Training plan generation method, apparatus, and electronic device

Abstract

The application provides a training plan generation method and device and electronic equipment, wherein the method comprises the following steps: performing semantic analysis on a training outline to obtain a candidate plan, performing constraint solving on the candidate plan based on resource constraints to obtain a benchmark plan; performing executability verification on the benchmark plan in a simulation sandbox; after the verification is passed, performing digital signature on abstract information of the benchmark plan to obtain an executable credential; and generating a target plan for execution control based on the benchmark plan and the executable credential. Through automatic semantic analysis on the training outline and solving based on resource constraints, the scientificity of plan making and resource utilization efficiency are improved; through executability verification on the benchmark plan in the simulation sandbox, potential conflicts can be avoided before plan execution; and based on the benchmark plan and the executable credential, the target plan is generated, the integrity and authority of the plan in the whole process of making and execution are guaranteed, and the traceability of the plan making basis is ensured.

Description

Technical Field

[0001] This invention relates to the field of training plan generation technology, and in particular to a training plan generation method, apparatus, and electronic device. Background Technology

[0002] In many complex business systems, it is often necessary to develop detailed and actionable execution plans based on guiding documents. The purpose of developing such plans is to ensure that all tasks strictly adhere to the outline requirements and make efficient and rational use of limited resources such as personnel, equipment, and facilities.

[0003] To meet the above needs, existing methods for generating training plans typically rely on manual development by instructors or managers with relevant domain knowledge. This process generally includes: manually interpreting the requirements of each clause in the training syllabus, and then manually arranging tasks and allocating resources based on available personnel qualifications, equipment status, and venue availability, ultimately forming an executable plan.

[0004] However, the aforementioned existing technical solutions have shortcomings. First, the efficiency and quality of the manual arrangement method are highly dependent on the personal experience of the planners. When faced with large-scale and complex planning tasks, the manual approach is not only time-consuming and labor-intensive, but also struggles to comprehensively coordinate numerous constraints, often leading to uneven resource allocation or potential conflicts, affecting the overall feasibility and resource utilization efficiency of the plan. Second, existing technologies lack objective and standardized verification methods for the feasibility and compliance of the generated plans. Whether a plan is truly feasible and meets all safety and regulatory requirements often relies solely on manual review and experience-based judgment, which carries the risk of oversight. Furthermore, due to the lack of systematic documentation of the planning process, it is difficult to effectively trace the basis for the plan's formulation when problems arise during execution or review. Summary of the Invention

[0005] This invention provides a training plan generation method, apparatus, and electronic device to address the shortcomings of existing technologies that rely on manual arrangement, resulting in efficiency and plan quality being limited by personal experience, difficulty in coordinating complex constraints, lack of objective verification methods and systematic recording, and deficiencies in executability review and accountability.

[0006] This invention provides a training plan generation method, comprising the following steps: Obtain the training outline, perform semantic parsing on the training outline to obtain candidate plans, and solve the constraints of the candidate plans based on resource constraints to obtain the baseline plan; The feasibility of the baseline plan was verified in a simulation sandbox, and the verification results were obtained. In response to the verification result being successful, the summary information of the benchmark plan is digitally signed to obtain an executable credential; Based on the baseline plan and the executable credentials, a target plan for execution control is generated.

[0007] According to a training plan generation method provided by the present invention, the resource constraints include a resource constraint knowledge graph; The step of solving the candidate plan based on resource constraints to obtain the baseline plan includes: Acquire multi-source heterogeneous resource data, and construct the resource constraint knowledge graph based on the multi-source heterogeneous resource data; the nodes in the resource constraint knowledge graph are resource entities, and the edges in the resource constraint graph are the constraint relationships between the resource entities, including mutual exclusion relationships, temporal relationships, and dependency relationships; The candidate plan and the resource constraint knowledge graph are input into the constraint solver to obtain the baseline plan output by the constraint solver.

[0008] According to a training plan generation method provided by the present invention, the step of inputting the candidate plan and the resource constraint knowledge graph into a constraint solver to obtain the baseline plan output by the constraint solver includes: The constraint solver transforms the planning events in the candidate plans and the edges in the resource constraint knowledge graph into a formal constraint satisfaction problem. The constraint solver optimizes the soft objective while satisfying the formal constraint satisfaction problem, thus obtaining the baseline plan; The soft targets include the achievement rate, task balance, and resource utilization rate.

[0009] According to a training plan generation method provided by the present invention, the step of semantically parsing the training outline to obtain candidate plans includes: Extract key elements from the training outline; these key elements include skill objectives, subject units, prerequisite relationships, achievement standards, and safety constraints. Encode the key elements into an intermediate representation; The candidate plan is generated based on the intermediate representation.

[0010] According to a training plan generation method provided by the present invention, the step of performing executability verification on the benchmark plan in a simulation sandbox to obtain verification results includes: The baseline plan is mapped to a simulation model consisting of discrete events, and the personnel, equipment, and site resources in the baseline plan are modeled as resource pools of the simulation model. In the simulation model, based on each preset time granularity, the discrete events are simulated to occupy and release the resource pool, and the conflict results are determined based on the occupation and release results. If the conflict result indicates that no resource conflict or security boundary violation has occurred, the verification result is determined to be verification passed. In response to the conflict result being a resource conflict or security boundary violation, the verification result is determined to be a verification failure, and a minimal set of causes is generated to guide the modification of the baseline plan.

[0011] According to a training plan generation method provided by the present invention, the executable credential includes a plan fingerprint, and a timestamp and validity period associated with the plan fingerprint; The plan fingerprint is obtained by digitally signing the summary information of the baseline plan; The process of generating a target plan for execution control based on the baseline plan and the executable credentials further includes: The plan fingerprint in the executable credential is verified using a preset public key to obtain the integrity verification result of the baseline plan; Verify the compliance of the timestamp and validity period in the executable credential to obtain the compliance verification result; If both the integrity verification result and the compliance verification result are successful, the target plan is executed.

