Automobile skill training closed-loop optimization method and system

Abstract

The application discloses a kind of automobile skill training closed-loop optimization method and system, it is related to automobile skill training data processing technical field, including obtaining the vehicle-mounted data generated in automobile skill training field, training equipment data, detection instrument data, maintenance workstation data, student terminal data and teacher review data, generate automobile skill training event stream;According to event stream, action fragment slicing is carried out, and state transmission type training state carrier is built;Training state carrier is corresponded with automobile skill training benchmark track, and associated deviation source stripping is executed, and training deviation source result is generated;According to training deviation source result and training deviation chain, layered linkage rectification control instruction is generated;Collect rectification feedback data and write state mark layer by layer, update student training state and training resource state.The application realizes training data eventization, deviation source stripping and rectification write closed-loop optimization.

Description

Technical Field

[0001] This invention relates to the field of automotive skills training data processing technology, specifically to a closed-loop optimization method and system for automotive skills training. Background Technology

[0002] As the automotive industry evolves towards intelligent connectivity, new energy, and complex mechatronics, the target audience for automotive skills training has expanded from traditional driving operations and basic maintenance to include complex skills scenarios such as vehicle inspection, fault diagnosis, new energy high-voltage safety, use of intelligent testing equipment, and collaborative operation at repair stations. Existing training systems typically collect training data through onboard data acquisition devices, driving simulators, fault diagnostic instruments, testing benches, repair station terminals, and teaching management platforms. This data is then combined with assessment rules, teacher evaluations, and equipment output results to record, score, and provide feedback on the trainees' training process. Related technologies are shifting from manual experience-based evaluation to data-driven, process-oriented, and platform-based management to improve the objectivity and traceability of automotive skills training.

[0003] Current automotive skills training methods often rely on final scores, deductions, or single-device outputs as evaluation criteria. They lack a unified event-driven processing mechanism for data generated during training, including vehicle-mounted data, testing equipment data, repair bay data, student-submitted data, and instructor-verified data. This makes it difficult to accurately recreate the actual operational process of trainees within the same training batch. Even when existing methods can collect some process data, they typically only use it as supplementary scoring information, lacking hierarchical modeling of training action segments, training resource status, operational execution processes, parameter response sources, and result confirmation relationships. This leads to equipment calibration issues, missing data, resource usage conflicts, invalid actions, and trainee judgment errors being easily conflated. Furthermore, current retraining methods often directly arrange repeated training based on deductions or instructor experience, failing to establish layered, interconnected corrective control based on the source of deviation. They also fail to write corrective feedback data back layer by layer to resource status, action validity status, parameter source status, and result confirmation status, resulting in a lack of closed-loop correlation between retraining results and subsequent training evaluation, training resource verification, and training benchmark updates. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by this invention is that existing automotive skills training evaluation and retraining optimization methods have the following problems: it is difficult to uniformly associate and reconstruct multi-source training data according to training batches; the relationship between training resource status, action execution process, parameter response source and result confirmation is not progressively distinguished, resulting in inaccurate identification of deviation sources; the retraining correction results are difficult to write back to the training state and resource state layer by layer, making it difficult to form a closed-loop optimization; and how to achieve integrated optimization of training evaluation, correction and writing back based on automotive skills training event flow, state-transfer type training state carrier and hierarchical linkage correction control.

[0006] To address the aforementioned technical problems, this invention provides the following technical solution: a closed-loop optimization method for automotive skills training, comprising acquiring vehicle-mounted data, training equipment data, testing instrument data, repair bay data, student terminal data, and teacher review data generated at the automotive skills training site; binding training batches, marking sources, synchronizing time, and unifying field formats to generate an automotive skills training event stream; based on action trigger signals, resource state changes, vehicle state changes, tool usage changes, testing parameter output signals, and result submission signals in the automotive skills training event stream, segmenting the training process of the same training batch into action segments, constructing a state-transferable training state carrier corresponding to each action segment; and transmitting the state... The progressive training state carrier corresponds to the benchmark trajectory of automotive skills training. It performs associative deviation source stripping processing in the order of resource state layer, action execution layer, parameter response layer, and result confirmation layer, generating training deviation source results. Based on the training deviation chain formed by the training deviation source results and the training state carrier, it generates hierarchical linkage correction control commands, which are then sent to the corresponding training terminals or training equipment. It collects correction feedback data, regenerates the correction training state carrier, compares and correlates it with the original training state carrier, and writes back resource state markers, action validity markers, parameter source markers, and result confirmation markers layer by layer, updating the trainee's training status and training resource status.

[0007] As a preferred embodiment of the closed-loop optimization method for automotive skills training described in this invention, the generation of the automotive skills training event stream includes writing student identifier, training project identifier, training batch identifier, data source identifier, device identifier, original timestamp, server receiving time, data field type, and data field value to onboard data, training equipment data, testing instrument data, repair bay data, student terminal data, and teacher review data, respectively; synchronizing data from different data sources according to a unified clock source, and correcting the time based on the offset relationship between the device's local time and the server's receiving time; generating an automotive skills training event stream from the data that has completed training batch binding, source marking, time synchronization, and field format unification, according to the event time sequence. The automotive skills training event stream includes resource status events, action trigger events, parameter response events, result submission events, and review confirmation events.

[0008] As a preferred embodiment of the closed-loop optimization method for automotive skills training described in this invention, the step of segmenting the training process of the same training batch into action segments includes using action triggering events, resource state change events, vehicle state change events, tool usage change events, detection parameter output events, or result submission events as the basis for segment boundary identification; when driving control actions, fault diagnosis actions, maintenance operation actions, detection actions, high-voltage safety confirmation actions, or result confirmation actions are formed in the automotive skills training event flow, the action segments are determined according to the start time, end time, and relationship with adjacent events; when the same event can correspond to multiple action segments, the attribution is determined according to the training batch identifier, equipment identifier, event time, action object, and the sequential relationship of adjacent action segments; if the attribution still cannot be determined, it is written into the event set to be assigned and is not used for training deviation judgment.

[0009] As a preferred embodiment of the closed-loop optimization method for automotive skills training described in this invention, the state-transfer type training state carrier includes a resource state layer, an action execution layer, a parameter response layer, a result confirmation layer, and a state transfer layer. The resource state layer records the states of the training vehicle, training equipment, testing instruments, repair bays, data acquisition devices, and teacher verification terminals during the execution of action segments. The action execution layer records action triggering events, action execution sequence, tool usage records, vehicle control actions, testing and connection actions, repair and disassembly actions, safety confirmation actions, and action completion records. The parameter response layer records parameter names, parameter values, parameter units, parameter acquisition time, parameter source device, parameter corresponding action segment, and device calibration information. The result confirmation layer records student submission results, device output results, teacher verification results, device retest results, and report confirmation records. The state transfer layer records resource state markers, action validity markers, parameter source markers, result confirmation markers, and disposal entry markers, such that the action validity marker inherits the resource state marker, the parameter source marker inherits the action validity marker, and the result confirmation marker inherits the parameter source marker.

[0010] As a preferred embodiment of the closed-loop optimization method for automotive skills training described in this invention, the automotive skills training baseline trajectory is generated from the training syllabus, vocational skills assessment rules, vehicle repair manual, training equipment instruction manual, equipment calibration records, and historical qualified training status carriers. When the training syllabus, vocational skills assessment rules, vehicle repair manual, or training equipment instruction manual contains operation sequences, detection parameters, completion conditions, or verification requirements, these are converted into action sequence benchmarks, parameter response benchmarks, result confirmation benchmarks, and verification formation benchmarks in the automotive skills training baseline trajectory. When the baseline trajectory involves numerical detection data, the equipment calibration error, acquisition resolution, and measurement unit are written into the automotive skills training baseline trajectory. When external documents do not provide a complete baseline, a reference baseline trajectory is extracted from historical qualified training status carriers that are from the same training project, the same vehicle type, the same training equipment type, the same training stage, and have been verified by the teacher.

[0011] As a preferred embodiment of the closed-loop optimization method for automotive skills training described in this invention, the execution-related deviation source stripping process includes generating a resource impact status when the resource status layer indicates that the training vehicle, training equipment, testing instruments, repair bay, data acquisition device, or teacher review terminal cannot support the current training evaluation, thus preventing the action execution layer, parameter response layer, and result confirmation layer from generating student skill deviation conclusions; when the resource status is marked as a resource evaluable status, the action execution layer is processed, and if the action segment lacks corresponding operation records, the action sequence is inconsistent with the automotive skills training baseline trajectory, or the safety confirmation action lacks confirmation records, then an action is generated. Missing bias, action sequence bias, or safety confirmation action missing bias; when an action is validly marked as action established, the parameter response layer is processed. If the parameter lacks source device, acquisition time, measurement unit, or corresponding action segment, a parameter source missing result is generated. If the parameter difference exceeds the range determined by the baseline trajectory and device error, a parameter response bias is generated. When the parameter source is marked as parameter source established, the result confirmation layer is processed. If the result submitted by the student is inconsistent with the baseline trajectory, device output result, or teacher review result, a result judgment bias is generated. If review is required but review data has not been generated, a result pending confirmation status is generated.

[0012] As a preferred embodiment of the closed-loop optimization method for automotive skills training described in this invention, the following steps are included: Associating and comparing the corrective training state carrier with the original training state carrier, and writing back resource state markers, action validity markers, parameter source markers, and result confirmation markers layer by layer: When the training deviation source result is a resource impact state, a resource review control command is generated to trigger the state review of the training vehicle, training equipment, testing instrument, repair bay, data acquisition device, or teacher review terminal; when the training deviation source result is an action missing deviation, action sequence deviation, or safety confirmation action missing deviation, an action lock correction control command is generated; when the training deviation source result is a parameter source missing result or parameter response deviation, an evidence completion control command or parameter identification correction control command is generated; when the training deviation source result is a result judgment deviation or a result pending confirmation state, a result review correction control command or review completion control command is generated; after receiving the correction feedback data, the corrective training state carrier is regenerated, and the state is written back in the order of resource state markers, action validity markers, parameter source markers, and result confirmation markers.

[0013] Another objective of this invention is to provide a closed-loop optimization system for automotive skills training. This system can generate hierarchical linkage correction control commands based on the training deviation chain formed by the training deviation source results and the training state carrier, and then send these hierarchical linkage correction control commands to the corresponding training terminals or training equipment. This solves the problem that current automotive skills training evaluation and retraining optimization methods have difficulty in forming closed-loop optimization because the retraining correction results are difficult to write back to the training state and resource state layer by layer.

