A method for determining driving eligibility based on a multi-point physiological characteristic input

Abstract

This invention provides a method for determining driving qualifications based on physical condition levels using multi-point physiological feature inputs, relating to the field of machine learning technology. The method comprises the following steps: S1, determining a first risk factor based on the driver's static physiological function limitation parameters; S2, acquiring driving behavior data generated by the driver while performing a preset driving task in a VR environment to determine a second efficiency factor; S3, calculating a compensation efficiency load index based on the first risk factor and the second efficiency factor; S4, generating a disturbance command for secondary disturbances based on the compensation efficiency load index; and S5, adjusting the dynamic safety rating associated with the driving qualification level based on the analysis results of micro-instability characteristics. This invention integrates abstract decision parameters into a dynamically evolving closed-loop control; and simultaneously drives the virtual environment and analysis algorithm to adaptively adjust the control concept, converging the complex dynamic evaluation process into a precise dynamic safety rating and establishing a clear mapping relationship with the driving license level.

Description

Technical Field

[0001] This invention relates to the field of machine learning technology, specifically to a method for determining driving qualifications based on the level of physical condition input from multiple physiological features. Background Technology

[0002] In the fields of traffic safety and public transportation, accurate and reliable assessment of drivers' qualifications, especially those engaged in commercial operations or with specific physical limitations, is a core element in ensuring public safety. With technological advancements, assessment methods are evolving from traditional static physiological indicator checks to comprehensive and dynamic assessments utilizing advanced technologies such as virtual reality (VR) and machine learning. This aims to obtain more objective and quantifiable assessment results that closely resemble real-world driving scenarios. By using the driver's static physiological indicators as input and combining them with virtual technology to simulate the driver's skill level, it facilitates the determination of driving license levels, particularly enabling accurate assessment of driving license levels for people with disabilities.

[0003] Existing technologies face the following main challenges in achieving the above objectives:

[0004] Current evaluation methods that employ machine learning models (such as deep neural networks) have improved the ability to perform nonlinear analysis on static physiological data (such as vision loss or limb amputation), but their evaluation basis remains a static snapshot detached from driving tasks. While these methods can establish a correlation between physiological conditions and qualification levels, they cannot reveal how these physiological limitations specifically affect driver behavior and compensatory abilities under dynamic driving loads, resulting in evaluation conclusions lacking validation support in dynamic scenarios.

[0005] Existing assessment systems, whether based on traditional rules (such as decision trees) or advanced models, essentially play the role of passive observers. They evaluate driver behavior by collecting data, but cannot actively apply controlled stress tests to detect the driver's safety margins and potential fragility characteristics as they approach their capability limits. Therefore, existing technologies lack effective means to detect critical states that appear normal on the surface but have extremely low safety redundancy. Summary of the Invention

[0006] The purpose of this invention is to provide a method for determining driving qualifications based on the level of physical condition input of multi-point physiological characteristics, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A method for determining driving qualifications based on physical condition levels using multi-point physiological feature input, comprising the following steps:

[0009] S1. Based on at least one static physiological function limitation parameter of the driver, a first risk factor is determined through a preset driving function fault tree model to characterize the theoretical failure risk in a preset driving task in a virtual reality environment.

[0010] S2. Obtain driving behavior data generated by the driver when performing a preset driving task in the VR environment, and analyze the driving behavior data through a compensation effectiveness analysis model to obtain an adaptive strategy effectiveness index, so as to determine the second effectiveness factor used to characterize the driver's actual compensation ability.

[0011] S3. Based on the first risk factor and the second efficiency factor, calculate the compensation efficiency load index, which characterizes the degree of compensatory effort required by the driver to achieve the observed driving performance.

[0012] S4. When the compensation efficiency load index meets the preset disturbance triggering conditions, a disturbance command is generated to inject secondary disturbances into the VR environment.

[0013] S5. In response to the generation of disturbance instructions, trigger the analysis of micro-instability characteristics of driving behavior data, and adjust the dynamic safety rating associated with the driving qualification level based on the analysis results of micro-instability characteristics.

[0014] Furthermore, the first risk factor in S1 is the top event failure rate of the preset driving task obtained by using the static physiological function limitation parameters as basic failure events through a driving function fault tree model and performing probability propagation calculations.

[0015] And / or, the compensation effectiveness analysis model in S2 is a pre-defined model for analyzing time-series data of driving behavior; the second effectiveness factor includes the success rate of the observed task calculated based on the driving behavior data, and the adaptive strategy effectiveness index output by the compensation effectiveness analysis model.

[0016] And / or, the specific steps in S3 for calculating the compensation efficiency load index are: calculating the compensation efficiency load index that characterizes the functional relationship between the second efficiency factor and the first risk factor;

[0017] And / or, the preset disturbance triggering condition in S4 is: the instantaneous value of the compensation efficiency load index in the current time window exceeds a preset multiple of the moving average value in the past time window;

[0018] And / or, the micro-instability features in S5 include: the increase in steering wheel vibration energy of the steering wheel angle signal within a preset frequency band, and the rate of change of the gaze entropy value of the driver's eye gaze point distribution.

[0019] Furthermore, the calculation steps for the first risk factor and the second effectiveness factor are as follows:

[0020] Input the driver's static physiological function limitation parameters and the driving behavior data of the driver recorded by the VR scene engine during the execution of basic events to form a multi-dimensional time series data segment;

[0021] Based on the input static physiological function limitation parameters, the corresponding baseline failure rate of the basic event is retrieved from the physiological limitation-baseline failure rate lookup table; the VR scene engine calculates the scene complexity factor based on the current task configuration.

[0022] Based on the baseline failure rate of the basic event and the scenario complexity factor, the failure rate after dynamic modulation is calculated; the failure rate after dynamic modulation is used as the probability of the basic event and input into the preset driving function fault tree model; the driving function fault tree model also contains basic events affected by different physiological limitations; the driving function fault tree model performs probability propagation calculation and finally outputs the top event failure rate, which is the first risk factor.

[0023] The recorded complete driving behavior data is input into the encoder of a pre-trained Siamese neural network; the behavior sequence embedding vector is output; the distance between the behavior sequence embedding vector and the standard behavior center vector of the base event is calculated, and the success rate of the observation task is calculated accordingly.

