Forklift driver training management platform and method based on data visualization analysis

Abstract

The application discloses a forklift driver training management platform and method based on data visualization analysis, and particularly relates to the technical field of forklift driving training management, which comprises collecting operation data, position data and subject result data of trainees in simulation training, performing action perception based on the operation data and the position data, determining corresponding action states at each time, and performing alignment and combination of the action states, the operation data, the position data and the subject result data according to the same trainee, the same subject and time sequence to generate a simulation training record table; action perception is performed on operation data in simulation training, and a deviation chain is extracted, then the deviation chain is mapped to real vehicle training records to count subsequent alarm, parking protection and coach takeover results to form a mistake amplification table, and real vehicle training access and teaching sequence are rearranged according to the mistake amplification table in combination with real vehicle training states.

Description

Technical Field

[0001] This invention relates to the field of forklift driver training management technology, and more specifically, to a forklift driver training management platform and method based on data visualization analysis. Background Technology

[0002] In forklift driver training, existing technologies mostly focus on improving teaching standardization, timely training feedback, and practical training safety. In practice, simulation training typically collects operational information such as direction, pedals, control handles, and line of sight coordination to identify and prompt non-standard actions. In real vehicle training, the trainee's performance is recorded by combining vehicle position, operating status, site boundaries, and voice alarm results. Simulation scores, real vehicle scores, error records, and training files are then compiled into a management platform for subsequent teaching evaluation and training arrangements. Taking the scenario where training institutions continuously organize students to complete simulated driving, field operation and road driving training as an example, the training process requires consistent assessment logic in the previous and subsequent stages, and also requires the real vehicle stage to maintain continuous training under limited site, existing vehicle configuration and real-time safety control conditions, and cannot rely on repeated manual intervention by the instructor for a long time. In this situation, the existing approach is prone to the following phenomenon: some trainees perform normally in the simulation phase, but frequently trigger voice warnings, parking protection, or instructor intervention after entering the real vehicle phase, resulting in a clear disconnect between the results before and after. Further analysis reveals that the problem is not that the platform does not collect action information, nor that there are no scoring methods for each phase, but that the existing practices mostly judge based on the real-time performance of each training phase. They cannot identify in advance that some seemingly minor operational deviations in the simulation phase will evolve into higher-risk errors after entering the real vehicle training due to vehicle inertia, load changes, site boundaries, and continuous work rhythm. As a result, the simulation training results are difficult to truly use for access control and sequence adjustment of real vehicle training. The technical problem this application aims to solve is: how to identify the subsequent amplification of operational deviations in simulation training during real-vehicle training in the process of forklift driver training based on data visualization analysis, and adjust the admission criteria and teaching arrangements for real-vehicle training accordingly. Summary of the Invention

[0003] To overcome the aforementioned deficiencies in the prior art, embodiments of the present invention provide a forklift driver training management platform and method based on data visualization analysis. By performing action perception and extracting deviation chains from the operational data in simulated training, the deviation chains are then mapped to the actual vehicle training records to statistically analyze subsequent alarms, parking protection, and instructor takeover results to form an error amplification table. Based on the error amplification table and the actual vehicle training status, the actual vehicle training admission and teaching sequence are rearranged to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a forklift driver training management method based on data visualization analysis, comprising: S1. Collect the student's manipulation data, position data and subject result data in the simulation training. Based on the manipulation data and position data, perform motion perception, determine the corresponding motion state at each moment, and perform alignment and combination of motion state, manipulation data, position data and subject result data according to the same student, the same subject and time sequence to generate a simulation training record table. S2. Based on the simulation training record table, extract the deviation actions, the operation segments where the deviations are located, and the sequential relationship between adjacent deviations in each subject according to time sequence, and connect the first and last deviation actions in sequence to generate a deviation linked list. S3. Retrieve the actual vehicle training records corresponding to the subject, and according to the positional correspondence of the same operation segment in the same subject, substitute each deviation action in the deviation chain into the operation segment corresponding to the actual vehicle training record, and count the alarm, parking protection and instructor takeover results caused by each deviation action in the subsequent operation segment, and generate a deviation consequence correspondence table. S4. Based on the deviation consequence correspondence table, accumulate the corresponding number of alarms, number of parking protections, and number of instructor takeovers for each deviation chain, and arrange each deviation chain from high to low according to the cumulative results to generate an error amplification table.

[0005] In a preferred embodiment, it further includes: S5. For the error amplification table corresponding to trainees waiting for actual vehicle training, and in combination with the motion perception results in actual vehicle training, rearrange the order of entry into actual vehicle training subjects and the order of operation practice within subjects according to the arrangement order of each deviation chain in the error amplification table. Write the rearranged actual vehicle training results and the error amplification table into the visual training archive to generate training management results.

[0006] In a preferred embodiment, S1 includes: S11. Perform simultaneous splicing of the steering wheel angle, pedal travel and control handle displacement in the control data according to the sampling time, and perform position correspondence on the vehicle position, vehicle orientation and fork position in the position data according to the same time, and generate a time action segment table. S12. Based on the action segment table, extract action change groups according to the changes in steering wheel angle, pedal travel, control handle displacement, vehicle body displacement and fork displacement at adjacent time points, and merge action change groups with the same direction of change and consecutive positions in multiple consecutive time points to generate an action segment table. S13. Based on the action segment table, determine the starting state, turning state, approach state, pick-up / placement state or adjustment state corresponding to each action segment according to the correspondence between the changes in steering wheel angle, pedal travel, control handle displacement and the changes in vehicle body displacement and fork displacement in each action segment, and write them into each time point in chronological order to generate the action state.

[0007] In a preferred embodiment, S2 includes: S21. Read the simulation training record table, determine the deviation actions corresponding to each moment according to the positional deviation, timing deviation and state switching deviation between the action status at each moment within the same subject and the corresponding subject result data, and generate a deviation action table according to the operation segment corresponding to the moment where each deviation action is located. S22. For the deviation action table, extract the time interval between each deviation action, the change in the operation segment, and the direction of action status continuation according to the time sequence. For deviation action pairs where there are no other deviation actions inserted between the end time of the previous deviation action and the start time of the next deviation action, and the operation segment where the next deviation action is located is a continuation of the operation segment where the previous deviation action is located or the next sequential operation segment, determine them as continuation deviation pairs and generate a deviation continuation table. S23. Based on the deviation connection table, take the deviation action without a preceding deviation pair as the first deviation action of the chain, and connect the subsequent deviation actions in sequence along the connection order of the deviation pairs until the deviation action with no subsequent deviation pair is reached, thus generating a deviation chain list.

[0008] In a preferred embodiment, S3 includes: S31. Retrieve the operation segment records, vehicle status records, alarm records, parking protection records, and instructor takeover records from the actual vehicle training records. According to the rule that the operation segment numbers are consistent, the start and end positions of the operation segments are corresponding, and the direction of movement of the operation segments are consistent in the same subject, each deviation action in the deviation chain list is mapped to the target operation segment in the actual vehicle training records. Then, the results are tracked along the sequential operation segments after the target operation segment to generate a deviation substitution table. S32. Based on the deviation substitution table, extract the operation segment number sequence, vehicle status sequence, alarm number sequence, parking protection number sequence, and instructor takeover number sequence corresponding to each deviation action. Determine the records with consecutive operation segment numbers, vehicle status change direction consistent with deviation action direction, and alarm number sequence, parking protection sequence, and instructor takeover number sequence corresponding to each other in the order of operation segments as valid result chains, and generate a deviation result chain list.

[0009] In a preferred embodiment, S3 further includes: S33. Based on the deviation result chain list, count the number of alarms, the number of parking protections, and the number of instructor takeovers for the same deviation action, and write them into the result location group according to the operation segment in which the alarm occurs, the operation segment in which the parking protection occurs, and the operation segment in which the instructor takeover occurs; merge the deviation result chains with the same result location group and the same number of times and write them into the deviation consequence correspondence table; split the deviation result chains with different result location groups or different numbers of times according to the operation segment number order and write them into the deviation consequence correspondence table.

[0010] In a preferred embodiment, S4 includes: S41. Retrieve the deviation consequences correspondence table, collect the corresponding alarm counts, parking protection counts, and instructor takeover counts according to the same deviation chain, and sequentially splice the alarm occurrence operation segment, parking protection occurrence operation segment, and instructor takeover occurrence operation segment corresponding to each deviation chain to generate a deviation chain result table. S42. Calculate the total number of alarms, the total number of parking protections, and the total number of instructor interventions for each deviation chain in the deviation chain result table, and sort them in the order of the total number of instructor interventions, the total number of parking protections, the total number of alarms, and the result location group number to generate a deviation chain sorting table. S43. Write each deviation chain in the deviation chain sorting table into the corresponding sorting position, total number of alarms, total number of parking protection times, and total number of instructor takeover times in the sorting order to generate an error amplification table.