[0014] As a preferred embodiment of the automotive skills training closed-loop optimization system described in this invention, it includes: a data event processing module, a state carrier and deviation stripping module, and a deviation correction control and closed-loop write-back module; the data event processing module is used to collect vehicle-mounted data, training equipment data, testing instrument data, repair station data, student terminal data, and teacher review data generated at the automotive skills training site, and to perform training batch binding, source marking, time synchronization, and field format unification to form an automotive skills training event flow; the state carrier and deviation stripping module is used to slice action segments according to the automotive skills training event flow and construct a state-transfer type training state corresponding to each action segment. The training state carrier maps the state transfer type training state carrier to the benchmark trajectory of automotive skill training. It performs associative deviation source stripping processing in the order of resource state layer, action execution layer, parameter response layer, and result confirmation layer to generate training deviation source results and training deviation chain. The deviation correction control and closed-loop write-back module is used to generate hierarchical linkage deviation correction control instructions based on the training deviation source results and training deviation chain, and send the instructions to the corresponding training terminal or training equipment. After collecting deviation correction feedback data, it regenerates the deviation correction training state carrier and writes back the resource state mark, action validity mark, parameter source mark, and result confirmation mark layer by layer to update the trainee training state and training resource state.

[0015] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program as steps to implement a closed-loop optimization method for automotive skills training.

[0016] A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of a closed-loop optimization method for automotive skills training.

[0017] The beneficial effects of this invention are as follows: The closed-loop optimization method for automotive skills training provided by this invention collects data from vehicle-mounted systems, training equipment, testing instruments, repair stations, trainee terminals, and instructors, and performs training batch binding, source marking, time synchronization, and field format unification. This achieves event-based reconstruction of multi-source data from the automotive skills training site, eliminating inconsistencies in time references and unclear batch attribution between data from different devices and terminals, thus forming a traceable training data foundation. By slicing the training event stream into action segments and constructing a training state carrier containing a resource state layer, action execution layer, parameter response layer, result confirmation layer, and state transmission layer, the training process is transformed from continuous mixed data to action segment-level state objects, accurately describing the resource, action, parameter, and result relationships corresponding to each training action. By mapping the training state carrier to the baseline trajectory and performing correlated deviation source stripping, progressive differentiation is achieved for resource influence, action failure, missing parameter sources, and result judgment deviations, avoiding misjudging equipment problems, missing data collection, or incomplete actions as insufficient trainee skills. By generating hierarchical, interconnected error correction control commands, the system achieves the corresponding issuance of error correction tasks and deviation sources, enabling resource verification, action locking, evidence completion, parameter identification, and result verification to be executed in the order of status transmission. By collecting error correction feedback data and writing back status markers layer by layer, a closed-loop linkage between training evaluation, error correction execution, and resource status updates is achieved, thereby improving the accuracy of training evaluation, the relevance of supplementary training, and the ability to continuously optimize the training process. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 The first embodiment of the present invention provides an overall flowchart of a closed-loop optimization method for automotive skills training. Detailed Implementation

[0020] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0021] Example 1, referring to Figure 1As an embodiment of the present invention, a closed-loop optimization method for automotive skills training is provided, comprising: S1: Acquire vehicle data, training equipment data, testing instrument data, repair station data, student terminal data, and teacher review data generated at the automotive skills training site, bind training batches, mark sources, synchronize time, and unify field formats to generate an automotive skills training event stream.

[0022] Furthermore, before the automotive skills training begins, the training management platform receives training start information from trainees. This start information includes trainee identification, training project identification, training batch identification, training vehicle identification, training equipment identification, training site identification, and training terminal identification. Based on this training start information, the training management platform sends data collection start commands to the vehicle data interface, training equipment interface, testing instrument interface, repair bay data acquisition interface, trainee terminal, and instructor review terminal. This ensures that vehicle data, equipment data, operational data, testing data, result submission data, and review data generated within the same training batch can be linked to that batch.

[0023] The training batch identifier is used to distinguish training tasks performed by the same trainee at different times, and also to distinguish training tasks performed by different trainees in the same training area or on the same training equipment. The training batch identifier is generated when the trainee confirms the start of training and is written into the data collection records of the vehicle-mounted data acquisition device, training equipment, testing instruments, maintenance station terminals, trainee terminals, and instructor review terminals. For devices that do not support actively writing training batch identifiers, the training management platform records the device's operational records from the start to the end of training and establishes a link between the device identifier, training time period, and training item identifier and the training batch identifier. This process avoids data confusion caused by the same device performing different training tasks.

[0024] In driving operation training scenarios, onboard data interfaces include the vehicle's OBD interface, CAN bus interface, or the output interface of the driving simulator, used to collect vehicle operating status and driving operation status. Vehicle operating status includes vehicle speed, engine speed, motor speed, gear position, parking status, braking status, turn signal status, vehicle lateral position, vehicle longitudinal position, and vehicle trajectory data. Driving operation status includes steering wheel angle, brake pedal opening, accelerator pedal opening, gear shift operation records, turn signal operation records, and parking operation records. For driving simulators, the collected data also includes the simulation scenario number, scenario loading time, virtual vehicle position, collision warning signal, line crossing warning signal, and parking area offset data. This data reflects the trainee's vehicle control process during driving skills training, rather than simply recording the final score after training.

[0025] In fault diagnosis, vehicle inspection, and maintenance training scenarios, training equipment interfaces include data interfaces for fault diagnostic instruments, testing benches, lifts, torque tools, insulation testing equipment, voltage testing equipment, and high-voltage safety training equipment for new energy vehicles. Fault diagnostic instruments are used to collect data on connection status, fault code reading results, data stream reading results, code clearing operation records, and re-inspection results. Testing benches are used to collect data on testing items, testing start time, testing parameters, testing conclusions, and equipment feedback status. Lifts are used to collect data on workstation occupancy status, lifting height status, and safety lock status. Torque tools are used to collect data on tool number, tightening object, tightening time, tightening torque, and tool calibration status. Insulation testing equipment and voltage testing equipment are used to collect insulation test values, voltage testing confirmation results, equipment connection status, and testing completion status. The above equipment data reflects whether trainees have actually completed the corresponding testing, diagnosis, disassembly, tightening, or safety confirmation operations.

[0026] In a maintenance workstation training scenario, the workstation data acquisition interface includes at least one of the following: an RFID tool identification device, a tool retrieval and placement detection device, a workstation status acquisition device, a maintenance terminal, and a video acquisition device. The RFID tool identification device and tool retrieval and placement detection device are used to collect data on tool retrieval time, tool return time, tool number, and tool type. The workstation status acquisition device is used to collect data on workstation occupancy status, component placement status, and workstation operation status. The maintenance terminal is used to collect disassembly and assembly confirmation records, maintenance step confirmation records, fault handling results, and re-inspection confirmation records submitted by trainees. The video acquisition device is used to assist in forming operation confirmation records; its collection results can serve as supplementary evidence for teacher review or workstation status confirmation, but are not the sole basis for trainee skill evaluation. Combining multiple data acquisition interfaces can reduce the untraceability issues in the training process caused by relying solely on manual recording.

[0027] On both the student terminal and the teacher review terminal, the student terminal is used to collect the test results, fault diagnosis results, repair results, driving training confirmation results, safety confirmation results, and training completion confirmation information submitted by the student. The teacher review terminal is used to collect the teacher's review object, review time, review conclusion, review reason, and the training action segment corresponding to the review result. The teacher review record is not an independent evaluation rule, but rather a review-type event in the training event stream, used together with vehicle data, equipment data, operational data, and result submission data to form the subsequent training status carrier.

[0028] After the collected multi-source data enters the data access layer of the training management platform, it first undergoes data source marking processing. Each data entry is written with a data source identifier, physical device identifier, training resource identifier, training batch identifier, student identifier, training project identifier, original timestamp, server reception time, data field type, data unit, and original data value. The data source identifier distinguishes between vehicle-mounted data, equipment data, workstation data, student terminal data, and teacher review data; the physical device identifier distinguishes between specific vehicles, testing instruments, tools, or terminals; and the data field type distinguishes between different types of data such as vehicle speed, turning angle, pedal status, testing parameters, tool status, fault codes, conclusion submissions, and review confirmations. This processing ensures that data from different sources has a unified data identity in subsequent processing.

[0029] For data from different devices, time synchronization is performed. If the vehicle-mounted data acquisition device, training equipment, testing instruments, or terminal supports receiving the unified clock of the training management platform, a data acquisition timestamp is generated using the unified clock. If the device only provides its local time, the correspondence between the device's local time and the training management platform's server time is recorded at the start and end of training, and the device's local time is corrected based on this correspondence. For data continuously uploaded during training, the local time is converted to a corrected time under the unified timeline according to the offset between the device's local time and the server time. If a device cannot provide its local time, the server's receiving time is used as the timestamp for the data, and the data is marked as a receiving time event. Receiving time events can be used to confirm the existence of data, but in subsequent action segment slices involving strict sequential order, auxiliary confirmation is required, combined with device status changes, action trigger records, or teacher review records. This time synchronization process reduces the impact of inconsistent data acquisition times from multiple devices on training action recognition.

[0030] After time synchronization is completed, a unified format processing is performed on multi-source data. For numerical data, it is standardized into parameter records containing the value, unit, acquisition precision, and device source. For status data, it is standardized into status records containing the status name, status value, start time, and end time. For action data, it is standardized into action records containing the action object, action type, action trigger time, end time, and source. For conclusion data, it is standardized into result records containing the submission object, submission content, submission time, and submission terminal. For review data, it is standardized into review records containing the review object, review method, review conclusion, review time, and reviewer identification. When different devices use different units or encoding methods, unit conversion and encoding mapping are performed according to the equipment manual, vehicle maintenance manual, or equipment interface documentation, retaining both the original and converted data values ​​to avoid data untraceability due to format conversion.

[0031] After completing source tagging, time synchronization, and format standardization, the training management platform sorts the multi-source data according to a unified timeline to generate an automotive skills training event stream. The training event stream consists of multiple training event records, each of which includes at least an event number, training batch identifier, student identifier, training project identifier, data source identifier, training resource identifier, event type, event time, event fields, event value, and data status flags.

