Multi-agent cooperation task decomposition and execution planning method and system

Abstract

The application discloses a multi-agent cooperation task decomposition and execution planning method and system, relates to the technical field of multi-agent planning, and comprises the following steps: acquiring task section data and execution return data of a target object; performing corresponding grouping on the execution return data and the task section data, generating breakpoint undertaking data, and screening a part in the breakpoint undertaking data which cannot be continuously undertaken but still needs to be attributed to a task according to the task section data, to generate hosting segment data. According to the scheme, the execution return data is written in the range limited by the hosting segment data, and then the execution return data is arranged with the hosting segment data, to generate back insertion candidate data, so that the execution return data which is late or recovered can only enter the writable task range defined by the hosting segment data, and the task stage dislocation caused by direct addition according to the communication arrival order is avoided.

Description

Technical Field

[0001] This invention relates to the field of multi-agent planning technology, specifically to a method and system for multi-agent collaborative task decomposition and execution planning. Background Technology

[0002] With the application of emergency rescue robots, drones, smoke extraction equipment control systems, vehicle-mounted command systems, and rescue personnel terminals in traffic accident handling, multi-agent collaborative task decomposition and execution planning technology has gradually become an important technical means in complex accident rescue scenarios. This type of technology usually obtains accident handling tasks through the command terminal, and decomposes tasks such as reconnaissance, sealing off, smoke extraction, and passage clearing and assigns them to different execution entities. Then, it updates the task plan based on the execution status returned by each execution entity, so that multiple intelligent agents can coordinate and execute in the same accident handling process.

[0003] Existing multi-agent collaborative task planning schemes typically involve a command platform acquiring rescue mission data, creating task queues based on task type, priority, executor capabilities, or location, and then distributing these queues to executors such as drones, inspection robots, fixed control equipment, or personnel terminals. During execution, each executor transmits its status back to the command center via communication links. The command center then updates the task queue based on the transmitted data, or reissues or reassigns tasks if an executor goes offline, a task is unconfirmed, or a communication timeout occurs. The advantages of this approach are continuous communication, real-time status feedback from executors, and smooth task execution. When the row node can be continuously confirmed, task decomposition, task distribution and status update can be completed relatively directly, and tasks of multiple execution entities can be centrally managed through task queues. However, in the intermittent communication environment of accident rescue in mountainous highway tunnel groups, if the execution feedback generated by different intelligent agents after going offline or after communication is restored is written into the task record according to the order of reception at the command end, it is difficult to determine the original rescue task segment to which it should return, and it is also difficult to confirm whether it still belongs to the existing rescue task main chain. This leads to uncertainty in the insertion position of execution feedback, non-unique task affiliation, and inaccurate constraints on the order of subsequent reconnaissance, sealing, smoke extraction and rescue channel clearing actions. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for multi-agent collaborative task decomposition and execution planning, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows:

[0006] In a first aspect, this invention discloses a method for multi-agent cooperative task decomposition and execution planning, applied to multi-agent task planning under conditions of interrupted communication during tunnel group accident rescue, comprising the following steps:

[0007] Obtain the task segment data and execution feedback data of the target object;

[0008] The execution feedback data and the task segment data are grouped accordingly to generate breakpoint acceptance data. Based on the task segment data, the parts of the breakpoint acceptance data that cannot be continuously accepted but still need to be assigned to tasks are screened to generate managed segment data.

[0009] After the execution feedback data is written within the scope of the managed fragment data, it is organized with the managed fragment data to generate back-insertion candidate data. The back-insertion candidate data and the breakpoint acceptance data are then judged to be consistent in terms of task. If they are consistent, the back-insertion candidate data is effectively retained.

[0010] The valid retention results are merged based on the breakpoint data to generate back-insertion verification data. The task segment data and the back-insertion verification data are then subjected to task order constraint processing under the same task to generate phase-locked data.

[0011] The phase-locked data, the back-insertion verification data, and the task segment data are subjected to task consistency verification to generate planning result data.

[0012] Secondly, this invention discloses a multi-agent cooperative task decomposition and execution planning system, comprising:

[0013] The data acquisition module is used to acquire task segment data and execution feedback data of the target object;

[0014] The blind zone hosting screening module is used to group the execution feedback data and the task segment data accordingly, generate breakpoint acceptance data, and screen the part of the breakpoint acceptance data that cannot be continuously accepted but still needs to be assigned a task according to the task segment data, and generate hosting segment data.

[0015] The consistency judgment module is used to organize the execution return data with the managed fragment data after limiting the writing range of the managed fragment data, generate back-insertion candidate data, and perform task consistency judgment between the back-insertion candidate data and the breakpoint acceptance data. If they are consistent, the back-insertion candidate data is effectively retained.

[0016] The task sequence constraint module is used to merge the valid retention results based on the breakpoint continuation data, generate back-insertion verification data, and perform task sequence constraint processing on the task segment data and the back-insertion verification data under the same task to generate phase-locked data.

[0017] The planning result generation module is used to perform task consistency verification on the phase-locked data, the back-insertion verification data, and the task segment data, and generate planning result data.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0019] 1. This solution generates back-insertion candidate data by limiting the write range of the execution return data after it has been processed by the managed fragment data. This ensures that late or resumed execution return data can only enter the writable task range limited by the managed fragment data, avoiding direct appending based on the communication arrival order and causing task stage misalignment. The back-insertion candidate data and the breakpoint acceptance data are judged for task consistency. If they are consistent, the back-insertion candidate data is effectively retained, allowing the retained objects to be traced back to the previous breakpoint acceptance relationship and excluding return objects that do not belong to the same task main chain.

[0020] 2. This solution determines the task affiliation of back-insertion candidate data and managed fragment data, thus limiting the back-insertion candidate data to the main chain task scope corresponding to the managed fragment data, preventing back-up objects from different managed fragments from entering subsequent processing. When the determination is valid, the back-insertion candidate data is retained and main chain affiliation data is generated, giving the back-insertable objects a clear basis for main chain affiliation. Then, it is determined whether the main chain affiliation data can be traced back to the breakpoint receiving data, allowing the retained objects to be further verified with the previous breakpoint receiving relationship, preventing late-returned objects from deviating from the original receiving path. When the determination is valid, a valid retention result is generated, ensuring that subsequent planning and processing only accepts task objects that simultaneously satisfy managed affiliation and breakpoint backtracking. Attached Figure Description

[0021] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts. Wherein:

[0022] Figure 1 A flowchart illustrating the steps of the multi-agent collaborative task decomposition and execution planning method provided by the present invention;

[0023] Figure 2 A flowchart illustrating the process of acquiring task segment data of a target object and executing feedback data provided by the present invention;

[0024] Figure 3 This is a schematic diagram of the process for generating breakpoint continuation data provided by the present invention.

[0025] Figure 4 This is a schematic diagram of the process for generating back-interpolation candidate data provided by the present invention;

[0026] Figure 5This is a schematic diagram of the module functions of the multi-agent collaborative task decomposition and execution planning system provided by the present invention. Detailed Implementation

[0027] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0028] Application Overview:

[0029] Existing multi-agent collaborative task planning schemes typically involve a command platform acquiring rescue task data and forming a task queue based on task type, priority, executor capabilities, or task location. The task queue is then updated based on execution status feedback from executors such as drones, inspection robots, fixed control equipment, or personnel terminals. This type of processing relies on continuous communication, real-time feedback of executor status, and continuous confirmation of task execution nodes. However, in the intermittent communication environment of accident rescue in mountainous highway tunnel clusters, different agents may generate delayed execution feedback after going offline or after communication is restored. If the command end simply writes the execution feedback into the task record according to the order of receipt, there is a lack of a basis for determining the connection between the execution feedback data and the task segment data, making it difficult to confirm whether the execution feedback still belongs to the existing rescue task main chain. Consequently, the task queue update process exposes problems such as uncertain insertion position of execution feedback, non-unique task affiliation, and inaccurate constraints on the order of subsequent tasks under the same task.

[0030] During the handling of traffic accidents in mountainous highway tunnel clusters, the command center breaks down the tasks of reconnaissance, closure, smoke extraction, and rescue channel clearing, and distributes them separately to drones, inspection robots, smoke extraction equipment control systems, and rescue personnel terminals. After the inspection robot enters the tunnel section, it fails to continuously report its execution status due to a communication link interruption. During this period, the command center continues to update the task queue based on the feedback results from other executing entities. When the inspection robot's communication is restored, its execution feedback arrives at the command center. If this feedback is only written to the task record in the order of receipt, the command center cannot determine whether the feedback corresponds to the original reconnaissance task segment, the pre-closure task segment, or a task already included in the subsequent rescue channel clearing sequence. Furthermore, it is difficult to determine whether the feedback still belongs to the existing main rescue task chain. This problem manifests in the following ways at the task record update location: the insertion position of the later-arriving feedback is uncertain, multiple possible task attributions exist, and the sequential constraints between subsequent reconnaissance, closure, smoke extraction, and rescue channel clearing tasks are shifted.

