A beidou grid space-time intelligent agent closed-loop scheduling control method

Abstract

The application provides a Beidou grid space-time intelligent agent closed-loop scheduling control method, comprising: obtaining an artificial confirmation voucher, extracting a task identifier and a confirmation event grid code, and identifying a voucher grid version and a voucher generation time; extracting a target migration record from a world model according to the task identifier, and identifying a migration stage after the artificial confirmation is completed; triggering the three conditions of the current target grid not belonging to the current allowed grid mask, the confirmation event grid code being separated from the current or historical allowed mask within the trace window, and the confirmation crossing the grid and the current mask not covering the migration path; any of the three conditions is established to enter trajectory verification and issue a block; analyzing a real-time motion trajectory, forming a no-entry conclusion only in the form of continuous separation from the mask coverage, and recording the back-and-forth crossing; issuing a control task by an intelligent agent and initiating an artificial review, issuing a new voucher bound to the current version or permanently terminating the task according to the review conclusion, and rewriting the world model to complete the closed-loop scheduling control.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a closed-loop scheduling and control method for a BeiDou grid-based spatiotemporal intelligent agent. Background Technology

[0002] BeiDou-based grid-based spatiotemporal intelligent scheduling and control is a key scheduling method that divides the land surface and airspace into spatial units using a unified code, and then continuously tracks the target position using BeiDou positioning. It is widely used in remote command scenarios for high-risk equipment, such as unmanned platform operations and coordinated response in dangerous areas. If a command is issued to the wrong location, it can cause irreparable damage. Therefore, before officially issuing a command, it is usually manually verified again to reduce the possibility of erroneous actions. However, current practices mostly treat manual verification as an isolated review process, only concerned with whether the review is completed or the process is finished, without taking into account the fact that the external world is still changing during the review period. In other words, the verification certificate is valid at the moment of generation, but it does not maintain continuous alignment with the subsequent actual target position and grid status, resulting in an invisible gap between the verification result and the actual situation. The root cause of this problem is that moving targets do not stop in place just because a person is watching the screen. A target may migrate from its initially bound grid to an adjacent grid at any time, causing a misalignment between the "event grid code" recorded on the verification certificate and the target's current grid location. As targets enter and exit the grid and the allowed passage range is adjusted, the grid state version in the world model is constantly updated. The version corresponding to the generated certificate can easily be surpassed by a newer version, resulting in version inconsistency. Target migration causes positional misalignment, which, coupled with version lag, creates a vicious cycle. This means that a control task initially deemed ready to be issued may no longer apply to the correct spatial object after tens of seconds. For example, an operator issues a "confirm release" order to a high-risk device tracking a vehicle. When the certificate is issued, the target is still in grid A. However, the confirmation process is time-consuming. By the time the order is about to be issued, the vehicle has already entered the adjacent grid B, but the certificate still records grid A. The allowed passage range of grid B in the system's world model does not actually cover the task. If the order is issued as before, it's like issuing a command to an incorrect grid. The longer the wait, the more thorough the investigation, but the more likely this misalignment will occur. Therefore, the key issue in ensuring the effectiveness of the closed-loop scheduling space is how to align the timing of target cross-grid migration, the timing of voucher generation, and the old and new grid status versions between manual confirmation and instruction issuance, so as to promptly identify high-risk tasks that have already experienced location drift and version mismatch after confirmation, and avoid releasing expired vouchers as valid vouchers. Summary of the Invention

[0003] This invention provides a closed-loop scheduling and control method for a BeiDou grid-based spatiotemporal intelligent agent, the method comprising: Obtain the manual confirmation voucher, extract the task identifier and confirmation event grid code, and identify the voucher grid version and voucher generation time through the confirmation event grid code; Based on the task identifier, extract the target migration record from the world model, obtain the target's current location and entry time, and identify the migration stage after manual confirmation; The target grid is obtained by identifying the grid affiliation of the current location of the target through Beidou grid coding. The migration path to the current target grid is traced starting from the grid indicated by the confirmed event grid coding, thus forming a confirmed cross-grid migration fact. Extract the current grid version and allowed grid mask from the current scheduling system, trace the execution status of the credential grid version along the version update chain, evaluate the in-service qualification of the credential version, and form a conclusion that the credential version has failed. Based on the confirmed cross-grid migration facts and the conclusion of invalid certificate version, the triggering conditions are determined. If any triggering condition is met, the control target is locked from the target migration record and identified as a spatial object, and the real-time motion trajectory of the spatial object is obtained. Analyze the real-time motion trajectory within the allowed mesh mask coverage area, identify the crossing patterns where the trajectory crosses the mask boundary, and form a no-entry conclusion for the spatial object based on the crossing patterns that continuously leave the mask coverage. Based on the task identifier corresponding to the spatial object that forms the prohibition conclusion, the control task associated with the spatial object is blocked from being issued, the interception result is recorded and a manual review is initiated, and control measures are executed according to the review conclusion to complete the closed-loop scheduling control.

[0004] Preferably, the step of obtaining the manual confirmation voucher, extracting the task identifier and confirmation event grid code, and identifying the voucher grid version and voucher generation time through the confirmation event grid code includes: Obtain the original message of the manual confirmation credential, extract the task identifier and confirmation event grid code, and verify the level and parent affiliation of the confirmation event grid code level by level. Retrieve the voucher grid version from the version registration table with grid code as the key, extract the voucher generation time and store it in the baseline element table.

[0005] Preferably, the step of extracting target migration records from the world model based on the task identifier, obtaining the target's current location and entry time, and identifying the migration stage after manual confirmation includes: Retrieve the target migration records associated with the task identifier from the target trajectory library of the world model. The target migration records record inter-cell jump events one by one in order of timestamp. Read the latest BeiDou positioning point from the target trajectory database and use that BeiDou positioning point itself as the target's current location; Events whose entry time is later than the voucher generation time are screened out and connected in sequence as the migration stage.

[0006] Preferably, the step of identifying the grid affiliation of the current target location through BeiDou grid coding to obtain the current target grid, and tracing the migration path to the current target grid starting from the grid indicated by the confirmed event grid coding to form a confirmed cross-grid migration fact, includes: The target address is mapped to a code at the same level as the confirmation event grid code to obtain the current target grid; Starting from the grid indicated by the confirmed event grid code, a migration path is formed by connecting the inter-grid transition event sequence to the grid grid code; In the events covered by the migration path, the jumps where the entry time is later than the voucher generation time are marked as confirmed cross-cell migration facts after the cross-cell node is confirmed.

[0007] Preferably, the step of extracting the current grid version and allowed grid mask from the current scheduling system, tracing the execution status of the credential grid version along the version update chain, evaluating the in-service qualification of the credential version, and forming a conclusion that the credential version has expired includes: Read the current mesh version and the bound allowed mesh masks from the mesh state library; The version update chain is formed by retrieving the iteration records of each credential grid version to the current grid version and its historical allowed grid masks. The tracing window is defined by the entry times of the first and last cross-grid nodes in the confirmed cross-grid migration facts; The historical allowed grid mask of the iteration item falling into the traceability window is compared with the confirmation event grid code. If there is an iteration item whose confirmation event grid code is removed from the historical allowed grid mask, the credential grid version is determined to be out of service.

[0008] Preferably, the step of determining the triggering condition based on the confirmed cross-grid migration fact and the conclusion of the invalidation of the certificate version, wherein if any triggering condition is met, the control target is locked from the target migration record and identified as a spatial object, and the real-time motion trajectory of the spatial object is obtained, includes: The system compares the current target mesh with the existing allowed mesh mask. If the current target mesh does not belong to the existing allowed mesh mask, it is determined to be a position drift. It also compares the confirmed event mesh code with the existing allowed mesh mask and the historical allowed mesh mask in the tracing window along the version update chain. If the confirmed event mesh code does not belong to the existing allowed mesh mask or is removed from the historical allowed mesh mask in the tracing window, it is determined to be a strategy change. The system compares the migration path recorded in the confirmed cross-mesh migration fact with the existing allowed mesh mask. If a cross-mesh migration occurs after confirmation and the existing allowed mesh mask does not cover the migration path, it is determined to be a path mismatch. The three triggering conditions are determined independently in parallel. When any one of them is met, the control target is locked as a tracking item and registered as a spatial object. A trajectory push request is initiated to the BeiDou positioning module, using the spatial object task identifier as the filtering condition, and the received positioning points are stitched together to form a real-time motion trajectory.