[0032] It should be noted that the event types include resource status events, action trigger events, parameter response events, result submission events, and verification and confirmation events. Resource status events are used to record the online status, connection status, calibration status, and occupancy status of vehicles, equipment, testing instruments, data acquisition devices, maintenance stations, and terminals; action trigger events are used to record driving control actions, tool retrieval actions, testing connection actions, disassembly and assembly actions, safety confirmation actions, and maintenance confirmation actions; parameter response events are used to record vehicle status parameters, testing parameters, maintenance measurement parameters, equipment feedback parameters, and measurement units; result submission events are used to record the testing conclusions, fault diagnosis conclusions, maintenance processing results, driving training results, and training confirmation information submitted by trainees; and verification and confirmation events are used to record the teacher's verification, equipment retesting, report confirmation, and result re-inspection.

[0033] Data status markers are used to indicate whether an event record has the basis for further processing. Data status markers include available events, events awaiting association, received time events, and events awaiting verification. Available events are those that have completed training batch binding, time synchronization, and field format standardization; events awaiting association are those where the data source and device identifier have been determined, but the training action segment to which they belong cannot be determined yet; received time events are those where the server's received time is used for marking due to a lack of local device acquisition time; events awaiting verification are those where there are source conflicts, inconsistent device statuses, or missing manual confirmation within the same time period, requiring subsequent verification by teachers or device retesting. The above data status markers only describe the processing status of event data and do not directly indicate whether the trainee's skills are up to standard.

[0034] When generating the training event stream, no final evaluation of the trainee's skill level is made, nor is a retraining conclusion directly generated. Instead, the vehicle operation process, equipment data collection process, maintenance operation process, changes in detection parameters, trainee result submission, and teacher review results are uniformly organized into a traceable data sequence. Subsequent steps can use this training event stream to identify training action segments, construct training state carriers, and perform deviation source stripping.

[0035] S2: Based on the action trigger signals, resource state changes, vehicle state changes, tool usage changes, detection parameter output signals, and result submission signals in the automotive skill training event flow, the training process of the same training batch is sliced ​​into action segments, and a state-transfer type training state carrier corresponding to each action segment is constructed.

[0036] Furthermore, after the automotive skills training event flow is formed, the training management platform aggregates the training event records within the same training batch according to the training batch identifier, trainee identifier, and training project identifier. Based on the event type, event time, event source, and event fields in the training event flow, it segments the training process into action fragments, generating multiple training action fragments. Each training action fragment represents a set of resource status events, action trigger events, parameter response events, result submission events, and review confirmation events formed by the trainee within a continuous time frame around the same automotive skills training objective.

[0037] The action segment segments in this step are not directly divided according to manually preset scoring items, nor are they simply segmented according to fixed time lengths. Instead, the segment boundaries are based on actual event changes occurring in the training environment. The boundaries of action segments are jointly determined by action trigger events, resource status events, parameter response events, result submission events, and verification confirmation events. For the same training item, if there is a clear action trigger event in the training event stream, that action trigger event is used as the starting boundary of the corresponding action segment. If there is no clear action trigger event in the training event stream, but there are changes in equipment status from disconnected to connected, tool status from unused to used, vehicle control status from untriggered to triggered, or detection instrument status from standby to acquisition, then that status change event is used as the starting boundary of the action segment. The ending boundary of an action segment is determined based on an action completion event, equipment feedback completion event, result submission event, verification confirmation event, or a mutually exclusive subsequent action trigger event. If a certain action segment lacks a clear ending event, its ending boundary is determined based on subsequent events in the same training batch, training terminal confirmation records, or teacher review records; if the ending boundary still cannot be determined, the action segment is marked as a boundary pending confirmation segment, and this segment will not be directly entered into the subsequent skill deviation judgment.

[0038] In driving operation training, driving action segments are divided according to changes in vehicle state and driving control actions. Starting action segments can be triggered by events such as releasing the vehicle from parking, changing gear position, changing accelerator pedal position, or the vehicle speed transitioning from a stationary state to a moving state. The ending boundary is the vehicle entering a stable driving state, the next driving action triggering an event, or a confirmation event from the training terminal. Braking action segments can be triggered by changes in brake pedal position, a decrease in vehicle speed, or a braking system feedback event. The ending boundary is the brake pedal being released, the vehicle speed reaching a stop, or the next driving action triggering an event. Steering action segments can be triggered by changes in steering wheel angle, turn signal operation recording, or changes in vehicle trajectory direction. The ending boundary is the steering wheel returning to center, the steering action being completed, the vehicle trajectory returning to a straight line, or the next driving action triggering an event. Parking action segments can be triggered by a decrease in vehicle speed, a change in parking state, an approach to a parking area event, or a parking detection event from the simulated driving equipment. The ending boundary is the establishment of a parking state, the submission of parking confirmation results, or teacher review and confirmation. In this way, the driving training process is divided into action segments that correspond to the actual movement of the vehicle and the student's control actions, rather than being recorded solely by the name of the training item or the final score.

[0039] In fault diagnosis training, fault diagnosis action segments are divided into sections based on the diagnostic instrument connection status, fault code reading status, data stream reading status, code clearing operation status, and re-inspection status. The diagnostic connection segment can be triggered by events such as the diagnostic instrument connecting to the vehicle interface, the diagnostic software establishing communication, or the diagnostic device uploading connection status, and ends with communication establishment completion, connection failure feedback, or student terminal confirmation. The fault code reading segment can be triggered by fault code reading commands, diagnostic instrument data request events, or fault code output events, and ends with fault code reading result generation, reading failure feedback, or subsequent data stream reading events. The data stream analysis segment can be triggered by data stream reading commands or continuous parameter data upload events, and ends with the student submitting their judgment, the instructor reviewing the record, or the testing equipment stopping data acquisition events. The code clearing and re-inspection segment can be triggered by code clearing operation records, equipment re-inspection initiation events, or fault code re-reading events, and ends with the re-inspection result formation or report confirmation events.

[0040] In automotive repair training, repair action segments are divided into segments based on events such as tool retrieval, component disassembly / assembly, tightening measurement, and re-inspection confirmation. Tool selection segments can be triggered by events such as RFID tool identification detecting tool retrieval, tool removal detection recording tool displacement, or tool selection confirmation from the repair terminal, ending with events such as tool return, tool use completion, or the next tool retrieval. Disassembly segments can be triggered by events such as component disassembly confirmation, changes in workstation status, changes in tool usage status, or video-assisted confirmation, ending with events such as component disassembly completion recording, repair terminal confirmation, or the next installation action triggering the next segment. Installation segments can be triggered by events such as component installation confirmation, tool usage recording, or changes in workstation status, ending with events such as installation completion recording, torque tool tightening, or re-inspection confirmation. Tightening segments can be triggered by events such as torque tool activation, fastening object binding, or torque data upload, ending with events such as torque value upload, tool cessation of use, or workstation terminal confirmation. The re-inspection segment can be triggered by the re-inspection initiation record, equipment retest record, or teacher review record, and the end boundary is the re-inspection result submission, report confirmation, or training end confirmation event.

[0041] In vehicle inspection training, inspection action segments are divided into sections based on events related to equipment connection, inspection item selection, inspection parameter output, and inspection report generation. The inspection connection segment can be triggered by events such as the connection of the inspection instrument to the vehicle, the start of the inspection bench, or the inspection equipment entering the data acquisition state, and ends with events such as connection confirmation, connection failure feedback, or inspection item selection. The inspection data acquisition segment can be triggered by events such as the start of an inspection item, the commencement of parameter acquisition, or the equipment entering the inspection state, and ends with events such as the formation of inspection parameters, the cessation of data acquisition by the inspection equipment, or the trainee submitting the inspection conclusion. The inspection report segment can be triggered by events such as the trainee submitting the inspection conclusion, the report generation command, or the report confirmation event, and ends with events such as the completion of report generation, instructor review and confirmation, or the formation of equipment retest results.

[0042] In high-voltage safety training for new energy vehicles, high-voltage safety action segments are divided into sections based on events such as power outage confirmation, voltage verification confirmation, insulation testing, protective equipment verification, and safety review. The power outage confirmation segment can be triggered by a power outage operation record, a change in the high-voltage system status, or a power outage confirmation event on the trainee's terminal, and ends with a power outage completion record, equipment status feedback, or instructor review confirmation. The voltage verification confirmation segment can be triggered by a voltage testing device connection, a voltage testing action, or a voltage testing result output event, and ends with the formation of the voltage testing result, an instructor review record, or the triggering of the next safety action. The insulation testing segment can be triggered by the start of the insulation testing equipment, the uploading of testing parameters, or the binding of the testing object, and ends with the formation of the insulation testing result, an equipment retest record, or a report confirmation event. The protective equipment verification segment can be triggered by a protective equipment identification terminal, a trainee confirmation record, or an instructor review record, and ends with the formation of the verification result or the start of a subsequent safety action.

[0043] When multiple related events exist within the same training batch, the training management platform assigns segments according to the principle of "resource status events covering action segments, action triggering events defining the main body of the segment, parameter response events belonging to the corresponding action segment, result submission events bound to the nearest attributable action segment, and verification confirmation events bound to their verification object." Resource status events can span multiple action segments and describe the status of vehicles, equipment, workstations, or data acquisition devices within that time frame; action triggering events determine the main body of the action segment; parameter response events are preferentially assigned to the action segment that generated the parameter; result submission events are assigned to the action segment corresponding to their submitted conclusion; and verification confirmation events are assigned to the action segment pointed to by teacher verification, equipment retesting, or report confirmation. If a parameter response event or result submission event can correspond to multiple action segments simultaneously, it is assigned based on time sequence, device identifier, training project identifier, action object, and terminal submission object; if it still cannot be assigned, it is marked as an event to be assigned and added to the event set to be assigned, and is not directly used for subsequent result judgment.

[0044] After the action segment is sliced, the training management platform constructs a state-transfer type training state carrier for each training action segment. The training state carrier is not simply an evaluation table, but a data object used to carry the multi-source events, hierarchical states, and subsequent processing entry points corresponding to that action segment. Each training state carrier includes at least the following: carrier identifier, training batch identifier, trainee identifier, training project identifier, action segment identifier, action segment type, segment start time, segment end time, segment source event set, resource state layer, action execution layer, parameter response layer, result confirmation layer, and state transfer layer.

[0045] The carrier identifier is used to uniquely identify the training state carrier formed by a training action segment; the action segment type is used to distinguish driving action segments, fault diagnosis action segments, maintenance operation action segments, detection action segments, high-voltage safety action segments, and result confirmation action segments; the segment source event set is used to store the original event number that formed the action segment, so that subsequent processing can trace back to the original vehicle data, equipment data, workstation data, terminal data, and review data. The segment start time and segment end time are derived from the aforementioned action segment slicing process; for boundary unconfirmed segments, their end time can temporarily use the time of the last attributable event, and a boundary unconfirmed marker is written in the state transfer layer.