[0031] If the above issues are not addressed, the subsequent execution feedback will propagate through the task record writing stage to the task queue update stage. This makes it difficult for the execution feedback data, which should correspond to a certain rescue task segment, to be stably included in the task scope defined by the task segment data. This deviation will continue to propagate to the subsequent task sequence constraint formation process, causing the sequential relationship between reconnaissance, containment, smoke removal, and rescue channel clearing actions to lose a reliable task attribution basis. The adverse consequence is that when the planning results are formed, task arrangements may be generated based on incorrect insertion positions or non-unique task attribution, resulting in link anomalies between the execution feedback in the same accident handling process and the existing rescue task main chain, thereby affecting the determination of task inheritance relationships, task attribution relationships, and task sequence relationships in the planning result data.

[0032] After introducing the basic concept of the present invention, the embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0033] Example 1:

[0034] Please see Figure 1 A multi-agent collaborative task decomposition and execution planning method, applied to multi-agent task planning under intermittent communication conditions during tunnel group accident rescue, includes the following steps:

[0035] Obtain the task segment data and execution feedback data of the target object;

[0036] The execution feedback data and task segment data are grouped accordingly to generate breakpoint acceptance data. Based on the task segment data, the parts of the breakpoint acceptance data that cannot be continuously accepted but still need to be assigned to tasks are screened to generate managed segment data.

[0037] After limiting the write range of the execution return data and the managed fragment data, the data is organized with the managed fragment data to generate back-insertion candidate data. The back-insertion candidate data and the breakpoint acceptance data are then judged to be consistent with the task. If they are consistent, the back-insertion candidate data is effectively retained.

[0038] The valid retained results are merged based on the breakpoint data to generate back-insertion verification data. The task segment data and back-insertion verification data are then subjected to task order constraint processing under the same task to generate phase-locked data.

[0039] Perform task consistency verification on phase-locked data, back-interpolation verification data, and task segment data to generate planning result data.

[0040] Among them, the task segment data refers to the complete task basic data object formed after unified expression, sequential limitation and segmentation of rescue tasks such as exploration, sealing, smoke extraction and rescue channel clearing under the condition of intermittent communication in tunnel group accident rescue, and is used to group the corresponding execution feedback data.

[0041] Execution feedback data refers to the complete data object used to characterize the actual correspondence between the task execution object and the task segment after the target object participating in the tunnel group accident rescue mission receives or executes the task defined by the corresponding task segment data;

[0042] Breakpoint takeover data refers to the complete intermediate data object formed by the task scheduling service after correspondingly grouping the task segment data and the execution feedback data. It is used to represent the task takeover status of the execution feedback data relative to the task segment data under the intermittent communication conditions of tunnel group accident rescue.

[0043] Managed fragment data refers to a complete intermediate data object formed under the condition of intermittent communication in tunnel group accident rescue. It is formed by defining the task affiliation of the part of the data that cannot be continuously connected but still belongs to the same main chain of rescue task from the task segment data. It is used to represent the task fragment that cannot be directly incorporated into the continuous task chain when communication is interrupted, the return is delayed or the return is restored, but can still be rewritten, verified and recycled by the subsequent execution return data.

[0044] Back-insertion candidate data refers to a data object formed under intermittent communication conditions, for execution back-transmission data that has been resumed to be received, firstly by limiting the write range that can return to the original task main chain with managed fragment data, and then by correspondingly processing the limited execution back-transmission data and managed fragment data, which is used to determine whether it can return to the original breakpoint to continue the path.

[0045] Effective retention results refer to the set of back-insertion candidate objects that are determined to continue to belong to the same rescue mission main chain, can be traced back to the original breakpoint inheritance relationship, and have not broken out of the writing range limited by the managed fragment data.

[0046] Back-insertion verification data refers to the data object formed after confirming the task ownership, limiting the back-insertion position, and verifying the consistency of execution back-transmission data after limiting the writing range of managed fragment data under intermittent communication conditions. It is used to indicate that the execution back-transmission data of late back-transmission, recovery back-transmission, or intermittent back-transmission has been reintegrated into the same task main chain and can be used as the verification basis for the subsequent generation of phase-locked data.

[0047] Phase-locked data refers to a complete data object formed by task segment data and back-insertion verification data, used to characterize the execution sequence constraints of each task object within the main chain of the same rescue mission.

[0048] Planning result data refers to the final data object formed under the intermittent communication conditions of tunnel group accident rescue, which is formed by processing task segment data, execution feedback data, breakpoint takeover data, managed segment data, back-insertion verification data and phase locking data through the same data main chain, and is used to limit the decomposition and execution order of multi-agent collaborative tasks.

[0049] This solution acquires task segment data and execution feedback data, allowing the task object to be executed and the actual feedback object from multiple agents to enter the same processing chain as complete data objects, providing task and execution sources for subsequent acceptance judgments. It generates breakpoint acceptance data by correspondingly grouping the execution feedback data and task segment data, ensuring that the execution feedback object is first limited to the task range corresponding to the task segment data, and then distinguishes between continuously acceptable objects and objects that cannot be continuously accepted but still require assignment confirmation. This avoids task assignment misalignment caused by directly merging execution feedback data into the task chain in the order of arrival. Based on the task segment data, it filters the parts of the breakpoint acceptance data that cannot be continuously accepted but still require task assignment to generate managed fragment data, ensuring that task objects that cannot be continuously accepted under intermittent communication remain within the same task source range, providing writable main chain objects for subsequent late or resumed feedback.

[0050] The scope of execution return data is limited by managed fragment data, and it is then combined with managed fragment data to generate back-insertion candidate data. This ensures that execution return data is no longer appended as a regular return record, but is instead confined to manageable and reclaimable task fragments, thereby reducing the possibility of erroneous writing to non-corresponding task chains. Task consistency is checked between the back-insertion candidate data and the breakpoint acceptance data, and valid results are retained. This ensures that the back-insertion candidate data must return to the preceding breakpoint acceptance relationship, preventing late returns from falling into managed fragments but deviating from the original acceptance path. Based on the breakpoint acceptance data, valid retained results are merged to generate back-insertion verification data, ensuring consistency with managed fragments and breakpoints. The returned data forms a unified verification result, providing an execution basis with eliminated erroneous attribution for subsequent task sequence constraints. The task segment data and back-insertion verification data are processed under the same task to generate phase-locked data, ensuring that the tasks of exploration, containment, smoke extraction, and rescue channel clearing do not enter the output stage in multiple parallel queues, but rather form a sequence constraint under the same task chain. The phase-locked data, back-insertion verification data, and task segment data are subjected to task consistency verification to generate planning result data, so that the final output is simultaneously constrained by the task source, back-insertion verification, and sequence constraints, thereby forming planning result data with a clear source, closed sequence, and uniqueness.

[0051] The above describes a complete scheme for multi-agent collaborative task decomposition and execution planning. The following section describes how to obtain task segment data and execution feedback data for the target object. Please refer to [link / reference]. Figure 2 , Figure 2 The flowchart illustrating the process of obtaining task segment data and execution feedback data of the target object provided in this application embodiment specifically includes:

[0052] Acquire rescue mission data and execution feedback data of the target object;

[0053] The rescue mission data is formatted and normalized to generate mission format data. The mission format data is then coded and segmented to generate segment candidate data.

[0054] The correspondence between the candidate data of the segment and the task format data is determined. If the correspondence is found, the candidate data of the segment is retained and the task segment data is generated.

[0055] Otherwise, exclude the corresponding segment candidate data.

[0056] Among them, rescue mission data refers to complete machine-side task data that is generated and output based on the current needs of accident rescue and disposal, and is used to characterize the initial decomposition objects of multi-agent collaborative tasks;

[0057] Task format data refers to data objects formed by format normalizing rescue task data, which are used to unify different rescue task objects into the same machine-processable expression form.

[0058] Candidate segment data refers to the candidate task segments used to represent each task object in the rescue mission data under a unified machine-side processing format.

[0059] The above content will be described in detail below:

[0060] The rescue mission data of the target objects is obtained through the rescue dispatch equipment. The rescue mission data includes, but is not limited to, the target objects for reconnaissance, the target objects for sealing off, the target objects for smoke removal, and the target objects for clearing rescue channels. It may also include the task category, the entity to which the execution is assigned, the section of the task execution, the sequence of tasks, and the status of the task processing in the main chain of the accident rescue of the mountain highway tunnel group.

[0061] The execution feedback data of the target object is obtained through the intelligent agent communication interface. The execution feedback data includes, but is not limited to, the execution response object, execution arrival object, execution interruption object, execution recovery object and execution feedback object of each execution subject for the tasks of surveying, sealing, smoke removal and rescue channel clearing. It may also include the feedback subject, feedback sequence, feedback attribution, feedback status and feedback connection relationship of the above execution objects under intermittent communication conditions.

