degree values ​​of each node from the transmission node topology graph, determine the node with no in-edge and out-edge as the source node candidate, and determine the source node corresponding to the original collection source if the relay terminal identifier string corresponding to the source node candidate matches the collection terminal identifier string of the confirmed collection device fingerprint. Traverse forward along the directed edges from the source node, recording the relay terminal identifier strings and the order of the directed edges, until a node with no outgoing edges is reached, thus obtaining the candidate relay path chain. The field file entry numbers mounted on each node in the path chain are concatenated along the path. The upstream and downstream connection positions of the directed edge records between adjacent nodes are verified to match the upstream node's back pointer field of the corresponding file entry. The connection lines between adjacent nodes that do not match are identified as breakpoint branches and removed, thus obtaining a relay transmission path with the acquisition start point.

6. The hierarchical backflow update method for the execution results of a generalized world model according to claim 1, characterized in that, The relay transmission path along the data acquisition starting point is used to trace the copies of the files from the same source, verify the source relationship of the copies, remove duplicate evidence, and classify the original field files directly obtained by each acquisition terminal into an independent evidence set, including: Tracing the copies within the same fission file set along the relay transmission path, attaching the field file entries to each copy to the corresponding nodes in the path chain according to the relay node they belong to, and defining the field file entry at the starting node of the path chain as the starting file; Read the fragment summaries of the remaining copies and compare them with the fingerprints of the acquisition devices and the starting file for verification. Mark the ones that match as the transfer copies. Remove the transferred copies after the starting node from the candidate file sequence and add the starting file to the independent evidence set.

7. The hierarchical backflow update method for the execution results of a generalized world model according to claim 1, characterized in that, The process of associating independent evidence with the execution results of the natural and social system, and backtracking duplicate evidence to the transmission node topology graph to determine the node and the order of relay, and associating it with the node connection position to form an evidence backflow mapping, includes: Extract each starting point file from the independent evidence set, read the on-site handling action and completion status fields, match the fingerprint of the collection device corresponding to the starting point file with the execution action entries registered in the execution result layer of the natural and social system according to the handling action category and collection coordinate position, and attach the completion status field to the matched execution result entries as independent observation values ​​to obtain the independent evidence attachment mapping of the execution result layer; The terminal transfer timestamp and transfer chain mark corresponding to the duplicate evidence are retrieved back to the transmission node topology map. The node is located according to the relay terminal identifier string. Adjacent nodes are linked according to the upstream node back pointer field and verified by the order of terminal transfer timestamps to determine the node and the relay order. The on-site document entry numbers of duplicate evidence are attached to the corresponding node connection positions marked by the directed edge, and merged and archived with the attachment mapping of independent evidence to form an evidence backflow mapping covering the execution result layer and the link layer.

8. The hierarchical backflow update method for the execution result of a generalized world model according to claim 1, characterized in that, The process of backflowing independent evidence sets to the execution result layer, backflowing duplicate evidence to the corresponding transmission node at the link layer, and supplementing missing items and revising inconsistencies based on the continuity of execution status and relay path, thereby completing the hierarchical backflow update control of execution results and relay transmission path records, includes: The execution result layer observation values ​​of each starting point file are backfilled into the execution result entries of the natural and social system according to the fingerprint of the acquisition device. The node connection position of duplicate evidence is backfilled into the corresponding transmission node record of the link layer according to the relay terminal identifier string. Read the completion status field of the execution result entries arranged along the collection time, mark the gap period where the progress jump exceeds the set increase range as the record missing item, and only fill in the independent evidence value that the collection time falls within the gap period and originates from other collection starting points. If there is no independent evidence that meets the conditions, mark the record missing item as unfillable. Read the upstream and downstream connection positions of the transmission node records arranged in the relay order, mark the upstream and downstream breakpoint positions as missing items in the path and fill them into the corresponding node connection positions to obtain the filling results of missing record items and missing path items.

9. The hierarchical backflow update method for the execution result of a generalized world model according to claim 4, characterized in that, The directional terminal relay sequence includes: Each relay terminal identifier string is abstracted as a node, and the upstream and downstream connection positions are abstracted as directed edges pointing from the upstream terminal to the downstream terminal. The corresponding terminal transfer timestamp and on-site file entry number are attached to the node, and the pointing record of the upstream node back pointer field is attached to the directed edge. A transmission node topology diagram with the collection starting point as the source node and the aggregation node as the sink node is drawn.

10. The hierarchical backflow update method for the execution result of a generalized world model according to claim 8, characterized in that, The results of filling in missing records and missing paths include: Each field in the world model is compared with the observed values ​​after reflow to determine if they are consistent. Fields with inconsistent values ​​are defined as inconsistencies. For inconsistencies at the execution result layer, the independent evidence value corresponding to the later acquisition time is used to overwrite the original field value corresponding to the earlier acquisition time, without using the terminal transfer timestamp as the basis for overwriting at the execution result layer. For inconsistencies at the link layer, the node connection position corresponding to the later terminal transfer timestamp is used to revise the earlier record, thus completing the hierarchical reflow update control of execution result records and relay transmission path records.

Images