[0046] The resource status layer records the status of the corresponding training resources during the execution of the action segment. The resource status layer includes at least the training vehicle status, training equipment status, testing instrument status, maintenance station status, data acquisition device status, instructor review terminal status, equipment calibration status, and data upload status. The training vehicle status indicates the vehicle model, vehicle connection status, vehicle operating status, and whether the vehicle is bound to this training batch. The training equipment status indicates whether the fault diagnostic instrument, testing bench, lift, torque tool, insulation testing equipment, or voltage testing equipment is in a data acquisition-ready state. The testing instrument status indicates whether the testing instrument has completed connection and testing output. The maintenance station status indicates whether the station is occupied by this training batch and whether the station status corresponds to this training item. The data acquisition device status indicates whether the RFID identification device, tool handling detection device, camera acquisition device, or terminal upload channel has generated data. The instructor review terminal status indicates whether the training segment requiring review has a review data entry point. The resource status layer only describes whether the resources can support the data acquisition and subsequent evaluation of the action segment; it does not directly determine whether the trainee's skills are qualified.

[0047] The action execution layer records the actual actions performed by the trainee within the action segment. The action execution layer includes at least the action triggering event, action execution sequence, action object, tool usage record, vehicle control actions, detection and connection actions, maintenance and disassembly actions, safety confirmation actions, action duration, and action completion record. For driving training, the action execution layer records vehicle control data corresponding to actions such as starting, braking, steering, shifting gears, reversing, and parking; for maintenance training, it records action data such as tool retrieval, component disassembly, component installation, torque tightening, and re-inspection confirmation; for detection training, it records action data such as connection of detection equipment, selection of detection items, detection data collection, and report confirmation; for high-voltage safety training of new energy vehicles, it records action data such as power outage confirmation, voltage verification confirmation, insulation testing, and protective equipment confirmation. The data in the action execution layer must originate from action triggering events, state change events, or terminal confirmation events in the training event stream and maintain a corresponding relationship with the event set from which the segment originates.

[0048] The parameter response layer records the parameter data generated by the action segment and its source relationships. The parameter response layer includes at least the parameter name, parameter value, parameter unit, parameter acquisition time, parameter source device, corresponding action segment, device calibration information, acquisition resolution, and original event number. For driving training, the parameter response layer can record data on vehicle speed changes, steering wheel angle changes, brake pedal changes, accelerator pedal changes, and parking position offset. For vehicle inspection training, the parameter response layer can record the test parameters output by the test bench, the data stream parameters read by the diagnostic tool, and fault code information. For maintenance training, the parameter response layer can record tightening torque, measurement parameters, re-inspection parameters, and equipment feedback parameters. For high-voltage safety training, the parameter response layer can record insulation test values, voltage detection results, and high-voltage system status parameters. The parameter response layer not only stores parameter values ​​but also the source relationships between parameters and corresponding action segments, enabling subsequent determination of whether a parameter was actually generated by that action segment.

[0049] The result confirmation layer records the results submitted by the trainee, the output results of the equipment, and the review results of the instructor. The result confirmation layer includes at least the trainee's submitted content, submission time, submission terminal, equipment output results, report confirmation record, instructor review record, equipment retest record, and the object of the result. Trainee-submitted content includes test conclusions, fault diagnosis conclusions, repair conclusions, driving operation confirmation results, or safety confirmation results; equipment output results include test instrument output conclusions, diagnostic instrument readings, torque tool feedback results, insulation testing equipment test results, and simulated driving equipment evaluation results; instructor review records include the object of review, review conclusions, review time, and review reasons. The result confirmation layer is used to indicate whether the action segment has formed a result expression and review basis, but it does not directly determine whether the result is correct in this step.

[0050] It should be noted that the state transmission layer is used to record the inheritance relationship between the resource state layer, action execution layer, parameter response layer, and result confirmation layer. The state transmission layer includes at least resource state flags, action open flags, action valid flags, parameter open flags, parameter source flags, result open flags, result confirmation flags, and disposal entry flags. Resource state flags indicate whether the resource state corresponding to the action segment has a basis for subsequent evaluation; the action open flag is determined based on the resource state flag; only when the resource state flag indicates that the action segment has an evaluation basis will the action execution layer open the subsequent action valid judgment; the parameter open flag is determined based on the action valid flag; only when the action valid flag indicates that the action segment is valid will the parameter response layer open the parameter source judgment; the result open flag is determined based on the parameter source flag; only when the parameter source can establish a correspondence with the action segment will the result confirmation layer open the result confirmation judgment.

[0051] In this step, the markers in the state transit layer are only initialized and do not directly generate the final deviation conclusion. Specifically, when the vehicle, equipment, workstation, acquisition device, and terminal data in the resource state layer can all be associated with the training batch, and there are no resource state conflicts that require immediate manual verification, the resource state marker is initialized as a resource that can proceed to subsequent processing. When the resource state layer has issues such as equipment not being bound to a training batch, missing equipment calibration status, vehicle models not corresponding to training items, acquisition devices not generating data, or workstation occupancy relationships not being confirmed, the resource state marker is initialized as a resource pending verification. In the resource pending verification state, the action open marker is not directly opened. For cases where there are many fragments with pending boundary confirmation, fragments with many pending events, or result submission objects that cannot correspond to action fragments, the corresponding pending confirmation marker is written in the state transit layer, and the process proceeds to the resource verification, evidence completion, or teacher verification path in subsequent steps.

[0052] The action open flag inherits from the resource status flag. When the resource status flag indicates that the resource can proceed to subsequent processing, the action open flag is set to "action can be processed"; when the resource status flag indicates that the resource is pending verification, the action open flag is set to "action not processed temporarily". The action validity flag is not checked for bias in this step; it is initialized to "action record has been formed" or "action record is pending completion" based solely on whether the action execution layer has generated an action trigger event, action completion record, or training terminal confirmation record. The parameter open flag inherits from the action validity flag. When the action record has been formed, the parameter response layer can proceed to parameter source matching; when the action record is pending completion, the parameter open flag is set to "parameter not processed temporarily". The parameter source flag is initialized in this step based on whether the parameter response layer has a parameter source device, parameter acquisition time, parameter unit, and corresponding action segment, recording "parameter source traceable" or "parameter source pending confirmation". The result open flag inherits from the parameter source flag. When the parameter source is traceable, the result confirmation layer can proceed to result confirmation processing; when the parameter source is pending confirmation, the result open flag is set to "result not processed temporarily". In this step, the result confirmation flag is initialized to indicate whether the result record has been formed, the result record is pending review, or the result record is missing, depending on whether the result confirmation layer has generated student-submitted results, device-output results, or teacher-reviewed records.

[0053] Through the aforementioned state transfer layer setup, a hierarchical constraint relationship is established in the training state carrier during the construction phase: action segments whose resource states have not been verified are not allowed to be judged as valid actions; segments whose action records have not been formed are not allowed to be judged as having parameter source; segments whose parameter sources cannot be traced are not allowed to be judged as having result confirmation; and segments whose result verification has not been formed are not allowed to be directly written to the final evaluation. This avoids resource judgment, action judgment, parameter judgment, and result judgment being performed independently in subsequent steps, ensuring that the identification of deviation sources and the control of deviation are all based on the results of the previous level of state.

[0054] After the training state carrier is constructed, the training management platform generates a training state carrier sequence according to training batches. The training state carrier sequence retains the temporal order, sequential relationship, and source event relationship between each action segment. For action segments with sequential dependencies within the same training batch, such as the power outage confirmation segment, voltage verification segment, and insulation detection segment in high-voltage safety training for new energy vehicles, the training state carrier sequence records their sequential relationship; for tool selection segment, disassembly segment, installation segment, and tightening segment in automotive repair training, the training state carrier sequence records their operational connection relationship; for starting, driving, steering, braking, and stopping segments in driving training, the training state carrier sequence records their temporal connection relationship.

[0055] S3: The state-transfer type training state carrier is mapped to the vehicle skill training baseline trajectory. The correlation-based deviation source stripping process is performed in the order of resource state layer, action execution layer, parameter response layer and result confirmation layer to generate training deviation source results.

[0056] Furthermore, after forming the training state carrier sequence, the training management platform maps each training state carrier to the automotive skills training baseline trajectory, performs correlation-based deviation source stripping processing, and generates training deviation source results. The purpose of this step is not to directly determine whether the trainee has passed the training, but to first determine whether the training deviation originates from training resources, action execution, parameter response, result judgment, or lack of verification evidence. It also limits the order of judgments at each level through a state transmission layer, avoiding directly attributing training result differences to insufficient trainee skills when resource status is not confirmed, action segments are not established, or parameter sources are unclear.

[0057] The automotive skills training baseline trajectory represents the reference training process that should be formed for a specific training item under the corresponding vehicle model, training equipment, training stage, and training scenario. The baseline trajectory is jointly formed by the training syllabus, vocational skills assessment rules, vehicle repair manuals, training equipment manuals, equipment calibration records, and historical qualified training status data. When the training syllabus, assessment rules, vehicle repair manuals, or equipment manuals clearly specify the operation sequence, testing items, testing parameters, completion conditions, or verification requirements, the corresponding content is converted into the action sequence baseline, parameter response baseline, result confirmation baseline, and verification formation baseline in the baseline trajectory, and the source document, version information, applicable vehicle model, applicable equipment, and applicable training item are recorded. When numerical data output from testing instruments, torque tools, insulation testing equipment, or on-board data acquisition devices is involved, the baseline trajectory also includes equipment calibration error, acquisition resolution, and measurement unit to distinguish between equipment measurement errors and trainee operational deviations during subsequent comparisons.

[0058] When external standard documents do not provide a complete benchmark, the training management platform extracts a reference benchmark trajectory from historical qualified training status carriers. The source of the reference benchmark trajectory is limited to training status carriers from the same training project, vehicle type, training equipment type, and training stage, and which have been verified as passed by the instructor. During extraction, the sequence of action segments, segment boundaries, resource status, parameter response, result confirmation, and instructor verification results from the historical carriers are retained. If a historical training status carrier has states where resources are pending verification, action boundaries are pending confirmation, parameter sources are pending confirmation, or result records are pending verification, then that historical carrier will not be used to form a reference benchmark trajectory. If the available historical carriers are insufficient to form a reference benchmark, the corresponding action segment enters the instructor verification path, and no skill deviation conclusion is automatically generated.