[0009] Preferably, the analysis of the real-time motion trajectory within the allowed mesh mask coverage area, identifying the crossing patterns where the trajectory exceeds the mask boundary, and forming a no-entry conclusion for the spatial object based on the continuously detached crossing patterns, includes: The positioning points of the real-time motion trajectory are mapped to grid codes, and the members of the grid code set that allows grid masking are determined. Points belonging to the set are considered inside points, and those not belonging to the set are considered outside points. The intersection of the line connecting adjacent positioning points and the mask boundary is taken as the boundary crossing node or return node to obtain the entry and exit transformation sequence; A node that crosses the boundary and has a paired return node is marked as a round trip; a node that does not have a paired return node is marked as continuously leaving the mask cover. The conclusion that the space object is prohibited from entering is formed by continuously detaching from the mask coverage, and the round trip is recorded and stored, but the control task is not blocked based on this.

[0010] Preferably, the prohibition conclusion for the spatial object includes: The intersection of the line connecting adjacent positioning points with the boundary of the allowable mesh mask is screened along the real-time motion trajectory. The intersection point that falls within the grid edge line segment is determined to be a valid intersection point. The valid intersection points from the outside to the inside are marked as the entry intersection points, and the valid intersection points from the inside to the outside are marked as the exit intersection points; these are collectively referred to as the intersection positions. The handover positions are linked together in the order of movement to form a handover position chain. After each point is crossed, it is checked whether an entry point is encountered. If an entry point is encountered, it is marked as having returned; otherwise, it is marked as not having returned. If the chain tail fails to return consecutively and continuously crosses the intersection point in the same direction, a no-entry conclusion is formed.

[0011] Preferably, the step of blocking the issuance of control tasks associated with the spatial object based on the task identifier corresponding to the spatial object that forms the prohibition conclusion, recording the interception result and initiating manual review, and executing control measures according to the review conclusion to complete closed-loop scheduling control includes: In the command issuance channel of the Beidou grid spatiotemporal intelligent agent, the task identifier is blocked, the issuance status of the control task is changed to the intercepted status, and the interception result associated with the task identifier is obtained; if the review is successful, a new manual confirmation certificate bound to the current grid version is issued; if the review fails, the control task associated with the task identifier is permanently terminated and the task termination status is written back to the world model.

[0012] Preferably, obtaining the interception result associated with the task identifier includes: The interception result, along with the field of the prohibition conclusion, is written into the interception log, with the task identifier as the primary key. A review work order is pushed to the manual review terminal. The review work order uses the task identifier as the binding key and carries the record snapshot in the interception log to complete the closed-loop scheduling control.

[0013] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: This invention discloses a closed-loop scheduling and control method for BeiDou grid-based spatiotemporal intelligent agents, which solves the problem of out-of-bounds control failure caused by the invalidation of manually confirmed credentials during target cross-grid migration due to grid version updates. When a controlled target migrates across grids after manual confirmation, the grid version on which the original confirmation credentials were based may no longer be in service. If the original scheduling is continued, the target will enter a restricted area. This invention identifies the credential generation time by extracting the confirmation event grid code and combining it with the target migration record in the world model to track the cross-grid migration path after confirmation. It traces the execution status of the credential grid version during the migration phase along the version update chain to determine its eligibility. Three conditions are triggered simultaneously: the current target grid drifts away from the current allowed grid mask; the confirmation event grid code deviates from the current or traced historical allowed mask strategy; and the confirmed cross-grid crossing occurs before the current mask covers the migration path. Upon any one of these conditions being met, the spatial object to be tracked is locked, and trajectory verification and blocking are initiated. Its real-time trajectory is continuously acquired. Only continuous non-returning at the end of the chain with a consistent direction of departure constitutes a prohibition conclusion. Round trips are recorded but not blocked. Finally, the agent blocks the issuance of control tasks associated with the object and initiates manual review. If the review passes, a new credential bound to the current grid version is issued; if the review fails, the task is permanently terminated and the world model is written back. This invention achieves the association identification and proactive interception of credential invalidation and target boundary crossing, effectively preventing boundary control risks in a dynamically updated grid version environment and improving the security and closed-loop management capabilities of time-space scheduling. Attached Figure Description

[0014] Figure 1 This is a flowchart of a closed-loop scheduling and control method for a BeiDou grid-based spatiotemporal intelligent agent according to the present invention.

[0015] Figure 2 This is a schematic diagram of a closed-loop scheduling and control method for a BeiDou grid-based spatiotemporal intelligent agent according to the present invention.

[0016] Figure 3 This is another schematic diagram of a closed-loop scheduling and control method for a BeiDou grid-based spatiotemporal intelligent agent according to the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0018] like Figures 1-3 This embodiment of a BeiDou grid spatiotemporal intelligent agent closed-loop scheduling and control method may specifically include: S101. During the closed-loop scheduling and control process of the Beidou grid spatiotemporal intelligent agent, obtain the manual confirmation certificate, extract the task identifier and confirmation event grid code, and identify the certificate grid version and certificate generation time through the confirmation event grid code.

[0019] During the closed-loop scheduling and control process of the BeiDou grid spatiotemporal intelligent agent, the original message of the manual confirmation certificate is acquired. The task identifier and confirmation event grid code are extracted according to the certificate's field structure. The task identifier undergoes length and character set verification. The confirmation event grid code is checked against its hierarchical structure according to the BeiDou two-dimensional grid location code, level by level, to obtain a certificate parsing result that passes the verification. Based on the confirmation event grid code in the certificate parsing result, the version registration table in the world model, with the grid code as the key, is retrieved. The certificate grid version bound to this code at the time of certificate issuance is read. The certificate generation time is extracted from the issuance field of the certificate parsing result. The task identifier, the confirmation event grid code, the certificate grid version, and the certificate generation time are associated and stored in the baseline element table to obtain the baseline certificate elements of the manual confirmation certificate.

[0020] In the credential access stage of the closed-loop scheduling and control of the Beidou grid spatiotemporal intelligent agent, the original message of the manually confirmed credential is organized with a field structure of fixed-length header plus variable-length payload. The header records the credential issuance field, and the payload carries the business field.

[0021] For example, the header includes an issuance timestamp and an issuer identifier, while the payload includes a task identifier and a confirmation event grid code. The variable-length payload uses a segmented layout. The first two bytes of each segment identify the field type, followed by two bytes indicating the field's length. The field values ​​are then read according to their lengths, and the payload is segmented and assigned to the corresponding fields accordingly. The segments are distinguished by type, length, and field values. After disassembling the document according to its field structure, the task identifier undergoes length and character set verification, and documents with incorrect formats are discarded.

[0022] Specifically, the confirmation event grid code is checked level by level against its parent level according to the hierarchical structure of the BeiDou two-dimensional grid location code. The BeiDou two-dimensional grid location code adopts a progressively subdivided encoding method, with higher-level codes corresponding to coarse-grained grids and lower-level codes corresponding to fine-grained grids. The prefix of a lower-level code must be consistent with the code of its parent grid. In one embodiment, the code increments the character segments level by level from level 1 to level 10. The nth-level code is formed by concatenating the character segments of the (n-1)th-level code as a prefix with the character segments of the current level. If the level declared by a confirmation event grid code does not match its actual number of character segments, or if its parent prefix is ​​inconsistent with the registered previous-level grid, the code is deemed invalid and discarded. If all level checks pass, a verified credential parsing result is obtained. Further, based on the confirmation event grid code in the credential parsing result, the version registration table with the grid code as the key in the world model is retrieved, and the credential grid version bound to the code at the time of credential issuance is read. The version registration table uses the grid code as the primary key and the version number, effective start time, and effective end time as fields to record the version corresponding to each grid refresh. The voucher generation time is extracted from the issuance field of the voucher parsing result. The voucher generation time is then used to locate the corresponding version number within the interval formed by the effective start time and effective end time; the version number corresponding to the matched interval is the bound voucher grid version. The task identifier, the confirmation event grid code, the voucher grid version, and the voucher generation time are linked row-wise and stored in the baseline element table. The baseline element table uses the task identifier as the primary key, and the remaining fields are attribute columns for that row. Only one baseline record is retained for the same task identifier, thus obtaining the baseline voucher elements for manually confirmed vouchers.

[0023] S102. Extract target migration records from the world model based on the task identifier, obtain the target's current location and entry time, and identify the migration stage after manual confirmation.

[0024] Based on the task identifier in the benchmark credential element, the target migration record associated with the task identifier is retrieved from the target trajectory library of the world model. The target migration record is ordered by timestamp, recording each inter-grid jump event generated by the target crossing the grid boundary. Each event includes an entering grid code, an exiting grid code, and an entering time. The latest reported BeiDou positioning point is read from the end of the trajectory chain in the target trajectory library. The latitude and longitude coordinates of this BeiDou positioning point are used as the target's current location, and its grid affiliation is consistent with the entering grid code of the latest inter-grid jump event. The entering time of the same event is also read to obtain the entering time bound to the target's current location. Using the credential generation time in the benchmark credential element as the time boundary for manual confirmation, inter-grid jump events with entering times later than the credential generation time are filtered from the target migration records. These events are then concatenated in chronological order by timestamp to form an event sequence, which is then used as the migration stage after manual confirmation.