[0011] In a preferred embodiment, S5 includes: S51. Based on the error amplification table corresponding to the trainees to be trained in the actual vehicle, the current actual vehicle training record and the action perception result, expand the number of alarms, the number of parking protections, the number of instructor takeovers, the target subjects and the target operation segments corresponding to each deviation chain into a deviation matrix in a fixed field order. Expand the current action state, the current deviation action and the completed operation segments in the current action perception result into a state matrix. Perform segment alignment, difference expansion and orthogonal projection on the deviation matrix and the state matrix to solve the segment deviation amount and subsequent continuation amount of each deviation chain to the current training stage and generate a deviation correction table. S52. Construct a real-vehicle training continuity graph based on the deviation correction table. Subject entry relationships form subject edges, intra-segment practice relationships form segment edges, completed operation segments prevent re-entry relationships of incomplete operation segments form blocking edges, and segment deviation and subsequent continuity amounts form correction edges. Perform graph traversal, conflict edge deletion, and sequential reconnection on the real-vehicle training continuity graph. After reconnection, encode each training path in sequence according to the number of times the instructor takes over, the number of times the parking protection is applied, the number of alarms, the segment deviation amount, and the operation segment number, and then sort them in lexicographical order. Output the rearranged training sequence.

[0012] In a preferred embodiment, S5 further includes: S53. Organize real vehicle training according to the rearranged training sequence and continuously receive motion perception results. Align the current operation segment sequence with the execution sequence of the rearranged training sequence. Retain the original continuation relationship for the consistent parts after alignment. Perform posterior consistency update on the alarm results, parking protection results and coach takeover results after conditionally decomposed the offset parts after alignment. Write the remaining operation segments after consistency update back to the real vehicle training continuation diagram and perform conflict edge deletion and sequential reconnection again until the remaining operation segment sequence completely corresponds to the current operation segment sequence. Then write the corresponding real vehicle training results, rearranged training sequence and error amplification table into the visual training archive to generate training management results.

[0013] A forklift driver training management platform based on data visualization analysis, comprising a filing module, a deviation concatenation module, a consequence mapping module, a sorting module, and a reordering control module: The filing module is used to collect the student's manipulation data, position data, and subject result data in the simulation training. Based on the manipulation data and position data, it performs motion perception to determine the corresponding motion state at each moment. It also performs alignment and combination of motion state, manipulation data, position data, and subject result data according to the same student, the same subject, and the same time sequence to generate a simulation training record table. The deviation chaining module, based on the simulation training record table, extracts the deviation actions, the operation segments where the deviations are located, and the sequential relationship between adjacent deviations in each subject according to time sequence, and chains the first and last deviation actions in sequence to generate a deviation chain list. The consequence mapping module is used to retrieve the actual vehicle training records corresponding to the subject, and according to the position correspondence of the same operation segment in the same subject, substitute each deviation action in the deviation chain into the operation segment corresponding to the actual vehicle training record, and count the alarm, parking protection and instructor takeover results caused by each deviation action in the subsequent operation segment, and generate a deviation consequence correspondence table. The sorting module, based on the deviation consequence correspondence table, accumulates the corresponding number of alarms, parking protection times, and instructor takeover times for each deviation chain, and sorts each deviation chain from high to low according to the cumulative results, generating an error amplification table. The reordering control module is used to reorder the order of entry into the actual training subjects and the order of operation practice within the subjects according to the order of each deviation chain in the error amplification table for trainees to be trained in the actual vehicle, combined with the action perception results in the actual vehicle training. The reordered actual vehicle training results are written into the visual training archive corresponding to the error amplification table to generate training management results.

[0014] The technical effects and advantages of this invention are as follows: 1. This solution connects the deviation actions in the simulation training into a deviation chain and maps them to the real vehicle training records of the same subject to statistically analyze the subsequent alarm, parking protection and instructor takeover results. This can identify the real vehicle amplification consequences of the simulation deviation in advance, so that the simulation training results can be used for real vehicle training access control and sequence adjustment. 2. By performing motion perception on manipulation data and position data, and grouping discrete sampling records into action states and action segments, continuous operation changes during training can be transformed into a unified and comparable state sequence, thereby relatively improving the completeness and consistency of deviation extraction in the simulation training phase. 3. Determine the deviation action based on position deviation, timing deviation and state switching deviation, and generate a deviation chain list based on the preceding and following intervals, changes in operation segments and succession relationships. This can expand a single point of failure into a continuous error process, thereby providing an input basis with causal order for subsequent real vehicle consequence tracking. 4. By mapping the deviation actions to the target operation segments in real vehicle training and tracking the alarm, parking protection, and instructor takeover results along the sequential operation segments, a correspondence between the deviation actions and the risk results of the real vehicle can be established, thereby relatively reducing the disconnect between simulation evaluation and real vehicle performance. 5. The number of alarms, parking protection times, and instructor takeover times corresponding to the deviation chain are accumulated and sorted to generate an error amplification table. This table can distinguish the degree of impact of different deviation chains on real vehicle training, thereby providing a basis for prioritizing the allocation of training resources and conducting pre-training for key aspects. Attached Figure Description

[0015] Figure 1 This is a flowchart outlining the method steps of the present invention; Figure 2 This is a schematic diagram of the platform module structure of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Refer to the instruction manual appendix Figures 1-2 The forklift driver training management method based on data visualization analysis of the present invention includes: S1. Collect the student's manipulation data, position data and subject result data in the simulation training. Based on the manipulation data and position data, perform motion perception, determine the corresponding motion state at each moment, and perform alignment and combination of motion state, manipulation data, position data and subject result data according to the same student, the same subject and time sequence to generate a simulation training record table. In this embodiment, S1 is used to convert the manipulation data and position data in the simulation training into continuously readable motion states, which serve as a unified input for subsequent deviation motion extraction. The processing order is as follows: first, the manipulation items and position items are matched according to the sampling time to form a time-motion segment; then, changes are extracted and merged into motion segments according to adjacent sampling times; finally, the motion state is determined based on the correspondence between manipulation changes and position changes within the motion segment. This implementation process includes the following steps: S11 is used to generate a complete action record at the same sampling time. Input data includes manipulation data and position data. Manipulation data includes at least the sampling time, steering wheel angle, pedal travel, and control handle displacement. Position data includes at least the sampling time, vehicle position, vehicle orientation, and fork position. During processing, the manipulation data is first grouped by sampling time. Steering wheel angle, pedal travel, and control handle displacement at the same sampling time are then concatenated into a single manipulation item according to a fixed field order. Next, the position data is grouped by sampling time, and vehicle position, vehicle orientation, and fork position at the same sampling time are concatenated into a single manipulation item according to a fixed field order. For each sampling time, the same-time position item is assigned a corresponding key. Then, the same-time operation item and the same-time position item are matched one-to-one to generate a time-action segment table. The time-action segment table includes at least the student identifier, subject identifier, sampling time, steering wheel angle, pedal travel, control handle displacement, vehicle position, vehicle orientation, and fork position, and is written to the buffer for S12 to read. If multiple operation data or multiple position data exist at the same sampling time, the last one is retained according to the writing order. If a sampling time only contains operation data or only contains position data, the sampling time is marked as a missing item time and is not written to the time-action segment table. S12 is used to merge discrete time records into continuous action segments; the input is a time action segment table; during processing, the time action segment table is arranged in ascending order according to the same student, the same subject, and the sampling time; for two adjacent sampling times, the changes in steering wheel angle, pedal travel, control handle displacement, vehicle body displacement, and fork displacement are calculated respectively; each change is obtained by subtracting the value of the previous sampling time from the value of the later sampling time; then the direction of change is determined according to the positive sign and zero value of each change; each change and its direction of change constitute an action change group. Subsequently, the continuity relationship between action change groups is checked sequentially along the sampling time. When multiple consecutive action change groups satisfy the following conditions: the steering wheel angle change direction is not reversed, the pedal travel change direction is not reversed, the control handle displacement change direction is not reversed, and the starting point of the vehicle body position corresponding to the subsequent action change group is equal to the ending point of the vehicle body position corresponding to the previous action change group, and the starting point of the fork position corresponding to the subsequent action change group is equal to the ending point of the fork position corresponding to the previous action change group, the multiple consecutive action change groups are merged into one action segment, and an action segment table is generated. The action segment table includes at least the action segment identifier, the start sampling time, the end sampling time, the steering wheel angle change sequence, the pedal travel change sequence, the control handle displacement change sequence, the vehicle body displacement change sequence, and the fork displacement change sequence, and is written to the storage area for S13 to read. If there is a missing time in an adjacent sampling time pair, the action change group corresponding to the adjacent sampling time pair is not calculated, and the missing time is used as the action segment cutoff position. If an action change group is reversed in any change direction from the previous action change group, or the position continuity relationship is not established, the action change group is used as the starting point of a new action segment and merged again. S13 is used to map action segments to action states; the input is the action segment table; during processing, each action segment in the action segment table is read one by one, and the directional combination relationship of the steering wheel angle change sequence, pedal travel change sequence, control handle displacement change sequence, vehicle body displacement change sequence, and fork displacement change sequence within the action segment is calculated; when the pedal travel change direction is positive, the vehicle body displacement change direction is positive, and the steering wheel angle change direction is zero, it is determined to be the starting state; when the steering wheel angle change direction is positive or negative, and the vehicle body displacement change direction also has displacement change, it is determined to be the turning state; when the vehicle body displacement change direction is positive, the fork displacement change direction is zero, and the distance between the vehicle body position and the target working position continues to decrease, it is determined to be the approach state; when the control handle displacement change direction is the same as the fork displacement change direction, it is determined to be the pick-up / place-down state. When there is a reciprocating switching in the steering wheel angle change, pedal travel change, or control handle displacement change, and the corresponding reciprocating change occurs in the vehicle body displacement or fork displacement change, it is determined to be an adjustment state. Then, according to the start and end sampling times of the action segment, the corresponding action state of the action segment is written into the action segment table time by time to generate the action state, and written into the simulation training record table for S2 to read. If the same action segment satisfies two types of state correspondence at the same time, the previous state is retained in a fixed order of pick-up and put-down state, steering state, starting state, approach state, and adjustment state. If an action segment does not satisfy any state correspondence, it is marked as an unclassified state and written into the simulation training record table. Through the above steps, the raw sampling data in the simulation training can be transformed into action states with clear boundaries and continuous time sequence, serving as a unified input for subsequent deviation action recognition. In practical applications: when trainees perform training in starting, turning, approaching the pallet, inserting and adjusting the forks, and lifting the forks to pick up goods, the platform first combines the steering wheel angle, pedal travel, control handle displacement, vehicle position, vehicle orientation, and fork position according to the sampling time to form a time action segment table; then, it calculates the changes according to adjacent sampling times and merges them into an action segment table; finally, it identifies forward movement as the starting state, direction adjustment as the turning state, approaching the pallet as the approaching state, fork lifting and inserting as the picking and placing state, and small repeated correction actions as the adjustment state, and writes them all into the simulation training record table.