[0062] The rescue mission data is formatted and normalized to generate mission format data. The format normalization process includes converting the different mission expression methods in the rescue mission data, which are oriented towards reconnaissance, blockade, smoke removal and rescue channel clearing, into a unified machine-side mission expression method, so that different types of missions can be processed in the same data processing main chain.

[0063] Task format data is encoded to generate task-coded data. Task encoding processing is used to convert each task object in the task format data into an encoded object that can be processed according to the same rescue task main chain. Task encoding processing takes the task format data as a complete input object and forms task-coded data through machine-side encoding.

[0064] The task coding data is divided into segments to generate segment candidate data.

[0065] As a preferred implementation, the segmentation process includes: taking task coding data as the object to be segmented, performing task object grouping processing on the task coding data to generate candidate group data; then performing segmentation processing on the candidate group data to generate segment candidate data; the candidate group data is used to limit the task object source of the segment candidate data, and the segment candidate data is used to limit the candidate range for subsequent correspondence judgment with task format data.

[0066] The process involves determining the correspondence between candidate segment data and task format data. Candidate segment data is used as the object to be determined, and task format data is used as the source verification object. Correspondence processing is performed to generate segment source relationship data. Specifically, the correspondence determination involves: determining whether the segment source relationship data indicates that the task object in the candidate segment data can be supported by the unified task expression object in the task format data; if the segment source relationship data indicates that the candidate segment data can return to the unified task expression range defined by the task format data, the determination result is "yes," and the corresponding candidate segment data is retained to generate task segment data; if the segment source relationship data indicates that the candidate segment data cannot return to the unified task expression range defined by the task format data, the determination result is "no," and the corresponding candidate segment data is excluded. Excluded candidate segment data does not participate in the formation of task segment data, nor does it enter the subsequent processing chain as a new output result.

[0067] This solution acquires rescue mission data and execution feedback data of the target object, enabling subsequent processing to have a data foundation of both the task objects to be decomposed and the actual execution feedback objects. This avoids the inability to handle subsequent feedback data when planning is based solely on pre-defined tasks. The rescue mission data is formatted to generate task format data, which converts tasks with different expressions such as reconnaissance, containment, smoke removal, and rescue channel clearing into unified machine-processable objects. This eliminates the inconsistency in data boundaries caused by differences in task category expressions in subsequent coding and segmentation.

[0068] Task format data is encoded and segmented to generate segment candidate data. This transforms rescue tasks, originally organized by task category, into segmented task objects that can establish a correspondence with execution feedback data. This provides a clear task scope for subsequent breakpoint acceptance, back-insertion verification, and planning result generation. The correspondence between segment candidate data and task format data is determined. When the correspondence is valid, the segment candidate data is retained and task segment data is generated. This allows task segment data to be traced back to the normalized task format data, preventing segment objects without a source or with incorrect segmentation from entering the subsequent main chain. When the correspondence is invalid, the corresponding segment candidate data is excluded. This prevents segment objects that cannot form a valid source correspondence with rescue task data from participating in subsequent processing, thereby reducing the risk of miscorrespondence between subsequent execution feedback data and incorrect task segments.

[0069] The above describes how to obtain the task segment data and execution feedback data of the target object. The following describes how to group the execution feedback data and task segment data accordingly to generate breakpoint continuation data. Please refer to [link / reference]. Figure 3 , Figure 3 This is a flowchart illustrating the process of generating breakpoint acceptance data provided in an embodiment of this application. Generating breakpoint acceptance data specifically includes:

[0070] The format of the execution return data is converted to generate the return data, and the return data is incorporated into the task segment data to generate the segment corresponding data.

[0071] Perform a continuity judgment on the data corresponding to the section and the data of the task section. If they are consecutive, output the data corresponding to the section as the consecutive data.

[0072] Otherwise, the data corresponding to the same segment will be judged to belong to the same task. If the task is found to be the same, the corresponding data of the segment will be output as the data to be managed and accepted.

[0073] Otherwise, the corresponding data of the relevant segment will be output as the main chain external data, and the corresponding data of the segment will be grouped according to the continuous acceptance data, the acceptance data to be hosted and the main chain external data to generate the breakpoint acceptance data.

[0074] Among them, the data received back refers to the complete data object formed by receiving the execution back data through the intelligent agent communication interface, and then performing interface receiving format unification, data expression form conversion and back object normalization on the execution back data.

[0075] Segment correspondence data refers to the data object used to represent the task segment attribution relationship and the basic relationship that can be continued between the execution feedback data after format reception and conversion and the task segment data obtained by the rescue mission data processing.

[0076] Continuous data refers to a complete data object that is determined to form a continuous correspondence with the task segment data within the same rescue mission main chain when performing continuous data inheritance judgment on the corresponding data of the segment and the task segment data.

[0077] Data to be managed and accepted refers to the set of data corresponding to the segments that, after continuous acceptance judgment of the corresponding data of the segments, cannot form a continuous acceptance relationship with the data of the task segments, but can still be attributed to the same rescue task main chain after the same task attribution judgment.

[0078] Main chain outbound data refers to excluded data objects that, after completing the continuous inheritance judgment and the same task affiliation judgment of the corresponding data of the segment, fail to form a continuous inheritance relationship with the task segment data and fail to be classified into the corresponding data of the same rescue task main chain.

[0079] The above content will be described in detail below:

[0080] The execution feedback data undergoes format conversion to generate feedback reception data. Specifically, the execution feedback data undergoes a unified data format conversion so that execution feedback objects returned by different execution entities are converted into feedback reception data that the task scheduling service can process together.

[0081] The entire returned data is incorporated into the task object scope defined by the task segment data to generate segment-corresponding data; specifically, the task segment data is used as the scope to be corresponded and the returned data is used as the object to be corresponded. Object correspondence processing is performed on the two to make the returned data correspond to the same rescue task main chain defined by the task segment data, thereby generating segment-corresponding data.

[0082] Continuous succession judgment is performed on segment-corresponding data and task segment data: Continuous judgment data is generated based on segment-corresponding data and task segment data. Specifically, segment-corresponding data is used as the object to be judged, and task segment data is used as the basis for the scope of the task main chain. The two are compared with previous and subsequent task objects, compared with the same task main chain, and compared with the order of task objects to generate continuous judgment data. Continuous judgment data is used to characterize whether segment-corresponding data can form a previous and subsequent succession relationship with existing task objects in the same rescue task main chain defined by task segment data.

[0083] If the continuous criterion data indicates that the data corresponding to the segment can form a continuous succession relationship with the data of the task segment, then the corresponding data of the segment will be output as continuous succession data; if the continuous criterion data indicates that the data corresponding to the segment cannot form a continuous succession relationship with the data of the task segment, then it will not be directly treated as invalid data, but the same task ownership judgment will continue to be performed on the data corresponding to the segment.

[0084] In the same task attribution determination, attribution criterion data is generated based on the corresponding data of segments that cannot form a continuous succession relationship and the task segment data. Specifically, the corresponding data of segments that cannot form a continuous succession relationship is taken as the object to be attributed, and the task segment data is taken as the basis for the task source range. The task source, main chain range, and object attribution are compared between the two to generate attribution criterion data. The attribution criterion data is used to characterize whether the corresponding data of segments that cannot form a continuous succession relationship still belong to the same rescue task main chain defined by the task segment data.

[0085] If the attribution criteria data indicates that the corresponding data in the corresponding segment still belongs to the same rescue mission main chain, then the corresponding data in the corresponding segment will be output as data to be hosted and accepted; if the attribution criteria data indicates that the corresponding data in the corresponding segment cannot be attributed to the same rescue mission main chain, then the corresponding data in the corresponding segment will be output as data to be discharged from the main chain.

[0086] After the continuous acceptance data, pending acceptance data, and main chain outbound data are formed, the corresponding data of the segment are grouped according to the above three data to generate breakpoint acceptance data. Specifically, the segment corresponding data is used as the object to be grouped, the continuous acceptance data is used as the basis for the continuous processing group, the pending acceptance data is used as the basis for the custody processing group, and the main chain outbound data is used as the basis for the exclusion processing group. The segment corresponding data is grouped to generate breakpoint acceptance data. Breakpoint acceptance data includes continuous acceptance data that can continue to be processed along the same task main chain, pending acceptance data that cannot be continuously accepted but still belongs to the same task main chain, and main chain outbound data that does not participate in the subsequent planning of the main chain.