[0025] In the migration perception stage of the BeiDou grid-based spatiotemporal intelligent agent closed-loop scheduling and control, the target trajectory library in the world model uses the task identifier as the primary key to aggregate target positioning sequences under the same control task into a trajectory chain for subsequent migration phase retrieval. Each trajectory chain in the target trajectory library is attached to a grid jump event table, with events arranged in ascending order by timestamp, forming a temporal structure that can be traced back by boundary. In one implementation, the task identifier is generated along with the credential when issued by the scheduling side, serving as a globally unique retrieval key. Based on the task identifier in the reference credential elements, a point lookup request is initiated to the target trajectory library, matching the trajectory chain corresponding to the task identifier, and then retrieving the attached target migration record. The target migration record is ordered by timestamp, recording the grid jump events generated each time the target crosses the grid boundary.

[0026] Specifically, the writing of inter-grid jump events is driven by the BeiDou positioning module. The BeiDou positioning module reports the latitude and longitude coordinates of the target at a sampling frequency of 1Hz to 10Hz. After receiving each coordinate report, the world model reverse-engineers it into a BeiDou two-dimensional grid position code and compares it with the grid code to which the previous report belongs. If the two grid codes are consistent, only the dwell time at the end of the trajectory chain is refreshed, and no new event is generated. If the two grid codes are inconsistent, it is determined that an inter-grid jump has occurred, and an inter-grid jump event is added to the target migration record. The event field includes the entering grid code, the exiting grid code, and the entering time. The entering grid code is taken from the grid to which the current report belongs, the exiting grid code is taken from the grid to which the previous report belongs, and the entering time is taken from the BeiDou time value currently reported. All events and the time of certificate generation uniformly use the BeiDou system time as the time reference.

[0027] It should be noted that although the grid out-of-grid encoding field does not directly participate in the subsequent target location process, it is retained as a complete field item for inter-grid jump events, used for continuous playback of the migration path and interpretation of cross-grid directions. Furthermore, the latest reported BeiDou positioning point is read from the end of the trajectory chain in the target trajectory library, and its latitude and longitude coordinates are used as the target location, without replacing the actual positioning point with the geometric center point of the grid, thus preserving the target's true observation position. The BeiDou positioning point is directly reported by the BeiDou positioning module. The grid division and grid boundary calculation of the BeiDou two-dimensional grid uniformly adopt the coordinate origin specified in GB / T 39409, with the origin longitude Lon0 being -180 degrees and the origin latitude Lat0 being -90 degrees, ensuring that the target location's positioning point coordinates and the grid boundary are on the same coordinate reference. The reverse calculation of the grid encoding is performed level by level, restoring the structure of the encoding: the level of the grid is determined by the encoding level segments, and then the longitude and latitude steps corresponding to each grid at that level are determined; the longitude and latitude offsets of the southwest corner of the grid relative to the origin are determined by the encoding positioning segments. The formulas for calculating the southwest corner longitude Lon1 and southwest corner latitude Lat1 are Lon1=Lon0+x×dx, Lat1=Lat0+y×dy, where x is the integer index of the longitude direction of the grid at that level relative to the origin, y is the integer index of the latitude direction, and dx and dy are the longitude and latitude steps of that level, respectively; the northeast corner longitude is equal to the southwest corner longitude plus the longitude step of that level, and the northeast corner latitude is equal to the southwest corner latitude plus the latitude step of that level. This reverse calculation is used for the geometric solution of the mask boundary grid edges and the grid range, and is no longer used for the coordinate positioning of the target site. The entry time of the same event is retrieved synchronously and bound to the target current address to obtain the entry time bound to the target current address.

[0028] Understandably, using the BeiDou positioning point itself as the target's current location preserves the target's true observation position and avoids positioning deviations introduced by replacing the true positioning point with the grid's geometric center. The grid generation and grid boundary calculation uniformly adopt the coordinate origin specified in GB / T 39409, ensuring that the positioning point coordinates and the spatial comparison of the allowable grid mask are kept on the same coordinate reference, avoiding comparison ambiguities caused by different coordinate references.

[0029] Preferably, the document generation time in the benchmark document element is used as the time boundary for the completion of manual confirmation. Both the document generation time and the entry time of the inter-cell transition event are normalized to the BeiDou system time, eliminating the need for time base conversion during comparison. Each target migration record is compared against the entry time of each inter-cell transition event. Events with entry times strictly later than the document generation time are filtered out, while events with entry times earlier than or equal to the document generation time are eliminated, preventing the inclusion of cell crossings occurring before confirmation in this stage of judgment. The filtered events are concatenated into an event sequence according to their timestamps. The first event in the event sequence corresponds to the first cell crossing of the target after manual confirmation, and the last event corresponds to the latest cell crossing. This entire event sequence is extracted as the migration stage after manual confirmation, thus defining the temporal sequence of the migration facts in this stage.

[0030] S103. Identify the grid affiliation of the target site through Beidou grid coding to obtain the current target grid. Take the grid indicated by the confirmed event grid coding as the starting point to trace the migration path to the current target grid. Analyze the chronological relationship between the voucher generation time and the grid entry time during the migration phase after manual confirmation to form the confirmed cross-grid migration fact.

[0031] Based on the latitude and longitude coordinates of the target location, and following the coding rules of the BeiDou 2D grid location code which subdivides by latitude and longitude, the coordinates are mapped to the grid codes of each level. The code at the same level as the grid code of the confirmation event is taken as the grid affiliation of the target location, thus obtaining the current target grid. Taking the grid indicated by the grid code of the confirmation event as the starting grid, the migration path is formed by sequentially tracing the inter-grid transition events during the migration phase after the manual confirmation. The grid code of the previous event entering the grid is adjacent to the grid code of the next event leaving the grid, and so on, until the grid code of the final event is consistent with the current target grid, thus forming the migration path from the starting grid to the current target grid. For each inter-grid transition event covered by the migration path, the chronological relationship between its entry time and the voucher generation time is determined. Transitions with entry times later than the voucher generation time are marked as cross-grid nodes after confirmation. All cross-grid nodes after confirmation are aggregated in chronological order to obtain the cross-grid migration facts after confirmation.

[0032] In the migration fact interpretation stage of the BeiDou grid-based spatiotemporal intelligent agent closed-loop scheduling and control, grid affiliation identification uses the latitude and longitude coordinates of the target's current location as input and the grid code at the same level as the confirmation event grid code as output, avoiding inconsistencies in affiliation scales due to different levels. The grid code of the confirmation event recorded when the manual confirmation certificate is issued has its level pre-set by the business side according to the control granularity. Level 10 corresponds to meter-level granularity, and Level 6 corresponds to hundred-meter-level granularity. Comparison at the same level is a prerequisite for forming the subsequent migration path.

[0033] Specifically, the BeiDou two-dimensional grid location code adopts a coding rule that subdivides the data level by latitude and longitude. The Earth's surface is divided into nested grids along the longitude and latitude directions. The nth level grid is further subdivided from the (n-1)th level grid according to longitude and latitude step sizes. Based on the latitude and longitude coordinates of the target location, the step-by-step mapping process is as follows: First, calculate the offset of the target location relative to the origin. Let the longitude of the target location be L, the longitude of the origin be L0, the longitude step size for this level be S1, and the longitude index be I. The calculation logic is: longitude difference D1 = L - L0, I = D1 / S1, taking the integer part. Let the latitude of the target location be B, the latitude of the origin be B0, the latitude step size for this level be S2, and the latitude index be J. The calculation logic is: latitude difference D2 = B - B0, J = D2 / S2, taking the integer part. The origin longitude L0 and origin latitude B0 are uniformly taken as the origin of the BeiDou two-dimensional grid coordinates as specified in GB / T 39409, maintaining the same coordinate reference as the aforementioned grid boundary calculation and mask comparison. For example, if the target's current location longitude is 116.5, the longitude step size for this level is 0.1, and the origin longitude L0 is uniformly taken as -180 degrees, then the longitude difference D1 is 296.5, and the longitude sequence number I is 2965. Subsequently, the longitude sequence number and latitude sequence number are merged into the encoding segment according to a predetermined splicing order to form the grid code for this level. The levels are advanced sequentially from high to low until the level declared by the confirmed event grid code is reached. The grid code under this level is taken as the grid affiliation of the target's current location, thus obtaining the current target grid.