[0012] According to a training plan generation method provided by the present invention, executing the target plan includes: During the execution of the target plan, monitor disturbance events triggered by changes in the personnel, equipment, or site resources; In response to the disturbance event reaching a preset threshold, incremental solution is performed within the time window and resource subgraph affected by the disturbance event to obtain a differential packet including the minimum disturbance set; Incremental simulation verification is performed on the differential package, and a new executable credential is generated after the verification is passed. The executable credential is then updated based on the new executable credential.

[0013] According to a training plan generation method provided by the present invention, the step of performing executability verification on the benchmark plan in a simulation sandbox to obtain verification results further includes: Each event of the baseline plan is subjected to compliance verification against the hard constraint rules in the versioned rule base to obtain the compliance verification result; If the compliance verification result indicates that any of the planned events violates the hard constraint rules, the baseline plan is prevented from entering the simulation sandbox for executability verification.

[0014] The present invention also provides a training plan generation apparatus, comprising the following units: The acquisition unit is used to acquire the training outline, perform semantic parsing on the training outline to obtain candidate plans, and perform constraint solving on the candidate plans based on resource constraints to obtain the baseline plan; The verification unit is used to perform executability verification on the benchmark plan in a simulation sandbox and obtain the verification results. A digital signature unit is used to digitally sign the summary information of the baseline plan in response to the verification result being successful, thereby obtaining an executable credential; The generation unit is used to generate a target plan for execution control based on the baseline plan and the executable credentials.

[0015] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the training plan generation method as described above.

[0016] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the training plan generation method as described above.

[0017] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the training plan generation method as described above.

[0018] The training plan generation method, apparatus, and electronic device provided by this invention perform semantic parsing on the training outline to obtain candidate plans, and then solve for the constraints of the candidate plans based on resource constraints to obtain a baseline plan. Next, the executability of the baseline plan is verified in a simulation sandbox. After successful verification, the summary information of the baseline plan is digitally signed to obtain an executable credential. Finally, a target plan for execution control is generated based on the baseline plan and the executable credential. On the one hand, by automating the semantic parsing of the training outline and solving for resource constraints, a baseline plan conforming to real-world conditions can be automatically generated from the training outline. This solves the problems of low efficiency and high dependence on personal experience in manual plan arrangement in existing technologies, improving the scientific nature of plan formulation and resource utilization efficiency. On the other hand, by verifying the executability of the baseline plan in a simulation sandbox, potential conflicts can be avoided before plan execution, preventing insufficient plan feasibility due to a lack of effective verification methods. Furthermore, generating a target plan based on the baseline plan and the executable credential ensures the integrity and authority of the plan throughout the entire process of formulation and execution, ensuring the traceability of the basis for plan formulation. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating the training plan generation method provided by the present invention.

[0021] Figure 2 This is a schematic diagram of the training plan generation device provided by the present invention.

[0022] Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0024] This invention provides a training plan generation method. Figure 1 This is a flowchart illustrating the training plan generation method provided by the present invention, as shown below. Figure 1 As shown, the method includes steps 110, 120, 130 and 140.

[0025] Step 110: Obtain the training outline, perform semantic parsing on the training outline to obtain candidate plans, and perform constraint solving on the candidate plans based on resource constraints to obtain the baseline plan.

[0026] Specifically, firstly, a training syllabus can be obtained. A training syllabus is a collection of information that reflects the top-level design and overall requirements of a training task. Training syllabi are usually formulated by higher-level command organizations or training management departments, specifying the training objectives to be achieved, the included subjects, the standards to be followed, and the safety regulations that must be observed, etc. This embodiment of the invention does not specifically limit this.

[0027] Here, the training outline can be one or more outline text documents, such as Word or PDF files, or structured table files, such as Excel spreadsheets, or a complex collection of documents containing text and attachments. This embodiment of the invention does not specifically limit this.

[0028] The training syllabus can be obtained in various ways. For example, it can be automatically received from an external command information system or archive system through a system interface; it can also be manually imported or entered into the system by operators, such as instructors or managers; or it can be read from a preset database or file server. This embodiment of the invention does not specifically limit the methods used.

[0029] After obtaining the training outline, semantic parsing can be performed on the training outline to obtain candidate plans, and constraint solving can be performed on the candidate plans based on resource constraints to obtain the baseline plan.

[0030] Semantic parsing refers to the process of using natural language processing techniques, especially the semantic understanding and reasoning capabilities of Large Language Models (LLMs), to perform in-depth analysis of the text or table content of the training syllabus. This transforms the outline-style expression, which is easily understood by humans but difficult for machines to directly execute, into structured information that can be computed and arranged by machines. The semantic parsing process aims to understand the inherent logic, sequence, and relationships among the requirements in the training syllabus.

[0031] In this context, candidate plans reflect a logically coherent sequence of training activities derived from the training syllabus, but without considering the feasibility of real-world resources. A candidate plan can be understood as a preliminary, conceptual draft. For example, a candidate plan might be a list of training events to be scheduled, each specifying the required tasks, the types of roles involved, and logical prerequisites, but without assigning specific personnel, equipment, or precise times. While a candidate plan provides a basis for subsequent constraint solving, it does not guarantee feasibility.

[0032] Resource constraints, in this context, reflect the limitations of all available resources in terms of quantity, time, status, and rules when executing a training plan. Resource constraints are the realistic boundary conditions that ensure the successful implementation of subsequent candidate plans. Here, resource constraints may include constraints on personnel resources, equipment resources, site and airspace resources, and other constraints such as weather conditions, safety regulations, and confidentiality provisions. This embodiment of the invention does not specifically limit these constraints.

[0033] Here, constraints on personnel resources may include personnel qualifications, levels, number, shift schedules, fatigue limits, etc.; constraints on equipment resources may include equipment models, quantity, availability, maintenance cycles, cool-down times, etc.; constraints on site and airspace resources may include site capacity, availability, airspace open windows, etc. This embodiment of the invention does not impose specific limitations on these.

[0034] It should be understood that constraint solving is a computational process whose goal is to find a feasible resource allocation scheme for each training event in the candidate plan, while satisfying all known resource constraints, thereby transforming the logical plan draft into a concrete, conflict-free execution plan. This process can be performed by one or more constraint solvers, which take the candidate plan and resource constraints as input, and output one or more solutions that meet the conditions through complex algorithmic calculations.