[0087] This solution converts the format of the execution feedback data to generate feedback reception data. This transforms different feedback objects from drones, inspection robots, smoke extraction equipment controllers, and rescue personnel terminals into unified data objects that can be recognized and processed by the same task scheduling service. This avoids inconsistencies in feedback formats from different execution entities, which could lead to subsequent inconsistencies in the same link correspondence. The solution also incorporates the feedback reception data into the task segment data, generating segment correspondence data. This prevents the execution feedback data from directly updating the task status based on arrival order. Instead, it first constrains the task object range defined by the task segment data, ensuring that feedback objects can establish a clear correspondence with established rescue task segments. This prevents feedback objects without a task source or outside the task range from entering the breakpoint relay chain. Finally, the solution performs a continuous relay judgment on the segment correspondence data and task segment data. When the judgment is valid, the segment correspondence data is output as continuous relay data. This ensures that feedback objects that maintain a continuous correspondence with the task segment data are retained as directly relayable objects, guaranteeing that feedback from completed continuous reconnaissance, containment, smoke extraction, or obstacle clearing operations can continue into subsequent planning chains.

[0088] When the continuous assignment judgment fails, the corresponding data of the segment is judged to belong to the same task. If the judgment is correct, data to be hosted is generated. This ensures that objects that cannot be continuously assigned due to intermittent communication but still belong to the same rescue task main chain are retained as objects that can be hosted and processed later. This avoids the direct discarding or incorrect redistribution of objects with communication interruptions. When the same task assignment judgment fails, main chain outgoing data is generated. This excludes the return objects that cannot be included in the current rescue task main chain from the breakpoint assignment chain. This prevents incorrectly assigned objects from affecting the formation of subsequent hosted segment data and back-insertion verification data. Finally, the corresponding data of the segment is grouped according to the continuous assignment data, data to be hosted, and main chain outgoing data to generate breakpoint assignment data. This ensures that continuous assignment objects, data to be hosted, and main chain outgoing objects form a clear technical destination in the same complete data object. This provides a clear basis for subsequent hosting segment generation, back-insertion verification, and the unique formation of planning results.

[0089] The above describes the process of grouping execution feedback data and task segment data to generate breakpoint continuation data. The following describes how to organize the execution feedback data, after limiting the write range of managed segment data, with the managed segment data to generate back-insertion candidate data. Please refer to [link / reference]. Figure 4 , Figure 4 This is a schematic diagram of the process for generating interpolation candidate data provided in an embodiment of this application. Generating interpolation candidate data specifically includes:

[0090] The overall object range of the returned data and managed fragment data corresponding to the executed returned data is compared to generate the returned range data. The returned range data and managed fragment data are matched for the same task to generate the corresponding data for back-insertion.

[0091] The correspondence between the backhaul range data and the corresponding back-insertion data is determined. If the correspondence is found, the backhaul range data is retained, and the retained backhaul range data and the corresponding back-insertion data are merged to generate back-insertion candidate data.

[0092] Among them, the return range data refers to the complete data object generated after comparing the overall object range of the return received data and the managed fragment data. It is used to limit whether the execution return data can enter the interpolable processing range defined by the managed fragment data after being received by the interface.

[0093] Back-insertion corresponding data refers to the complete intermediate data object formed after the back-in range data has already indicated that the back-in received data corresponding to the executed back-in data is within the writable range defined by the managed fragment data, by further matching the back-in range data and the managed fragment data according to the task objects in the same rescue mission main chain.

[0094] The above content will be described in detail below:

[0095] Using task segment data as the source limiting object of the rescue mission main chain, the part of the breakpoint acceptance data that cannot be continuously accepted but still needs to be processed for mission attribution is screened; specifically, the breakpoint acceptance data is used as the object to be screened, and the task segment data is used as the basis for limiting the main chain attribution, and the breakpoint acceptance data and the task segment data are processed to correspond to the whole object; when the object in the breakpoint acceptance data can correspond to the same rescue mission main chain limited by the task segment data, the object is assigned to the manageable processing path and managed fragment data is generated.

[0096] The returned received data and the managed fragment data are compared as a complete object to generate the overall object range data. The overall object range comparison includes: using the manageable task object defined by the managed fragment data as the comparison benchmark, and using the machine-side returned object corresponding to the returned received data as the comparison object, to determine whether the returned received data can enter the task back-insertion range defined by the managed fragment data; the returned objects that can enter the task back-insertion range form the back-access range data.

[0097] The same task matching is performed on the returned range data and the managed fragment data to generate the corresponding data for back-insertion. Specifically, the returned range data is regarded as the returned object that has passed the range limitation, and the managed fragment data is regarded as the main chain task object to be back-inserted. The object matching processing under the same rescue task main chain is performed on the two. If the returned object corresponding to the returned range data can form a consistent task ownership relationship with the same task object in the managed fragment data, the corresponding data for back-insertion is generated.

[0098] A correspondence determination is performed on the return range data and the corresponding insertion data. This determination is based on the return range data and the corresponding insertion data, and is based on the writable range result represented by the return range data and the same task matching result represented by the corresponding insertion data. The determination is successful when the same return range data can find a corresponding same task matching object in the corresponding insertion data, and the corresponding object still belongs to the task main chain limited by the managed fragment data. The determination is unsuccessful when the return range data cannot form the same task matching object in the corresponding insertion data, or the corresponding object is removed from the task main chain limited by the managed fragment data. When the determination is successful, reserved range data is generated. The reserved range data is formed by the return range data after the correspondence determination.

[0099] By separating the return range data and the corresponding insertion data and then determining the correspondence, this implementation method makes the writable range and the same task match into two mutually verified complete data objects, avoiding the direct merging of execution return data into insertion candidate data based on a single path.

[0100] The retained range data and the corresponding back-insertion data are merged to generate back-insertion candidate data. This merging process uses the retained range data as the back-transmission object that has been limited by the write range, and the corresponding back-insertion data as the task belonging object that has been matched by the same task. The two are merged under the same task main chain to generate back-insertion candidate data.

[0101] In one preferred implementation, when an inspection robot in a mountainous highway tunnel group accident rescue completes a local survey in a tunnel communication blind spot and then delays the transmission of execution data, it forms managed segment data based on the breakpoint continuation data and task segment data. Subsequently, the delayed execution data is converted into transmission reception data and compared with the managed segment data to generate transmission range data. Then, the transmission range data and managed segment data are matched for the same task to generate corresponding back-insertion data. When the transmission range data and the corresponding back-insertion data can form a correspondence under the same task main chain, reserved range data is generated, and candidate back-insertion data is generated from the reserved range data and the corresponding back-insertion data.

[0102] This solution generates return range data by comparing the overall object scope of the received return data corresponding to the executed return data with the managed fragment data. This first limits late or recovered executed return objects to the recyclable task scope represented by the managed fragment data, preventing executed return data from directly entering the task chain according to the interface arrival order. This excludes return objects that do not belong to the current managed fragment from the data scope level, ensuring that subsequent back-insertion candidates have clear write boundaries. The solution then performs same-task matching on the return range data and managed fragment data to generate corresponding back-insertion data, further confirming whether the return objects already within the scope are related to the managed fragment. The system ensures that the same rescue mission objects are matched within the same data segment, preventing incorrect writing of return objects that only overlap in scope but have different mission affiliations into the main chain. It performs a correspondence check on the return scope data and the corresponding return data, establishing consistency between the scope limitation result and the mission matching result, excluding return objects that only meet a single condition. If the check is successful, the return scope data is retained, and the retained return scope data is merged with the corresponding return data to generate candidate return data. This ensures that the candidate return data simultaneously possesses both a writeable scope basis and a basis for corresponding to the same mission, providing clear-boundary and clearly affiliated candidate data objects for subsequent return verification.

[0103] The above describes how the execution feedback data, after being written to the managed fragment data within a defined range, is processed together with the managed fragment data to generate back-insertion candidate data. The following describes how to perform a task consistency check between the back-insertion candidate data and the breakpoint inherited data; if they match, the back-insertion candidate data is effectively retained. Specifically, this includes:

[0104] Perform task attribution determination on the candidate data for back-insertion and the managed fragment data; otherwise, exclude the candidate data for back-insertion.

[0105] If yes, the candidate data for back-insertion is retained, the main chain ownership data is generated, and it is determined whether the main chain ownership data can be traced back to the breakpoint data. If yes, the main chain ownership data is retained, and a valid retention result is generated.

[0106] Otherwise, exclude the main chain ownership data.

[0107] Among them, the main chain attribution data refers to the complete data object formed by retaining the back-insertion candidate data that can be classified into the same rescue mission main chain as defined by the managed fragment data after judging the task attribution of the back-insertion candidate data and the managed fragment data.

[0108] The above content will be described in detail below:

[0109] Task attribution criterion data is generated based on managed fragment data. Specifically, managed fragment data is used as the attributable scope object. The managed fragment data is processed by limiting the object scope, organizing the main chain inheritance relationship, and retaining recyclable objects to generate task attribution criterion data. The task attribution criterion data is used to characterize the task main chain scope, attributable object scope, and recyclable inheritance relationship in the managed fragment data that are allowed to receive back-insertion candidate data.