[0034] In one possible implementation, if the latitude and longitude coordinates of the target location happen to fall on the hierarchical boundary, the adjacent grid closest to the origin is taken as the belonging grid to maintain the uniqueness of the belonging result. Further, using the grid indicated by the confirmed event grid code as the starting grid, the path is sequentially connected along the inter-grid jump events in the migration phase after the manual confirmation. The path connection is advanced using adjacency verification: the out-of-grid grid code of the current event to be connected is taken and compared with the in-grid grid code of the already connected end event. If they match, the event is connected; if they do not match, it is determined that there is a missed cross-grid between the event and the already connected end event. A second pull is performed from the original trajectory flow database using the in-grid time interval of both events as the query window and the spatial bounding box of their grid codes as the spatial range to obtain the underlying inter-grid jump events that have not successfully attached the task identifier. The supplemented events are inserted before the event to be connected by timestamp and then adjacency verification is performed again until the comparison passes. The in-grid codes of each event are sequentially connected to the serial sequence until the in-grid code of the end event is consistent with the current target grid, thus forming a migration path from the starting grid to the current target grid.

[0035] It should be noted that the migration path is expressed in the form of a grid-coded sequence. Each node in the sequence corresponds to one grid crossing, and the adjacency relationship between nodes is verified by the grid codes of the out-of-grid and in-grid events before and after the migration.

[0036] Preferably, for each inter-grid transition event covered by the migration path, the chronological relationship between its entry time and the voucher generation time is determined under the same BeiDou system time reference. If the entry time of an event is later than the voucher generation time, the cross-grid node corresponding to that event is marked as a confirmed cross-grid node; if the entry time of an event is earlier than or equal to the voucher generation time, the node is not marked and exits the current round of judgment. All confirmed cross-grid nodes are collected in chronological order, and the node set records the entry time of the first cross-grid node, the entry time of the last cross-grid node, and the number of nodes, forming a confirmed cross-grid migration fact. The confirmed cross-grid migration fact is stored in parallel with the baseline voucher element, using the task identifier as the binding key.

[0037] It is understandable that after the grid affiliation identification, migration path connection and time sequence determination are sequentially linked, the misalignment between the original bound grid and the target's current grid, as recorded in the certificate, is recorded one by one in the confirmed cross-grid migration facts in the form of inter-grid jump nodes.

[0038] S104. Extract the current grid version and allowed grid mask from the current scheduling system, trace the execution status of the credential grid version in the migration stage corresponding to the confirmed cross-grid migration fact along the version update chain, evaluate the in-service qualification of the credential version, and form a conclusion that the credential version has failed.

[0039] The version number of the current grid version and its bound allowed grid mask are read from the grid state database of the current scheduling system. The allowed grid mask expresses the currently allowed spatial range in the form of a set of grid codes. Tracing back from the current grid version along the version update chain, the version iteration records between the certificate grid version in the benchmark certificate element and the current grid version are obtained. Each iteration record is accompanied by the historical allowed grid mask bound to that version, forming a version update chain. Starting from the entry time of the first cross-grid node in the confirmed cross-grid migration fact, the entry time of the last cross-grid node is defined as the starting point and the ending point as the ending point. The iteration items whose effective time falls into the tracing window are screened from the version update chain. For each iteration item, its bound historical allowed grid mask is compared with the confirmation event grid code in the benchmark certificate element to form a coverage result sequence at the iteration item level. The coverage result sequence is summarized item by item. If the comparison result of any iteration item in the sequence shows that the confirmation event grid code is separated from the historical allowed grid mask bound to the iteration item, then the iteration item is marked as a mismatch node, and the certificate grid version is determined to have lost its in-service qualification, thus forming the conclusion that the certificate version is invalid.

[0040] In the version timeliness evaluation stage of the Beidou grid spatiotemporal intelligent agent closed-loop scheduling and control, the grid version is constantly refreshed as the allowed passage range is adjusted. The grid version corresponding to the certificate at the time of issuance will be updated to the new version over time. If the relationship between the certificate grid version at the time of issuance and subsequent updates is not explicitly traced, the certificate's eligibility for service during the period between manual confirmation and instruction issuance cannot be objectively determined.

[0041] In one possible implementation, the current scheduling system's grid state database organizes grid version snapshots using version numbers as the primary key. Each version snapshot carries the allowed grid mask corresponding to that version number. The allowed grid mask expresses the currently permitted spatial range in the form of a grid code set, where each element is a BeiDou two-dimensional grid location code.

[0042] For example, a mesh mask can cover 1,000 to 5,000 mesh codes at the meter level in a certain work area, or it can cover dozens of mesh codes at the hundred-meter level. The specific granularity depends on the pre-set control granularity by the business side.

[0043] Specifically, the currently active version number is read from the grid state database of the current scheduling system as the current grid version, and the allowed grid mask bound to it is retrieved. The current grid version then serves as the starting point for reverse tracing of the version update chain.

[0044] It should be noted that the system generates a baseline credential element during the initial registration of the target. This baseline credential element includes the credential grid version and the confirmation event grid code corresponding to the target's position at the time of the initial confirmation event. The version update chain is organized in a linked list structure, with each node carrying a version iteration record. The node fields include the version number, the effective start time, a pointer to the previous version number, and a set of added and deleted items relative to the previous version. The set of added and deleted items records the grid codes added and deleted in the current version relative to the previous version. When traversing backward from the current grid version to any historical version, the added and deleted items of each node along the way are reversed in the direction of reversal, that is, the added items are deducted from the current allowed grid mask, and the deleted items are added back, thereby restoring the complete allowed grid mask corresponding to the historical version. Tracing backward from the current grid version along the version update chain, the version iteration records between the credential grid version and the current version in the baseline credential element are obtained, forming a version update chain. The version update chain is chronological, covering all version refresh events that have occurred since the credential was issued. Furthermore, a tracing window is defined starting from the entry time of the first cross-grid node in the confirmed cross-grid migration fact and ending at the entry time of the last cross-grid node. The time span of the tracing window is consistent with the active phase of the target's confirmed cross-grid migration, and the version validity period is limited to the time period during which the target's actual location changes. Iteration items whose effective start time falls within the tracing window are screened from the version update chain. If the effective start time of an iteration item does not fall within the tracing window, that iteration item is not included in this round of version validity period determination. The screened iteration items are arranged in ascending order of their effective start time, and each iteration item carries its bound historical allowed grid mask. For each screened iteration item, its bound historical allowed grid mask is compared with the confirmation event grid code in the benchmark credential element. The attribution comparison adopts a set membership determination method: the confirmed event grid code is used as the query key, and a membership search is performed in the set of grid codes corresponding to the historical allowed grid masks of the iteration item. If the search is successful, the comparison result of the iteration item is recorded as within the coverage; if the search is unsuccessful, the comparison result of the iteration item is recorded as out of coverage. The comparison results of all filtered iteration items are arranged in order of their effective start time to form a coverage result sequence at the iteration item level.

[0045] Preferably, the coverage result sequence adopts a binary record format, with each element corresponding to a version iteration item, and the value being either within coverage or out of coverage. Elements are concatenated with the effective start time as the index. This record format facilitates centralized determination by subsequent sequence scanning, eliminating the need to repeatedly retrieve the historical allowed grid mask for each iteration item. The coverage result sequence is summarized item by item, and mismatch node labeling is completed using a single linear scan: if an element is encountered with the value of being out of coverage during the scan, the iteration item corresponding to that element is labeled as a mismatch node, and the version number and effective start time of that iteration item are recorded. After the scan is completed, if the number of mismatch nodes is 0, it is determined that the certificate grid version is always under the coverage of the allowed grid mask within the traceability window, and the certificate grid version retains its in-service qualification; if the number of mismatch nodes is greater than or equal to 1, the item with the earliest effective start time among all mismatch nodes is taken as the first mismatch node, and the effective start time of the first mismatch node is recorded as the failure start point. From the time corresponding to the failure start point, the certificate grid version is determined to have lost its in-service qualification. In one implementation, the system pre-creates an evaluation result table in the storage medium. This table has a pre-defined task identifier primary key and multiple data columns for recording credential status and migration facts. The resulting credential version failure conclusion is stored in the form of a structured record, with fields including the task identifier, credential grid version, failure start point, mismatch node set, and current grid version from the baseline credential elements. The credential version failure conclusion, using the task identifier as the binding key, is stored in parallel with the confirmed cross-grid migration facts in the pre-created evaluation result table.

[0046] It is understandable that the reverse tracing of the version update chain and the comparison of the coverage result sequence within the tracing window are interconnected. The coverage status of the credential grid version at each version refresh within the tracing window is recorded item by item, and finally, a closed-loop determination is made on the credential's eligibility for service in the form of a credential version invalidation conclusion.