[0024] The weight distribution is extracted from the self-attention layer of the network, and the attention weight distribution entropy is calculated. Based on the attention weight distribution entropy, the effectiveness index of the adaptive strategy is calculated.

[0025] Output the first risk factor and the second effectiveness factor.

[0026] Furthermore, the calculation steps for the compensation efficiency load index are as follows:

[0027] Input the first risk factor and the second effectiveness factor to observe the task success rate and adaptive strategy effectiveness index;

[0028] The product of the input observation task success rate and the adaptive strategy effectiveness index is multiplied and then the square root is taken to calculate the comprehensive compensation performance score; the efficiency contribution nonlinear index is obtained; the comprehensive compensation performance score is used as the base and the efficiency contribution nonlinear index is used as the exponent, and the power operation is performed to obtain the load numerator.

[0029] Obtain the first risk factor; subtract the first risk factor from the constant 1 to obtain the median value representing the safety margin; obtain the risk aversion nonlinear index; use the safety margin as the base and the risk aversion nonlinear index as the exponent, and perform exponentiation; add a very small smoothing constant to the result; obtain the final load denominator.

[0030] Divide the obtained load numerator by the obtained load denominator to obtain the preliminary compensation efficiency load index;

[0031] Set the lower and upper limits of the output; if the calculated value is less than the lower limit, the final output is the lower limit; if it is greater than the upper limit, the output is the upper limit; otherwise, output the original calculated value; the final output is the compensation efficiency load index after amplitude limiting.

[0032] Furthermore, the calculation steps for the dynamic safety rating are as follows:

[0033] At the beginning of each time step, input the compensation efficiency load index and continuously monitor the health status of the data streams for the first risk factor and the second efficiency factor;

[0034] Check the latest timestamp of the first risk factor. If the difference between the current time and the timestamp exceeds the information source interruption timeout threshold, it is determined that the acquisition of the first risk factor has been interrupted, and the system immediately jumps to the first degraded mode.

[0035] Check the latest timestamp of the data stream of the second performance factor; if the difference between the current time and the timestamp exceeds the source interruption timeout threshold, it is determined that the acquisition of the second performance factor is interrupted, and immediately jump to the second degradation mode;

[0036] If both the first risk factor and the second efficiency factor are obtained normally, continue; store the latest compensation efficiency load index into the short-term and long-term sliding windows; calculate the long-term moving average and standard deviation of the compensation efficiency load index; calculate the adaptive trigger multiple; if the instantaneous value of the compensation efficiency load index is greater than the long-term moving average multiplied by the adaptive trigger multiple, the trigger condition is determined to be met, and proceed to the next step; otherwise, end this loop.

[0037] Calculate the overshoot of the compensation efficiency load index, which is the instantaneous value minus the long-term moving average multiplied by the adaptive triggering factor; calculate the specific disturbance intensity value based on the overshoot and disturbance intensity mapping factor; generate a disturbance command containing the disturbance type, the calculated intensity, and the preset duration.

[0038] The disturbance command is sent to the VR scene engine actuator in the technical environment; after the disturbance command is issued, the high-frequency data analysis mode is immediately entered; the steering wheel vibration energy increment and the rate of change of gaze entropy value are calculated in parallel; and the final attitude instability comprehensive score is calculated by fusion.

[0039] Based on the comprehensive attitude instability score, a penalty deduction is applied to the rating; the penalty deduction is equal to the maximum dynamic safety rating multiplied by the comprehensive attitude instability score; the dynamic safety rating is obtained by subtracting the penalty deduction from the maximum dynamic safety rating; the current cycle ends.

[0040] The first downgrade mode involves stopping the calculation of S3 and S4, and directly calculating the downgraded dynamic security rating based on the second performance factor output by S2; the final report clearly states that the assessment results are based on performance data, do not include theoretical risk analysis, and have reduced confidence.

[0041] The second downgrade mode is to stop the calculation of S3 and S4; based on the first risk factor output by S1, directly calculate the downgraded dynamic security rating; and clearly indicate in the final report that the assessment results are based on risk projection and do not include actual performance verification, thus reducing the confidence level.

[0042] The final output of each loop is either an updated dynamic security rating or a suboptimal rating with explicit degrading instructions.

[0043] Furthermore, based on the dynamic safety rating, the obtainable driver's license level is determined as follows:

[0044] If the dynamic safety rating is ≥0.85, then you can obtain an A1 driver's license for large passenger vehicles;

[0045] If the dynamic safety rating is 0.82 or less and the rating is less than 0.85, then a driver's license for city buses (A3) can be obtained.

[0046] If the dynamic safety rating is 0.80 or less and the rating is less than 0.82, then an A2 driver's license can be obtained for towing trailers.

[0047] If the dynamic safety rating is 0.75 or less and the rating is less than 0.80, then a B1 driver's license for medium-sized passenger vehicles can be obtained.

[0048] If the dynamic safety rating is 0.70 or less and the rating is less than 0.75, then a B2 driver's license for large trucks can be obtained.

[0049] If the dynamic safety rating is 0.65 or less and the rating is less than 0.70, then a C1 or C2 driver's license for small cars can be obtained.

[0050] If the dynamic safety rating is 0.60 ≤ Dynamic Safety Rating < 0.65, then a Class D driver's license for ordinary three-wheeled motorcycles can be obtained.

[0051] If the dynamic safety rating is less than 0.60, no driver's license can be obtained.

[0052] Furthermore, the static physiological function limitation parameters are obtained through a human body model device capable of acquiring physical condition limitation parameters. The human body model device includes a human body model body with assessment key parts divided and set according to driving operation needs, switch components respectively set on each assessment key part, indicator lights set on each assessment key part and controlled by the corresponding switch components, an interactive module for user operation and interaction, and a control system module for overall control. The switch components, indicator lights, and interactive module are connected to the control system module for signal or data transmission. The user's static physical function limitation parameters are obtained by the user operating the switch components or the interactive module on the human body model body.

[0053] Furthermore, it also includes a testing module for testing the user's physical functions, which is connected to the control system module for signal or data transmission.