[0035] Here, the baseline plan is used to reflect a detailed and feasible training execution plan that has been validated and optimized for resource constraints. The baseline plan is a detailed schedule down to specific execution units, such as weeks, days, or even sorties, specifying when, where, by whom, and what equipment will be used to carry out each training activity.

[0036] Step 120: Perform executability verification on the baseline plan in the simulation sandbox and obtain the verification results.

[0037] Specifically, after obtaining the baseline plan, the feasibility of the baseline plan can be verified in the simulation sandbox to obtain the verification results.

[0038] A simulation sandbox is a software system that provides an isolated, controlled virtual environment to simulate the execution of a baseline plan. Within the sandbox environment, all relevant real-world entities, such as personnel, equipment, sites, and their interaction rules, are modeled as computer models. For example, personnel and equipment can be modeled as resource pools with specific attributes and quantity limits, and planned events are modeled as operations involving the occupancy and release of these resource pools. It should be understood that simulation sandboxes allow for iterative, fine-grained simulations of the baseline plan without affecting the real world, thereby exposing problems early on.

[0039] The executability verification is the specific process run in the simulation sandbox. The system takes the baseline plan as the input script and advances it according to the time axis specified in the plan in the simulation environment. At each simulation time step (e.g., in minutes), the system checks whether the events in the baseline plan can be executed successfully, such as whether the requested resources are available, whether there are resource conflicts, and whether safety boundaries have been violated, such as two aircraft being too close. This embodiment of the invention does not specifically limit the scope of these checks.

[0040] Here, the verification results are used to reflect the feasibility assessment conclusions of the baseline plan under simulated execution. Verification results can include verification pass or verification fail. Verification pass indicates that all events of the baseline plan were executed smoothly throughout the simulation process, and no resource conflicts or rule violations were found. Verification fail indicates that at least one conflict or violation was found during the simulation, suggesting that the baseline plan may not be able to execute as expected in reality.

[0041] Step 130: In response to the verification result being successful, digitally sign the summary information of the baseline plan to obtain an executable credential.

[0042] Specifically, in response to a successful verification result, the summary information of the baseline plan is digitally signed to obtain an executable credential.

[0043] The summary information is a compact representation reflecting the core content of the baseline plan. It serves as a unique digital fingerprint identifying the plan's content and version. For example, the summary information could be a fixed-length string obtained by hashing the entire baseline plan. Any minor modification to the baseline plan will result in a significant change to its summary information.

[0044] Here, digital signatures are implemented using asymmetric encryption technology. An entity with issuing authority, such as a simulation sandbox system or a trusted authentication authority, uses its private key to encrypt the digest information of the baseline plan, forming a digital signature. The executable credential is the final product of the digital signature process, serving as valid proof that the plan has been verified and approved for execution. The executable credential is essentially a structured data packet containing the digital signature and the original digest information. Any verifier can decrypt the digital signature using the public key corresponding to the signing private key and compare it with the digest information. If the comparison matches, it proves that the executable plan has not been tampered with since its issuance, possesses integrity, and that the issuing entity is authentic and trustworthy, possessing non-repudiation.

[0045] Step 140: Based on the baseline plan and the executable credentials, generate a target plan for execution control.

[0046] Specifically, based on the baseline plan and the executable credentials, a target plan for execution control is generated.

[0047] The target plan is used for execution and control during actual training. It includes a baseline plan and executable credentials, ensuring that the plan issued to the execution layer has undergone complete process verification and authorization. During execution and monitoring, relevant systems can verify the validity of the executable credentials at any time to prevent unauthorized plan changes or the execution of falsified plans.

[0048] The method provided in this invention performs semantic parsing on a training outline to obtain candidate plans, and then solves for constraints on the candidate plans based on resource constraints to obtain a baseline plan. Next, the executability of the baseline plan is verified in a simulation sandbox. After successful verification, the summary information of the baseline plan is digitally signed to obtain an executable credential. Finally, a target plan for execution control is generated based on the baseline plan and the executable credential. On the one hand, by automating semantic parsing and solving for resource constraints on the training outline, a baseline plan conforming to real-world conditions can be automatically generated from the training outline. This solves the problems of low efficiency and high dependence on personal experience in manual plan arrangement in existing technologies, improving the scientific nature of plan formulation and resource utilization efficiency. On the other hand, by verifying the executability of the baseline plan in a simulation sandbox, potential conflicts can be avoided before plan execution, preventing insufficient plan feasibility due to a lack of effective verification methods. Furthermore, generating a target plan based on the baseline plan and the executable credential ensures the integrity and authority of the plan throughout the entire process of formulation and execution, guaranteeing the traceability of the basis for plan formulation.

[0049] Based on the above embodiments, the resource constraints include a resource constraint knowledge graph; Step 110, which involves solving for resource constraints on the candidate plan to obtain the baseline plan, includes: Step 1101: Obtain multi-source heterogeneous resource data, and construct the resource constraint knowledge graph based on the multi-source heterogeneous resource data; the nodes in the resource constraint knowledge graph are resource entities, and the edges in the resource constraint graph are the constraint relationships between the resource entities, including mutual exclusion relationships, temporal relationships, and dependency relationships; Step 1102: Input the candidate plan and the resource constraint knowledge graph into the constraint solver to obtain the baseline plan output by the constraint solver.

[0050] Specifically, in a preferred embodiment, resource constraints may include a resource constraint knowledge graph.

[0051] Accordingly, the system first acquires multi-source heterogeneous resource data and constructs a resource constraint knowledge graph based on this data. Nodes in the resource constraint knowledge graph represent resource entities, and edges represent the relationships between these entities, including mutual exclusion, temporal, and dependency relationships. For example, concurrency limits and confidentiality domain restrictions are expressed in structured relational forms. These constraints form the core basis for constraint solving in subsequent planning and verification processes. The system labels each constraint with hard / soft categories and priority attributes, thus forming a unified and standardized data foundation for planning and verification.

[0052] Multi-source heterogeneous resource data refers to the raw data set used to reflect the status, attributes, and availability of various resources involved in the training. This data can include data from multiple different business systems, such as personnel management systems, equipment management systems, site reservation systems, and meteorological service platforms. For example, the system can automatically aggregate and model multi-source information through a southbound adapter, including personnel qualifications and scheduling from the personnel management system, equipment health and maintenance windows from the equipment management system, site availability and capacity from the site management system, airspace control information from the airspace management system, weather forecasts from the meteorological service platform, and safety regulations from the regulatory system.