[0047] S105. Triggered by three conditions in parallel: position drift, strategy change and cross-grid mismatch after confirmation. Once any condition is met, trajectory verification and blocking are initiated. The control target that meets any trigger condition is identified as a spatial object, and the real-time motion trajectory of the spatial object is continuously acquired.

[0048] The following three triggering conditions are used for parallel judgment. If any one condition is met, the trajectory verification and blocking of the corresponding control target will commence: First, if the current target grid does not belong to the grid code set corresponding to the current allowed grid mask, it indicates that the target has drifted out of the allowed passage range, and is judged as position drift; Second, if the confirmation event grid code does not belong to the current allowed grid mask, or if there is an iterative item deviating from the historical allowed grid mask within the tracing window defined by the entry time of the first and last cross-grid node in the confirmed cross-grid migration fact, it indicates that the allowed passage policy has changed, and is judged as a policy change; Third, if the confirmed cross-grid migration fact indicates that a cross-grid occurred after manual confirmation, and the current allowed grid mask does not cover the grid traversed by the migration path, it is judged as a confirmed cross-grid mismatch. Control targets that meet any of the triggering conditions are locked as trackable items. Based on the task identifier, using the task identifier corresponding to the trackable item as the binding key, the control target is registered as a spatial object, and the task identifier bound to the spatial object is written into the real-time trajectory subscription list. The real-time trajectory subscription list uses the task identifier as the primary key and records the subscription effective time. According to the real-time trajectory subscription list, a trajectory push request is initiated to the Beidou positioning module with the task identifier as the filter condition. The latitude and longitude positioning points reported by the spatial object are continuously received at a sampling frequency of 1Hz to 10Hz. The positioning points are then spliced ​​together in the order of the reporting time to form the real-time motion trajectory of the spatial object.

[0049] In the object locking and trajectory access stage of the BeiDou grid-based spatiotemporal intelligent agent closed-loop scheduling and control, if any of the following conditions is met during the time period between credential issuance and instruction delivery: position drift, strategy change, or cross-grid mismatch after confirmation, it indicates a misalignment between the original confirmation credential and the actual situation. This control target constitutes the spatial object source for subsequent trajectory verification and instruction delivery blocking. Object locking is triggered by the parallel determination of the above three triggering conditions. Each determination is based on the task identifier as the binding key to achieve field-level connectivity. The occurrence of any single condition immediately initiates trajectory verification and instruction delivery blocking; multiple conditions are not required to be met simultaneously.

[0050] Specifically, the three triggering conditions are determined as follows: Position drift determination: The current target grid obtained in step S103 is read, and a member search is performed in the set of grid codes corresponding to the current allowed grid mask. If no match is found, the current target grid is determined not to belong to the current allowed grid mask. Strategy change determination: The confirmation event grid code in the baseline credential element is read. First, a member search is performed in the current allowed grid mask. Then, along the version update chain, a member search is performed on each of the historical allowed grid masks bound to each iteration item within the traceability window defined by the entry time of the first and last cross-grid node in the confirmed cross-grid migration fact. If a separation exists, the confirmation event grid code is determined to be separated from the current or historical allowed grid mask within the traceability window. Confirmed cross-grid mismatch determination: The migration path recorded in the confirmed cross-grid migration fact is read, and the intersection of each grid code traversed by the migration path with the set of grid codes corresponding to the current allowed grid mask is checked. If the intersection is empty, the current allowed grid mask is determined not to cover the migration path. The three conditions are determined independently and take effect in parallel. If any condition is met, the item to be tracked is locked, and trajectory verification and blocking are initiated.

[0051] It should be noted that the adoption of the same BeiDou system time base avoids time base conversion errors when comparing time across modules, and allows the certificate generation time, cross-grid node entry time, version iteration effective time and trajectory reporting time to be directly judged by numerical order, providing a unified time base for the parallel judgment of the three triggering conditions.

[0052] In one possible implementation, the control target is registered as a spatial object using the task identifier corresponding to the item to be tracked as the binding key. The registration action is completed by adding a new record to the spatial object registration table. The registration table uses the task identifier as the primary key and includes fields such as the spatial object registration time, task identifier, associated credential grid version, and registration reason. The registration reason field is always filled with the information that the credential has expired and is still in the migration stage. After registration, the control target is referenced as a spatial object in subsequent processes.

[0053] Specifically, the task identifier bound to the spatial object is written into the real-time trajectory subscription list. The real-time trajectory subscription list uses the task identifier as the primary key, and the fields include the subscription effective time, subscription status, and sampling frequency range. The subscription status takes one of two values: active or paused, and the sampling frequency range is 1Hz to 10Hz.

[0054] For example, the subscription activation time is taken as the spatial object registration time, and the subscription status is set to active upon registration completion. The system architecture pre-deploys a BeiDou positioning module responsible for receiving and broadcasting satellite positioning signals. The sampling frequency value is pre-set by the service side based on the target mobility. The number of grid jumps is counted using a unified decision level adopted by the grid mask. Let the current time be t, then the statistical window is (t-300 seconds, t], and the number of grid jump events within the window is denoted as C5; when C5≥3, the sampling frequency is 10 Hz, and when C5<3, the sampling frequency is 1 Hz. Further, based on the real-time trajectory subscription list, a trajectory push request is initiated to the BeiDou positioning module, using the task identifier as the filtering condition. After receiving the trajectory push request, the BeiDou positioning module filters out the positioning points bound to the task identifier from its internally maintained positioning point broadcast stream, and samples them at a sampling frequency from 1 Hz to 10 Hz. The selected positioning points are then pushed to the scheduling side. The pushed content includes the latitude and longitude positioning points reported by the spatial object, the reporting time, and the task identifier. In one implementation, the scheduling side maintains a trajectory cache for each spatial object. The trajectory cache uses the task identifier as the key and the values ​​as a sequence of positioning points arranged in ascending order of reporting time. Because the BeiDou positioning module continuously pushes positioning points in chronological order, each time a pushed positioning point is received, the scheduling side directly appends it to the end of the positioning point sequence, thereby efficiently piecing together the real-time motion trajectory of the spatial object.

[0055] It is understandable that the real-time motion trajectory of the spatial object is formed by stitching together the positioning points in the order of their reported times. The real-time motion trajectory is continuously extended in the trajectory cache as the positioning points are continuously pushed, and the end point always corresponds to the latest position of the spatial object, thus completing the continuous acquisition of the real-time motion trajectory of the spatial object.

[0056] S106. Analyze the real-time motion trajectory within the allowed mesh mask coverage area, identify the crossing pattern of the trajectory beyond the mask boundary, and only form a prohibition conclusion for the spatial object based on the crossing pattern of continuously leaving the mask coverage. Round trip crossings are recorded but do not form a prohibition conclusion.

[0057] Each positioning point in the real-time motion trajectory of the spatial object is sequentially mapped to its corresponding grid code according to the encoding rules of the BeiDou 2D grid position code. A set membership determination is then performed between this grid code and the set of grid codes corresponding to the allowed grid mask: if the grid code to which the positioning point belongs belongs to the set, it is marked as an inside point; otherwise, it is marked as an outside point, resulting in a travel path with inside and outside markings. Along the travel path, the markings of adjacent positioning points are scanned sequentially according to the reported time. The intersection of the line connecting two adjacent positioning points and the mask boundary is taken as either a boundary crossing node or a return node. The former is the boundary crossing position from inside to outside point, and the latter is the boundary crossing position from outside to inside point. Consecutive similar marking segments are considered as the same inside or outside segment, resulting in an entry / exit conversion sequence. Based on the sequential pairing of outbound nodes and return nodes in the entry-exit transition sequence, the crossing pattern of the spatial object trajectory beyond the mask boundary is identified: if there is a temporally paired return node after the outbound node, it is marked as a round trip crossing and the duration of this crossing in the outer segment is recorded; if there is no paired return node after the outbound node, it is marked as continuously leaving the mask coverage; the prohibition conclusion is formed only by the continuous leaving the mask coverage, and the round trip crossing is only recorded and not used as the basis for issuing the blocking control task.

[0058] In the spatial movement interpretation stage of the BeiDou grid-based spatiotemporal intelligent agent closed-loop scheduling and control, a grid mask is allowed to carry the permitted passage range in the form of a grid code set. The mask boundary is formed by the common grid edges between grids within the set and grids outside the set. The boundary is not a regular polygonal outline, but a zigzag grid edge sequence naturally formed according to the shape of the grid code set. Explicitly interpreting the movement of the spatial object inside and outside this boundary is the direct basis for forming a prohibition conclusion.