[0054] Furthermore, the testing module includes a strength testing module for testing the strength of the user's palms and feet to obtain strength parameters of key assessment points for evaluation and judgment;

[0055] And / or, including activity sensing devices worn on key assessment points of the user's body to obtain activity trajectory parameters for assessment and judgment;

[0056] And / or, VR driving simulation devices used to allow users to simulate driving operations in order to obtain driving behavior data for evaluation and judgment.

[0057] Furthermore, a light-transmitting housing is provided for the key assessment parts and the areas defined according to the driving operation needs, and the indicator lights are set inside the light-transmitting housing;

[0058] And / or, the interaction module is set on a display screen device on the main body of the human body model or is a separate display screen device.

[0059] Compared with the prior art, the beneficial effects of the present invention are:

[0060] This invention uses a driving function fault tree model based on deductive reasoning to transform the driver's static physiological function limiting parameters into quantifiable first risk factors in a preset driving task, thus constituting a priori assessment of theoretical risk. In a highly immersive virtual reality driving environment, a compensation efficacy analysis model based on deep learning is used to conduct in-depth analysis of the driver's actual driving behavior time series data and extract a second efficacy factor that can characterize driving skills and strategy efficiency.

[0061] This invention does not simply combine these two factors linearly, but instead fuses them into a novel decision parameter with profound physical meaning through a nonlinear power-law model: the Compensation Performance Load Index. The underlying logic of the Compensation Performance Load Index is to quantify the degree of compensatory effort a driver makes to achieve the observed driving performance, which is disproportionate to the theoretical risk. A rapidly escalating Compensation Performance Load Index even under excellent performance is a strong signal that the driver is in a brittle compensation state. When the instantaneous value of the Compensation Performance Load Index fluctuates abnormally, it actively triggers two subsequent key steps: instructing the VR environment engine to inject a precisely calculated secondary disturbance related to the degree of index anomaly, switching the evaluation from the normal cruise mode to the stress test mode; and simultaneously switching the Compensation Performance Load Index behavior analysis algorithm to a high-frequency sampling mode, focusing on analyzing whether the driver exhibits micro-instability characteristics in steering wheel operation or visual search after the disturbance occurs.

[0062] Ultimately, the complex dynamic assessment process is converged into a precise dynamic safety rating, and a clear mapping relationship is established between the dynamic safety rating and the specific driving license level. This realizes the transformation of assessment results from abstract indicators to specific operational instructions, improves the objectivity and operability of decision-making, achieves refined risk stratification of driver capability boundaries, and contributes to the optimal allocation of social human resources. Attached Figure Description

[0063] Figure 1 This is a schematic diagram of the physical simulation of the present invention;

[0064] Figure 2 This is a schematic diagram of the overall method flow of the present invention;

[0065] Figure 3 This is a schematic diagram illustrating the calculation steps of the first risk factor and the second effectiveness factor of the present invention.

[0066] Figure 4 This is a schematic diagram illustrating the calculation steps of the compensation efficiency load index of the present invention;

[0067] Figure 5 This is a schematic diagram illustrating the calculation steps of the dynamic security rating of this invention;

[0068] Figure 6 This is a schematic diagram of the calculation steps for the degradation mode processing of the present invention;

[0069] Figure 7 This is a schematic diagram of a human model for inputting static physiological function limitation parameters according to the present invention.

[0070] In the diagram: 1. Human body model body; 11. Key assessment parts; 2. Switch components; 3. Indicator lights; 31. Transparent shell; 4. Interactive module. Detailed Implementation

[0071] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0072] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0073] Example 1:

[0074] Please see Figures 1 to 7 This invention provides a technical solution: a method for determining driving qualifications based on multi-point physiological feature input, comprising the following steps:

[0075] S1. Based on at least one static physiological function limitation parameter of the driver, a first risk factor is determined through a preset driving function fault tree model to characterize the theoretical failure risk in a preset driving task in a virtual reality environment.

[0076] The first risk factor in S1 is the top event failure rate of the preset driving task, which is obtained by using the static physiological function limitation parameters as the basic failure events through the driving function fault tree model and performing probability propagation calculation.

[0077] During the initialization phase, a structured driving function fault tree model, constructed by ergonomics and traffic safety experts, is loaded. The driving function fault tree model is a special decision tree that takes the safe completion of the driving task as the top event, each sub-driving action as an intermediate gate, and independent physiological functions as basic events. When S1 is executed, the input static physiological function limitation parameters of the driver are converted into the probability of occurrence of the corresponding basic events. Then, the probability is calculated from bottom to top through the logic gates (AND gate, OR gate) of the fault tree to finally obtain the probability of occurrence of the top event, that is, the theoretical probability of task failure, which is used as the first risk factor.

[0078] Static physiological function limitation parameters are entered through the human body model 1. That is, the physical condition of the licensed driver is obtained as needed. Switch components 2 and indicator lights 3 are set on various limbs and parts of the human body model 1. When the licensed driver's disability or physical abnormality is obtained, the switch component 2 is pressed at the corresponding position of the human body model 1, which makes the indicator light 3 light up and transmits the corresponding data to the driving function fault tree model. Different positions correspond to different physical conditions, including static physiological function limitation parameters such as amputation and disease.

[0079] S2. Obtain driving behavior data generated by the driver when performing a preset driving task in the VR environment, and analyze the driving behavior data through a compensation effectiveness analysis model to obtain an adaptive strategy effectiveness index, so as to determine the second effectiveness factor used to characterize the driver's actual compensation ability.

[0080] In S2, the compensation effectiveness analysis model is a pre-defined model used to analyze time-series data of driving behavior; the second effectiveness factor includes the success rate of the observed task calculated based on the driving behavior data, and the adaptive strategy effectiveness index output by the compensation effectiveness analysis model.

[0081] During S2 execution, driving behavior data collected in real time from the VR system (in this embodiment, including time-series data such as steering wheel angle, pedal opening, vehicle trajectory, driver eye movement, and head movement) is input into a pre-trained compensation effectiveness analysis model. In this embodiment, a temporal convolutional network or Transformer is used. The compensation effectiveness analysis model compares and analyzes the input data with a standard driving behavior database and outputs two results: a standardized observation task success score and an adaptive strategy effectiveness index that quantifies whether the driver uses an effective compensation strategy (in this embodiment, compensating for visual field defects by increasing head rotation). Together, these constitute the second effectiveness factor.