[0053] Furthermore, this embodiment of the invention cleans, standardizes, and deduplicates multi-source heterogeneous resource data through a resource accessor, and solves entity alignment problems such as "multiple codes for the same person / multiple names in the same event" through a master data strategy, assigns credibility tags to missing fields, and lays the foundation for building a unified resource view.

[0054] It should be understood that the formats, structures, and semantics of data in multi-source heterogeneous resource data are different. For example, personnel qualifications in multi-source heterogeneous resource data may be a record in a database, equipment status may come from a real-time monitoring interface, and site schedules may be a spreadsheet.

[0055] Here, the resource-constrained knowledge graph is a model used to uniformly describe all resources and their complex constraints in a structured, machine-readable manner. The resource-constrained knowledge graph cleanses, integrates, and models the aforementioned multi-source heterogeneous resource data into a graph-structured network.

[0056] Resource entities are used to reflect specific, independent resources that can be planned for use. For example, an instructor named Zhang San, a simulator numbered J-101, a training airspace A, and a classroom number 201 can all be defined as resource entities.

[0057] Here, mutual exclusion is used to reflect situations where multiple training activities cannot simultaneously occupy the same or the same set of resources. For example, a simulator can only be used by one student at a time. Therefore, if two different training events overlap in time, they cannot be assigned to the same simulator entity, and a mutual exclusion relationship is formed between these two events based on that resource entity.

[0058] Temporal relationships are used to reflect the constraints of resource usage in the order of time. Temporal relationships can include preconditions and postconditions. For example, equipment maintenance events must be completed before equipment usage events, which is a temporal dependency; another example is that a high-power radar needs to cool down for 30 minutes after use before it can be used again, and the cooling process constitutes a temporal constraint on subsequent radar usage events.

[0059] Here, dependencies are used to characterize the necessary association between resource entities in terms of function or use. For example, training in an advanced course requires that a resource entity of type "Advanced Instructor" be associated with a resource entity of type "Beginner Student," which constitutes a dependency between resources.

[0060] After obtaining the resource constraint knowledge graph, the candidate plans and the resource constraint knowledge graph can be input into the constraint solver to obtain the baseline plan output by the constraint solver.

[0061] The constraint solver is a software module whose core function is to solve constraint satisfaction problems. The constraint solver receives a set of variables, the range of values ​​for each variable, and a set of constraints defining the relationships between these variables. It then uses specific algorithms, such as backtracking and forward checking, to search for and assign values ​​to variables that satisfy all constraints.

[0062] The method provided in this invention constructs a unified resource constraint knowledge graph from scattered and heterogeneous multi-source heterogeneous resource data, making it possible to perform automated and global planning using constraint solvers. This avoids omissions and conflicts that are prone to occur when manually comparing resources, greatly improving the intrinsic quality and feasibility of the baseline plan, and enhancing the accuracy and reliability of the baseline plan determination.

[0063] Based on the above embodiments, step 1102 includes: Step 1102-1: The constraint solver transforms the planning events in the candidate plans and the edges in the resource constraint knowledge graph into a formal constraint satisfaction problem. Step 1102-2: The constraint solver optimizes the soft objective while satisfying the formal constraint satisfaction problem to obtain the baseline plan; The soft targets include the achievement rate, task balance, and resource utilization rate.

[0064] Specifically, firstly, the constraint solver transforms the planning events in the candidate plans and the edges, i.e. the relationships, in the resource constraint knowledge graph into a formal constraint satisfaction problem.

[0065] In this context, a planned event is the basic unit of a candidate plan, representing a specific training activity that needs to be scheduled. For example, trainee A's simulator flight training is a planned event.

[0066] Here, planned events are transformed into a formal constraint satisfaction problem. Specifically, each planned event's undetermined attributes, such as start time, personnel, and equipment used, are defined as variables, and each variable is assigned a reasonable range of values. Simultaneously, constraints are constructed based on relationships within a resource constraint knowledge graph, such as site capacity limits, resource exclusivity rules, and time conflict restrictions. For example, event E1 must satisfy the constraint "Event E1.Site.Capacity ≥ Event E1.Number of Personnel Required". Through this transformation, the originally complex and variable planning problem, reliant on human experience, is formalized into a mathematical model that can be systematically searched and reasoned using constraint-solving algorithms. This allows for the automatic search for feasible or optimized solutions while satisfying all constraints, rather than directly guaranteeing the existence of a unique, exact solution.

[0067] Then, the constraint solver optimizes the soft objectives while satisfying the formal constraint satisfaction problem, i.e., satisfying all hard constraints, such as time windows, qualification matching, capacity and mutual exclusion, cooling and concurrency, to obtain the baseline plan. Soft objectives may include, but are not limited to, achievement rate, task balance, and resource utilization. In some implementations, soft objectives may further include deployment costs, personnel fatigue, and equipment maintenance matching, etc., but this embodiment of the invention does not specifically limit these aspects.

[0068] Among them, soft objectives are indicators used to evaluate the merits of a feasible plan. Unlike hard constraints that must be met 100%, such as pilots needing to hold valid driver's licenses, soft objectives are the optimization directions that the system strives to achieve after satisfying all hard constraints.

[0069] Here, the achievement rate reflects the degree to which the baseline plan meets the objectives specified in the training syllabus. For example, if the training syllabus requires the completion of 100 lessons, and a plan can cover 95 of them, the achievement rate is 95%. Optimizing this objective means that the system will generate a baseline plan that covers as much training content as possible.

[0070] Task balance reflects the evenness of the distribution of training tasks among different resource entities. Optimizing this objective can prevent some personnel from being overburdened and overworked, or some equipment from operating at excessive capacity while others are idle, thus making the resource load more balanced. Resource utilization reflects the proportion of time that various resources are effectively used within the planned execution cycle. Optimizing this objective aims to reduce resource idle time and maximize the effectiveness of limited training resources, such as expensive simulators and scarce facilities.