[0059] Specifically, each positioning point in the real-time motion trajectory of the spatial object is sequentially mapped to its corresponding grid code according to the encoding rules of the BeiDou two-dimensional grid position code. The mapping process uniformly adopts the encoding division according to the level corresponding to the grid code in the allowed grid mask. If the allowed grid mask contains grid codes of multiple levels, the positioning point code and the mask code are first uniformly converted to a preset judgment level, and then the set membership judgment is performed. The mapped grid code is used as a query key to perform set membership judgment in the set of grid codes corresponding to the allowed grid mask: if the query matches, the positioning point is marked as an inside point; if the query does not match, the positioning point is marked as an outside point. All positioning points are arranged according to the order of reporting time to obtain the travel path with inside and outside labels.

[0060] It should be noted that the set member determination is implemented using a hash index, which allows the set of grid codes corresponding to the grid mask to be built into a hash table during loading. The time cost of querying a single grid code is approximately constant, which can support real-time interpretation in high-frequency trajectory push scenarios.

[0061] In one possible implementation, the labels of adjacent positioning points are scanned along the travel path in the order of the reported times. The following judgment is performed on the combination of labels of adjacent positioning points: if both are inside points or both are outside points, they are merged into the currently existing inside segment or at the end of the outside segment, and no boundary crossing event occurs; if the previous positioning point is inside and the next positioning point is outside, a boundary crossing from inside to outside is determined; if the previous positioning point is outside and the next positioning point is inside, a boundary crossing from outside to inside is determined. For each boundary crossing, the latitude and longitude of the two adjacent positioning points are taken to form a straight line segment. Candidate grid edges intersecting the bounding box of the latitude and longitude of this straight line segment are screened from the zigzag grid edge sequence of the allowed grid mask boundary. After calculating the intersections one by one, the intersection point closest to the previous positioning point along the direction of travel of the straight line segment is taken as the landing point of this boundary crossing. The grid edge is a line segment parallel to the longitude or latitude direction. Solving for the intersection point is reduced to calculating the intersection of a straight line segment and an axial line segment: Let the coordinates of the first fixed point be longitude x1 and latitude y1, and the coordinates of the second fixed point be longitude x2 and latitude y2. Define the longitude difference dx = x2 - x1, and the latitude difference dy = y2 - y1. For the transverse grid edge along the latitude direction, its latitude value y0 is known and its longitude is within the interval xa to xb. Define the target latitude difference dy0 = y0 - y1. When dy = 0, the trajectory is determined to be parallel to the latitude direction and has no intersection point; when ≠ 0, the intersection point is not found. Latitude is taken as y0, and the longitude of the intersection point is x = x1 + dx × dy0 / dy. If the calculated longitude x falls within the interval xa to xb, the intersection point is a valid intersection point. For the longitudinal grid edge along the longitude direction, its longitude value x0 is known and the latitude is within the interval ya to yb. The target longitude difference dx0 = x0 - x1 is defined. When dx = 0, the trajectory is determined to be parallel to the longitude direction and there is no intersection point. When dx ≠ 0, the longitude of the intersection point is taken as x0, and the latitude of the intersection point is y = y1 + dy × dx0 / dx. If the calculated latitude y falls within the interval ya to yb, the intersection point is a valid intersection point. Valid intersection points corresponding to crossings from the inside out are uniformly marked as boundary crossing nodes, and valid intersection points corresponding to crossings from the outside in are uniformly marked as return nodes to avoid ambiguity caused by other terms. Positioning points with consecutive similar annotations are grouped into the same inner segment or the same outer segment. Out-of-bounds nodes and return nodes are inserted at their corresponding segment boundaries according to their occurrence time, forming an entry-exit transformation sequence. Further, based on the sequential pairing of out-of-bounds nodes and return nodes in the entry-exit transformation sequence, the traversal pattern of the spatial object's trajectory crossing the mask boundary is identified. Pairing is advanced using the closest temporal match method: scanning sequentially from the sequence start point, each out-of-bounds node encountered is pushed onto a pairing stack; each return node encountered is paired with the top out-of-bounds node on the stack and popped. Successfully paired out-of-bounds nodes and return nodes form a round-trip traversal pair. The duration of this traversal in the outer segment is recorded, which is obtained by subtracting the occurrence time of the out-of-bounds node from the occurrence time of the return node.

[0062] In one embodiment, if the pairing stack is still not empty after the sequence scan, each unpaired out-of-bounds node in the stack corresponds to a boundary crossing that has not yet been returned to. The last unpaired out-of-bounds node is the starting point where the spatial object has not yet returned to the mask-covered area. At this time, the crossing pattern is marked as continuously leaving the mask coverage. Conversely, if the stack is empty, the spatial object has re-entered the mask-covered area from outside the mask within the current judgment window, and the crossing pattern consists only of round-trip crossings. In this case, no prohibition conclusion is formed, and each round-trip crossing is only recorded and not used to block the issuance of control tasks.

[0063] Preferably, the resulting prohibition conclusion is stored in the form of a structured record. Fields include the spatial object's task identifier, traversal type, set of outbound nodes, set of returning nodes, the continuous departure start time in the case of continuous departure from the mask cover, the set of consecutive non-returning outbound intersections, and the average azimuth angle of the continuous outbound direction. The traversal type takes one of two values: round-trip traversal or continuous departure from the mask cover. The continuous departure start time field in the case of round-trip traversal is set to null. The continuous departure start time in the case of continuous departure from the mask cover is the time when the last unpaired outbound node occurs. The set of consecutive non-returning outbound intersections consists of the last unpaired outbound node and its subsequent outbound nodes. The average azimuth angle of the continuous outbound direction is denoted as A, calculated as the sum of the azimuth angles of all nodes in the set of consecutive non-returning outbound intersections divided by the total number of nodes N.

[0064] It is understandable that the explicit solution of boundary intersections is interconnected with the temporal pairing of outbound nodes and return nodes, allowing the movement status inside and outside the mesh mask to be fully recorded in the prohibition conclusions of the spatial object in a structured field.

[0065] The trajectory segments where the real-time motion trajectory intersects with the boundary of the allowed mesh mask are captured. The intersection positions of each trajectory segment entering the allowed mesh mask coverage area and crossing the mask boundary are marked. The intersection positions are connected in sequence according to the movement order, and it is checked whether the trajectory after crossing the boundary returns to the allowed mesh mask coverage area. Based on the continuous crossing direction and return status, the prohibition conclusion of the spatial object is formed.

[0066] Using the real-time motion trajectory of the spatial object as the processing object, the intersection screening is performed along the trajectory in the order of the reported time for the jagged grid edge sequence corresponding to the boundary of the allowable grid mask of each pair of adjacent positioning points. If the intersection point of the connecting line with a certain grid edge falls within the range of the grid edge segment, the intersection point is determined to be a valid intersection point. The sub-segment formed by the two adjacent positioning points with valid intersection points is extracted as the trajectory segment to be processed, resulting in a set of trajectory segments intersecting with the mask boundary. For each trajectory segment in the set of trajectory segments, the inside and outside classification of the positioning points at both ends of the segment relative to the coverage area of ​​the allowable grid mask is used as the criterion. Valid intersection points from the outside to the inside are marked as entry intersection points, and valid intersection points from the inside to the outside are marked as exit intersection points. The two types of intersection points are collectively referred to as the intersection position, and the latitude and longitude, occurrence time and crossing direction of the intersection position are recorded. The intersection positions are linked together in the order of movement to form an intersection position chain. Each intersection point is checked to see if an entry intersection point appears after it is crossed. If an entry intersection point appears after crossing, the crossed intersection point is marked as a returned state. If no entry intersection point appears, it is marked as a non-returned state. For crossed intersection points that have not returned, their crossing direction is taken as the continuous crossing direction. If the number of consecutive non-returned crossed intersection points at the end of the chain reaches two or more and the continuous crossing directions are consistent, then the conclusion that the space object is prohibited is formed.

[0067] In the optimization stage of the restricted access determination in the closed-loop scheduling and control of the Beidou grid spatiotemporal intelligent agent, the aforementioned entry and exit conversion judgment based on the positioning point as the node is applicable to the boundary recognition of conventional trajectories. However, when the trajectory sampling interval is large, the line connecting two adjacent positioning points may cross the boundary of the allowed grid mask, and both ends are on the same side. The attribution of the endpoints inside and outside cannot expose such crossings. Therefore, it is necessary to use the intersection relationship between the trajectory segment and the boundary as the direct criterion to carry out more granular judgment.

[0068] Specifically, taking the real-time motion trajectory of the spatial object as the processing object, the line connecting each pair of adjacent positioning points along the trajectory in the order of the reported time is used to perform intersection screening with the jagged grid edge sequence corresponding to the boundary of the allowed grid mask. The intersection screening uses the grid edge as the smallest comparison unit, and the grid edge is a line segment along the longitude or latitude direction.