[0059] When performing deviation source identification, the platform first aligns the training state carrier with the baseline trajectory. During alignment, the applicable baseline trajectory is determined based on the training project identifier, vehicle model identifier, training equipment identifier, and training stage identifier. Then, based on the action segment type, segment start time, segment end time, action trigger event, and result submission object, a correspondence is established between the training state carrier and the reference segment in the baseline trajectory. If a training state carrier can correspond to multiple baseline segments, the correspondence is determined first based on the action trigger event and training equipment source. If this still cannot be determined, the time sequence of the previous and subsequent training state carriers is used for confirmation. If a clear correspondence still cannot be established, the segment is added to the set of segments to be reviewed and is not directly entered into the student skill deviation judgment.

[0060] Deviation source identification proceeds progressively in the order of resource status layer, action execution layer, parameter response layer, and result confirmation layer. First, the resource status layer is processed. The training management platform reads the status of training vehicles, training equipment, testing instruments, repair stations, data acquisition devices, instructor verification terminals, equipment calibration, and data upload, and matches them with the applicable vehicle type, applicable equipment, data acquisition source, and verification conditions in the baseline trajectory. If the training vehicle model does not correspond to the training item, the training equipment is not bound to this training batch, the testing instrument is not connected, the equipment calibration status cannot support this data acquisition, the repair station corresponds to other training batches in the same time period, the data acquisition device has not generated an upload record, or the instructor verification terminal has not generated a usable entry point, then the resource status is marked as being in the resource impact status. Action segments in the resource impact status do not enter the action execution deviation judgment; their data in the action execution layer, parameter response layer, and result confirmation layer are only retained data and are not directly used to determine the trainee's skill deficiency. If the resource status layer can establish a consistent correspondence with this training batch, training item, and baseline trajectory, then the resource status is marked as being in the resource evaluable status, and the action execution layer processing is enabled.

[0061] Secondly, under the premise that the resource status is marked as an evaluable resource, the action execution layer is processed. The training management platform reads the action trigger events, action execution sequence, action object, tool usage records, vehicle control actions, detection and connection actions, maintenance and disassembly actions, safety confirmation actions, action duration, and action completion records, and aligns them with the action sequence benchmark and action completion conditions in the baseline trajectory. For driving training, if the corresponding vehicle control event is missing for action segments such as starting, braking, steering, shifting, reversing, or parking, an action missing deviation is generated; if the sequence between action segments is inconsistent with the baseline trajectory, and the sequence difference does not belong to the action connection relationship allowed by the training project, an action sequence deviation is generated. For maintenance training, if the tool retrieval record, disassembly record, installation record, tightening record, or re-inspection confirmation record is missing, an action missing deviation is generated; if the tightening segment appears before the installation segment is established, or the re-inspection segment appears before the maintenance result is submitted, an action sequence deviation is generated. For high-voltage safety training for new energy vehicles, if the corresponding confirmation record or verification record is missing for power outage confirmation, power verification confirmation, insulation detection, or protective equipment confirmation, a safety confirmation action missing deviation is generated, and subsequent operation segments are blocked from entering the pass state.

[0062] The action execution layer also processes the action timing. Action timing requirements do not use fixed thresholds without a source, but rather a reference timing range derived from vocational skills assessment rules, equipment operation specifications, vehicle maintenance manuals, or historical qualified training data. If external documents and historical qualified training data do not support this timing requirement, the action timing is not automatically used as a basis for student deviation, but only as a reference for teacher review. If the action duration does not meet the timing requirements with a source, and the resource status has been confirmed as evaluable, an action timing deviation is generated. When action missing deviation, action sequence deviation, safety confirmation action missing deviation, or action timing deviation occurs, the action validity marker is written to the action invalid state. In the action invalid state, the parameter response layer only retains parameter records and does not directly generate parameter response deviations; the parameter response layer only opens processing when the action validity marker is in the action valid state.

[0063] It should be noted that the parameter response layer is processed only after the action is effectively marked as established. The training management platform reads the parameter name, parameter value, parameter unit, parameter acquisition time, parameter source device, corresponding action segment, device calibration information, acquisition resolution, and original event number to determine whether the parameter can establish a source relationship with an established action segment. The parameter source relationship can be confirmed at least by the parameter acquisition time falling within the action segment time range, the parameter source device being consistent with the device used in the action segment, the parameter name being consistent with the training project requirements, the parameter unit being consistent with the baseline trajectory, and the original event number being traceable. If the parameter lacks a source device, parameter unit, acquisition time, or corresponding action segment, the parameter source is marked as missing and enters the evidence completion entry; in the missing parameter source state, no bias judgment is generated for the results submitted by the trainee.

[0064] Once the parameter source is established, the platform compares the acquired parameters with the parameter response benchmark in the baseline trajectory. For numerical parameters, it first confirms whether the parameter units are consistent; if the units are inconsistent, it performs unit conversion according to the equipment manual or vehicle maintenance manual, and retains the values ​​before and after conversion. Then, it compares the acquired parameters with the baseline parameters, while introducing the error coverage range formed by equipment calibration error, acquisition resolution, and measurement units. If the difference can be covered by equipment error, the parameter source marker is written to the parameter source established status, and no parameter response deviation is generated; if the difference exceeds the error coverage range, and the baseline trajectory has a clear source, a parameter response deviation is generated, and the parameter source marker is written to the parameter deviation status.

[0065] Finally, under the premise that the parameter source is marked as either "parameter source established" or "parameter deviation," the result confirmation layer is processed. The training management platform reads the content submitted by the trainee, the submission time, the submission terminal, the equipment output results, the report confirmation record, the teacher review record, the equipment retest record, and the result object to determine whether the trainee's submitted result has a corresponding parameter source and action source. If the trainee submits a test conclusion, fault diagnosis conclusion, repair result, driving training result, or safety confirmation result, but the parameter source is still marked as "parameter source missing," no result judgment deviation is generated. Instead, a result basis insufficient record is generated, and the evidence completion entry is returned. If the parameter source is established, and the trainee's submitted result is consistent with the standard result, historical sample label, equipment output result, or teacher review result in the baseline trajectory, the result confirmation mark is written to the result consistency status; if inconsistent, the result confirmation mark is written to the result judgment deviation status.

[0066] For motion segments requiring teacher review, equipment retesting, or report confirmation, even if the student's submitted result matches the conditions in the baseline trajectory, it's still necessary to determine whether review data has been generated. If no teacher review record, equipment retesting record, or report confirmation record has been generated in the result confirmation layer, the result confirmation marker is written to the result pending confirmation status and is not directly written to the final training evaluation result. If review data is generated and the review object corresponds to the current motion segment, the result confirmation marker is updated based on the review result; if the teacher review result or equipment retesting result is inconsistent with the student's submitted result, a review inconsistency record is generated and enters the result review correction entry point.

[0067] After the above processing, each training state carrier outputs a set of training bias source results, including at least action segment identifier, resource state marker, action validity marker, parameter source marker, result confirmation marker, bias source type, bias evidence event, blocking location, and subsequent processing entry point. Bias source types include resource impact, missing action bias, action sequence bias, missing safety confirmation action bias, action timing bias, missing parameter source, parameter response bias, insufficient result evidence, result judgment bias, result pending confirmation, and inconsistent verification. The blocking location indicates at which layer the current training state carrier stops propagating forward, such as resource state layer blocking, action execution layer blocking, parameter response layer blocking, or result confirmation layer blocking. The subsequent processing entry point directs to resource verification, action locking correction, evidence completion, parameter identification correction, result verification correction, or state pass write-back.

[0068] For multiple training state carriers within the same training batch, the training management platform forms a training deviation chain based on the temporal sequence and dependencies of action segments. If a previous action segment is under resource influence, subsequent data segments generated from the same resource will not directly enter the trainee skill deviation judgment; if a previous action segment is in an action invalid state, subsequent result segments that depend on the parameters generated by that action segment will not directly generate result judgment deviations; if a parameter response segment is in a parameter source missing state, the result submission segment supported by that parameter will enter the evidence completion entry point. Through the training deviation chain, the impact range of a deviation on subsequent training actions, parameters, and results can be identified, avoiding the repeated identification of the same cause as multiple independent deviations.

[0069] S4: Based on the training deviation source results and the training deviation chain formed by the training state carrier, generate hierarchical linkage correction control instructions and send the hierarchical linkage correction control instructions to the corresponding training terminal or training device.

[0070] Furthermore, after obtaining the results of training deviation sources and the training deviation chain, the training management platform generates hierarchical and linked correction control instructions based on the resource status markers, action validity markers, parameter source markers, result confirmation markers, blocking locations, and subsequent handling entry points in each training status carrier. These correction control instructions do not simply push supplementary training prompts to trainees, nor do they generate training tasks independently based on a single deviation. Instead, they determine the correction sequence based on the status transmission relationship, ensuring that the correction processing remains consistent with the deviation source identification results formed in the previous step. Specifically, before resource status is eliminated, action correction instructions are not generated; before action validity status is established, parameter identification correction instructions are not generated; before parameter sources are clarified, result judgment correction instructions are not generated; and before results form verification evidence, they are not written into the training completion status.

[0071] The hierarchical linkage correction control command includes at least the following: command number, training batch identifier, trainee identifier, training project identifier, action segment identifier, deviation source type, target execution object, control action, opening condition, locking condition, feedback field to be collected, review entry point, write-back location, and command status. The target execution object includes at least one of the following: trainee terminal, driving simulator, training vehicle data acquisition device, maintenance station terminal, testing equipment, fault diagnostic instrument, torque tool, insulation testing equipment, electrical testing equipment, teacher review terminal, and training management platform. Control actions include reopening the action segment, locking the subsequent training entry point, triggering equipment self-test, requesting equipment retesting, pushing parameter identification tasks, requiring resubmission of training conclusions, triggering teacher review, and writing to resource review records. The feedback field to be collected is used to limit the data content that must be returned after the correction is executed, so that the correction result can be reintegrated into the training status carrier, rather than simply recording "retrained" or "completed".