[0110] Task attribution is determined for both the candidate data for back-insertion and the managed fragment data. Specifically, the candidate data for back-insertion is used as the object to be judged, and the task attribution criterion data is used as the basis for judgment. The two are compared accordingly. When the execution feedback object corresponding to the candidate data for back-insertion can enter the same rescue task main chain defined by the task attribution criterion data, and can form a corresponding relationship with the range of recyclable objects in the task attribution criterion data, the task attribution is determined to be valid. When the candidate data for back-insertion cannot enter the same rescue task main chain defined by the task attribution criterion data, or cannot form a corresponding relationship with the range of recyclable objects in the task attribution criterion data, the task attribution is determined to be invalid.

[0111] When the task assignment is established, the candidate data for back-insertion is retained, and the main chain assignment data is generated.

[0112] Based on the breakpoint acceptance data, acceptance backtracking criteria data is generated. Specifically, the breakpoint acceptance data is used as the basis for accepting the preceding task. The breakpoint acceptance data is processed by organizing the acceptance objects, limiting the preceding and following acceptance relationships, and retaining the main chain source to generate acceptance backtracking criteria data. The acceptance backtracking criteria data is used to characterize the acceptance path, the range of traceable objects, and the preceding and following acceptance relationships formed after the task segment data and the execution feedback data are first substantially coupled.

[0113] The process involves determining whether the main chain's attribution data can be traced back to the breakpoint's successor data. Specifically, the main chain's attribution data is used as the object to be traced back, and the successor tracing criteria data is used as the basis for judgment. The main chain's attribution data and the successor tracing criteria data are compared accordingly. If the task object corresponding to the main chain's attribution data exists within the traceable object range defined by the successor tracing criteria data, and the task object corresponding to the main chain's attribution data maintains a sequential relationship with the successor object in the successor tracing criteria data under the same rescue mission main chain, it is determined that the data can be traced back to the breakpoint's successor data. If the task object corresponding to the main chain's attribution data does not exist within the traceable object range defined by the successor tracing criteria data, or does not maintain a sequential relationship with the successor object in the successor tracing criteria data under the same rescue mission main chain, it is determined that the data cannot be traced back to the breakpoint's successor data. If the data can be traced back, the main chain's attribution data is retained, generating a valid retention result.

[0114] The valid retained results are merged based on the breakpoint acceptance data to generate back-insertion verification data. Specifically, the valid retained results are used as the objects to be merged, and the breakpoint acceptance data are used as the acceptance criteria. Object correspondence processing, acceptance order limitation processing, and duplicate object exclusion processing are performed on the valid retained results and the breakpoint acceptance data. When the valid retained result can form a correspondence with the acceptance object in the breakpoint acceptance data under the same rescue mission main chain, the valid retained result is included in the back-insertion verification data. When the valid retained result cannot form a correspondence with the acceptance object in the breakpoint acceptance data under the same rescue mission main chain, the valid retained result is excluded from the back-insertion verification data formation path.

[0115] In one scenario-based embodiment, after a traffic accident occurs in a mountainous highway tunnel complex, the inspection robot has already generated managed fragment data through previous steps before entering the tunnel's communication blind zone. The breakpoint continuation data has recorded the sequential relationship between the inspection robot's task and the task segment data. After communication is restored, the inspection robot returns execution feedback data through the intelligent agent communication interface. The previous feedback processing generates back-insertion candidate data. Based on the managed fragment data, task attribution criterion data is generated, and the back-insertion candidate data is compared with the task attribution criterion data. If the task object corresponding to the back-insertion candidate data belongs to the same rescue task main chain defined by the managed fragment data, main chain attribution data is generated. Subsequently, based on the breakpoint continuation data, continuation backtracking criterion data is generated, and the main chain attribution data is compared with the continuation backtracking criterion data. If the task object corresponding to the main chain attribution data can return to the continuation path defined by the breakpoint continuation data, a valid retention result is generated. Finally, based on the breakpoint continuation data, the valid retention result is subjected to object correspondence, continuation order limitation, and duplicate object exclusion to generate back-insertion verification data.

[0116] This solution performs task attribution determination on back-insertion candidate data and managed fragment data. By first limiting back-insertion candidate data to whether it belongs to the same retrievable task fragment within the communication blind zone using managed fragment data, it can prevent back-insertion candidate data formed by late or recovered backhaul from directly entering the main chain, thereby excluding backhaul objects that have no corresponding relationship with the current rescue task's main chain. If the determination is not valid, back-insertion candidate data is excluded, so that candidate objects without managed attribution are eliminated before entering subsequent backtracking determination, preventing them from continuing to occupy breakpoints and carry out determination paths and causing incorrect attribution. If the determination is valid, back-insertion candidate data is retained and main chain attribution data is generated, transforming candidate objects that have been confirmed through managed fragment attribution into complete data objects with clear main chain attribution, providing stable input for subsequent backtracking determination.

[0117] The process involves determining whether the main chain ownership data can be traced back to the breakpoint acceptance data. Further verification confirms that the main chain ownership data still corresponds to the previous breakpoint acceptance relationship, ensuring that the back-insertion object not only belongs to the managed segment but also returns to the acceptance path formed by the initial coupling of the task segment data and the execution feedback data. If the backtracking is successful, the main chain ownership data is retained, and a valid retention result is generated. This ensures that the final retained back-insertion object simultaneously satisfies both the managed segment ownership and breakpoint acceptance backtracking conditions, guaranteeing that subsequent back-insertion verification data only accepts objects with clear sources and stable ownership. If the backtracking is unsuccessful, the main chain ownership data is excluded to prevent objects with only superficial task ownership but unable to return to the breakpoint acceptance path from entering the subsequent planning result chain, thus avoiding erroneous acceptance objects affecting the unique formation of the final planning result data.

[0118] The above describes the task consistency judgment between back-insertion candidate data and breakpoint continuation data. If they match, the back-insertion candidate data is effectively retained. The following describes the task sequence constraint processing of task segment data and back-insertion verification data under the same task to generate phase-locked data, specifically including:

[0119] The back-insertion verification data and task segment data are mapped to the same task to generate phase correspondence data. The phase correspondence data are then arranged in order according to the sequential relationship in the same task to generate phase sequence data.

[0120] Phase sequence data is mapped to interpolation verification data and task segment data respectively, and the corresponding parts are retained to generate phase constraint data. Then, the phase constraint data is verified for correspondence and continuity based on phase correspondence data and phase sequence data to generate phase lock data.

[0121] Among them, phase correspondence data refers to the survey tasks, blockade tasks, smoke extraction tasks and rescue channel clearing tasks in the task segment data, which have been mapped to the same execution main chain defined by the back-insertion verification data. This makes the generation of subsequent phase sequence data no longer based on the original task category of the task segment data, but on the data objects arranged in order based on the data inheritance relationship that has been completed by back-insertion verification.

[0122] Phase sequence data refers to the complete intermediate data object formed by sequentially arranging the phase correspondence data according to the sequential relationship between the interpolation verification data and the task segment data in the same task, after the phase correspondence data has established a correspondence relationship between the interpolation verification data and the task segment data in the same task.

[0123] Phase-constrained data refers to a set of task objects that have completed the same task in a sequential order and simultaneously satisfy the back-insertion verification relationship and the task source constraint relationship.

[0124] The above content will be described in detail below:

[0125] Same task criterion data is generated based on task segment data and back-insertion verification data. Same task criterion data is generated jointly by task segment data and back-insertion verification data. The generation method is as follows: the task objects defined by the task segment data and the execution main chain objects defined by the back-insertion verification data are processed to correspond as a whole, and objects that can be simultaneously limited by the task source range of the task segment data and the execution main chain range of the back-insertion verification data are formed into same task criterion data.

[0126] The back-insertion verification data and task segment data are processed according to the same task criterion data to generate phase correspondence data. The objects of this correspondence processing are complete back-insertion verification data and complete task segment data. The processing basis is the same task criterion data. The processing actions include: matching the execution main chain objects in the back-insertion verification data that fall within the scope of the same task criterion data with the task source objects in the task segment data that fall within the scope of the same task criterion data; and generating phase correspondence data for the objects that have been matched.

[0127] The succession order criterion data is generated based on phase correspondence data. The source of the succession order criterion data is phase correspondence data, and its generation method is as follows: the succession relationship of objects that are already in the same task scope in the phase correspondence data is processed to determine the succession order of each object in the same rescue task main chain, and the succession order is formed into succession order criterion data. For task objects in continuous time sequence, the succession relationship already formed in the phase correspondence data is used as the processing object, and the possibility of delayed or restored backhaul limited by the previous backhaul verification data is allowed. For task objects in discrete event form, the order of task objects in the phase correspondence data is used as the processing object, so that discrete task objects can enter the succession sequence of the same task.

[0128] The phase-corresponding data are arranged sequentially according to the succession order criterion data to generate phase order data.