[0018] S2. Based on the simulation training record table, extract the deviation actions, the operation segments where the deviations are located, and the sequential relationship between adjacent deviations in each subject according to time sequence, and connect the first and last deviation actions in sequence to generate a deviation linked list. In this embodiment, S2 is used to extract the deviation actions formed during the training of each subject from the simulated training record table, and further solve the sequential relationship between each deviation action to finally generate a deviation linked list as the direct input for subsequent real vehicle training result mapping. Its processing principle is as follows: First, the action state at each moment is compared with the corresponding subject result data moment by moment to solve the position deviation, timing deviation, and state switching deviation, and based on this, the deviation action and its corresponding operation segment are determined; then, it is judged whether adjacent deviation actions meet the succession conditions along the time sequence to form succession deviation pairs; finally, with the deviation action without a preceding succession deviation pair as the head of the chain, subsequent deviation actions are concatenated along the succession deviation pair sequence to output the deviation linked list. This implementation process includes the following steps: S21 is used to determine the deviation actions corresponding to each moment and form a deviation action table. Its working mechanism is to separate abnormal actions from continuous action states by comparing the moment-by-moment differences between the action state and the subject result data. The input is a simulation training record table, which includes at least the student identifier, subject identifier, sampling time, action state, vehicle position, vehicle orientation, fork position, operation segment number, and subject result data. The subject result data includes at least the target position, target timing sequence, and target state switching sequence. During processing, the simulation training record table is first read in ascending order according to the same student, the same subject, and the sampling time. For each sampling time, the position deviation, timing deviation, and state switching deviation are calculated respectively. The position deviation is obtained by the difference between the vehicle position or fork position corresponding to the current sampling time and the target position corresponding to the sampling time. The timing deviation is obtained by the difference between the sequence of the current action state at the sampling time and the target timing sequence that the action state should be in. The state switching deviation is obtained by the difference between the switching sequence of the current action state and the target state switching sequence. When the position deviation, timing deviation, or state switching deviation is not zero, the sampling time is determined as the deviation time, and it is marked as a position deviation action, timing deviation action, or state switching deviation action according to the source of the deviation. Then, the operation segment number corresponding to the deviation time is used as the operation segment to generate a deviation action table. The deviation action table includes at least the deviation action identifier, deviation category, sampling time, operation segment, action status, position deviation, timing deviation, and state switching deviation, and is written to the storage area for S22 to read. The abnormal or missing handling is as follows: when any field of target position, target timing sequence, or target state switching sequence is missing at a certain sampling time, the sampling time does not participate in the deviation action determination, and a missing item mark is recorded in the deviation action table. When there are two or more types of deviations at the same sampling time, multiple deviation action records are written in a fixed order of position deviation action, timing deviation action, and state switching deviation action. S22 is used to solve the connection relationship between deviation actions and form a deviation connection table. Its working mechanism is to screen out continuously traceable deviation action pairs from all deviation actions by jointly verifying the time sequence, operation segment sequence, and action state continuation direction. The input is the deviation action table. During processing, the deviation action table is first arranged in ascending order according to the same student, the same subject, and the sampling time. For any previous deviation action and the next deviation action, the time interval between the two actions, the change in the operation segment, and the action state continuation direction are extracted. The time interval between the two actions is obtained by subtracting the sampling time of the previous deviation action from the sampling time of the next deviation action. The change in the operation segment is obtained by subtracting the operation segment number of the previous deviation action from the operation segment number of the next deviation action. The action state continuation direction is determined by the switching order from the action state corresponding to the previous deviation action to the action state corresponding to the next deviation action. Then, check if there are any other deviation actions inserted between the end time of the previous deviation action and the start time of the next deviation action. If other deviation actions are inserted, the deviation action pair is not recorded as a successor deviation pair. If no other deviation actions are inserted, check if the operation segment to which the next deviation action belongs is a continuation of the current operation segment of the operation segment to which the previous deviation action belongs or a subsequent sequential operation segment. The current operation segment continuation is determined by the same operation segment number, and the subsequent sequential operation segment is determined by the operation segment number to which the next deviation action belongs being equal to the operation segment number to which the previous deviation action belongs plus one. When both of the above conditions are met, the deviation action pair is determined as a successor deviation pair, and a new deviation action pair is generated. The deviation inheritance table includes at least the preceding deviation action identifier, the following deviation action identifier, the interval between the preceding and following actions, the preceding operation segment number, the following operation segment number, and the action status inheritance direction. This table is written to a buffer for S23 to read. Abnormal or missing data is handled as follows: when the same following deviation action corresponds to multiple preceding deviation actions, the preceding deviation action pairs are retained in ascending order of the preceding and following interval durations. If the preceding and following interval durations are the same, the preceding deviation action pairs are retained in ascending order of the preceding operation segment number, and the remaining deviation action pairs are deleted. When the same preceding deviation action is followed by multiple following deviation actions, the preceding deviation action pairs are retained in ascending order of the following sampling time. S23 is used to concatenate the received deviation pairs into a deviation chain and generate a deviation chain list. Its working mechanism is to transform the discrete receiving relationship into a continuously traceable deviation chain by determining the chain head, connecting each pair, and stopping the chain tail. The input is the deviation receiving table. During processing, firstly, the deviation action without a preceding received deviation pair is searched in the deviation receiving table, and the deviation action without a preceding received deviation pair is determined as the chain head deviation action. Then, starting from the chain head deviation action, the subsequent deviation actions are sequentially concatenated along the connection order of the preceding deviation action identifier and the subsequent deviation action identifier in the deviation receiving table. Each subsequent deviation action is concatenated and its position in the chain is recorded once. When the current deviation action does not have a corresponding subsequent deviation action in the deviation receiving table, the concatenation stops, and the current deviation action is determined as the chain tail deviation action. Subsequently, a deviation chain is generated for each concatenation result, forming a deviation chain list. The deviation chain list includes at least the deviation chain identifier, the order within the chain, the deviation action identifier, the operation segment to which it belongs, the sampling time, the identifier of the preceding deviation action, and the identifier of the following deviation action. This list is then written into the subsequent real vehicle training mapping unit for S3 to read. The handling of anomalies or missing values ​​is as follows: when a deviation action has neither a preceding nor a following deviation pair, a separate deviation chain is generated for that deviation action. When a loop connection appears in the deviation list, the next connection in the loop is cut off in ascending order of the operation segment number, and the deviation action before the cut-off position is taken as the tail deviation action of the chain. Through the above steps, the deviation actions in the simulation training record table can be transformed from single-point abnormal records into a deviation chain list with sequential relationships, providing continuous input for subsequent deviation actions to be substituted into the real vehicle training record and the consequences to be statistically analyzed. At the same time, it eliminates the problems of unclear sources of deviation actions, unstable connection of deviation actions, and multiple candidate sequential relationships in parallel. In practical applications: When a trainee experiences three types of anomalies during a training session—early steering, approach position deviation, and delayed switching between pick-up and drop-off states—the platform first calculates the position deviation, timing deviation, and state switching deviation at the corresponding sampling times, generating a deviation action table. Then, it checks whether other deviation actions are inserted between the three types of deviation actions according to the sampling time sequence, and whether the operation segment number is a continuation of the current segment or the next sequential operation segment, generating a deviation inheritance table. Finally, it uses the first deviation action without a preceding inheritance pair as the chain head, and sequentially strings together subsequent deviation actions along the inheritance pair sequence to generate a complete deviation chain, which is then written into the deviation chain table for subsequent statistical analysis of the alarm, parking protection, and instructor takeover results caused by this deviation chain during actual vehicle training.