[0072] When the resource status in the training deviation source results is marked as a resource impact status, or when the training deviation chain shows that multiple action segments are affected by the same training vehicle, the same training equipment, the same data acquisition device, or the same maintenance station, the training management platform prioritizes generating a resource verification control instruction. This instruction is sent to the training equipment, testing instruments, maintenance station terminals, data acquisition devices, or teacher verification terminals to verify the training vehicle model, equipment binding relationship, equipment calibration status, data acquisition device connection status, station occupancy status, and verification terminal availability. The feedback fields of the resource verification control instruction include the resource verification object, verification time, verification result, equipment status record, calibration status record, connection status record, and teacher verification record. If the feedback result shows that the resource cannot support the current training evaluation, the corresponding training segment is excluded from the trainee skill evaluation sample, and the resource issue is written back to the training resource status record. If the feedback result shows that the resource can support the evaluation, the resource status mark of the training status carrier is updated to a resource evaluable status, and the process returns to S3 to re-execute the deviation source stripping processing of the action execution layer, parameter response layer, and result confirmation layer.

[0073] When a resource status is marked as evaluable and an action validity is marked as invalid, the training management platform generates an action lock correction control command. The action lock correction control command determines the control object and opening method based on action missing deviation, action sequence deviation, safety confirmation action missing deviation, or action timing deviation. For driving training, the control command can be sent to the driving simulator or student terminal to reload the corresponding starting, braking, steering, shifting, reversing, or parking training scenarios, and lock subsequent training segments that have a sequential dependency on this action segment; only after the current action segment re-forms the action trigger event, vehicle control data, and action completion record can subsequent action segments be written to the pass status. For automotive repair training, the control command can be sent to the repair bay terminal, RFID tool identification device, or torque tool to reopen tool selection, component disassembly, component installation, torque tightening, or re-inspection confirmation actions, and requires the return of tool retrieval records, disassembly / installation confirmation records, torque upload records, and repair terminal confirmation records. For high-voltage safety training of new energy vehicles, control commands are sent to high-voltage safety training equipment, voltage testing equipment, insulation testing equipment, or teacher verification terminals to reopen power outage confirmation, voltage testing confirmation, protective equipment confirmation, or insulation testing actions; before the above safety confirmation actions form confirmation records and verification records, subsequent high-voltage system disassembly, testing, or maintenance actions shall not be recorded as passed.

[0074] When a resource status is marked as "evaluable" and an action as "established," but the parameter source is marked as "missing source," the training management platform generates an evidence completion control instruction. This instruction is used to supplement the parameter source device, parameter acquisition time, parameter unit, original event number, equipment retest record, or instructor review record. This instruction does not require trainees to immediately retrain; instead, it first determines whether the parameter has a traceable source. For testing training, the evidence completion control instruction can trigger the testing equipment to re-upload the testing items, testing parameters, testing time, and testing object. For maintenance training, it can trigger torque tools, diagnostic instruments, or maintenance terminals to supplement parameter source records. For driving training, it can require the simulated driving equipment or onboard data acquisition device to supplement the action segment identifier corresponding to the vehicle status parameters. If the supplemented parameter can establish a source relationship with an already established action segment, the parameter source mark is updated to "established source," and the process returns to S3 for further result confirmation. If a source relationship still cannot be established after supplementation, the action segment is marked as "invalid evaluation segment" and is not used as a direct basis for the trainee's pass or fail.

[0075] When the resource status is marked as "evaluable," the action is marked as "action established," and the parameter source is marked as "parameter deviation," the training management platform generates a parameter identification and correction control instruction. This instruction is sent to the trainee terminal, testing equipment, fault diagnostic instrument, maintenance station terminal, or instructor review terminal to guide trainees to re-read parameters, confirm parameter sources, select reasons for parameter changes, and submit conclusions. The instruction calls upon the baseline trajectory, historical qualified parameter trajectory, and historical erroneous parameter trajectory corresponding to the current training project, vehicle model, equipment, and action segment. However, it does not directly use historical trajectories as fixed scoring rules; instead, it serves as reference data for trainees' parameter identification training and instructor review. The feedback fields of the parameter identification and correction control instruction include the reread parameter value, parameter source equipment, parameter unit, trainee's explanation of the reason, trainee's resubmitted conclusion, equipment retest results, and instructor review results. If the re-acquired parameter response falls within the error coverage range allowed by the baseline trajectory, or if the teacher verifies that the original parameter deviation was caused by equipment error, acquisition conditions, or changes in training resources, then the deviation will not be recorded as a student's skill deficiency. If the parameter deviation still exists and cannot be explained by equipment error or resource problems, then the parameter deviation status will be retained, and parameter identification and correction control instructions will continue to be generated or transferred to the teacher for review.

[0076] When the resource status is marked as "resource evaluable," the action as "action established," the parameter source as "parameter source established," and the result confirmation as "result judgment deviation," the training management platform generates a result review and correction control instruction. This instruction requires trainees to resubmit test results, fault diagnosis results, repair results, driving training results, or safety confirmation results, and simultaneously triggers teacher review, equipment retesting, or report confirmation. This instruction must be bound to the data source of the corresponding parameter response layer, enabling trainees to make judgments based on already confirmed parameter data when resubmitting results. If the trainee's resubmitted result matches the standard result in the baseline trajectory, the equipment output result, or the teacher's review result, the result confirmation mark is updated to "result consistency." If it still doesn't match, the result judgment deviation status is retained, and this deviation is used as the basis for generating subsequent personalized training content.

[0077] When the result confirmation mark is in the pending confirmation state or the review inconsistency state, the training management platform generates a review completion control instruction. This instruction is sent to the teacher review terminal, equipment retest terminal, or report confirmation terminal to supplement the teacher review record, equipment retest record, or report confirmation record. The feedback fields of the review completion control instruction include the review object, review method, review conclusion, review time, and reviewer identifier or retest device identifier. Before review data is generated, the corresponding training segment cannot be written to the final evaluation status; after review data is generated, the result confirmation layer updates the result confirmation mark according to the correspondence between the review object and the current action segment. If the review object cannot correspond to the current action segment, the review data is written to the pending review record and is not directly used to determine whether the trainee's result is correct.

[0078] It should be noted that when multiple blocking points exist within the same training deviation chain, the training management platform generates corrective control instructions according to the state transmission sequence. If both resource impact and action missing deviations exist simultaneously in the same training batch, a resource review control instruction is generated first; only after the resource review is completed and the resource is confirmed to be evaluable is an action lock corrective control instruction generated. If both action invalidity and parameter deviation records exist in the same action segment, action lock correction is executed first; after the action is re-established, it is then determined whether the parameter deviation still exists. If both parameter source missing and result judgment deviation exist in the same result submission, an evidence completion control instruction is executed first; after the parameter source is established, it is then determined whether result review correction is needed. This sequential control avoids different corrective tasks being triggered independently, ensuring consistency in the hierarchical relationship between corrective control and deviation source decoupling.

[0079] After generating hierarchical linkage correction control instructions, the training management platform writes the control instructions into the correction task queue and distributes them to the corresponding target execution objects. Each correction control instruction has states such as pending execution, executing, feedback pending, feedback received, requiring review, and closed. After receiving the instruction, the target execution object executes the correction action according to the action segment identifier, opening conditions, and feedback fields to be collected in the instruction, and returns the correction process data to the training management platform. After receiving the feedback data, the training management platform does not directly determine that the correction is complete, but instead rewrites the feedback data into the training event stream as the basis for subsequent S5 reconstruction of the correction training state carrier and execution state write-back.

[0080] S5: Collect correction feedback data, regenerate the correction training state carrier, compare and correlate the correction training state carrier with the original training state carrier, and write back the resource state marker, action validity marker, parameter source marker and result confirmation marker layer by layer to update the trainee training state and training resource state.

[0081] Furthermore, after the hierarchical linkage correction control command is issued, the training management platform continues to collect vehicle data, training equipment data, maintenance station data, testing instrument data, student terminal data, and teacher review terminal data during the student's execution of the correction task, generating a correction training event stream. The correction training event stream uses the same data structure as the initial training event stream, including training batch identifier, student identifier, training item identifier, action segment identifier, data source identifier, equipment identifier, event time, event type, event field, event value, and data status flag. The difference lies in that the correction training event stream also includes the corresponding correction control command number, original deviation source type, original state blocking position, and correction execution round, used to establish a correspondence between the correction feedback data and the original training deviation source results.

[0082] The data collection targets for corrective feedback are determined based on the type of corrective control instruction. When executing a resource verification control instruction, the following data are collected: training vehicle model verification results, equipment binding results, equipment calibration status, data acquisition device connectivity status, maintenance bay occupancy status, and teacher verification terminal status. When executing an action lock corrective control instruction, the following data are collected: re-executed action trigger event, action completion record, vehicle control data, tool usage record, maintenance confirmation record, safety confirmation record, and training terminal confirmation record. When executing an evidence completion control instruction, the following data are collected: supplementary uploaded parameter source equipment, parameter acquisition time, parameter unit, original event number, equipment retest record, and teacher verification record. When executing a parameter identification corrective control instruction, the following data are collected: reread parameter values, parameter source description, student reason selection, equipment retest results, and teacher verification results. When executing a result verification corrective control instruction or a verification completion control instruction, the following data are collected: student resubmitted results, teacher verification conclusion, equipment retest conclusion, and report confirmation record.

[0083] After receiving the correction feedback data, the training management platform performs source marking, time synchronization, and format unification according to step S1, and then reconstructs the correction training state carrier according to step S2. The correction training state carrier includes at least the correction carrier identifier, original action segment identifier, correction action segment identifier, correction instruction number, resource state layer, action execution layer, parameter response layer, result confirmation layer, and state transmission layer. The correction training state carrier does not replace the original training state carrier, but rather establishes a relationship with it to determine whether the original state blocking point has been removed and whether a new state blocking point has appeared.

[0084] Before writing back the status, the training management platform first compares and correlates the corrective training status carrier with the original training status carrier. This correlation comparison includes comparisons of action segment correspondences, training resource correspondences, parameter source correspondences, and result verification correspondences. Action segment correspondence comparisons confirm whether the corrective action is executed for the original deviated action segment; training resource correspondence comparisons confirm whether the training vehicles, equipment, testing instruments, or maintenance stations used during the correction process are consistent with the original training segment, or whether they are alternative resources confirmed through resource verification; parameter source correspondence comparisons confirm whether the corrected parameters originate from an established action segment; and result verification correspondence comparisons confirm whether teacher verification, equipment retesting, or report confirmation points to the current corrective action segment. If the corrective feedback data cannot establish a correspondence with the original deviated action segment, the corrective feedback is written to the pending correction record and is not directly used to update the trainee's training status.