[0069] It should be noted that before the intersection screening, a spatial index based on the latitude and longitude bounding box is first established for the jagged grid edge sequence. When connecting two adjacent positioning points one after another, only the candidate grid edges that intersect with their bounding boxes are retrieved during the screening, avoiding traversal of all grid edges.

[0070] In one possible implementation, for each candidate grid edge, the intersection point of the line connecting two adjacent positioning points and the line containing the grid edge is calculated. If the directions of the two lines are not parallel (i.e., the determinant of the simultaneous equations of the line segments is not zero), a unique intersection point is obtained. It is then checked whether this intersection point falls between the two endpoints of the connecting line and the two endpoints of the grid edge simultaneously. If both conditions are met, it is considered a valid intersection point; otherwise, it is considered an invalid intersection point and discarded. If the directions of the two lines are parallel and do not coincide, there is no intersection point, and the candidate grid edge is discarded directly. If the directions of the two lines are parallel and coincident, the two endpoints of the candidate grid edge are used as candidate valid intersection points in the degenerate case, and those points located within the interval between the two endpoints of the connecting line are included in the set of valid intersection points. The sub-segment formed by two adjacent positioning points with at least one valid intersection point is extracted as the trajectory segment to be processed. All trajectory segments to be processed are aggregated according to the reporting time sequence into a set of trajectory segments intersecting the mask boundary. Furthermore, for each trajectory segment in the set of trajectory segments, the intersection position is determined by the classification of the positioning points at both ends of the segment relative to the area covered by the allowable grid mask. Each positioning point at both ends of the segment performs a set membership determination within the grid code set of the allowable grid mask according to the grid code to which its BeiDou two-dimensional grid position code belongs. Those that match are classified as "inside," and those that do not match are classified as "outside." If the front end of the segment is outside and the end end is inside, then the valid intersection point in that segment is marked as an "entry intersection point"; if the front end of the segment is inside and the end end is outside, then the valid intersection point in that segment is marked as an "exit intersection point." When multiple valid intersection points appear in the same segment, the intersection points are sorted according to their distance along the segment's direction of travel. The nearest intersection point undertakes the first inside / outside switch, and each subsequent intersection point undertakes a switch in the opposite direction, thus alternately marking them as either "entry intersection point" or "exit intersection point." The entry point and the exit point are collectively referred to as the intersection position. The intersection position includes fields for latitude and longitude, time of occurrence, and crossing direction. The crossing direction can be either from outside to inside or from inside to outside. The time of occurrence is obtained by linear interpolation based on the distance ratio of the valid intersection point to its segment. Specifically, the coordinates of the front-end positioning point of the segment are taken as longitude x1, latitude y1, and reporting time t1; the coordinates of the end positioning point are taken as longitude x2, latitude y2, and reporting time t2; and the coordinates of the valid intersection point are taken as longitude xc and latitude yc, and the distance ratio is used. The time of the handover is tc = t1 + k × (t2 - t1).

[0071] In one embodiment, all handover positions are concatenated in ascending order of occurrence time to form a handover position chain. Each node on the chain is a handover position. The node type is either entering the handover point or crossing the handover point. Adjacent nodes of the chain correspond to the sequential actions of the target entering the allowed mesh mask coverage area and crossing the mask boundary, respectively.

[0072] Preferably, the intersection position chain is scanned sequentially according to the occurrence time. For each exit intersection point, adjacent nodes are traversed: if there is an entry intersection point with a later temporal position after the exit intersection point, the exit intersection point is marked as a returned state, and the difference between the occurrence time of the corresponding entry intersection point and the occurrence time of the exit intersection point is recorded as the duration of being outside; if no entry intersection point appears after traversing the chain to the end, the exit intersection point is marked as a non-return state. For all exit intersection points marked as non-return states, their crossing direction is taken as the continuous exit direction. The continuous outward movement is defined by azimuth angles from the inside out, with true north as 0 degrees and increasing clockwise to 360 degrees. Let x1 be the longitude and y1 of a point before the outward intersection, and x2 be the longitude and y2 of a point after the outward intersection. Let dx = x2 - x1 and dy = y2 - y1. The azimuth angle θ = atan2(dx, dy) is calculated in radians and converted to degrees. Negative values ​​are added to 360 degrees and placed within the 0 to 360 degree range. If the number of consecutive outward intersections without return at the end of the chain reaches two or more, and the difference in azimuth angle between adjacent non-returning outward intersections does not exceed 30 degrees, then the spatial object is determined to have continuously moved out of the allowed mesh mask coverage area along a consistent direction, resulting in a no-entry conclusion for the spatial object. In one implementation, the resulting prohibition conclusion is stored in the form of a structured record, with fields including the task identifier of the spatial object, the set of consecutive non-returning cross-intersection points, the average azimuth angle of the consecutive cross-intersection directions, the departure time from the starting point, and the determination time. The departure time from the starting point is the earliest occurrence time among the consecutive non-returning cross-intersection points, and the determination time is the current time when the prohibition conclusion is formed.

[0073] S107. Based on the task identifier corresponding to the spatial object that forms the prohibition conclusion, the Beidou grid spatiotemporal intelligent agent issues a control task to block the spatial object association, records the interception result associated with the task identifier and initiates manual review. If the review is successful, a new manual confirmation certificate bound to the current grid version is issued, or if the review fails, the task is permanently terminated and the world model is written back to complete the closed-loop scheduling control.

[0074] Based on the task identifier corresponding to the spatial object that forms the prohibition conclusion, the task identifier is blocked in the command issuance channel of the Beidou grid spatiotemporal intelligent agent. The issuance status of the control task associated with the task identifier is changed from pending issuance to intercepted issuance. The issuance status can be one of four values: pending issuance, issued issuance, intercepted, or terminated. The interception result associated with the task identifier is obtained. The interception result, along with all fields of the prohibition conclusion, is written into the interception log. The interception log uses the task identifier as the primary key, and the fields include the interception time, the issuance status before interception, the set of consecutive non-returned crossover intersections, the average azimuth angle of the consecutive crossover directions, the departure time from the starting point, and the judgment time. A review work order is pushed to the manual review terminal. The work order uses the task identifier as the binding key and carries a snapshot of the record in the interception log to complete the closed-loop scheduling control.

[0075] In the command blocking stage of the BeiDou grid spatiotemporal intelligent agent closed-loop scheduling and control, the command issuance channel organizes the issuance queues of each control task using the task identifier as the routing key. The issuance status of each task in the queue takes one of four values: pending issuance, issued, intercepted, or terminated. Based on the task identifier corresponding to the spatial object that forms the prohibition conclusion, the issuance queue node corresponding to the task identifier is located in the command issuance channel, and a blocking operation is performed on it: the node is removed from the pending issuance queue, transferred to and attached to the intercepted queue, and its issuance status field is simultaneously changed from pending issuance to intercepted, thus obtaining the interception result associated with the task identifier.

[0076] Specifically, the interception result, along with all fields of the prohibition conclusion, is written into the interception log. The interception log uses the task identifier as the primary key, and the fields include the interception time, the issuance status before interception, the set of consecutive non-returning crossover intersections, the average azimuth angle of the consecutive crossover directions, the departure time from the starting point, the judgment time, the review status, the review conclusion, and the post-review handling result. The post-review handling result takes one of two values: new voucher issuance or permanent termination of the task.

[0077] For example, the interception time is the current time when the blocking operation is performed, the pre-interception status is the value of the node before the blocking, the initial value of the review status is set to pending review when it is written, and the review conclusion field is left for manual review to write back.

[0078] Preferably, a review work order is pushed to the manual reviewer. The work order uses the task identifier as the binding key and carries a snapshot of the record in the interception log. After receiving the work order, the manual reviewer performs a second reading of the interception result. The review conclusion obtained from the reading is written back to the review conclusion field of the interception log along the work order, and the review status is changed from pending review to reviewed. Manual review must produce control consequences and must not remain at the record level: When the review conclusion is pass, the manual reviewer issues a new manual confirmation certificate bound to the current grid version. The confirmation event grid code of the new certificate is re-valued according to the current target grid. Based on the new certificate, the distribution status of the task identifier is restored from the intercepted state to the pending state, the distribution block is lifted, and the control task re-enters the distribution queue with the new certificate. When the review conclusion is fail, the control task associated with the task identifier is permanently terminated, the distribution status of the task identifier is changed to the terminated state, the task termination status along with the review conclusion is written back to the target trajectory library of the world model and the interception ledger, the task identifier exits the closed-loop scheduling control, and the closed-loop scheduling control is completed accordingly.