[0082] In this embodiment, the key parameters are defined as follows:

[0083] The parameter sign of the baseline failure rate is: In a standardized driving scenario with minimal complexity, the prior probability of failure of a basic driving action due to specific physiological limitations.

[0084] Calibration was achieved through large-scale offline experiments. The experimental design was as follows: a group of drivers with specific single physiological function limitations (i.e., the degree of physical disability) were recruited and repeatedly performed tasks in a standardized VR driving scenario of minimum complexity (in this embodiment, driving in response to commands on a straight road without interference). The baseline failure rate between physiological limitations and basic events was obtained by statistically analyzing the ratio of the number of failures on the corresponding basic actions to the total number of failures. The results were stored in a lookup table of physiological limitations and baseline failure rates.

[0085] The parameter sign of the scene complexity factor is A dimensionless factor quantifies the overall difficulty increment of the current VR driving task relative to a standard scene; this factor is dynamically generated by the VR scene engine based on the current task configuration; each preset driving task is associated with a set of complexity attributes, such as road curvature, traffic density level, weather conditions, and event urgency in this embodiment. The quantified values ​​of these attributes are combined using a simple linear weighted model to generate the final scene complexity factor.

[0086] The calculation logic is as follows: Obtain the road curvature value, traffic density level, and crosswind level for the current task; assign weights to each attribute respectively. In this embodiment, the road curvature weight is 0.5, the traffic density weight is 0.3, and the crosswind weight is 0.2; multiply the normalized value of each attribute by its corresponding weight; sum all the product results to obtain the scene complexity factor. Scene complexity factor The value range is limited to between 0 and 1, with 0 representing the lowest complexity;

[0087] The parameter sign for the failure rate after dynamic modulation is: The baseline failure rate of the basic event Based on the scenario complexity factor After dynamic adjustment, the actual predicted failure rate of the basic event is obtained in the current specific scenario;

[0088] The calculation logic is as follows: First, obtain the baseline failure rate of the basic event. Obtain the scene complexity factor ; The scene complexity factor Multiplying by a preset modulation slope parameter, in this embodiment, the modulation slope parameter is used to control the degree to which risk increases dramatically with complexity, and is set to a value of 4; the scene complexity factor is then... The result multiplied by the preset modulation slope parameter is added to the bias term, which is used to adjust the inflection point of the S-curve. The bias term is calculated by applying the inverse function of the logistic function, so that when the complexity factor is zero, the output probability is close to the baseline failure rate. The result is then inversely taken; the inverse is then subjected to the natural exponential function operation; the result is incremented by one, and then divided by the result by the constant 1. The resulting value is the failure rate after dynamic modulation. ;

[0089] The parameter sign for the top event failure rate is: Given a fault tree structure for driving functions, the top-event failure rate of the entire composite driving task is calculated through logic gate probabilistic operations based on all relevant dynamic modulation failure rates. This is the first risk factor.

[0090] The parameter notation of the behavior sequence embedding vector is The mathematical vector obtained by mapping a multidimensional driving behavior time-series data (steering wheel angle, pedal opening, head movement posture, etc. in this embodiment) to a high-dimensional feature space through the encoder of the Siamese neural network is a mathematical vector that can condense the behavior style and pattern of the segment.

[0091] The parameter sign of the standard behavior center vector is In a pre-built database containing a large number of behavioral samples of excellent drivers under the same task, the cluster centers or average vectors of the behavioral sequence embedding vectors of all samples represent an ideal or standard execution mode of the task.

[0092] The symbol for the parameter of observation mission success is: This is an indicator that quantifies the gap between the current driver's actual performance and their ideal performance. The calculation logic is as follows: obtain the current driver's behavior sequence embedding vector. Retrieve the standard behavior center vector for the corresponding task from the pre-stored database. ; Calculate the embedding vector of the behavior sequence and standard behavior center vector The Euclidean distance between the two targets is calculated. This Euclidean distance is then mapped using a negative exponential function, which smoothly converts the distance into a success score between 0 and 1; the smaller the distance, the closer the score is to 1. The result of mapping the Euclidean distance using the negative exponential function is then normalized to ensure the output is strictly within the range of 0 to 1. The resulting value represents the success score of the observation task. ;

[0093] The parameter sign of the attention weight distribution entropy is: The value obtained by calculating the information entropy of the weight values ​​assigned to each time step of the input behavior sequence by the self-attention mechanism quantifies whether the key control points of the driver are concentrated in a few moments or dispersed throughout the entire process when performing the task.

[0094] The parameter symbol of the adaptive strategy effectiveness index is: This study quantifies the efficiency and accuracy of driver control strategies. The calculation logic is as follows: The time-step weight distribution for the current behavior sequence is extracted from the self-attention layer of the Siamese neural network. Based on the Shannon entropy definition in information theory, the information entropy of the time-step weight distribution is calculated to obtain the attention weight distribution entropy. Subtract the currently calculated entropy value from the preset maximum possible entropy value (i.e., the entropy under uniform distribution); divide the result by the maximum possible entropy value and normalize; the normalized value is the adaptive strategy effectiveness index. Attention weight distribution entropy The lower the value (the more focused and effective the strategy), the higher the adaptive strategy effectiveness index. The closer it is to 1, the closer it is to 0.