[0085] The status write-back is performed in the following order: resource status marker, action validity marker, parameter source marker, and result confirmation marker. When the resource status marker of the original training status carrier is in a resource-affected state, and the resource review result in the corrective training status carrier shows that the training vehicle, training equipment, testing instrument, acquisition device, maintenance station, or teacher review terminal can support training evaluation, the training management platform updates the resource status marker corresponding to the original action segment to a resource-evaluable state, and returns the action segment to step S3 to re-execute the deviation source stripping process of the action execution layer, parameter response layer, and result confirmation layer. If the resource review result shows that the training resources still cannot support training evaluation, the action segment is marked as a resource-unevaluable segment and excluded from the trainee skill evaluation sample. At the same time, the reason for the resource's unevaluability, the equipment involved, the vehicle involved, the workstation involved, and the review result are written into the training resource status record.

[0086] When the original training state carrier's action is marked as "action invalid," and the correction training state carrier shows that the corresponding action segment has re-formed the action trigger event, action execution record, and action completion record, and the action sequence can establish a correspondence with the baseline trajectory, the training management platform will update the action valid mark to "action valid" and open the parameter response layer for processing. If, after correction, there is still a lack of action completion record, the action sequence still cannot correspond to the baseline trajectory, or the safety confirmation action has not yet formed a confirmation record and review record, then the "action invalid" state will be retained, and action lock correction control commands will continue to be generated. For high-voltage safety training of new energy vehicles, before the states corresponding to power outage confirmation, power verification confirmation, protective equipment confirmation, or insulation detection are updated to "action valid," subsequent high-voltage system operation segments must not be written to the "pass" state.

[0087] When the parameter source marker in the original training state carrier is marked as missing, and the correction training state carrier supplements the parameter source device, parameter acquisition time, parameter unit, corresponding action segment, and original event number, the training management platform updates the parameter source marker to the established parameter source state and opens the result confirmation layer for processing. If the parameter source still cannot be confirmed after supplementation, or the parameter cannot be correlated with the established action segment, the missing parameter source state is retained, and evidence supplementation control instructions continue to be generated. If the correction training state carrier shows that the parameter source is established, but there is still a difference between the re-acquired parameters and the parameter response benchmark in the baseline trajectory, and this difference cannot be explained by device calibration error, acquisition resolution, or measurement unit conversion error, the parameter source marker is updated to the parameter deviation state, and parameter identification and correction control instructions are generated.

[0088] When the original training status carrier's result confirmation marker is in a result judgment deviation state, a result pending confirmation state, or a review inconsistency state, and new student-submitted results, device output results, teacher review results, or report confirmation records are generated in the corrected training status carrier, the training management platform will re-align the corrected result confirmation layer with the baseline trajectory, device output results, and review results. If the student's resubmitted result is consistent with the standard result, device output result, or teacher review result in the baseline trajectory, and the action segment requiring review has formed review evidence, then the result confirmation marker will be updated to a result consistency state. If the student's resubmitted result is still inconsistent with the standard result, device output result, or teacher review result, then the result judgment deviation state will be retained, and result review correction control instructions will continue to be generated. If review data has not yet been generated, then the result pending confirmation state will be retained, and review completion control instructions will continue to be generated.

[0089] It should be noted that after completing the state write-back at each level, the training management platform determines whether a training item can enter the state update stage based on the state transmission layer. Only when the resource status of the corresponding action segment is marked as "resource evaluable," the action as "action established," the parameter source as "parameter source established," or parameter deviation handling has been completed, and the result confirmation is marked as "result consistent," is the action segment written into the "passed" or "pending consolidation" state. If any level still has a "resource impact," "action invalid," "parameter source missing," "parameter deviation," "result judgment deviation," or "result pending confirmation" state, the action segment is not written into the "passed" state but retains the corresponding handling entry point. This ensures that training state updates must inherit the layer-by-layer processing results of resources, actions, parameters, and results, avoiding the direct change of training state based solely on manual confirmation or a single result submission for error correction feedback.

[0090] For multiple action segments within the same training batch, the training management platform performs a chained write-back based on the training status carrier sequence and training deviation chain. If a preceding action segment is updated from an invalid to an valid state after correction, subsequent parameter segments and result segments dependent on that action segment are reopened for evaluation. If the preceding action segment remains in an invalid state, subsequent dependent segments remain in a temporarily unevaluated state. If the review result of a resource shows that the same training device or acquisition device is unusable for evaluation in multiple consecutive action segments, the platform marks the affected segments as unevaluable resources and synchronizes the resource status to the training resource status record. If a missing parameter source is corrected after completion, the result submission segment supported by that parameter re-enters the result confirmation process. Through chained write-back, the impact on subsequent actions, parameters, and results can be avoided by updating only a single action segment.

[0091] After completing the status write-back for an individual trainee, the training management platform further summarizes the status interruption points formed under the same training project, vehicle model, equipment, and action segment. Status interruption points include resource-affected status, action invalid status, missing parameter source status, parameter deviation status, result judgment deviation status, and result pending confirmation status. When multiple training batches form similar status interruption points under the same conditions, and the teacher's review record shows that the interruption point was not caused by a single trainee's accidental operation, the platform generates a baseline trajectory update candidate record or a data acquisition configuration update candidate record. The baseline trajectory update candidate record includes the training project, vehicle model, equipment, action segment, original baseline trajectory, status interruption point type, involved training batches, teacher review conclusion, and suggested update content; the data acquisition configuration update candidate record includes missing fields, involved equipment, involved interfaces, missing occurrence scenario, and suggested supplementary data acquisition fields.

[0092] The baseline trajectory update candidate records and data acquisition configuration update candidate records do not take effect automatically. They require teacher review or confirmation through the training quality management process before updating the corresponding baseline trajectory, motion segment slice conditions, state transfer layer initialization conditions, or correction control instruction feedback fields. After the update takes effect, the platform records the previous version, the updated version, the reason for the update, the effective time, and the applicable scope. For historical training batches that have already been evaluated, the training results will not be automatically changed due to the baseline trajectory update; for training batches that have not yet completed the correction loop, they can be re-executed based on the updated baseline trajectory.

[0093] Example 2, one embodiment of the present invention, provides a closed-loop optimization method for automotive skills training. To verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0094] First, high-voltage safety training for new energy vehicles, fault code diagnosis training, braking system testing training, and torque tightening re-inspection training at a certain automotive vocational skills training center were used as experimental subjects. For each type of training, two sets of comparisons were set up: traditional training recording methods and the method of this invention. Traditional training recording methods mainly used on-site teacher recordings, one-way equipment output results, and manual review results after training as evaluation criteria. The method of this invention generates a training batch identifier through the training management platform at the start of training and connects to the on-board data interface, fault diagnostic instrument, testing bench, torque tool, high-voltage safety training equipment, repair station terminal, student terminal, and teacher review terminal. During training, the platform uniformly collects data on vehicle speed, pedal status, diagnostic instrument connection status, fault code reading results, test parameters, torque upload records, power-off confirmation records, power-on confirmation records, student-submitted conclusions, and teacher review conclusions, and completes training batch binding, source marking, time synchronization, and field format standardization to generate an automotive skills training event stream.

[0095] After forming the training event flow, the platform segments the training process within the same training batch into action segments based on action trigger signals, resource status changes, vehicle status changes, tool usage changes, detection parameter output signals, and result submission signals. For example, in high-voltage safety training, power outage confirmation, power verification confirmation, insulation testing, and teacher review are each segmented into corresponding action segments; in fault code diagnosis training, diagnostic tool connection, fault code reading, data stream analysis, and result submission are segmented into corresponding action segments; and in torque tightening re-inspection training, tool retrieval, component installation, torque tightening, and re-inspection confirmation are segmented into corresponding action segments. A state-transfer type training state carrier is constructed for each action segment, recording the resource status layer, action execution layer, parameter response layer, result confirmation layer, and state transfer layer. Subsequently, the training state carrier is matched with the corresponding automotive skill training baseline trajectory, and a correlation-based deviation source stripping is performed according to the progressive relationship between resource status, action validity status, parameter source status, and result confirmation status. For segments with unconfirmed resource status, the student's skill is not directly considered insufficient; for segments with invalid action segments, parameter deviation is not directly judged; for segments with missing parameter sources, the student's result is not directly judged as incorrect. Finally, based on the source of deviation, layered linkage control instructions are generated, including resource verification, action locking correction, evidence completion, parameter identification correction, and result verification correction. After correction, feedback data is re-collected, and the training status and resource status are written back layer by layer.

[0096] Table 1 Experimental Data

[0097] As shown in Table 1, among the four types of automotive skills training subjects, the method of this invention demonstrates significant improvements over traditional training recording methods in terms of multi-source event completeness correlation rate, action segment recognition completeness rate, resource influence segment extraction accuracy rate, trainee skill deviation recognition accuracy rate, and one-time closed-loop completion rate of corrective control. Traditional methods primarily rely on single-device output, teacher on-site observation, and post-training result recording, which easily leads to inaccurate correspondence between vehicle data, equipment data, workstation data, and verification data. For example, in high-voltage power outage testing training, the multi-source event completeness correlation rate of the traditional method is 72.5%, while the method of this invention reaches 96.3%, indicating that by binding training batches, synchronizing time, and constructing event streams, the data source relationships within the same training batch can be more completely restored.

[0098] Furthermore, traditional methods achieve an accuracy rate of only 61.9% to 67.5% in separating resource-related factors, indicating a difficulty in distinguishing the relationship between equipment status, workstation occupancy, missing data collection, and actual skill deviations in trainees. The method of this invention, through a state-transfer type training state carrier, processes the resource state layer before the action execution layer. It does not directly generate conclusions about trainee skill deviations when the resource state is not confirmed. Therefore, the accuracy rate of separating resource-related factors is increased to 90.7% to 94.1%, while the rate of mistakenly including resource issues in trainee insufficiency is reduced to 3.1% to 4.4%. This data demonstrates that this invention does not simply increase data collection, but rather solves the technical problem in traditional evaluations where "equipment problems, data collection problems, and resource conflicts are misjudged as trainee insufficiency" through state transfer and deviation source separation mechanisms.