[0071] Understandably, by introducing a two-stage solution strategy of satisfying hard constraints and optimizing soft objectives, we can retain the readability of human-style lesson plans while achieving industrial-grade feasibility and globality, thus avoiding the one-sided shortcomings caused by relying solely on heuristics or solvers.

[0072] The method provided in this invention introduces a two-stage solution strategy of satisfying hard constraints and optimizing soft objectives, which elevates the plan generation process from a simple combination and arrangement to an intelligent optimization process oriented towards multi-objective collaborative optimization, thereby significantly improving the scientificity, efficiency and balance of the benchmark plan.

[0073] Based on the above embodiments, step 110, which involves semantic parsing of the training outline to obtain candidate plans, includes: Step 111: Extract key elements from the training outline; the key elements include skill objectives, subject units, prerequisite relationships, achievement standards, and safety constraints. Step 112: Encode the key elements into an intermediate representation; Step 113: Generate the candidate plan based on the intermediate representation.

[0074] Specifically, firstly, key elements are extracted from the training syllabus. These key elements may include skill objectives, subject units, prerequisite relationships, achievement standards, and safety constraints, etc. This embodiment of the invention does not specifically limit these elements.

[0075] Here, the key elements extracted from the training outline can be accomplished by a large language model. For example, the large language model can segment the clauses of the training outline, disambiguate terms, and sort out the hierarchical relationships.

[0076] Here, key elements refer to the structured information units embedded in the natural language description of the training syllabus that play a decisive role in the planning and arrangement. Skill objectives reflect a specific ability or knowledge point that trainees should master upon completion of training. Subject units reflect the specific training activities or courses designed to achieve specific skill objectives. Prerequisite relationships reflect the logical dependencies between different subject units in terms of learning sequence; for example, theoretical learning subjects must be completed before practical operation subjects can be performed. Prerequisite relationships typically form a directed acyclic graph (DAG).

[0077] The standards for achieving the objectives are specific quantitative or qualitative indicators used to measure whether a subject unit has been completed or a skill objective has been achieved, such as a completion time of less than 20 minutes or no errors in the operational process. Safety constraints reflect the safety regulations or restrictions that must be followed when performing a specific subject, such as a flight altitude not lower than 500 meters.

[0078] Next, the extracted key elements are encoded into an intermediate representation (IR). For example, the system can solidify key elements into versioned intermediate representations, such as skill maps and directed acyclic graphs, subject cards, and core constraint sets. Simultaneously, for ambiguous clauses in the training syllabus, the system can automatically complete operable parameters, such as number of repetitions, conditions, and thresholds.

[0079] Finally, candidate plans are generated based on the intermediate representations.

[0080] Intermediate Representation (IR) is a standardized, structured data format used to carry key elements parsed from the training syllabus. For example, IR can be a collection of multiple objects, where "skill map" objects describe skill objectives and their dependencies, and "subject card" objects define the attributes of each subject unit in detail (including achievement standards, safety constraints, etc.). The advantage of using IR is that once the syllabus is converted to IR, all subsequent processing steps can be based on this stable, clear, and unambiguous structured data, achieving decoupling from the original syllabus text.

[0081] Based on the above embodiments, step 120 includes: Step 121: Map the baseline plan to a simulation model consisting of discrete events, and model the personnel, equipment and site resources in the baseline plan as a resource pool of the simulation model; Step 122: In the simulation model, based on each preset time granularity, the occupation and release of the resource pool by the discrete events are simulated, and the conflict result is determined based on the occupation and release results. Step 123: In response to the conflict result being that no resource conflict or security boundary violation occurred, determine that the verification result is verified as passed; Step 124: In response to the conflict result being a resource conflict or security boundary violation, determine the verification result as verification failure, and generate a minimal set of causes to guide the modification of the baseline plan.

[0082] Specifically, firstly, the baseline plan is mapped to a simulation model composed of discrete events, while the personnel, equipment, and site resources in the baseline plan are modeled as resource pools in the simulation model. The simulation model is a dynamic abstraction of the execution process of the baseline plan. In the simulation model, a continuous training activity is broken down into a series of key discrete events occurring at specific points in time, such as 09:00 resource locking, 09:15 entry check, and 11:15 cooling-off exit. This modeling approach allows the computer to simulate event-driven system evolution.

[0083] Here, a resource pool is an abstract representation of a collection of similar resources in the simulation model. For example, all available pilots can be modeled as a pilot resource pool, with attributes including total number and currently available number. When a discrete event requires a pilot, it requests one unit of resource from this pool.

[0084] Secondly, in the simulation model, based on each preset time granularity, the occupation and release of the resource pool by discrete events are simulated, and the conflict results are determined based on the occupation and release results.

[0085] Finally, if the conflict result is that no resource conflict or security boundary violation occurred, the verification result is determined to be verification passed; if the conflict result is that a resource conflict or security boundary violation occurred, the verification result is determined to be verification failed, and a minimal set of causes is generated to guide the modification of the baseline plan.

[0086] The time granularity refers to the time step in the simulation, such as 1 minute. The simulation system advances the virtual time according to this step. At each time point, the system checks whether any discrete events from the baseline plan have occurred. Occupation and release are the basic operations for the interaction between discrete events and the resource pool; occupation means that an event consumes a resource, and release means that an event returns a resource.

[0087] Conflict results are the output of the simulation system's checks at each time granularity, reflecting the execution status of the plan at the current moment. For example, when an event attempts to occupy a resource pool that is already full, the system determines that a resource conflict has occurred.

[0088] The minimal cause set is used to precisely pinpoint the direct cause of failure, rather than providing a vague conclusion. For example, the minimal cause set could include information such as: "{Conflict type: resource mutual exclusion; time: 10:30; conflict events: [event A, event B]; conflicting resource: simulator S01}", which clearly indicates that at 10:30, events A and B simultaneously competed for simulator S01, resulting in a conflict.

[0089] In addition, the minimum cause set outputs a minimum repair suggestion containing infeasible sub-intervals, which is fed back to the plan generator for iteration until it becomes feasible, greatly improving the efficiency of plan debugging and optimization and ensuring the robustness of the target plan.