[0019] S3. Retrieve the actual vehicle training records corresponding to the subject, and according to the positional correspondence of the same operation segment in the same subject, substitute each deviation action in the deviation chain into the operation segment corresponding to the actual vehicle training record, and count the alarm, parking protection and instructor takeover results caused by each deviation action in the subsequent operation segment, and generate a deviation consequence correspondence table. In this embodiment, S3 is used to map each deviation action in the deviation chain list to the actual vehicle training record of the same subject, and to track the alarm, parking protection, and instructor takeover results along the sequential operation segments following the corresponding operation segment, generating a deviation consequence correspondence table as the direct input for the subsequent error amplification table generation. Its processing principle is as follows: First, based on the correspondence between the operation segment number, start and end positions, and direction of travel, the deviation action is mapped to the target operation segment in the actual vehicle training record, forming a deviation substitution table; then, various sequences are extracted from the deviation substitution table, and their continuity, directional consistency, and result sequence correspondence are checked to filter out valid result chains, forming a deviation result chain list; finally, the number of executions, position aggregation, merging, and splitting of the result chain corresponding to the same deviation action are performed, and the deviation consequence correspondence table is output. This implementation process includes the following steps: S31 is used to match deviation actions with target operation segments in the actual vehicle training records and establish a basis for subsequent result tracking. Its mechanism is to project deviation actions formed in simulated training into similar operation processes in actual vehicle training by using operation segment correspondence rules under a unified subject. The input quantities are a deviation list and actual vehicle training records. The deviation list includes at least the subject identifier, deviation chain identifier, chain order, deviation action identifier, operation segment, sampling time, and deviation action direction. The actual vehicle training records include at least the operation segment record, vehicle status record, alarm record, parking protection record, and instructor takeover record. The operation segment record includes at least the subject identifier, operation segment number, start position, end position, direction of travel, start time, and end time. The vehicle status record includes at least the recording time, vehicle position, vehicle orientation, fork position, and vehicle operating status. The alarm record, parking protection record, and instructor takeover record include at least the result number, the operation segment in which the deviation occurred, and the time of occurrence. During processing, first, read the deviation list and operation segment records for the same subject. For each deviation action in the deviation list, extract the operation segment number, start and end positions of the operation segment, and direction of the deviation action. Then, search for operation segment records in the actual vehicle training records that have the same operation segment number, corresponding start and end positions, and consistent travel directions. Among them, the correspondence between start and end positions is determined by the start position of the operation segment being the same as the start position of the operation segment record and the end position of the operation segment being the same as the end position of the operation segment record. The consistency of travel directions is determined by the travel direction of the operation segment being the same as the travel direction field value of the operation segment record. After the matching is completed, the operation segment record is determined as the target operation segment, and the result tracking is carried out along the sequential operation segments whose numbers increase sequentially after the target operation segment. The vehicle status record, alarm record, parking protection record, and instructor takeover record corresponding to each sequential operation segment are extracted to generate a deviation substitution table. The deviation substitution table includes at least the deviation action identifier, target operation segment number, sequential operation segment number, vehicle status record, alarm record, parking protection record, and instructor takeover record, and is written to the buffer for S32 to read. The abnormal or missing handling is as follows: when the same deviation action corresponds to multiple target operation segments, the operation segment with the first position is retained in ascending order of the start time of the target operation segment; when a deviation action does not find a corresponding target operation segment in the real vehicle training record, the deviation action is written to the uncorresponding record table and not written to the deviation substitution table; when there is no sequential operation segment after the target operation segment, only the record corresponding to the target operation segment is retained and written to the deviation substitution table. S32 is used to filter out the result chain that satisfies the continuity and correspondence relationship from the deviation substitution table. Its working mechanism is to fix the actual vehicle training results that may be caused by the deviation action into a verifiable and valid result chain by jointly verifying the operation segment sequence, the direction of vehicle state change, and the occurrence sequence of the three types of results. The input is the deviation substitution table. During processing, the deviation substitution table is first read in groups according to the deviation action identifier. For each deviation action, the operation segment number sequence, vehicle state sequence, alarm number sequence, parking protection number sequence, and instructor takeover number sequence corresponding to the target operation segment and its subsequent sequential operation segments are extracted. Among them, the operation segment number sequence is arranged in order of the sequential operation segment number, the vehicle state sequence is arranged in order of the vehicle state record corresponding to the sequential operation segment, and the alarm number sequence, parking protection sequence, and instructor takeover number sequence are arranged in ascending order of the occurrence time. Then, perform a joint verification on the above sequence: First, check whether the operation segment number sequence is continuous. If the number of the next operation segment is equal to the number of the previous operation segment plus one, then the adjacent operation segment numbers are continuous. Second, check whether the direction of vehicle status change in the vehicle status sequence is consistent with the direction of deviation action. The direction of vehicle status change is obtained by subtracting the vehicle status value corresponding to the previous sequential operation segment from the vehicle status value corresponding to the next sequential operation segment. The vehicle status value uses the field corresponding to the operation segment to which the deviation action belongs, which is the direction of change of vehicle body position, direction of change of vehicle body orientation, or direction of change of fork position. Third, check whether the alarm number sequence, parking protection number sequence, and coach takeover number sequence correspond sequentially according to the operation segment order. If the operation segment in which a result record occurs is consistent with the corresponding sequential operation segment number in the operation segment number sequence, it is considered to be sequentially corresponding. When all three conditions are met, the corresponding record is identified as a valid result chain, and a deviation result chain list is generated. The deviation result chain list includes at least the deviation action identifier, operation segment number sequence, vehicle status sequence, alarm number sequence, parking protection number sequence, and instructor takeover number sequence, and is written to the storage area for S33 to read. The abnormal or missing handling is as follows: when any of the three types of result sequences is empty, the empty sequence is retained as an empty field, which does not affect the verification of the other non-empty sequences. When the same deviation action corresponds to multiple valid result chains that meet the conditions, the first valid result chain is retained in ascending order of the first number of the operation segment number sequence, and the other valid result chains are deleted. When the operation segment number sequence is not continuous or the direction of vehicle status change is inconsistent with the direction of deviation action, the record is not written to the deviation result chain list. S33 is used to merge the valid result chains in the deviation result chain list into a deviation consequence correspondence table. Its working mechanism is to fix the subsequent directly cumulative consequence records by performing unified statistics and unified writing on the result count and result position of the same deviation action. The input is the deviation result chain list. During processing, the deviation result chain list is first read in groups according to the deviation action identifier. For each deviation result chain corresponding to the same deviation action, the number of alarms, the number of parking protections, and the number of instructor takeovers are counted. Among them, the number of alarms is the number of numbers in the alarm number sequence, the number of parking protections is the number of numbers in the parking protection number sequence, and the number of instructor takeovers is the number of numbers in the instructor takeover number sequence. Then, the alarm occurrence operation segment, the parking protection occurrence operation segment, and the instructor takeover occurrence operation segment are extracted respectively, and concatenated into a result position group in a fixed field order. The fixed field order is alarm occurrence operation segment, parking protection occurrence operation segment, and instructor takeover occurrence operation segment. After completing the statistics, multiple deviation result chains for the same deviation action are merged and split: when the result location group is the same and the number of alarms, parking protection times, and instructor intervention times are the same, the multiple deviation result chains are merged into one record and written into the deviation consequence correspondence table; when the result location groups are different, or the result location groups are the same but any one of the alarms, parking protection times, and instructor intervention times is different, the deviation result chains are split into multiple records and written into the deviation consequence correspondence table in ascending order of the first digit of the operation segment number sequence; the deviation consequence correspondence table includes at least the deviation action identifier, alarm count, parking protection times, instructor intervention times, and result location group, and is written into the subsequent accumulation unit for S4 to read; the abnormal or missing handling is as follows: when all three counts of a deviation result chain are zero, it is still written into the deviation consequence correspondence table, and the result location group is recorded as an empty group; when the same deviation action corresponds to only one deviation result chain, it is directly written into the deviation consequence correspondence table, and no merging process is performed; Through the above steps, a fixed correspondence can be established between each deviation action in the deviation chain and the subsequent results in the actual vehicle training process, forming a deviation consequence correspondence table that can be directly accumulated and sorted, providing a unified input for the subsequent generation of the error amplification table, while eliminating the problems of unclear target operation segments, non-unique result chains, and inconsistent result writing standards. In practical applications: After a trainee forms a deviation chain containing steering deviation and approach deviation actions in the same subject during simulated training, the platform first searches for the target operation segment in the real vehicle training record that has the same operation segment number, corresponding start and end positions, and consistent direction of travel. Then, it tracks the alarm records, parking protection records, and instructor takeover records along the sequential operation segments following the target operation segment to generate a deviation substitution table. Next, it extracts the operation segment number sequence, vehicle status sequence, and three types of result sequences from the deviation substitution table, and performs a joint verification of the correspondence between the operation segment continuity, the direction of vehicle status change, and the result sequence to generate a deviation result chain list. Finally, it counts the number of alarms, parking protection times, and instructor takeover times corresponding to the same deviation action, and generates result position groups according to the operation segments where each type of result occurs. Records with the same result position group and the same number of times are merged and written into the deviation consequence correspondence table. Records with different result position groups or different numbers of times are split according to the operation segment number order and written into the deviation consequence correspondence table for direct reading in the subsequent generation of the error amplification table.