[0099] Furthermore, the method of this invention achieves a one-loop completion rate of 84.4% to 90.7% for error correction control, significantly higher than the 52.9% to 58.7% of the traditional method. This is because the invention does not simply arrange repetitive training based on deduction items, but rather generates error correction control instructions for resource review, action locking correction, evidence completion, parameter identification correction, and result review based on different error sources such as resource impact, missing actions, missing parameter sources, parameter response deviation, and result judgment deviation. This ensures a clear correspondence between error correction tasks and error sources. The average number of error correction rounds is reduced from 2.5 to 2.9 times in the traditional method to 1.4 to 1.6 times, and the time spent on manual review by teachers is reduced from 15.2 to 18.0 minutes per person to 7.1 to 8.7 minutes per person. This indicates that after the error correction feedback data re-enters the training state carrier and is written back layer by layer, it can reduce the workload of ineffective repetitive training and manual evidence searching. The final training pass rate increased to 91.2% to 95.0%, indicating that the present invention achieved a synergistic effect among training data eventification, action segment-level modeling, progressive stripping of bias sources, hierarchical correction control, and state write-back closed loop in the embodiments. Compared with the prior art, it has higher evaluation accuracy, more targeted correction, and continuous optimization capability in the training process.

[0100] Example 3, one embodiment of the present invention, provides a closed-loop optimization system for automotive skills training, including a data event processing module, a state carrier and deviation stripping module, and a deviation correction control and closed-loop write-back module.

[0101] The data event processing module collects on-board data, training equipment data, testing instrument data, repair station data, student terminal data, and teacher review data generated at the automotive skills training site. It then binds training batches, marks sources, synchronizes time, and standardizes field formats to form an automotive skills training event flow. The state carrier and deviation stripping module slices action segments based on the automotive skills training event flow, constructs state-transfer training state carriers corresponding to each action segment, maps these state-transfer training state carriers to the automotive skills training baseline trajectory, and performs associative deviation source stripping processing in the order of resource state layer, action execution layer, parameter response layer, and result confirmation layer, generating training deviation source results and training deviation chains. The deviation correction control and closed-loop write-back module generates layered linkage deviation correction control commands based on the training deviation source results and training deviation chains, and sends the commands to the corresponding training terminals or training equipment. After collecting deviation correction feedback data, it regenerates the deviation correction training state carrier and writes back resource state markers, action validity markers, parameter source markers, and result confirmation markers layer by layer, updating the student training status and training resource status.

[0102] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, 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 of the various embodiments of this 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.

[0103] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0104] More specific examples (a non-exhaustive list) of computer-readable media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0105] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc. It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

[0106] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A closed-loop optimization method for automotive skills training, characterized in that, include: Acquire vehicle data, training equipment data, testing instrument data, repair station data, student terminal data, and teacher review data generated at the automotive skills training site, bind training batches, mark sources, synchronize time, and unify field formats to generate automotive skills training event streams; Based on the action triggering signals, resource state changes, vehicle state changes, tool usage changes, detection parameter output signals, and result submission signals in the automotive skills training event flow, the training process of the same training batch is sliced ​​into action segments to construct a state-transfer type training state carrier corresponding to each action segment. The state-transfer type training state carrier is mapped to the benchmark trajectory of automotive skill training. Correlational deviation source stripping is performed in the order of resource state layer, action execution layer, parameter response layer and result confirmation layer to generate training deviation source results. Based on the training deviation chain formed by the training deviation source results and the training state carrier, a hierarchical linkage correction control command is generated and sent to the corresponding training terminal or training device. Collect correction feedback data, regenerate the correction training state carrier, compare and correlate the correction training state carrier with the original training state carrier, and write back the resource state marker, action validity marker, parameter source marker and result confirmation marker layer by layer to update the trainee training state and training resource state.

2. The closed-loop optimization method for automotive skills training as described in claim 1, characterized in that: The process of generating an automotive skills training event stream involves writing the following information into the vehicle data, training equipment data, testing instrument data, repair bay data, student terminal data, and teacher review data: student identifier, training project identifier, training batch identifier, data source identifier, device identifier, original timestamp, server receiving time, data field type, and data field value. Data from different data sources is synchronized using a unified clock source, and time correction is performed based on the offset between the device's local time and the time received by the server. The data that has completed training batch binding, source tagging, time synchronization, and field format unification is used to generate an automotive skills training event stream in the order of event time. The automotive skills training event stream includes resource status events, action trigger events, parameter response events, result submission events, and review confirmation events.

3. The closed-loop optimization method for automotive skills training as described in claim 2, characterized in that: The process of slicing action segments during the training process of the same training batch includes using action triggering events, resource status change events, vehicle status change events, tool usage change events, detection parameter output events, or result submission events as the basis for segment boundary identification of action segments. When driving control actions, fault diagnosis actions, maintenance operation actions, testing actions, high-voltage safety confirmation actions, or result confirmation actions are generated in the automotive skills training event flow, the action segments are determined according to the start time, end time, and relationship with adjacent events of the corresponding events. When the same event can correspond to multiple action segments, the attribution is determined based on the training batch identifier, device identifier, event time, action object, and the relationship between adjacent action segments; If the attribution still cannot be determined, it will be added to the event set to be assigned and will not be used for training bias judgment.

4. The closed-loop optimization method for automotive skills training as described in claim 3, characterized in that: The state-transfer type training state carrier includes a resource state layer, an action execution layer, a parameter response layer, a result confirmation layer, and a state transfer layer. The resource status layer is used to record the status of training vehicles, training equipment, testing instruments, maintenance stations, data acquisition devices, and teacher review terminals during the execution of action segments; The action execution layer is used to record action triggering events, action execution sequence, tool usage records, vehicle control actions, detection and connection actions, maintenance and disassembly actions, safety confirmation actions, and action completion records. The parameter response layer is used to record parameter name, parameter value, parameter unit, parameter acquisition time, parameter source device, parameter corresponding action segment, and device calibration information; The result confirmation layer is used to record the results submitted by students, the results output by the devices, the results reviewed by teachers, the results of device retesting, and the report confirmation records. The state transmission layer is used to record resource state markers, action validity markers, parameter source markers, result confirmation markers, and processing entry markers, so that the action validity marker inherits the resource state marker, the parameter source marker inherits the action validity marker, and the result confirmation marker inherits the parameter source marker.

5. The closed-loop optimization method for automotive skills training as described in claim 4, characterized in that: The automotive skills training baseline trajectory is generated from the training syllabus, vocational skills assessment rules, vehicle repair manual, training equipment instruction manual, equipment calibration records, and historical qualified training status data. When the training syllabus, vocational skills assessment rules, vehicle maintenance manual, or training equipment instruction manual contains operation sequence, testing parameters, completion conditions, or verification requirements, these are converted into action sequence benchmarks, parameter response benchmarks, result confirmation benchmarks, and verification formation benchmarks in the automotive skills training benchmark trajectory. When the reference trajectory involves numerical detection data, the equipment calibration error, acquisition resolution, and measurement unit are written into the automotive skills training reference trajectory; When external documents do not provide a complete baseline, a reference baseline trajectory is extracted from the historical qualified training status carriers of the same training project, the same vehicle type, the same training equipment type, the same training stage, and which have been reviewed and approved by the teacher.

6. The closed-loop optimization method for automotive skills training as described in claim 5, characterized in that: The execution-related deviation source stripping process includes generating a resource impact status when the resource status layer shows that the training vehicle, training equipment, testing instruments, maintenance station, data acquisition device, or teacher review terminal cannot support the current training evaluation, thereby preventing the action execution layer, parameter response layer, and result confirmation layer from generating student skill deviation conclusions. When the resource status is marked as a resource that can be evaluated, the action execution layer is processed. If the action segment lacks a corresponding operation record, the action sequence is inconsistent with the vehicle skill training baseline trajectory, or the safety confirmation action lacks a confirmation record, then an action missing deviation, action sequence deviation, or safety confirmation action missing deviation is generated. When an action is validly marked as an established action, the parameter response layer is processed. If the parameter lacks a source device, acquisition time, measurement unit, or corresponding action segment, a parameter source missing result is generated. If the parameter difference exceeds the range determined by the baseline trajectory and device error, a parameter response deviation is generated. When the parameter source is marked as being in the parameter source established state, the result confirmation layer is processed. If the result submitted by the student is inconsistent with the baseline trajectory, the device output result, or the teacher's review result, a result judgment deviation is generated. If review is required but no review data has been generated, a result pending confirmation state is generated.

7. The closed-loop optimization method for automotive skills training as described in claim 6, characterized in that: The process of associating and comparing the correction training state carrier with the original training state carrier, and writing back the resource state marker, action validity marker, parameter source marker, and result confirmation marker layer by layer, includes generating a resource review control command when the training deviation source result is a resource impact state, which is used to trigger the status review of the training vehicle, training equipment, testing instrument, maintenance station, acquisition device, or teacher review terminal. When the training bias source is a missing action bias, a missing action sequence bias, or a missing safety confirmation action bias, an action lock correction control command is generated. When the training bias source is a missing parameter source result or a parameter response bias, generate an evidence completion control command or a parameter identification correction control command. When the source of training bias is result judgment bias or result pending confirmation, generate result review and correction control command or review and completion control command. After receiving the correction feedback data, the correction training state carrier is regenerated, and the state is written back in the order of resource state marker, action valid marker, parameter source marker, and result confirmation marker.

8. A system employing the closed-loop optimization method for automotive skills training as described in any one of claims 1 to 7, characterized in that: It includes a data event processing module, a state carrier and deviation stripping module, and a deviation correction control and closed-loop write-back module; The data event processing module is used to collect vehicle data, training equipment data, testing instrument data, repair station data, student terminal data, and teacher review data generated at the automotive skills training site, and to bind training batches, mark sources, synchronize time, and unify field formats to form an automotive skills training event stream. The state carrier and deviation stripping module is used to slice action segments according to the automotive skills training event flow, construct state-transfer type training state carriers corresponding to each action segment, correspond the state-transfer type training state carriers to the automotive skills training baseline trajectory, and perform correlation-based deviation source stripping processing in the order of resource state layer, action execution layer, parameter response layer and result confirmation layer to generate training deviation source results and training deviation chain. The error correction control and closed-loop write-back module is used to generate hierarchical linkage error correction control instructions based on the training error source results and training error chain, and send the instructions to the corresponding training terminal or training device; after collecting error correction feedback data, it regenerates the error correction training status carrier, and writes back the resource status marker, action validity marker, parameter source marker and result confirmation marker layer by layer, and updates the trainee training status and training resource status.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the closed-loop optimization method for automotive skills training as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the closed-loop optimization method for automobile skills training as described in any one of claims 1 to 7.

Images