[0095] The calculation steps for the first risk factor and the second efficiency factor are as follows: input the driver's static physiological function limitation parameters and the driving behavior data of the driver recorded by the VR scene engine during the execution of basic events to form a multi-dimensional time series data segment;

[0096] Based on the input static physiological function limitation parameters, the corresponding baseline event failure rate is retrieved from the physiological limitation-benchmark failure rate lookup table. The VR scene engine calculates the scene complexity factor based on the current task configuration. ;

[0097] Based on the baseline failure rate and scene complexity factor The failure rate after dynamic modulation was calculated. The failure rate after dynamic modulation The probabilities of the basic events are input into a pre-defined driving function fault tree model; the driving function fault tree model also contains basic events affected by different physiological limitations; the driving function fault tree model performs probability propagation calculations and finally outputs the failure rate of the top event. This is the first risk factor;

[0098] The recorded complete driving behavior data is input into the encoder of a pre-trained Siamese neural network; the output is a behavior sequence embedding vector. ; Calculate the embedding vector of the behavior sequence Standard behavior center vector of the basic event The distance between them was used to calculate the success rate of the observation mission. ;

[0099] The weight distribution is extracted from the self-attention layer of the network, and the attention weight distribution entropy is calculated. According to the attention weight distribution entropy Calculate the effectiveness index of adaptive strategies ;

[0100] Output the first risk factor and the second effectiveness factor.

[0101] S3. Based on the first risk factor and the second efficiency factor, calculate the compensation efficiency load index, which characterizes the degree of compensatory effort required by the driver to achieve the observed driving performance.

[0102] The specific steps for calculating the compensation efficiency load index in S3 are as follows: calculate the compensation efficiency load index that characterizes the functional relationship between the second efficiency factor and the first risk factor;

[0103] During S3 execution, the first risk factor from S1 and the second performance factor from S2 are Z-score normalized to eliminate the influence of dimensions. The performance index of the compensation performance load index is the brittle failure prediction accuracy, which is the percentage of times micro-instability characteristics are successfully induced and detected after the stress test is triggered by the compensation performance load index.

[0104] In this embodiment, the key parameters are defined as follows:

[0105] The parameter symbol for the first risk factor is: ; represents the probability that the driver will theoretically fail due to physiological limitations in the current driving task, with a value range of [0,1].

[0106] The second performance factor is determined by the success rate of the observation task. and the effectiveness index of adaptive strategies Composition; representing the quality of the driving task completion result and the efficiency of the process, respectively, with a value range of [0,1];

[0107] The parameter symbol for the comprehensive compensation performance score is: It is a single comprehensive performance score that integrates the task completion result and process quality, with a value range of [0,1].

[0108] The calculation logic is as follows: Obtain the success rate of the observation task. and the effectiveness index of adaptive strategies ; Observation mission success rate and the effectiveness index of adaptive strategies Multiply; then take the square root of the product, and the resulting value is the comprehensive compensation performance score. ;

[0109] The parameter sign of the nonlinear exponent of efficiency contribution is: Controlled comprehensive compensation performance score The power exponent of the contribution reflects the nonlinear impact of performance improvement on the compensation efficiency load index.

[0110] The parameter sign of the risk aversion nonlinear index is: The power exponent of the contribution of the control safety margin (derived from the first risk factor) reflects the change in the sensitivity of the compensation efficiency load index as risk increases.

[0111] The calculation logic is as follows: A non-linear index for performance contribution is set. and risk aversion nonlinear index The search range in this embodiment is from 0.5 to 2.5, with a step size of 0.1, forming a parameter grid;

[0112] For each pair in the parameter grid The value is used to iterate through the entire training dataset and calculate the corresponding compensation efficiency load index for each sample. The calculated compensation efficiency load index The series values ​​are used as a single feature, and a gradient boosting decision tree classifier model (LightGBM or XGBoost in this embodiment) is used to train and cross-validate the label of whether micro instability has occurred.

[0113] Record in the present Below, the area under the receiver operating characteristic curve (AUC) achieved by the gradient boosting decision tree classifier model on the cross-validation set; AUC serves as a performance metric, with higher values ​​representing a higher compensation efficiency load index. The stronger the discriminative power of a predictive feature;

[0114] Repeat the calculation until all parameter grids have been traversed; select the pair that maximizes the cross-validation AUC value. , which is the parameter value ultimately used in this embodiment;

[0115] The parameter symbol for the compensation efficiency load index is: A dimensionless index that quantifies the cognitive-motor system load expended by the driver to achieve the current performance;

[0116] Compensation efficiency load index The calculation steps are as follows:

[0117] Enter the first risk factor Success rate of the observation mission of the second effectiveness factor and the effectiveness index of adaptive strategies ;

[0118] Success rate of the observation task on the input and the effectiveness index of adaptive strategies Perform the operation of taking the square root of the product to calculate the comprehensive compensation performance score. ; Nonlinear index of efficiency contribution In this embodiment, the efficiency contribution nonlinear index The comprehensive compensation performance is 1.2; As the base, the nonlinear exponent of efficiency contribution As an exponent, it undergoes exponentiation to obtain an intermediate result, which is called the load numerator;

[0119] Obtain the first risk factor Subtract the first risk factor from a constant of 1. This yields the median value representing the safety margin; and obtains the risk aversion nonlinear index. In this embodiment, the risk aversion nonlinear index The value is 1.5; using the safety margin as the base, the risk aversion nonlinear index... As an exponent, it performs exponentiation;

[0120] To prevent the first risk factor When the value approaches 1, the safety margin approaches 0, leading to a calculation overflow due to a zero denominator. A very small smoothing constant is added to the result. The smoothing constant is used to ensure the numerical stability of the calculation. In this embodiment, the smoothing constant is set to one-thousandth. The final load denominator is obtained.

[0121] Dividing the obtained load numerator by the obtained load denominator yields the preliminary compensation efficiency load index. ;

[0122] Set an output lower limit and an output upper limit; if the calculated value is less than the output lower limit, the final output is the output lower limit; if it is greater than the output upper limit, the output is the output upper limit; otherwise, output the original calculated value. In this embodiment, the output lower limit is 0 and the output upper limit is 100. The final output is the compensation efficiency load index after amplitude limiting. .

[0123] S4. When the compensation efficiency load index meets the preset disturbance triggering conditions, a disturbance command is generated to inject secondary disturbances into the VR environment.

[0124] The preset disturbance trigger condition in S4 is: the instantaneous value of the compensation efficiency load index in the current time window exceeds the preset multiple of the moving average value in the past time windows;

[0125] During S4 execution, two time series of the compensation efficiency load index are continuously maintained: a short-term (set to 3 seconds in this embodiment) instantaneous value series and a long-term (set to 30 seconds in this embodiment) moving average series. When the instantaneous value is detected to exceed 1.5 times the long-term moving average, the disturbance trigger condition is met, and a specific disturbance instruction that can be parsed by the VR engine is generated. The disturbance instruction includes parameters such as disturbance type, intensity and duration.