[0129] Phase sequence data is mapped to back-insertion verification data and task segment data respectively, and the corresponding parts are retained to generate phase constraint data. Specifically, this includes: using phase sequence data as the object to be constrained, using back-insertion verification data as the execution main chain verification object, and using task segment data as the task source range object, performing a first mapping process between phase sequence data and back-insertion verification data to generate execution main chain corresponding data; simultaneously performing a second mapping process between phase sequence data and task segment data to generate task source corresponding data; then performing a joint retention process on execution main chain corresponding data and task source corresponding data, generating phase constraint data from objects that exist simultaneously in both execution main chain corresponding data and task source corresponding data.

[0130] The phase constraint data is verified based on the phase correspondence data to generate corresponding verification data. The verification object is the phase constraint data, and the judgment basis is the phase correspondence data; when an object in the phase constraint data can be traced back to the same task correspondence defined by the phase correspondence data, corresponding verification data is generated.

[0131] The succession relationship of the corresponding verification data is verified based on the phase sequence data to generate succession verification data. The object of succession relationship verification is the corresponding verification data, and the judgment basis is the phase sequence data. When the objects in the corresponding verification data can maintain the sequential succession relationship defined by the phase sequence data, succession verification data is generated.

[0132] Phase-locked data is generated based on phase constraint data, corresponding verification data, and succession verification data. The generation condition for phase-locked data is that objects in the phase constraint data can be confirmed through corresponding verification data to still belong to the same task correspondence defined by the phase correspondence data, and can be confirmed through succession verification data to still maintain the sequential succession relationship defined by the phase order data. When both of the above conditions are met, phase-locked data is generated.

[0133] This solution generates phase-correspondence data by mapping back-insertion verification data to task segment data within the same task. This establishes a shared task affiliation between the execution main chain that has completed back-insertion verification and the task objects defined by the task segment data, preventing back-inserted execution objects from being mistakenly merged into other reconnaissance, containment, smoke extraction, or rescue channel clearing tasks. The phase-correspondence data is then arranged sequentially according to the order of execution within the same task, generating phase sequence data. This ensures a clear order of execution for different stages within the same task, preventing smoke extraction, containment, and clearing tasks from entering subsequent planning in parallel queues and causing uncertainty in the execution order. Sequence data is mapped to both the back-injection verification data and the task segment data, retaining the corresponding parts simultaneously to generate phase constraint data. This ensures that subsequent task objects participating in the planning are simultaneously constrained by both the back-injection verification results and the task segment source, excluding incorrectly assigned objects or objects without a source that exist only in a single set of data. Based on the phase correspondence data and phase sequence data, the phase constraint data is verified for correspondence and succession relationships, generating phase-locked data. This ensures that task objects in the phase constraint data maintain both the same task correspondence and the sequential succession relationship within the same task, thereby providing a phase constraint basis for subsequent planning result data that is consistent in source, clear in sequence, and closed in succession.

[0134] The above describes the task sequence constraint processing of task segment data and back-injection verification data under the same task to generate phase-locked data. The following describes the task consistency verification of phase-locked data, back-injection verification data, and task segment data to generate planning result data, specifically including:

[0135] Perform step-by-step backtracking and consistency verification on phase-locked data, interpolation verification data, breakpoint continuation data, and task segment data to generate closed-loop basic data;

[0136] The closed-loop basic data, phase-locked data, back-intrusion verification data, and breakpoint continuation data are constrained and converged under the same task to generate planning result data.

[0137] Closed-loop basic data refers to data objects that are formed by step-by-step backtracking and consistency verification, using phase-locked data, back-interpolation verification data, breakpoint continuation data, and task segment data as input objects, and are used to limit the range of planning result data generation.

[0138] The above content will be described in detail below:

[0139] The phase-locked data and the back-insertion verification data are compared to determine their corresponding tasks, generating phase backtracking data. Specifically, phase task-corresponding data is generated based on the phase-locked data, and back-insertion task-corresponding data is generated based on the back-insertion verification data. Then, the phase task-corresponding data and the back-insertion task-corresponding data are compared for the same task object. The phase task-corresponding data represents the set of task objects in the phase-locked data that have completed phase constraints, and the back-insertion task-corresponding data represents the set of task objects in the back-insertion verification data that have completed back-insertion verification. The phase task-corresponding data and the back-insertion task-corresponding data are compared item by item. When a task object in the phase task-corresponding data can be found in the back-insertion task-corresponding data, and both correspond to the same execution main chain, phase backtracking data is generated.

[0140] In the process of generating the aforementioned phase backtracking data, the following comparison formula can be used to explain the task correspondence between the phase-locked data and the back-interpolation verification data:

[0141] ;

[0142] In the formula, Indicates the first The phase backtracking value of each task object is valid. This indicates the first data in the phase task corresponding to the data. The task-corresponding value for each task object. This indicates the first data in the data corresponding to the back-insertion task. The task-corresponding value for each task object. This indicates the first data in the phase task corresponding to the data. The execution main chain value corresponding to each task object This indicates the first data in the data corresponding to the back-insertion task. The execution main chain value corresponding to each task object Indicates and, It means "or".

[0143] like This indicates that the task object corresponding to the phase-locked data can return to the same execution main chain defined by the back-insertion verification data, and phase backtracking data is generated based on this task object. This indicates that the task object cannot form a phase backtracking relationship.

[0144] The phase backtracking data and breakpoint continuation data are used to verify the continuation relationship and generate continuation backtracking data. Specifically, the phase backtracking data is used as the object to be verified, and the breakpoint continuation data is used as the basis for the preceding continuation path. First, phase continuation object data is generated from the phase backtracking data, and breakpoint continuation object data is generated from the breakpoint continuation data. Then, the task affiliation and continuation relationship of the phase continuation object data and the breakpoint continuation object data are compared. When the task object in the phase continuation object data belongs to the same task defined by the breakpoint continuation object data, and the task object maintains a continuation relationship with the continuation object in the breakpoint continuation object data, continuation backtracking data is generated.

[0145] In a preferred embodiment, the conditions for receiving backtracking data can be explained by the following formula:

[0146] ;

[0147] In the formula, Indicates the first The value of the backtracking of the task object is established. This indicates the first phase in the data of the receiving object. The task attribution value of each task object. This indicates the breakpoint in the data of the receiving object. The task attribution value of each task object. This indicates the first phase in the data of the receiving object. The preceding successor position value corresponding to each task object. This indicates the breakpoint in the data of the receiving object. The successor position value corresponding to each task object.

[0148] The source relationship between the backtracking data and the task segment data is verified to generate source closure data. Specifically, the backtracking data is used as the source verification object, and the task segment data is used as the basis for the task source range. First, the backtracking data is used to generate the source object data, and the task segment data is used to generate the segment source object data. Then, the task range and corresponding task are compared between the backtracking data and the segment source object data. When the task object corresponding to the backtracking data belongs to the task range defined by the segment source object data and can be traced back to the corresponding task in the task segment data, source closure data is generated.

[0149] The source closed data, phase backtracking data, and receiving backtracking data are mapped to the same task relationship to generate closed basic data. Specifically, source task relationship data is first generated from the source closed data, phase task relationship data is generated from the phase backtracking data, and receiving task relationship data is generated from the receiving backtracking data. Then, the source task relationship data, phase task relationship data, and receiving task relationship data are compared for the same task object. When the same task object exists in the source task relationship data, phase task relationship data, and receiving task relationship data at the same time, and the task object satisfies the source closed relationship, phase backtracking relationship, and receiving backtracking relationship, closed basic data is generated.

[0150] Output limitation data is generated by limiting the phase-locked data based on closed-loop basic data. Specifically, phase output object data is first generated from the phase-locked data, and closed task range data is generated from the closed-loop basic data. Then, the task range is compared between the phase output object data and the closed task range data. When the task object corresponding to the phase output object data belongs to the task range limited by the closed task range data, the output limitation data is generated.

[0151] The output constraint data and the back-insertion verification data are checked together to generate output verification data. Specifically, output constraint task data is first generated from the output constraint data, and back-insertion verification task data is generated from the back-insertion verification data. Then, the output constraint task data and the back-insertion verification task data are compared together. When the task object corresponding to the output constraint task data belongs to the same task defined by the back-insertion verification task data, the output verification data is generated.

[0152] The output verification data and breakpoint acceptance data are subjected to backtracking verification to generate output backtracking data. Specifically, output verification acceptance data is first generated from the output verification data, and breakpoint output acceptance data is generated from the breakpoint acceptance data. Then, the output verification acceptance data and breakpoint output acceptance data are compared with the same task and the previous and subsequent acceptance relationships. When the task object corresponding to the output verification acceptance data belongs to the same task defined by the breakpoint output acceptance data, and maintains the previous and subsequent acceptance relationship with the breakpoint acceptance data, output backtracking data is generated.