[0020] S4. Based on the deviation consequence correspondence table, accumulate the corresponding number of alarms, number of parking protections and number of instructor interventions for each deviation chain, and arrange each deviation chain from high to low according to the cumulative results to generate an error amplification table. In this embodiment, S4 is used to aggregate the scattered consequence records in the deviation consequence correspondence table to the deviation chain level, forming an error amplification table that can be directly used for real vehicle training rearrangement. Its processing principle is as follows: First, according to the same deviation chain, the number of alarms, the number of parking protections, and the number of instructor interventions corresponding to each deviation action are aggregated, and the operation segments in which various results occur are simultaneously aggregated to form a deviation chain result table; then, the sum of the three types of result counts is calculated for each deviation chain, and the results are sorted according to a fixed field order to form a deviation chain sorting table; finally, the sorting results are written into the error amplification table for subsequent reading by S5. This implementation process includes the following steps: S41 is used to aggregate the consequences of deviation actions in the deviation consequence correspondence table to the deviation chain level. Its working mechanism is to uniformly convert the consequences caused by a single deviation action into deviation chain consequences through the correspondence between deviation actions and deviation chains. The input quantities are the deviation consequence correspondence table and the deviation chain table. The deviation consequence correspondence table includes at least the deviation action identifier, alarm count, parking protection count, instructor takeover count, and result location group. The deviation chain table includes at least the deviation chain identifier, chain order, and deviation action identifier. During processing, the deviation consequence correspondence table and the deviation chain table are first connected according to the deviation action identifier, and each deviation consequence record belonging to the same deviation chain is grouped into the same deviation chain set. Then, the alarm count, parking protection count, and instructor takeover count are collected for each deviation chain set. The alarm occurrence operation segment, parking protection occurrence operation segment, and instructor takeover occurrence operation segment are extracted from the result location group and concatenated in the fixed field order of the alarm occurrence operation segment, parking protection occurrence operation segment, and instructor takeover occurrence operation segment to generate the deviation chain result table. The deviation chain result table includes at least the deviation chain identifier, alarm count set, parking protection count set, instructor takeover count set, alarm occurrence operation segment set, parking protection occurrence operation segment set, and instructor takeover occurrence operation segment set, and is written to the storage area for S42 to read; the abnormal or missing handling is as follows: when a deviation action does not find a corresponding deviation chain in the deviation chain list, the deviation action is written to the unreturned chain record table and not written to the deviation chain result table; when the result position group of a deviation consequence record is an empty group, it is still returned to the corresponding deviation chain set, and the alarm occurrence operation segment set, parking protection occurrence operation segment set, and instructor takeover occurrence operation segment set are recorded as empty sets respectively; S42 is used to perform cumulative summation and sorting of deviation chains. Its mechanism is to fix the subsequent impact of each deviation chain on real vehicle training into a comparable order by unifying the cumulative summation rules and the sorting rules. The input is the deviation chain result table. During processing, the deviation chain result table is read one by one according to the deviation chain identifier. For each deviation chain, the total number of alarms, the total number of parking protections, and the total number of instructor interventions are calculated. The total number of alarms is the sum of the values ​​in the alarm count set corresponding to the deviation chain, the total number of parking protections is the sum of the values ​​in the parking protection count set corresponding to the deviation chain, and the total number of instructor interventions is the sum of the values ​​in the instructor intervention count set corresponding to the deviation chain. Then, the alarm occurrence operation segment set, the parking protection occurrence operation segment set, and the instructor intervention occurrence operation segment set are concatenated in a fixed field order to form a result position group. The result position group is written with the result position group number according to the generation order in the deviation chain result table. Then, all deviation chains are sorted according to the fields of total number of instructor interventions, total number of parking protections, total number of alarms, and result location group number. If the values ​​of the preceding fields are the same, the following fields are compared until a unique sorting order is obtained, generating a deviation chain sorting table. The deviation chain sorting table includes at least the deviation chain identifier, total number of instructor interventions, total number of parking protections, total number of alarms, and result location group number, and is written to the buffer for S43 to read. The exception or missing handling is as follows: when multiple deviation chains are the same in the fields of total number of instructor interventions, total number of parking protections, total number of alarms, and result location group number, the sorting order is retained in ascending order of the deviation chain identifier; when the total number of all three types of occurrences of a deviation chain is zero, it still participates in the sorting and is entered into the deviation chain sorting table according to the above field order. S43 is used to solidify the arrangement results into an error amplification table. Its mechanism is to transform the consequence intensity of the deviation chain into a training rearrangement basis that can be directly called upon later by writing the sorting position and cumulative results. The input is the deviation chain sorting table. During processing, each deviation chain record is read one by one according to the sorting order in the deviation chain sorting table, and the sorting position is written in the order of reading. The corresponding total number of coach intervention times, total number of parking protection times, and total number of alarm times are also written to generate the error amplification table. The error amplification table includes at least the deviation chain identifier, sorting position, total number of coach intervention times, total number of parking protection times, and total number of alarm times, and is written to the subsequent training rearrangement unit for S5 to read. The abnormal or missing handling is as follows: when the deviation chain sorting table is empty, an empty error amplification table is generated and an empty table mark is written; when there are duplicate deviation chain identifiers in the deviation chain sorting table, the first record is retained according to the sorting order, and the remaining duplicate records are deleted. Through the above steps, the single consequence record in the deviation consequence correspondence table can be transformed into the cumulative consequence result of the deviation chain level, and the error amplification table can be generated in a unified order. This provides a direct basis for the subsequent order of entering the actual vehicle training subjects and the order of operation practice within the subjects. At the same time, it eliminates the problems of inconsistent collection criteria, non-unique cumulative results, and non-closed sorting rules between deviation actions and deviation chains. In practical application: A trainee forms two deviation chains in the same subject. The first deviation chain corresponds to three alarms, one parking protection, and one instructor intervention. The second deviation chain corresponds to two alarms, zero parking protections, and zero instructor interventions. The platform first aggregates the consequences of each deviation action according to the deviation chain to form a deviation chain result table. Then, it calculates the total number of alarms, parking protections, and instructor interventions for both deviation chains and arranges them in the order of the total number of instructor interventions, parking protections, alarms, and result position group number to generate a deviation chain sorting table. Finally, the first deviation chain is written into the first sorting position, and the second deviation chain is written into the second sorting position to form an error amplification table, which can be directly read for subsequent real-vehicle training reordering.