[0126] S5. In response to the generation of disturbance commands, trigger the analysis of micro-instability characteristics of driving behavior data, and adjust the dynamic safety rating associated with the driving qualification level based on the analysis results of micro-instability characteristics.

[0127] The micro-instability characteristics in S5 include: the increase in steering wheel vibration energy of the steering wheel angle signal within the preset frequency band, and the rate of change of the gaze entropy value of the driver's eye gaze point distribution;

[0128] When S5 is executed, in response to the disturbance command generated by S4, a fast Fourier transform is performed on the steering wheel angle signal in high-frequency sampling mode, and the energy in the frequency band of unconscious human muscle tremors (set to 6-12Hz in this embodiment) is calculated; if the energy value increases after the disturbance occurs, it is determined that a micro-instability feature has appeared.

[0129] When an interruption in the data flow of the first risk factor or the second effectiveness factor is detected, S3 and S4 are automatically bypassed, and the system switches to the preset degradation assessment mode based on the uninterrupted data source.

[0130] In this embodiment, the key parameters are defined as follows:

[0131] The parameter sign for short-term window length is: Used to calculate the compensation efficiency load index The length of the sliding time window for the instantaneous value;

[0132] The symbol for the long-term window length parameter is: Used to calculate the compensation efficiency load index The length of the sliding time window for the long-term moving average and standard deviation;

[0133] Short window length Designed to capture rapid load changes, the timeframe is set to 3 seconds; 3 seconds is sufficient to encompass a typical driving micro-operation, which in this embodiment is a slight correction of direction, yet short enough to reflect instantaneous changes; the long-term window length... The aim is to establish a stable baseline, with the value set at 30 seconds; 30 seconds covers multiple driving events, which can effectively smooth out short-term noise and form a stable assessment of the driver's recent state.

[0134] The symbol for the reference trigger multiple is: In the ideal scenario where the driver's historical performance is extremely stable (i.e., volatility is zero), the compensation efficiency load index is used to determine the compensation efficiency load index. Is the base magnification abnormal?

[0135] The parameter symbol for the volatility adjustment coefficient is: Used to adjust the driver's historical compensation efficiency load index The coefficient representing the degree to which volatility affects the trigger multiple;

[0136] The parameter symbol for the adaptive trigger multiplier is: A personalized multiplier for triggering disturbances, dynamically adjusted based on the driver's individual historical volatility.

[0137] The calculation logic is as follows: Obtain the baseline trigger multiple. In this embodiment, the value is taken as 1.5, which is used to calculate the long-term window length. Within the window, the compensation efficiency load index Standard deviation of time series; obtaining volatility adjustment coefficient This controls the increase in the standard deviation relative to the benchmark multiple; in this embodiment, it is set to 0.8 through offline calibration. The obtained standard deviation is then compared with the volatility adjustment coefficient. Multiply; multiply the base trigger factor Adding the product together yields the adaptive trigger multiplier. ;

[0138] The parameter sign of the disturbance intensity mapping factor is The compensation efficiency load index The degree exceeding the trigger threshold is linearly mapped to a proportional coefficient of the intensity of the disturbance event in the virtual reality environment;

[0139] Different intensities of disturbances were artificially introduced, and the overshoot-disturbance intensity relationship that effectively distinguished between brittle and robust drivers was observed. Finally, a linear mapping relationship was determined, with the slope representing the disturbance intensity mapping factor. In this embodiment, the disturbance intensity mapping factor The compensation efficiency load index is set to 0.2. For every unit increase in overshoot, the disturbance intensity in the VR environment (in this embodiment, the disturbance is set to crosswind speed) increases by 0.2 m / s;

[0140] The parameter symbol for the lower limit of the frequency band for flutter analysis is: The parameter sign for the upper limit of the flutter analysis frequency band is... Together, they defined the target frequency band for analyzing steering wheel vibration energy.

[0141] The symbol for the parameter of steering wheel vibration energy increment is: The normalized value of the energy increment of the steering wheel angle signal in the target frequency band relative to the energy before the disturbance within a specific time period after the disturbance occurs.

[0142] The calculation logic is as follows: Set the lower limit of the flutter analysis frequency band. The upper limit of the flutter analysis frequency band is 6 Hz. The frequency is 12 Hz, which closely matches the frequency of involuntary muscle tremors in humans under stress or fatigue. Steering wheel angle time-series data were extracted one second before and one second after the disturbance. Short-time Fourier transforms were performed on both segments of the steering wheel angle time-series data to obtain their respective power spectral densities. The integrals of the two power spectra within the 6-12 Hz frequency band were calculated to obtain the energy before the disturbance. and energy after disturbance ; using the energy after perturbation Energy before disturbance Divide the result by the energy before the perturbation. The energy increment ratio is obtained; this ratio is then mapped and normalized using the Sigmoid function to obtain the final value range of steering wheel vibration energy increment in [0,1]. ;

[0143] The notation for the parameter of the grid size for the gaze region is: When performing visual entropy analysis, the driver's instantaneous field of vision is divided into the dimensions of a two-dimensional grid;

[0144] The sign of the parameter for the rate of change of gaze entropy is: The rate of change of information entropy of the driver's gaze point distribution on the two-dimensional grid within a specific time period after the disturbance occurs;

[0145] The calculation logic is as follows: Set the mesh size for the viewing area. The driver's field of vision is divided into 100 regions, with a resolution of 10×10. The number of times the driver's eye fixation point falls on each region is counted one second before and one second after the disturbance, forming two frequency histograms. The Shannon entropy is calculated for each of the two frequency histograms to obtain the entropy value before the disturbance. and the entropy value after perturbation ; using the perturbed entropy value Subtract the entropy value before the disturbance The entropy increment is obtained; the entropy increment is divided by the entropy value before the perturbation. The rate of change is obtained; similarly, the rate of change is normalized using the Sigmoid function to obtain the rate of change of the fixation entropy value. ;

[0146] The parameter symbols for the attitude instability comprehensive score are as follows: A dimensionless score for quantifying the overall stability of a driver after being disturbed, which integrates motion control instability and cognitive state instability, with a value range of [0,1].