[0153] The output constraint data, output verification data, and output backtracking data are compared for consistency under the same task to generate planning result data. Specifically, output constraint relationship data is first generated from the output constraint data, output verification relationship data is generated from the output verification data, and output backtracking relationship data is generated from the output backtracking data. Then, the output constraint relationship data, output verification relationship data, and output backtracking relationship data are compared for consistency under the same task object. When the same task object exists in the output constraint relationship data, output verification relationship data, and output backtracking relationship data simultaneously, and the task object simultaneously satisfies closure constraint, back-interpolation verification, and breakpoint continuation backtracking, planning result data is generated.

[0154] As a scenario-based example, in an accident tunnel within a mountainous highway tunnel complex, an inspection robot, after completing the clearance and confirmation of the rescue passage within the tunnel, experienced a delayed data transmission due to a communication interruption. Simultaneously, the smoke extraction equipment controller transmitted the smoke extraction task execution status, and the rescue personnel's terminal transmitted the sealing execution result. The preceding steps had already generated task segment data based on the rescue task data, and breakpoint continuation data was generated based on the task segment data and the execution feedback data. Back-insertion verification data was also generated based on the back-insertion verification process. Upon receiving the phase-locked data, it first confirmed whether the rescue passage clearance task in the phase-locked data existed within the same execution main chain defined by the back-insertion verification data. If it did, phase backtracking data was generated. Then, it was confirmed whether the task could return to the breakpoint. The system first establishes a sequential relationship based on the inherited data. If a return to the previous data is possible, it generates backtracking data. Then, it confirms whether the task falls within the task range defined by the task segment data. If so, it generates source closure data. Finally, it performs a consistency comparison of the source closure data, phase backtracking data, and inherited backtracking data for the same task object, generating closure base data. The closure base data is then used to define the phase lock data, generating output constraint data. The output constraint data is then verified using back-interpolation verification data, generating output verification data. The output verification data is then backtracked using breakpoint inherited data, generating output backtracking data. Finally, the output constraint data, output verification data, and output backtracking data are compared for consistency with the same task, generating unique planning result data.

[0155] This solution generates closed-loop basic data by performing step-by-step backtracking and consistency verification on phase-locked data, back-insertion verification data, breakpoint continuation data, and task segment data. This allows the task objects defined by the phase-locked data to sequentially return to the back-insertion verification relationship defined by the back-insertion verification data, the task continuation relationship defined by the breakpoint continuation data, and the task source range defined by the task segment data. This eliminates task objects that only meet the phase sequence but cannot return to the back-insertion verification relationship, task objects that have completed the back-insertion verification but cannot return to the breakpoint continuation relationship, and task objects that have a continuation relationship but cannot return to the source range of the task segment data. As a result, the data objects entering the subsequent output formation process simultaneously possess task source consistency, continuation path consistency, and back-insertion verification consistency.

[0156] The closed-loop basic data, phase-locked data, back-intrusion verification data, and breakpoint continuation data are constrained and converged under the same task to generate planning result data. This ensures that the planning result data is not directly formed from any single data among the phase-locked data, back-intrusion verification data, or breakpoint continuation data. Instead, it is simultaneously constrained by the execution order of the phase-locked data, the back-transmission attribution of the back-intrusion verification data, and the sequential continuation of the breakpoint continuation data within the same task scope confirmed by the closed-loop basic data. This avoids the separate parallel output paths of the reconnaissance, containment, smoke extraction, and rescue channel clearing tasks, ensuring that the final planning result data remains consistent with the main chain of the same rescue task.

[0157] The above describes the task consistency verification of phase-locked data, back-interpolation verification data, and task segment data to generate planning result data. The following describes the step-by-step backtracking and consistency verification of phase-locked data, back-interpolation verification data, breakpoint continuation data, and task segment data to generate closed-loop basic data, specifically including:

[0158] The phase-locked data and the back-interpolation verification data are compared according to the task to generate phase backtracking data;

[0159] When the phase backtracking data belongs to the same task as the breakpoint continuation data and maintains a sequential relationship with the breakpoint continuation data, continuation backtracking data is generated.

[0160] When the backtracking data belongs to the task range of the task segment data and can be traced back to the corresponding task of the task segment data, source closed data is generated;

[0161] The source closed data, phase backtracking data, and receiving backtracking data are mapped to the same task to generate closed basic data.

[0162] Among them, phase backtracking data refers to the data object generated by comparing phase locking data and back-insertion verification data according to the correspondence of task objects, which is used to characterize that the task object in the phase locking data can return to the verified execution main chain defined by the back-insertion verification data.

[0163] The backtracking data refers to the complete data object jointly generated by the phase backtracking data and the breakpoint acceptance data after comparing the phase backtracking data and the breakpoint acceptance data. This ensures that the task object corresponding to the phase backtracking data still belongs to the same rescue task main chain defined by the breakpoint acceptance data, and that the task object maintains a sequential relationship with the preceding and subsequent task objects in the breakpoint acceptance data.

[0164] Source closure data refers to data objects formed by comparing the task range and task source of the backtracking data and the task segment data. It is used to represent task objects that have completed phase backtracking and backtracking, still belong to the task range defined by the task segment data, and can be mapped back to the same task object in the task segment data.

[0165] This part has already been described in detail above, so I will not repeat it here.

[0166] This solution generates phase backtracking data by comparing the phase-locked data and backtracking verification data according to the task correspondence. This allows phase-constrained tasks such as exploration, containment, smoke extraction, and rescue channel clearing to return to the scope of the task objects that have completed backtracking verification, thus preventing phase-locked data from becoming isolated execution sequences due to delayed verification results. When the phase backtracking data belongs to the same task as the breakpoint acceptance data and maintains a sequential relationship with the breakpoint acceptance data, acceptance backtracking data is generated. This allows the task objects that have already passed phase backtracking to return to the original task acceptance path formed under intermittent communication conditions, thereby eliminating acceptance path errors caused by misconnection, reconnection, or cross-task access of task objects after communication is restored.

[0167] When the received backtracking data falls within the task scope of the task segment data and can be traced back to the corresponding task in the task segment data, source closure data is generated. This ensures that the task objects that have completed the backtracking are still limited by the source scope of the rescue task, thereby preventing task objects without a source or those that have deviated from the task segment data from entering the final planning process. The source closure data, phase backtracking data, and received backtracking data are mapped to the task relationship to generate closed basic data. This ensures that the same task object simultaneously satisfies the three types of relationships: source limitation, phase constraint, and received backtracking. This provides a unified, traceable, and non-parallel closed basis for subsequent planning result data.

[0168] The above describes the step-by-step backtracking and consistency verification of phase-locked data, back-intrusion verification data, breakpoint continuation data, and task segment data to generate closed-loop basic data. The following describes the constraint convergence of the closed-loop basic data, phase-locked data, back-intrusion verification data, and breakpoint continuation data under the same task to generate planning result data, specifically including:

[0169] When the phase-locked data falls within the task scope of closed basic data, output constraint data is generated, and when the output constraint data corresponds to the same task of back-insertion verification data, output verification data is generated.

[0170] When the output verification data belongs to the same task as the breakpoint acceptance data and maintains a sequential relationship with the breakpoint acceptance data, output backtracking data is generated. The output limit data, output verification data, and output backtracking data are then compared for consistency under the same task to generate planning result data.

[0171] Among them, output constraint data refers to the data objects used to characterize phase constraint objects in phase-locked data that can be confirmed by closed basic data to belong to the same task scope;

[0172] Output verification data refers to the verification data formed by comparing the output limiting data and the back-insertion verification data under the same task. It is used to characterize whether the output objects that have entered the scope of the closed basic data task belong to the same task main chain confirmed by the back-insertion verification data.

[0173] Output backtracking data refers to the complete data object generated after comparing the same task affiliation and the preceding and following succession relationships based on the output verification data and breakpoint succession data. It is used to represent that the data object, which has passed the closed range limitation and back-insertion verification, can still return to the original task succession path defined by the breakpoint succession data.

[0174] This part has already been described in detail above, so I will not repeat it here.

[0175] This scheme generates output constraint data when phase-locked data falls within the task scope of closed-loop basic data. This ensures that phase-locked data entering the planning result data formation path is first constrained by the scope of closed-loop basic data, thereby excluding phase objects that have not completed closure verification or do not belong to the current task scope, and preventing phase constraint objects from detaching from the closed main chain. When the output constraint data corresponds to the same task as the back-insertion verification data, output verification data is generated. This further subjectes output objects that have passed the scope constraint to task attribution verification by the back-insertion verification data, thereby eliminating incorrectly assigned objects caused by late or resuming back-insertion, ensuring that the output object still corresponds to the same verified task. The process involves generating output backtracking data when the output verification data belongs to the same task as the breakpoint acceptance data and maintains a sequential relationship with it. This allows output objects that have completed backtracking verification to continue backtracking to the sequential chain defined by the breakpoint acceptance data, preventing output objects from deviating from the task sequence formed before and after the interrupted communication. Finally, the output limit data, output verification data, and output backtracking data are compared for consistency under the same task to generate planning result data. This ensures that the planning result data simultaneously meets the three data conditions of closed-range limitation, backtracking attribution verification, and breakpoint acceptance backtracking, thus forming unique planning result data with consistent source, consistent attribution, and consistent sequence relationship.