[0021] S5. For the error amplification table corresponding to trainees waiting for actual vehicle training, and in combination with the motion perception results in actual vehicle training, the order of entry into actual vehicle training subjects and the order of operation practice within subjects are rearranged according to the arrangement order of each deviation chain in the error amplification table. The rearranged actual vehicle training results are written into the visual training archive corresponding to the error amplification table to generate training management results. In this embodiment, S5 is used to dynamically rearrange the order of entry into the actual training subjects and the order of operation practice within the subjects for trainees based on the error amplification table and the current actual training status, and write the rearranged execution results back to the visual training archive. The processing principle is as follows: First, the deviation chain information in the error amplification table and the current motion perception results are uniformly expanded into a calculable deviation matrix and state matrix. Through segment alignment, difference expansion, and orthogonal projection, the segment deviation and subsequent continuity of each deviation chain relative to the current training stage are solved, forming a deviation correction table. Then, an actual training continuity diagram is constructed based on the deviation correction table, and the rearranged training sequence is output through simultaneous generation, conflict deletion, and sequential reconnection. Finally, during actual training, motion perception results are continuously received, and sequence alignment and consistency updates are performed on the current operation segment sequence and the rearranged training sequence, and the updated results are written back to the visual training archive. This implementation process includes the following steps: S51 is used to convert the error amplification table and the current motion perception result into a deviation correction table that can be directly used in the reordering calculation. Its working mechanism is to quantify the impact of the deviation chain on the current training stage into the segment deviation amount and the subsequent continuation amount through unified field expansion and matrix operation. The inputs are the error amplification table corresponding to the trainee to be trained in the actual vehicle, the current actual vehicle training record, and the motion perception result. Among them, the error amplification table includes at least the deviation chain identifier, sorting position, total number of alarms, total number of parking protection times, total number of instructor takeover times, target subject, and target operation segment. The current actual vehicle training record includes at least the current subject, current operation segment, and completed operations. The action perception result includes at least the current action state, the current deviation action, and the current operation segment, based on the segment set and the current training time. During processing, the total number of alarms, the total number of parking protections, the total number of instructor takeovers, the target subject, and the target operation segment are first expanded into a deviation matrix in ascending order of deviation chain identifiers and fixed field order. Each row of the deviation matrix represents a deviation chain, and each column represents a type of fixed field. Then, the current action state, the current deviation action, and the completed operation segment are expanded into a state matrix in ascending order of the current training time and fixed field order. Each row of the state matrix records the current training state, and each column represents a type of fixed field. Then, according to the rule that the target operation segment and the current operation segment number are consistent, same-segment alignment is performed. The deviation matrix rows and status matrix rows of the same operation segment are written into the alignment table. Deviation matrix rows that fail to align with the same segment are kept in their original positions according to the target operation segment number, and status matrix rows that fail to align with the same segment are kept in their original positions according to the current operation segment number. Subsequently, the alignment table is expanded by difference. The segment deviation value is obtained by subtracting the current operation segment number from the target operation segment number, the subject deviation value is obtained by subtracting the current subject number from the target subject number, and the completion deviation value is determined by whether the target operation segment belongs to the completed operation. Once the segment set is determined, it is recorded as one when it belongs and zero when it does not. Then, the column containing the current operation segment field in the state matrix is ​​used to form the projection basis vector. Orthogonal projection is performed on the target operation segment column, the total number of alarms column, the total number of parking protections column, and the total number of coach takeovers column in the deviation matrix to obtain the segment deviation and subsequent continuation amount of each deviation chain in the current training stage. The segment deviation is the difference between the projection value corresponding to the target operation segment column and the projection value corresponding to the current operation segment column after projection. The subsequent continuation amount is the projection length in the direction of the unfinished operation segment after projection. The final deviation correction table is generated, which includes at least the deviation chain identifier, target subject, target operation segment, segment deviation amount, subsequent continuation amount, total number of alarms, total number of parking protection times, and total number of instructor takeover times. This table is written to a buffer for S52 to read. Abnormal or missing data is handled as follows: when a deviation chain with a missing target subject or target operation segment exists in the error amplification table, the deviation chain is written to the supplementary field table but not to the deviation matrix; when the action perception result lacks the current action state or current operation segment, the generation of the current round of deviation correction table is paused, and the previous round's deviation correction table is retained as the current round's input; when the same deviation chain corresponds to multiple target operation segments, the previous target operation segment is retained in ascending order of its number. S52 is used to generate a real-vehicle training sequence graph based on the deviation correction table and output a rearranged training sequence. Its mechanism involves expressing subject entry relationships, segment practice relationships, re-entry prohibition relationships, and deviation correction relationships through a graph structure, and then obtaining a unique training order through conflict edge deletion and path reconnection. The inputs are the deviation correction table and the current real-vehicle training record. During processing, a node set is first constructed according to the target subject and target operation segment, generating one subject node for each target subject and one operation segment node for each target operation segment. Then, four types of edges are generated according to rules, where subject entry relationships form subject edges, and... The body consists of target subject nodes with higher rankings in the same deviation chain pointing to target subject nodes with lower rankings; intra-segment practice relationships form segment edges, specifically, the next operation segment node within the same target subject whose operation segment number is incremented by one is pointed to by the previous operation segment node; the non-reentrancy relationship between completed operation segments and incomplete operation segments forms blocking edges, specifically, the reverse non-connection between completed operation segment nodes and corresponding incomplete operation segment nodes; the segment deviation amount and subsequent continuation amount form correction edges, specifically, the current operation segment node points to the target operation segment node, and the edge attributes are written with the segment deviation amount and subsequent continuation amount; Subsequently, a graph traversal is performed on the actual vehicle training sequence graph. Starting with the current subject node and the current operation segment node, candidate paths are expanded along the subject edge, segment edge, and correction edge. During the traversal, conflicting edges are deleted. Conflicting edges include edges where the same node points to two different subsequent nodes, edges that form a reverse re-entry relationship with the blocking edge, and correction edges that point to completed operation segment nodes. After deleting conflicting edges, the disconnected preceding nodes and the retained subsequent nodes are reconnected sequentially to generate reconnected training paths. Then, each reconnected training path is encoded in the order of the fields: total number of instructor takeovers, total number of parking protections, total number of alarms, segment deviation, and operation segment number. The encoded strings are directly concatenated according to the above field order, and all encoded strings are sorted lexicographically. The training path that appears first is output as the reordered training sequence. The rearranged training sequence includes at least the sequence position, target subject, target operation segment, corresponding deviation chain identifier, and corresponding encoding string, and is written to the storage area for S53 to read; the abnormal or missing handling is as follows: when a node has no subsequent connectable edges, the node is marked as the end of the path and the traversal of the path ends; when all candidate paths in the graph are cut off by blocking edges, the correction edge is regenerated with the incomplete operation segment with the closest number after the current operation segment and the traversal is repeated; when multiple training path encoding strings are exactly the same, the previous training paths are retained in ascending order of the first operation segment number; S53 is used to dynamically correct and rearrange the training sequence and generate training management results during the execution of real-vehicle training. Its mechanism is to ensure that the training order continuously corresponds to the current training state through sequence alignment, conditional independence decomposition, posterior consistency update, and graph write-back reconnection. The inputs are the rearranged training sequence, motion perception results, real-vehicle training continuity graph, and error amplification table. During processing, real-vehicle training is first organized according to the rearranged training sequence, and motion perception results are continuously received. The current operation segments in the motion perception results are arranged into the current operation segment sequence according to the time sequence. Then, the current operation segment sequence is aligned with the execution sequence of the rearranged training sequence. The alignment rule is: if the operation segment numbers of the same position in the sequence are the same, it is recorded as consistent segment position; if the operation segment numbers of the same position are different, it is recorded as segment position offset. For the aligned segments with consistent characteristics, the continuity relationship in the original real-vehicle training continuity diagram is retained without adjustment. For the aligned segments with offset characteristics, the execution conditions are independently decomposed according to the alarm results, parking protection results, and instructor takeover results under the current operation segment conditions. Specifically, under the condition that the current operation segment number is fixed, the alarm results, parking protection results, and instructor takeover results corresponding to the operation segment are read separately, and the three types of results are used as independent input items. Then, the posterior consistency is updated based on the sum of the alarm counts, parking protection counts, and instructor takeover counts of the corresponding deviation chains in the error amplification table. The initial consistency value is taken from the consistency result of the previous round, and one is taken if there is no consistency result of the previous round. When the type of the result corresponding to the current operation segment is consistent with the cumulative result of the deviation chain, the corresponding consistency item is incremented by one. When the types are inconsistent, the corresponding consistency item is decremented by one. The sum of the three types of consistency items is used as the updated consistency of the deviation chain under the current operation segment. Then, the remaining operation segments after consistency updates are written back to the actual vehicle training sequence diagram. Correction edges that are no longer retained after the update are deleted, and correction edges that need to be retained are reconnected. Conflicting edge deletion and sequential reconnection are performed again until the remaining operation segment sequence completely corresponds to the current operation segment sequence. Finally, the corresponding actual vehicle training results, rearranged training sequences, and error amplification tables are written into the visual training archive according to student identifier, training date, subject identifier, and writing time to generate training management results. The visual training archive includes at least student identifier, rearranged training sequence, current operation segment sequence, actual vehicle training results, error amplification table identifier, and training management results. Abnormal or missing data is handled as follows: when the current operation segment sequence is empty, the previous round of rearranged training sequence is retained without updating; when the motion perception result is missing twice consecutively, the current operation segment in the current actual vehicle training record is used to replace the current operation segment in the motion perception result; when multiple remaining operation segment sequences still exist after consistency updates, the previous remaining operation segment sequence is retained in descending order of updated consistency and ascending order of operation segment number. Through the above steps, the error amplification table, the current real vehicle training status, and the motion perception results can be uniformly transformed into a dynamically updatable rearranged training sequence. During the real vehicle training process, the correspondence between this sequence and the current training process can be continuously corrected, and finally, a traceable training management result can be generated. At the same time, it solves the problems of unclear connection between the target subject and the current training stage, non-unique parallel training paths, and the inability to recall the sequence after training execution. In practical application: A trainee corresponds to three deviation chains in the error amplification table. The first deviation chain corresponds to the stacking subject and target operation segment five; the second deviation chain corresponds to the picking subject and target operation segment three; and the third deviation chain corresponds to the transfer subject and target operation segment two. The platform first expands the alarm count, parking protection count, instructor takeover count, target subject, and target operation segment of the three deviation chains into a deviation matrix, and expands the current action status, current deviation action, and completed operation segment into a status matrix. Through segment alignment, difference expansion, and orthogonal projection, the segment deviation and subsequent continuation amount of the three deviation chains are solved, and a deviation correction table is generated. Then, based on the deviation... The off-calibration table is used to construct a real-vehicle training continuity graph. Conflict edges are deleted, sequential reconnection is performed, and lexicographical ordering is applied to the training path, outputting a rearranged training sequence. Subsequently, during the trainee's real-vehicle training, motion perception results are continuously received. The current operation segment sequence is aligned with the rearranged training sequence. The segment offset is updated based on the alarm result, parking protection result, and instructor takeover result, with posterior consistency. The updated remaining operation segments are then written back to the real-vehicle training continuity graph for reconnection until the current operation segment sequence completely corresponds to the rearranged training sequence. Finally, the final real-vehicle training results, rearranged training sequences, and error amplification table are all written into the visualized training archive.