[0147] The calculation logic is as follows: Obtain the energy increment of steering wheel vibration. and the rate of change of gaze entropy The weighting coefficients are respectively the first weighting coefficient. Second weighting coefficient The result was obtained through an offline-trained logistic regression classifier, which uses the energy increment of steering wheel vibration. and the rate of change of gaze entropy The model is trained using the characteristics of whether or not it ultimately leads to the failure of the driving task as a label, and the resulting model coefficients are used as weights; in this embodiment, the first weight coefficient... The second weighting coefficient is 0.6. A value of 0.4 indicates that loss of steering wheel control is a stronger predictor of eventual failure than visual confusion; this represents the increase in steering wheel vibration energy. With the first weighting coefficient Multiply; the rate of change of gaze entropy With the second weighting coefficient Multiply; add the two products together, and the resulting value is the overall attitude instability score. ;

[0148] The parameter symbol for the source interruption timeout threshold is: Determine the maximum allowable period of no data to determine if the data source has been interrupted;

[0149] The calculation steps for dynamic security rating are as follows:

[0150] The initial input compensation efficiency load index at each time step (every 100 milliseconds in this embodiment) And continuously monitor the first risk factor. The health status of the data stream of the second performance factor;

[0151] Check the first risk factor The latest timestamp, if the difference between the current time and the timestamp exceeds the source interruption timeout threshold. In this embodiment, the source interruption timeout threshold If set to 500 milliseconds, then the first risk factor is determined. If the acquisition is interrupted, immediately jump to the first downgrade mode;

[0152] Check the latest timestamp of the data stream for the second performance factor; if the difference between the current time and the timestamp exceeds the source interruption timeout threshold... If the acquisition of the second performance factor is interrupted, the system will immediately switch to the second downgrade mode.

[0153] If the first risk factor If both the second performance factor and the second performance factor are normal, continue; then use the latest compensation performance load index. Store short-term and long-term sliding windows; calculate the compensation efficiency load index. The long-term moving average and standard deviation; calculate the adaptive trigger factor. If the compensation efficiency load index The instantaneous value is greater than the long-term moving average multiplied by the adaptive trigger factor. If the condition is met, proceed to the next step; otherwise, end the current loop.

[0154] Calculate the compensation efficiency load index The overshoot is the instantaneous value minus the long-term moving average multiplied by the adaptive trigger factor. Based on overshoot and disturbance intensity mapping factor The specific disturbance intensity value is calculated; a disturbance command containing the disturbance type, the calculated intensity, and the preset duration (in this embodiment, the duration is 0.8 seconds) is generated.

[0155] The disturbance command is sent to the VR scene engine actuator in the technical environment; immediately after the disturbance command is issued, the system enters high-frequency data analysis mode; and calculates the increase in steering wheel vibration energy in parallel. and the rate of change of gaze entropy The final attitude instability comprehensive score is calculated by fusion. ;

[0156] Based on the comprehensive score of posture instability The value of the rating is used to penalize the rating; in this embodiment, the penalty deduction is equal to the maximum dynamic safety rating (in this embodiment, the maximum dynamic safety rating is 100) multiplied by the attitude instability comprehensive score. The dynamic security rating is obtained by subtracting the penalty deduction from the maximum dynamic security rating; the current loop ends.

[0157] The first downgrade mode involves stopping the calculation of S3 and S4, and directly calculating the downgraded dynamic security rating based on the second performance factor output by S2 (in this embodiment, the dynamic security rating is proportional to the comprehensive compensation performance score). The final report explicitly states that the evaluation results are based on performance data and do not include theoretical risk analysis, thus reducing the confidence level.

[0158] The second downgrade mode involves stopping the calculation of S3 and S4; based on the first risk factor output by S1, the dynamic security rating of the downgrade is directly calculated (in this embodiment, the dynamic security rating is inversely proportional to the first risk factor). The final report explicitly states that the assessment results are based on risk projections and do not include actual performance verification, thus reducing the confidence level.

[0159] The final output of each loop is either an updated dynamic security rating or a suboptimal rating with explicit degradation instructions.

[0160] Based on the dynamic safety rating, the available driver's license levels are determined as follows:

[0161] If the dynamic safety rating is ≥0.85, then you can obtain an A1 driver's license for large passenger vehicles;

[0162] If the dynamic safety rating is 0.82 or less and the rating is less than 0.85, then a driver's license for city buses (A3) can be obtained.

[0163] If the dynamic safety rating is 0.80 or less and the rating is less than 0.82, then an A2 driver's license can be obtained for towing trailers.

[0164] If the dynamic safety rating is 0.75 or less and the rating is less than 0.80, then a B1 driver's license for medium-sized passenger vehicles can be obtained.

[0165] If the dynamic safety rating is 0.70 or less and the rating is less than 0.75, then a B2 driver's license for large trucks can be obtained.

[0166] If the dynamic safety rating is 0.65 or less and the rating is less than 0.70, then a C1 or C2 driver's license for small cars can be obtained.

[0167] If the dynamic safety rating is 0.60 ≤ Dynamic Safety Rating < 0.65, then a Class D driver's license for ordinary three-wheeled motorcycles can be obtained.

[0168] If the dynamic safety rating is less than 0.60, no driver's license can be obtained.

[0169] Figure 1 The isometric view on the left depicts the physical implementation scenario, in which the driver is sitting in a driving simulator that includes a steering wheel, seat, and pedals, and wearing a VR headset, performing a preset driving task in a fully immersive manner; it forms the basis for all subsequent data collection and analysis.

[0170] The dual-channel risk and effectiveness assessment flowchart corresponds to steps S1 and S2; information from the two channels is processed in parallel: based on the driver's static physiological function limitation parameters, the first risk factor characterizing the theoretical failure risk is determined through a preset driving function fault tree model; the driver's driving behavior data in the VR environment is acquired in real time, and the second effectiveness factor characterizing the actual compensation ability is determined through a compensation effectiveness analysis model.