[0090] Based on the above embodiments, the executable credential includes a plan fingerprint, and a timestamp and validity period associated with the plan fingerprint; The plan fingerprint is obtained by digitally signing the summary information of the baseline plan; Step 140, followed by: Step 141: Verify the plan fingerprint in the executable credential using a preset public key to obtain the integrity verification result of the baseline plan; Step 142: Verify the compliance of the timestamp and validity period in the executable credential to obtain the compliance verification result; Step 143: In response to the fact that both the integrity verification result and the compliance verification result are verified as passed, execute the target plan.

[0091] Specifically, in a preferred embodiment, the executable credential may include a plan fingerprint, as well as a timestamp and validity period associated with the plan fingerprint. The plan fingerprint is obtained by digitally signing a summary of the baseline plan. The plan fingerprint serves to prevent counterfeiting and tampering in the executable credential.

[0092] The timestamp is used to record the issuance time of the executable credential, while the validity period is used to define the start and end time window in which the executable credential is considered valid. In some implementations, the plan token may also include information such as random numbers and sandbox identifiers; however, this embodiment of the invention does not specifically limit this.

[0093] Accordingly, after step 140, the method may further include a step for performing pre-verification: First, the plan fingerprint in the executable credential is verified using a preset public key to obtain an integrity verification result. Second, the compliance of the timestamp and validity period in the executable credential is verified to obtain a compliance verification result. Finally, in response to both the integrity verification result and the compliance verification result being passed, the execution of the target plan is authorized. This verification mechanism effectively prevents "unendorsed on-the-spot changes." Any plan changes without a valid credential will be intercepted by the system and a re-simulation will be required, forming a strong constraint of "plan verifiability."

[0094] Here, the public key is a key paired with the private key used for signing, and the public key can be publicly distributed to any system or person that needs to verify credentials.

[0095] The integrity verification result reflects whether the plan content has been modified since its issuance. The compliance verification result reflects whether the current execution time is within the validity period of the voucher. Only when both results pass verification can the execution of the target plan be authorized.

[0096] The method provided in this invention ensures that the target plan is not tampered with during transmission and storage, guaranteeing its content integrity. Simultaneously, through timestamp and validity period verification, it prevents the target plan from being executed during expired or invalid periods, controlling execution risks from both timeliness and compliance perspectives. Finally, the target plan is executed only after both verifications are passed, achieving end-to-end trust and security control.

[0097] Based on the above embodiments, the execution of the target plan in step 143 includes: Step 1431: During the execution of the target plan, monitor disturbance events triggered by changes in the personnel, equipment, or site resources; Step 1432: In response to the disturbance event reaching a preset threshold, perform incremental solution within the time window and resource subgraph affected by the disturbance event to obtain a differential packet including the minimum disturbance set; Step 1433: Perform incremental simulation verification on the differential package, and generate a new executable credential after the verification is passed. Update the executable credential based on the new executable credential.

[0098] Specifically, during the execution of the target plan, continuous monitoring is conducted for disturbances triggered by changes in personnel, equipment, or site resources.

[0099] When the severity of a disturbance event reaches a preset threshold, the system performs an incremental solution only within the local time window and resource subgraph affected by the disturbance event, and obtains a differential packet containing the minimum disturbance set.

[0100] Subsequently, incremental simulation verification was performed on the differential package. After the verification was passed, a new executable credential was generated for it, and the old executable credential was updated with this new executable credential.

[0101] Disturbances refer to unexpected situations that occur during the execution of a plan and deviate from the original plan. Examples include a key instructor taking temporary leave, a core piece of equipment malfunctioning suddenly, or a training ground being urgently requisitioned.

[0102] A preset threshold is a value used to determine whether a disturbance event requires replanning. For example, it can be set to trigger replanning only when a core resource becomes unavailable or when more than 10% of non-core resources change, thereby avoiding overreacting to minor disturbances.

[0103] Here, incremental solving is an optimization method for replanning, the core of which lies in avoiding global recalculation and only making local adjustments to the parts affected by the disturbance. This method significantly reduces the computational cost of replanning by constraining the time window of the calculation and the scope of the resource subgraph, for example, limiting the calculation to a specific period after the disturbance occurs and to the associated resource subgraph.

[0104] The differential package is a set that describes the minimum changes to the original target plan. Its contents are the minimum set of adjustment operations required to satisfy the disturbance constraints, such as adjusting the timing of events or replacing execution resources, rather than generating a complete new plan.

[0105] Incremental simulation verification is a lightweight verification mechanism that simulates only the changed parts involved in the differential package and their directly related impact range, in order to efficiently assess the feasibility of the changes and the risk of conflict.

[0106] Based on the above embodiments, step 120 further includes the following prior steps: Step 120-1: Perform compliance verification on each event of the baseline plan against the hard constraint rules in the versioned rule base to obtain the compliance verification result; Step 120-2: In response to the compliance verification result that any of the planned events violates the hard constraint rules, the baseline plan is prevented from entering the simulation sandbox for executability verification.

[0107] Specifically, each planned event in the baseline plan is compared with the hard constraint rules in a versioned rule base for compliance verification to obtain the compliance verification result.

[0108] Then, in response to the compliance verification result that any planned event violates the hard constraint rule, the baseline plan is immediately prevented from entering the subsequent simulation sandbox for executability verification.

[0109] Among them, hard constraint rules refer to mandatory provisions with the highest priority that are absolutely inviolable, such as laws and regulations, safety red lines, and core operating procedures. The versioned rule base is a database that records and manages hardware constraint rules. Its versioning feature ensures that all rule changes are fully recorded, making each compliance verification traceable to the specific rule version it was based on. Here, the versioned rule base can cover outline clauses, regulations, aircraft standard operating procedures, confidential domain rules, and airspace / site specifications, and each rule is marked with its source, effective window, priority, and hard / soft attributes.

[0110] Here, compliance verification refers to the process of systematically comparing planned events with entries in the versioned rule base to detect whether the planned arrangements violate any hard constraints. The blocking mechanism is embodied in the veto principle, meaning that once a violation occurs, the plan approval process will be interrupted and a corresponding error message will be returned, preventing it from proceeding to the computationally intensive simulation verification stage. At this point, the system will not only block the process but also provide equivalent alternatives, such as changing the time slot, venue, instructor, or using a simulator. Furthermore, each judgment generates an evidence card and binds it to the rule base version fingerprint and the plan version fingerprint for audit traceability.