[0176] Example 2:

[0177] Please see Figure 5 A multi-agent collaborative task decomposition and execution planning system, including:

[0178] The data acquisition module is used to acquire task segment data and execution feedback data of the target object;

[0179] The blind zone hosting screening module is used to group the execution feedback data and task segment data accordingly, generate breakpoint acceptance data, and screen the part of the breakpoint acceptance data that cannot be continuously accepted but still needs to be assigned a task based on the task segment data, and generate hosting segment data.

[0180] The consistency judgment module is used to organize the execution feedback data and the managed fragment data after limiting the write range, generate back-insertion candidate data, and perform task consistency judgment between the back-insertion candidate data and the breakpoint acceptance data. If they are consistent, the back-insertion candidate data is effectively retained.

[0181] The task sequence constraint module is used to merge the valid retained results based on the breakpoint continuation data, generate back-insertion verification data, and perform task sequence constraint processing on the task segment data and back-insertion verification data under the same task to generate phase-locked data.

[0182] The planning result generation module is used to perform task consistency verification on phase-locked data, back-interpolation verification data, and task segment data, and generate planning result data.

[0183] This embodiment has the same technical effects as Embodiment 1.

[0184] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. The data mentioned in this application have undergone normalization and other preprocessing to unify dimensions during formula calculations.

[0185] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A multi-agent cooperative task decomposition and execution planning method, applied to multi-agent task planning under intermittent communication conditions during tunnel group accident rescue, characterized in that... Includes the following steps: Obtain the task segment data and execution feedback data of the target object; The execution feedback data and the task segment data are grouped accordingly to generate breakpoint acceptance data. Based on the task segment data, the parts of the breakpoint acceptance data that cannot be continuously accepted but still need to be assigned to tasks are screened to generate managed segment data. The execution feedback data is processed by the managed fragment data after the write range is limited, and then the managed fragment data is organized to generate back-insertion candidate data. The back-insertion candidate data and the breakpoint acceptance data are judged to be consistent in terms of task. If they are consistent, the back-insertion candidate data is effectively retained. The valid retention results are merged based on the breakpoint data to generate back-insertion verification data. The task segment data and the back-insertion verification data are then subjected to task order constraint processing under the same task to generate phase-locked data. The phase-locked data, the back-insertion verification data, and the task segment data are subjected to task consistency verification to generate planning result data.

2. The multi-agent cooperative task decomposition and execution planning method according to claim 1, characterized in that: Obtaining the task segment data and execution feedback data of the target object specifically includes: Acquire rescue mission data and execution feedback data of the target object; The rescue mission data is format-normalized to generate mission format data, and the mission format data is then coded and segmented to generate segment candidate data. The correspondence between the candidate segment data and the task format data is determined. If the correspondence is found, the candidate segment data is retained, and task segment data is generated. Otherwise, the corresponding candidate data for that segment is excluded.

3. The multi-agent cooperative task decomposition and execution planning method according to claim 1, characterized in that: The specific steps for generating breakpoint continuation data include grouping the execution feedback data and the task segment data into corresponding groups: The execution return data is converted to a new format to generate return data, and the return data is incorporated into the task segment data to generate segment-corresponding data. A continuous connection determination is made between the data corresponding to the segment and the data of the task segment. If the connection is found, the data corresponding to the segment is output as continuous connection data. Otherwise, the data corresponding to the segment is judged to belong to the same task. If the task is found, the corresponding data of the segment is output as data to be managed and accepted. Otherwise, the corresponding data of the aforementioned segment will be output as the main chain outgoing data, and the corresponding data of the segment will be grouped according to the continuous acceptance data, the pending hosting acceptance data and the main chain outgoing data to generate breakpoint acceptance data.

4. The multi-agent cooperative task decomposition and execution planning method according to claim 3, characterized in that: The execution return data, after being written within the scope of the managed fragment data, is then processed together with the managed fragment data to generate back-insertion candidate data, specifically including: The overall object range of the returned data and the managed fragment data corresponding to the executed returned data is compared to generate returned range data. Then, the returned range data and the managed fragment data are matched for the same task to generate corresponding back-insertion data. The correspondence between the backhaul range data and the corresponding back-insertion data is determined. If the correspondence is found, the backhaul range data is retained, and the retained backhaul range data and the corresponding back-insertion data are merged to generate back-insertion candidate data.

5. The multi-agent cooperative task decomposition and execution planning method according to claim 1, characterized in that: The process of determining the consistency between the candidate back-insertion data and the breakpoint continuation data, and then effectively retaining the candidate back-insertion data if the consistency is found, specifically includes: The task affiliation of the candidate data for back-insertion and the managed fragment data is determined; otherwise, the candidate data for back-insertion is excluded. If so, the candidate data for back-insertion is retained, main chain ownership data is generated, and it is determined whether the main chain ownership data can be traced back to the breakpoint receiving data. If so, the main chain ownership data is retained, and a valid retention result is generated. Otherwise, exclude the main chain attribution data.

6. The multi-agent cooperative task decomposition and execution planning method according to claim 1, characterized in that: The process of performing task sequence constraint processing on the task segment data and the back-insertion verification data under the same task to generate phase-locked data specifically includes: The back-insertion verification data and the task segment data are mapped to the same task to generate phase correspondence data, and the phase correspondence data are arranged in order according to the sequential relationship in the same task to generate phase sequence data. The phase sequence data is mapped to the interpolation verification data and the task segment data respectively, and the corresponding parts are retained to generate phase constraint data. The phase constraint data is then verified for correspondence and continuity based on the phase correspondence data and the phase sequence data to generate phase lock data.

7. The multi-agent cooperative task decomposition and execution planning method according to claim 6, characterized in that: The process of performing task consistency verification on the phase-locked data, the interpolation verification data, and the task segment data to generate planning result data specifically includes: The phase-locked data, the back-insertion verification data, the breakpoint continuation data, and the task segment data are backtracked and consistency verified step by step to generate closed basic data. The closed basic data, the phase-locked data, the back-insertion verification data, and the breakpoint continuation data are constrained and converged under the same task to generate planning result data.

8. The multi-agent cooperative task decomposition and execution planning method according to claim 7, characterized in that: The process of performing step-by-step backtracking and consistency verification on the phase-locked data, the interpolation verification data, the breakpoint continuation data, and the task segment data to generate closed-loop basic data specifically includes: The phase-locked data and the back-insertion verification data are compared according to the task to generate phase backtracking data; When the phase backtracking data belongs to the same task as the breakpoint continuation data and maintains a sequential relationship with the breakpoint continuation data, continuation backtracking data is generated. When the received backtracking data belongs to the task range of the task segment data and can be traced back to the corresponding task of the task segment data, source closed data is generated; The source closed data, the phase backtracking data, and the receiving backtracking data are mapped to the same task to generate closed basic data.

9. The multi-agent cooperative task decomposition and execution planning method according to claim 7, characterized in that: The closed-loop basic data, the phase-locked data, the back-interpolation verification data, and the breakpoint continuation data are constrained and converged under the same task to generate planning result data, specifically including: When the phase-locked data belongs to the task range of the closed basic data, output limiting data is generated, and when the output limiting data corresponds to the same task as the interpolation verification data, output verification data is generated. When the output verification data belongs to the same task as the breakpoint acceptance data and maintains a sequential relationship with the breakpoint acceptance data, output backtracking data is generated. The output limiting data, the output verification data, and the output backtracking data are then compared for consistency under the same task to generate planning result data.

10. A multi-agent collaborative task decomposition and execution planning system, characterized in that, include: The data acquisition module is used to acquire task segment data and execution feedback data of the target object; The blind zone hosting screening module is used to group the execution feedback data and the task segment data accordingly, generate breakpoint acceptance data, and screen the part of the breakpoint acceptance data that cannot be continuously accepted but still needs to be assigned a task according to the task segment data, and generate hosting segment data. The consistency judgment module is used to organize the execution return data with the managed fragment data after limiting the writing range of the managed fragment data, generate back-insertion candidate data, and perform task consistency judgment between the back-insertion candidate data and the breakpoint acceptance data. If they are consistent, the back-insertion candidate data is effectively retained. The task sequence constraint module is used to merge the valid retention results based on the breakpoint continuation data, generate back-insertion verification data, and perform task sequence constraint processing on the task segment data and the back-insertion verification data under the same task to generate phase-locked data. The planning result generation module is used to perform task consistency verification on the phase-locked data, the back-insertion verification data, and the task segment data, and generate planning result data.

Images