[0022] Furthermore, the present invention also includes a forklift driver training management platform based on data visualization analysis, the platform comprising a filing module, a deviation concatenation module, a consequence mapping module, a sorting module, and a reordering control module: The filing module is used to collect the student's manipulation data, position data, and subject result data in the simulation training. Based on the manipulation data and position data, it performs motion perception to determine the corresponding motion state at each moment. It also performs alignment and combination of motion state, manipulation data, position data, and subject result data according to the same student, the same subject, and the same time sequence to generate a simulation training record table. The deviation chaining module, based on the simulation training record table, extracts the deviation actions, the operation segments where the deviations are located, and the sequential relationship between adjacent deviations in each subject according to time sequence, and chains the first and last deviation actions in sequence to generate a deviation chain list. The consequence mapping module is used to retrieve the actual vehicle training records corresponding to the subject, and according to the position correspondence of the same operation segment in the same subject, substitute each deviation action in the deviation chain into the operation segment corresponding to the actual vehicle training record, and count the alarm, parking protection and instructor takeover results caused by each deviation action in the subsequent operation segment, and generate a deviation consequence correspondence table. The sorting module, based on the deviation consequence correspondence table, accumulates the corresponding number of alarms, parking protection times, and instructor takeover times for each deviation chain, and sorts each deviation chain from high to low according to the cumulative results, generating an error amplification table. The reordering control module is used to reorder the order of entry into the actual training subjects and the order of operation practice within the subjects according to the order of each deviation chain in the error amplification table for trainees to be trained in the actual vehicle, combined with the action perception results in the actual vehicle training. The reordered actual vehicle training results are written into the visual training archive corresponding to the error amplification table to generate training management results.

[0023] Working Principle: This solution first collects data on the trainee's steering wheel, pedals, control handles, vehicle body, and fork positions during the simulation training phase. Through motion perception, the continuous operation process is identified as starting, turning, approaching, picking up / placing, and adjusting actions. Abnormal actions with incorrect positions, incorrect sequences, or incorrect state transitions are then extracted and linked into deviation chains according to their sequence. Subsequently, these deviation chains are mapped to the actual vehicle training records for the same subject, and the number of alarms, parking protection, and instructor interventions caused by each deviation chain in subsequent operations is counted. This yields the impact of each deviation chain in the actual vehicle phase, and an error amplification table is generated accordingly. Finally, the platform combines the trainee's motion perception results from the current actual vehicle training to determine which operation segment the trainee is currently in and which deviation chains need to be addressed first. The entry order of the actual vehicle training subjects and the practice order within each subject are then dynamically rearranged, and the rearrangement results and training results are written into a visualized training archive, forming a continuously updated training management result. For example, when a trainee completes pallet handling and stacking exercises in a simulation, the platform detects three types of deviations: premature steering, approach deviation, and fork adjustment lag. These three deviations are then linked together as a deviation chain. Later, in the actual vehicle training records for the same subject, the platform finds that this deviation chain often triggers voice alarms, parking safety measures, or even instructor intervention. Therefore, this deviation chain is placed at the top of the error amplification table. When the trainee enters actual vehicle training, the platform, based on the current motion perception results, discovers that while the trainee has completed the first few operation segments, they are still prone to repeating the same types of deviations in the approach and adjustment phases. Therefore, it automatically schedules relevant subjects and operation segments in advance, allowing the trainee to focus on practicing these parts first. If the trainee's actual operation improves during training, the platform updates the training sequence and training records accordingly. This way, instructors see not only the performance level but also which problems are most likely to be amplified in actual vehicle training, what should be practiced first, and whether the risks have been truly eliminated after training.

[0024] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A forklift driver training management method based on data visualization analysis, characterized in that, include: S1. Collect the student's manipulation data, position data and subject result data in the simulation training. Based on the manipulation data and position data, perform motion perception, determine the corresponding motion state at each moment, and perform alignment and combination of motion state, manipulation data, position data and subject result data according to the same student, the same subject and time sequence to generate a simulation training record table. S2. Based on the simulation training record table, extract the deviation actions, the operation segments where the deviations are located, and the sequential relationship between adjacent deviations in each subject according to time sequence, and connect the first and last deviation actions in sequence to generate a deviation linked list. S3. Retrieve the actual vehicle training records corresponding to the subject, and according to the positional correspondence of the same operation segment in the same subject, substitute each deviation action in the deviation chain into the operation segment corresponding to the actual vehicle training record, and count the alarm, parking protection and instructor takeover results caused by each deviation action in the subsequent operation segment, and generate a deviation consequence correspondence table. S4. Based on the deviation consequence correspondence table, accumulate the corresponding number of alarms, number of parking protections, and number of instructor takeovers for each deviation chain, and arrange each deviation chain from high to low according to the cumulative results to generate an error amplification table.