[0079] Obviously, those skilled in the art can make various modifications and variations to the embodiments of this application without departing from the spirit and scope of the embodiments of this application. Therefore, if these modifications and variations to the embodiments of this application fall within the scope of the claims of this application and their equivalents, this application also intends to include these modifications and variations.

Claims

1. A closed-loop scheduling and control method for a BeiDou grid-based spatiotemporal intelligent agent, characterized in that, The method includes: Obtain the manual confirmation voucher, extract the task identifier and confirmation event grid code, and identify the voucher grid version and voucher generation time through the confirmation event grid code; Based on the task identifier, extract the target migration record from the world model, obtain the target's current location and entry time, and identify the migration stage after manual confirmation; The target grid is obtained by identifying the grid affiliation of the current location of the target through Beidou grid coding. The migration path to the current target grid is traced starting from the grid indicated by the confirmed event grid coding, thus forming a confirmed cross-grid migration fact. Extract the current grid version and allowed grid mask from the current scheduling system, trace the execution status of the credential grid version along the version update chain, evaluate the in-service qualification of the credential version, and form a conclusion that the credential version has failed. Based on the confirmed cross-grid migration facts and the conclusion of invalid certificate version, the triggering conditions are determined. If any triggering condition is met, the control target is locked from the target migration record and identified as a spatial object, and the real-time motion trajectory of the spatial object is obtained. Analyze the real-time motion trajectory within the allowed mesh mask coverage area, identify the crossing patterns where the trajectory crosses the mask boundary, and form a no-entry conclusion for the spatial object based on the crossing patterns that continuously leave the mask coverage. Based on the task identifier corresponding to the spatial object that forms the prohibition conclusion, the control task associated with the spatial object is blocked from being issued, the interception result is recorded and a manual review is initiated, and control measures are executed according to the review conclusion to complete the closed-loop scheduling control.

2. The BeiDou grid spatiotemporal intelligent agent closed-loop scheduling and control method according to claim 1, characterized in that, The process of obtaining manual confirmation credentials, extracting task identifiers and confirmation event grid codes, and identifying the credential grid version and credential generation time through the confirmation event grid codes includes: Obtain the original message of the manual confirmation credential, extract the task identifier and confirmation event grid code, and verify the level and parent affiliation of the confirmation event grid code level by level. Retrieve the voucher grid version from the version registration table with grid code as the key, extract the voucher generation time and store it in the baseline element table.

3. The BeiDou grid spatiotemporal intelligent agent closed-loop scheduling and control method according to claim 1, characterized in that, The process of extracting target migration records from the world model based on task identifiers, obtaining the target's current location and entry time, and identifying the migration stage after manual confirmation includes: Retrieve the target migration records associated with the task identifier from the target trajectory library of the world model. The target migration records record inter-cell jump events one by one in order of timestamp. Read the latest BeiDou positioning point from the target trajectory database and use that BeiDou positioning point itself as the target's current location; Events whose entry time is later than the voucher generation time are screened out and connected in sequence as the migration stage.

4. The BeiDou grid spatiotemporal intelligent agent closed-loop scheduling and control method according to claim 1, characterized in that, The process of identifying the grid affiliation of the target location through BeiDou grid coding to obtain the current target grid, and tracing the migration path to the current target grid starting from the grid indicated by the confirmed event grid code, thus forming a confirmed cross-grid migration fact, includes: The target address is mapped to a code at the same level as the confirmation event grid code to obtain the current target grid; Starting from the grid indicated by the confirmed event grid code, a migration path is formed by connecting the inter-grid transition event sequence to the grid grid code; In the events covered by the migration path, the jumps where the entry time is later than the voucher generation time are marked as confirmed cross-cell migration facts after the cross-cell node is confirmed.

5. The BeiDou grid spatiotemporal intelligent agent closed-loop scheduling and control method according to claim 1, characterized in that, The process of extracting the current grid version and allowed grid masks from the current scheduling system, tracing the execution status of the credential grid version along the version update chain, evaluating the in-service status of the credential version, and forming a conclusion that the credential version has expired includes: Read the current mesh version and the bound allowed mesh masks from the mesh state library; The version update chain is formed by retrieving the iteration records of each credential grid version to the current grid version and its historical allowed grid masks. The tracing window is defined by the entry times of the first and last cross-grid nodes in the confirmed cross-grid migration facts; The historical allowed grid mask of the iteration item falling into the traceability window is compared with the confirmation event grid code. If there is an iteration item whose confirmation event grid code is removed from the historical allowed grid mask, the credential grid version is determined to be out of service.

6. The BeiDou grid spatiotemporal intelligent agent closed-loop scheduling and control method according to claim 1, characterized in that, The process involves determining triggering conditions based on the confirmed cross-grid migration facts and the invalidation of the credential version. If any triggering condition is met, the control target is locked from the target migration record and identified as a spatial object. The real-time motion trajectory of the spatial object is then obtained, including: The current target mesh is compared with the existing allowed mesh mask. If the current target mesh does not belong to the existing allowed mesh mask, it is determined to be a position drift. The confirmation event grid code is compared with the current allowed grid mask and the historical allowed grid mask in the trace window along the version update chain. If the confirmation event grid code does not belong to the current allowed grid mask or is removed from the historical allowed grid mask in the trace window, it is determined to be a policy change. The migration path recorded after the confirmed cross-grid migration is compared with the current allowable grid mask. If the cross-grid migration occurs after confirmation and the current allowable grid mask does not cover the migration path, it is determined to be a path mismatch. The three triggering conditions are determined independently in parallel. When any one of them is met, the control target is locked as a trackable item and registered as a spatial object. A trajectory push request is initiated to the BeiDou positioning module, using the spatial object task identifier as the filtering condition, and the received positioning points are stitched together to form a real-time motion trajectory.

7. The BeiDou grid spatiotemporal intelligent agent closed-loop scheduling and control method according to claim 1, characterized in that, The analysis of real-time motion trajectories within the allowed mesh mask coverage area identifies trajectories that cross mask boundaries, and uses these trajectories that continuously detach from mask coverage to form a no-entry conclusion for the spatial object, including: Map each positioning point of the real-time motion trajectory to a grid code, and determine the membership of the grid code set that allows grid masking. Points belonging to the set are considered inside points, and those not belonging to the set are considered outside points. The intersection of the line connecting adjacent positioning points and the mask boundary is taken as the boundary crossing node or return node to obtain the entry and exit transformation sequence; A node that crosses the boundary and has a paired return node is marked as a round trip; a node that does not have a paired return node is marked as continuously leaving the mask cover. The conclusion that the space object is prohibited from entering is formed by continuously detaching from the mask coverage, and the round trip is recorded and stored, but the control task is not blocked based on this.

8. The BeiDou grid spatiotemporal intelligent agent closed-loop scheduling and control method according to claim 7, characterized in that, The prohibition conclusion that forms the spatial object includes: The intersection of the line connecting adjacent positioning points with the boundary of the allowable mesh mask is screened along the real-time motion trajectory. The intersection point that falls within the grid edge line segment is determined to be a valid intersection point. The valid intersection points from the outside to the inside are marked as the entry intersection points, and the valid intersection points from the inside to the outside are marked as the exit intersection points; these are collectively referred to as the intersection positions. The handover positions are linked together in the order of movement to form a handover position chain. After each point is crossed, it is checked whether an entry point is encountered. If an entry point is encountered, it is marked as having returned; otherwise, it is marked as not having returned. If the chain tail fails to return consecutively and continuously crosses the intersection point in the same direction, a no-entry conclusion is formed.

9. The BeiDou grid spatiotemporal intelligent agent closed-loop scheduling and control method according to claim 1, characterized in that, Based on the task identifier corresponding to the spatial object that forms the prohibition conclusion, the issuance of control tasks associated with the spatial object is blocked, the interception result is recorded and a manual review is initiated, and control measures are executed according to the review conclusion to complete the closed-loop scheduling control, including: In the command issuance channel of the Beidou grid spatiotemporal intelligent agent, the task identifier is blocked, the issuance status of the control task is changed to the intercepted status, and the interception result associated with the task identifier is obtained; if the review is successful, a new manual confirmation certificate bound to the current grid version is issued; if the review fails, the control task associated with the task identifier is permanently terminated and the task termination status is written back to the world model.

10. The BeiDou grid spatiotemporal intelligent agent closed-loop scheduling and control method according to claim 9, characterized in that, The process of obtaining the interception result associated with the task identifier includes: The interception result, along with the field of the prohibition conclusion, is written into the interception log, with the task identifier as the primary key. A review work order is pushed to the manual review terminal. The review work order uses the task identifier as the binding key and carries the record snapshot in the interception log to complete the closed-loop scheduling control.

Images