[0111] The method provided in this invention, by adding a fast filtering mechanism based on hard constraint rules before simulation verification, can quickly screen and exclude plans that are obviously non-compliant and lack execution basis with low computational cost, avoids wasting valuable simulation resources on invalid plans, thereby significantly improving the efficiency of the overall verification process and ensuring that only plans that pass the basic compliance check can enter the subsequent dynamic simulation and deduction stage.

[0112] The training plan generation apparatus provided by the present invention is described below. The training plan generation apparatus described below can be referred to in correspondence with the training plan generation method described above.

[0113] Based on any of the above embodiments, the present invention provides a training plan generation apparatus. Figure 2 This is a schematic diagram of the training plan generation device provided by the present invention, as shown below. Figure 2 As shown, the device includes: The acquisition unit 210 is used to acquire a training outline, perform semantic parsing on the training outline to obtain candidate plans, and perform constraint solving on the candidate plans based on resource constraints to obtain a baseline plan; The verification unit 220 is used to perform executability verification on the benchmark plan in a simulation sandbox and obtain the verification result; The digital signature unit 230 is used to digitally sign the summary information of the baseline plan in response to the verification result being successful, to obtain an executable credential. The generation unit 240 is used to generate a target plan for execution control based on the baseline plan and the executable credentials.

[0114] The apparatus provided in this invention performs semantic parsing on a training syllabus to obtain candidate plans, and solves for constraints on the candidate plans based on resource constraints to obtain a baseline plan. Then, it verifies the executability of the baseline plan in a simulation sandbox. After successful verification, it digitally signs the summary information of the baseline plan to obtain an executable credential. Finally, it generates a target plan for execution control based on the baseline plan and the executable credential. On the one hand, by automating semantic parsing and solving for resource constraints on the training syllabus, a baseline plan conforming to real-world conditions can be automatically generated from the training syllabus. This solves the problems of low efficiency and high dependence on personal experience in manual plan arrangement in existing technologies, improving the scientific nature of plan formulation and resource utilization efficiency. On the other hand, by verifying the executability of the baseline plan in a simulation sandbox, potential conflicts can be avoided before plan execution, preventing insufficient plan feasibility due to a lack of effective verification methods. Furthermore, generating a target plan based on the baseline plan and the executable credential ensures the integrity and authority of the plan throughout the entire process of formulation and execution, guaranteeing the traceability of the basis for plan formulation.

[0115] Based on any of the above embodiments, the resource constraints include a resource constraint knowledge graph; The acquisition unit 210 specifically includes: A data acquisition unit is used to acquire multi-source heterogeneous resource data and construct the resource constraint knowledge graph based on the multi-source heterogeneous resource data; the nodes in the resource constraint knowledge graph are resource entities, and the edges in the resource constraint graph are the constraint relationships between the resource entities, including mutual exclusion relationships, temporal relationships, and dependency relationships; The constraint solving unit is used to input the candidate plan and the resource constraint knowledge graph into the constraint solver to obtain the baseline plan output by the constraint solver.

[0116] Based on any of the above embodiments, the constraint solving unit is specifically used for: The constraint solver transforms the planning events in the candidate plans and the edges in the resource constraint knowledge graph into a formal constraint satisfaction problem. The constraint solver optimizes the soft objective while satisfying the formal constraint satisfaction problem, thus obtaining the baseline plan; The soft targets include the achievement rate, task balance, and resource utilization rate.

[0117] Based on any of the above embodiments, the acquisition unit 210 is specifically used for: Extract key elements from the training outline; these key elements include skill objectives, subject units, prerequisite relationships, achievement standards, and safety constraints. Encode the key elements into an intermediate representation; The candidate plan is generated based on the intermediate representation.

[0118] Based on any of the above embodiments, the verification unit 220 is specifically used for: The baseline plan is mapped to a simulation model consisting of discrete events, and the personnel, equipment, and site resources in the baseline plan are modeled as resource pools of the simulation model. In the simulation model, based on each preset time granularity, the discrete events are simulated to occupy and release the resource pool, and the conflict results are determined based on the occupation and release results. If the conflict result indicates that no resource conflict or security boundary violation has occurred, the verification result is determined to be verification passed. In response to the conflict result being a resource conflict or security boundary violation, the verification result is determined to be a verification failure, and a minimal set of causes is generated to guide the modification of the baseline plan.

[0119] Based on any of the above embodiments, the executable credential includes a plan fingerprint, and a timestamp and validity period associated with the plan fingerprint; The plan fingerprint is obtained by digitally signing the summary information of the baseline plan; It also includes a verification unit, which specifically includes: An integrity verification unit is used to verify the plan fingerprint in the executable credential using a preset public key, and obtain the integrity verification result of the baseline plan; A compliance verification unit is used to verify the compliance of the timestamp and validity period in the executable credential and obtain a compliance verification result; The dual verification unit is used to execute the target plan in response to both the integrity verification result and the compliance verification result being verified as passed.

[0120] Based on any of the above embodiments, the dual verification unit is specifically used for: During the execution of the target plan, monitor disturbance events triggered by changes in the personnel, equipment, or site resources; In response to the disturbance event reaching a preset threshold, incremental solution is performed within the time window and resource subgraph affected by the disturbance event to obtain a differential packet including the minimum disturbance set; Incremental simulation verification is performed on the differential package, and a new executable credential is generated after the verification is passed. The executable credential is then updated based on the new executable credential.

[0121] Based on any of the above embodiments, an executable verification unit is further included, wherein the executable verification unit is specifically used for: Each event of the baseline plan is subjected to compliance verification against the hard constraint rules in the versioned rule base to obtain the compliance verification result; If the compliance verification result indicates that any of the planned events violates the hard constraint rules, the baseline plan is prevented from entering the simulation sandbox for executability verification.