2. The forklift driver training management method based on data visualization analysis according to claim 1, characterized in that, Also includes: S5. For the error amplification table corresponding to trainees waiting for actual vehicle training, and in combination with the motion perception results in actual vehicle training, rearrange the order of entry into actual vehicle training subjects and the order of operation practice within subjects according to the arrangement order of each deviation chain in the error amplification table. Write the rearranged actual vehicle training results and the error amplification table into the visual training archive to generate training management results.

3. The forklift driver training management method based on data visualization analysis according to claim 2, characterized in that, S1 includes: S11. Perform simultaneous splicing of the steering wheel angle, pedal travel and control handle displacement in the control data according to the sampling time, and perform position correspondence on the vehicle position, vehicle orientation and fork position in the position data according to the same time, and generate a time action segment table. S12. Based on the action segment table, extract action change groups according to the changes in steering wheel angle, pedal travel, control handle displacement, vehicle body displacement and fork displacement at adjacent time points, and merge action change groups with the same direction of change and consecutive positions in multiple consecutive time points to generate an action segment table. S13. Based on the action segment table, determine the starting state, turning state, approach state, pick-up / placement state or adjustment state corresponding to each action segment according to the correspondence between the changes in steering wheel angle, pedal travel, control handle displacement and the changes in vehicle body displacement and fork displacement in each action segment, and write them into each time point in chronological order to generate the action state.

4. The forklift driver training management method based on data visualization analysis according to claim 3, characterized in that, S2 includes: S21. Read the simulation training record table, determine the deviation actions corresponding to each moment according to the positional deviation, timing deviation and state switching deviation between the action status at each moment within the same subject and the corresponding subject result data, and generate a deviation action table according to the operation segment corresponding to the moment where each deviation action is located. S22. For the deviation action table, extract the time interval between each deviation action, the change in the operation segment, and the direction of action status continuation according to the time sequence. For deviation action pairs where there are no other deviation actions inserted between the end time of the previous deviation action and the start time of the next deviation action, and the operation segment where the next deviation action is located is a continuation of the operation segment where the previous deviation action is located or the next sequential operation segment, determine them as continuation deviation pairs and generate a deviation continuation table. S23. Based on the deviation connection table, take the deviation action without a preceding deviation pair as the first deviation action of the chain, and connect the subsequent deviation actions in sequence along the connection order of the deviation pairs until the deviation action with no subsequent deviation pair is reached, thus generating a deviation chain list.

5. The forklift driver training management method based on data visualization analysis according to claim 4, characterized in that, S3 includes: S31. Retrieve the operation segment records, vehicle status records, alarm records, parking protection records, and instructor takeover records from the actual vehicle training records. According to the rule that the operation segment numbers are consistent, the start and end positions of the operation segments are corresponding, and the direction of movement of the operation segments are consistent in the same subject, each deviation action in the deviation chain list is mapped to the target operation segment in the actual vehicle training records. Then, the results are tracked along the sequential operation segments after the target operation segment to generate a deviation substitution table. S32. Based on the deviation substitution table, extract the operation segment number sequence, vehicle status sequence, alarm number sequence, parking protection number sequence, and instructor takeover number sequence corresponding to each deviation action. Determine the records with consecutive operation segment numbers, vehicle status change direction consistent with deviation action direction, and alarm number sequence, parking protection sequence, and instructor takeover number sequence corresponding to each other in the order of operation segments as valid result chains, and generate a deviation result chain list.

6. The forklift driver training management method based on data visualization analysis according to claim 5, characterized in that, S3 also includes: S33. Based on the deviation result chain list, count the number of alarms, the number of parking protections, and the number of instructor takeovers for the same deviation action, and write them into the result location group according to the operation segment in which the alarm occurs, the operation segment in which the parking protection occurs, and the operation segment in which the instructor takeover occurs; merge the deviation result chains with the same result location group and the same number of times and write them into the deviation consequence correspondence table; split the deviation result chains with different result location groups or different numbers of times according to the operation segment number order and write them into the deviation consequence correspondence table.

7. The forklift driver training management method based on data visualization analysis according to claim 6, characterized in that, S4 includes: S41. Retrieve the deviation consequences correspondence table, collect the corresponding alarm counts, parking protection counts, and instructor takeover counts according to the same deviation chain, and sequentially splice the alarm occurrence operation segment, parking protection occurrence operation segment, and instructor takeover occurrence operation segment corresponding to each deviation chain to generate a deviation chain result table. S42. Calculate the total number of alarms, the total number of parking protections, and the total number of instructor interventions for each deviation chain in the deviation chain result table, and sort them in the order of the total number of instructor interventions, the total number of parking protections, the total number of alarms, and the result location group number to generate a deviation chain sorting table. S43. Write each deviation chain in the deviation chain sorting table into the corresponding sorting position, total number of alarms, total number of parking protection times, and total number of instructor takeover times in the sorting order to generate an error amplification table.

8. The forklift driver training management method based on data visualization analysis according to claim 7, characterized in that, S5 includes: S51. Based on the error amplification table corresponding to the trainees to be trained in the actual vehicle, the current actual vehicle training record and the action perception result, expand the number of alarms, the number of parking protections, the number of instructor takeovers, the target subjects and the target operation segments corresponding to each deviation chain into a deviation matrix in a fixed field order. Expand the current action state, the current deviation action and the completed operation segments in the current action perception result into a state matrix. Perform segment alignment, difference expansion and orthogonal projection on the deviation matrix and the state matrix to solve the segment deviation amount and subsequent continuation amount of each deviation chain to the current training stage and generate a deviation correction table. S52. Construct a real-vehicle training continuity graph based on the deviation correction table. Subject entry relationships form subject edges, intra-segment practice relationships form segment edges, completed operation segments prevent re-entry relationships of incomplete operation segments form blocking edges, and segment deviation and subsequent continuity amounts form correction edges. Perform graph traversal, conflict edge deletion, and sequential reconnection on the real-vehicle training continuity graph. After reconnection, encode each training path in sequence according to the number of times the instructor takes over, the number of times the parking protection is applied, the number of alarms, the segment deviation amount, and the operation segment number, and then sort them in lexicographical order. Output the rearranged training sequence.

9. The forklift driver training management method based on data visualization analysis according to claim 8, characterized in that, S5 also includes: S53. Organize real vehicle training according to the rearranged training sequence and continuously receive motion perception results. Align the current operation segment sequence with the execution sequence of the rearranged training sequence. Retain the original continuation relationship for the consistent parts after alignment. Perform posterior consistency update on the alarm results, parking protection results and coach takeover results after conditionally decomposed the offset parts after alignment. Write the remaining operation segments after consistency update back to the real vehicle training continuation diagram and perform conflict edge deletion and sequential reconnection again until the remaining operation segment sequence completely corresponds to the current operation segment sequence. Then write the corresponding real vehicle training results, rearranged training sequence and error amplification table into the visual training archive to generate training management results.

10. A forklift driver training management platform based on data visualization analysis, used to implement the forklift driver training management method based on data visualization analysis as described in any one of claims 1-9, the platform comprising a filing module, a deviation concatenation module, a consequence mapping module, a sorting module, and a reordering control module, characterized in that: The filing module is used to collect the student's manipulation data, position data, and subject result data in the simulation training. Based on the manipulation data and position data, it performs motion perception to determine the corresponding motion state at each moment. It also performs alignment and combination of motion state, manipulation data, position data, and subject result data according to the same student, the same subject, and the same time sequence to generate a simulation training record table. The deviation chaining module, based on the simulation training record table, extracts the deviation actions, the operation segments where the deviations are located, and the sequential relationship between adjacent deviations in each subject according to time sequence, and chains the first and last deviation actions in sequence to generate a deviation chain list. The consequence mapping module is used to retrieve the actual vehicle training records corresponding to the subject, and according to the position correspondence of the same operation segment in the same subject, substitute each deviation action in the deviation chain into the operation segment corresponding to the actual vehicle training record, and count the alarm, parking protection and instructor takeover results caused by each deviation action in the subsequent operation segment, and generate a deviation consequence correspondence table. The sorting module, based on the deviation consequence correspondence table, accumulates the corresponding number of alarms, parking protection times, and instructor takeover times for each deviation chain, and sorts each deviation chain from high to low according to the cumulative results, generating an error amplification table. The reordering control module is used to reorder the order of entry into the actual training subjects and the order of operation practice within the subjects according to the order of each deviation chain in the error amplification table for trainees to be trained in the actual vehicle, combined with the action perception results in the actual vehicle training. The reordered actual vehicle training results are written into the visual training archive corresponding to the error amplification table to generate training management results.

Images