[0122] Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention, such as... Figure 3As shown, the electronic device may include a processor 310, a communications interface 320, a memory 330, and a communication bus 340, wherein the processor 310, communications interface 320, and memory 330 communicate with each other via the communication bus 340. The processor 310 can invoke logical instructions in the memory 330 to execute a training plan generation method. This method includes: obtaining a training outline; performing semantic parsing on the training outline to obtain candidate plans; and performing constraint solving on the candidate plans based on resource constraints to obtain a baseline plan; performing executability verification on the baseline plan in a simulation sandbox to obtain a verification result; in response to the verification result being successful, digitally signing the summary information of the baseline plan to obtain an executable credential; and generating a target plan for execution control based on the baseline plan and the executable credential.

[0123] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0124] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the training plan generation method provided by the above methods. The method includes: obtaining a training outline; performing semantic parsing on the training outline to obtain candidate plans; and performing constraint solving on the candidate plans based on resource constraints to obtain a baseline plan; performing executability verification on the baseline plan in a simulation sandbox to obtain a verification result; in response to the verification result being a successful verification, digitally signing the summary information of the baseline plan to obtain an executable credential; and generating a target plan for execution control based on the baseline plan and the executable credential.

[0125] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a training plan generation method provided by the above methods. The method includes: obtaining a training outline; performing semantic parsing on the training outline to obtain candidate plans; and performing constraint solving on the candidate plans based on resource constraints to obtain a baseline plan; performing executability verification on the baseline plan in a simulation sandbox to obtain a verification result; in response to the verification result being successful, digitally signing the summary information of the baseline plan to obtain an executable credential; and generating a target plan for execution control based on the baseline plan and the executable credential.

[0126] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0127] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0128] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A training plan generation method, characterized in that, include: Obtain the training outline, perform semantic parsing on the training outline to obtain candidate plans, and solve the constraints of the candidate plans based on resource constraints to obtain the baseline plan; The feasibility of the baseline plan was verified in a simulation sandbox, and the verification results were obtained. In response to the verification result being successful, the summary information of the benchmark plan is digitally signed to obtain an executable credential; Based on the baseline plan and the executable credentials, a target plan for execution control is generated.

2. The training plan generation method according to claim 1, characterized in that, The resource constraints include a resource constraint knowledge graph; The step of solving the candidate plan based on resource constraints to obtain the baseline plan includes: Acquire multi-source heterogeneous resource data, and construct the resource constraint knowledge graph based on the multi-source heterogeneous resource data; the nodes in the resource constraint knowledge graph are resource entities, and the edges in the resource constraint graph are the constraint relationships between the resource entities, including mutual exclusion relationships, temporal relationships, and dependency relationships; The candidate plan and the resource constraint knowledge graph are input into the constraint solver to obtain the baseline plan output by the constraint solver.

3. The training plan generation method according to claim 2, characterized in that, The step of inputting the candidate plan and the resource constraint knowledge graph into the constraint solver to obtain the baseline plan output by the constraint solver includes: The constraint solver transforms the planning events in the candidate plans and the edges in the resource constraint knowledge graph into a formal constraint satisfaction problem. The constraint solver optimizes the soft objective while satisfying the formal constraint satisfaction problem, thus obtaining the baseline plan; The soft targets include the achievement rate, task balance, and resource utilization rate.

4. The training plan generation method according to any one of claims 1 to 3, characterized in that, The step of semantically parsing the training outline to obtain candidate plans includes: Extract key elements from the training outline; these key elements include skill objectives, subject units, prerequisite relationships, achievement standards, and safety constraints. Encode the key elements into an intermediate representation; The candidate plan is generated based on the intermediate representation.

5. The training plan generation method according to any one of claims 1 to 3, characterized in that, The executability verification of the baseline plan in the simulation sandbox, and the resulting verification results, include: The baseline plan is mapped to a simulation model consisting of discrete events, and the personnel, equipment, and site resources in the baseline plan are modeled as resource pools of the simulation model. In the simulation model, based on each preset time granularity, the discrete events are simulated to occupy and release the resource pool, and the conflict results are determined based on the occupation and release results. If the conflict result indicates that no resource conflict or security boundary violation has occurred, the verification result is determined to be verification passed. In response to the conflict result being a resource conflict or security boundary violation, the verification result is determined to be a verification failure, and a minimal set of causes is generated to guide the modification of the baseline plan.

6. The training plan generation method according to claim 5, characterized in that, The executable credential includes a plan fingerprint, as well as a timestamp and validity period associated with the plan fingerprint; The plan fingerprint is obtained by digitally signing the summary information of the baseline plan; The process of generating a target plan for execution control based on the baseline plan and the executable credentials further includes: The plan fingerprint in the executable credential is verified using a preset public key to obtain the integrity verification result of the baseline plan; Verify the compliance of the timestamp and validity period in the executable credential to obtain the compliance verification result; If both the integrity verification result and the compliance verification result are successful, the target plan is executed.

7. The training plan generation method according to claim 6, characterized in that, The execution of the target plan includes: During the execution of the target plan, monitor disturbance events triggered by changes in the personnel, equipment, or site resources; In response to the disturbance event reaching a preset threshold, incremental solution is performed within the time window and resource subgraph affected by the disturbance event to obtain a differential packet including the minimum disturbance set; Incremental simulation verification is performed on the differential package, and a new executable credential is generated after the verification is passed. The executable credential is then updated based on the new executable credential.

8. The training plan generation method according to any one of claims 1 to 3, characterized in that, The step of performing executability verification on the baseline plan in a simulation sandbox and obtaining verification results includes, prior to: Each event of the baseline plan is subjected to compliance verification against the hard constraint rules in the versioned rule base to obtain the compliance verification result; If the compliance verification result indicates that any of the planned events violates the hard constraint rules, the baseline plan is prevented from entering the simulation sandbox for executability verification.

9. A training plan generation device, characterized in that, include: The acquisition unit is used to acquire the training outline, perform semantic parsing on the training outline to obtain candidate plans, and perform constraint solving on the candidate plans based on resource constraints to obtain the baseline plan; The verification unit is used to perform executability verification on the benchmark plan in a simulation sandbox and obtain the verification results. A digital signature unit is used to digitally sign the summary information of the baseline plan in response to the verification result being successful, thereby obtaining an executable credential; The generation unit is used to generate a target plan for execution control based on the baseline plan and the executable credentials.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the training plan generation method as described in any one of claims 1 to 8.

Images