[0171] The intrinsic load quantification and decision-making flowchart correspond to step S3; the first risk factor and the second efficiency factor obtained in the previous stage are fused through a nonlinear model to calculate the compensation efficiency load index, which is used to characterize the degree of compensatory effort required by the driver to achieve the current performance; the compensation efficiency load index is the core basis for subsequent decision-making.

[0172] The active disturbance and stress test flowchart corresponds to step S4; decisions are made based on the dynamic changes of the compensation efficiency load index; when the compensation efficiency load index meets the preset disturbance triggering conditions, the system no longer passively observes, but actively generates a disturbance command to apply a controllable secondary disturbance for stress testing to the driver in the VR environment.

[0173] The micro-instability analysis and rating update flowchart corresponds to step S5; in response to the generation of the disturbance command, the driver's driving behavior data after the disturbance is immediately analyzed at high frequency to capture micro-instability characteristics, and based on the analysis results of the micro-instability characteristics, the dynamic safety rating associated with the driving qualification level is finally adjusted to complete a complete closed-loop assessment.

[0174] Furthermore, based on the usage requirements of the above-mentioned method, the present invention further improves the human body model body for obtaining the user's static body function limitation parameters, thereby constructing a human body model device that is more conducive to the accurate acquisition of static body function limitation parameters, such as... Figure 7 As shown. The core of the human body model device lies in providing an intuitive and efficient tool for obtaining physical condition limitation parameters. For example... Figure 1 As shown, the device includes a human body model body 1, a switch component 2, an indicator light component 3, an interaction module 4, and a control system module (not shown in the figure). To further enhance the device's functionality and obtain more physical condition limitation parameters beneficial for driver qualification assessment, this embodiment also includes a testing module (not shown in the figure). The specific structural setup and positional connections are described in detail below with reference to the accompanying drawings.

[0175] The human body model body 1 is as follows Figure 1As shown, the model is in a standard human standing posture, with a stable support base at the bottom. The main body 1 of the human model includes a head module, torso module, left upper limb module, right upper limb module, left lower limb module, and right lower limb module. The modules can be firmly assembled by plugging, snapping, or screws. The main body 1 of the human model can be made of lightweight and sturdy materials such as ABS plastic, and the names of key parts can be printed on it. According to the needs of driving operation (such as operating the steering wheel, shifting gears, observing instruments and rearview mirrors, and pressing the accelerator, brake, and clutch pedals), the main body 1 of the human model is divided and set with multiple key evaluation parts 11. These parts are mainly concentrated in the joints of motor function and important sensory organs, such as: the left and right turning parts of the head and neck (corresponding to observing the rearview mirror), shoulder joints, elbow joints, wrist joints, finger joints (corresponding to holding the steering wheel and operating buttons), hip joints, knee joints, ankle joints (corresponding to pedaling actions), as well as eyes and ears.

[0176] The switch components 2 are respectively installed on each key assessment area 11 of the human body model body 1, and the indicator lights 3 are installed corresponding to each key assessment area 11 and controlled by the corresponding switch component 2. As shown in the figure, a switch component 2 is installed at each key assessment area. The switch component 2 can be a self-locking switch with light, a tactile switch, or a membrane switch, etc. The indicator lights 3 can be LED lights, installed next to each switch component 2 or integrated together. In addition, in order to improve the lighting effect of the indicator lights 3, a structure that can achieve a large-area brightening effect can also be adopted. For example, a light-transmitting shell 31 can be set for each key assessment area 11 and for the area that needs to be divided according to driving operation. The indicator lights 3 are set inside the light-transmitting shell 31, and the light is uniform and soft, and the indication range is clearer.

[0177] The interactive module 4 is for user interaction and can be used to display operation guidelines, current acquisition status, test results, and provide menus for evaluators to perform advanced operations such as mode selection, data confirmation, and report generation. It can be a display device mounted on the main body 1 of the human body model, such as a touch screen integrated into the front of the model's torso module. Alternatively, it can be a standalone display device, such as a tablet computer or computer terminal.

[0178] The control system module is used to realize overall operation control. The specific structure and principle of the control system module can be designed and developed according to the product functions. This embodiment provides a simple example. For example, the core of the module can be a circuit board integrating a microcontroller (such as an STM32 series microcontroller). The microcontroller is responsible for the logic control of the system. The signal lines of all trigger switches 2 are connected to the input ports (GPIO) of the microcontroller, and all indicator lights 3 are connected to the output ports of the microcontroller. The entire system is powered by a power module (such as a built-in lithium battery with a charging circuit, or an external DC power interface). The microcontroller scans the status of all trigger switches 2 in real time. The control system module may also include a wireless unit, a remote module, a host computer, etc. Thus, the switch components 2, indicator lights 3, and interaction module 4 are connected to the control system module for signal or data transmission. By having the user operate the switch components or the interaction module 4 on the human body model body 1, the system can obtain static physical function limitation parameters of the user for driving qualification assessment.

[0179] The testing module is used for testing the user's physical functions and is connected to the control system module for signal or data transmission. The testing module may include one or more of the following: a strength testing module, a motion sensing device, a VR driving simulator, etc., which can be selected and configured according to the product design functions. The strength testing module can be used to test the user's hand and foot strength to obtain strength parameters for evaluation of key assessment points 11, such as using existing similar products like grip strength testers or tensile strength testers; the motion sensing device can be worn on the user's body to obtain motion trajectory parameters for evaluation of key assessment points 11, and can use inertial sensors with high sensing accuracy, such as wireless six-axis IMU inertial sensors, which not only have high sensing accuracy but are also more convenient to use wirelessly; the VR driving simulator can be used to allow the user to simulate driving operations to obtain driving behavior data for evaluation. For example, the usage process is as follows: Connect the main body 1 of the human body model to the computer of the driving qualification assessment system via a data cable; after the system starts, the model initializes and all indicator lights turn off; the assessor or user determines the physiological function limitation area based on the physical examination report, on-site inquiry, or the user's own situation, finds the corresponding area on the model, presses the switch component 2, and the indicator light 3 at that location lights up; repeat the steps until all limitation areas are obtained; the assessment system receives the data sent by the model in real time and inputs it as "static physiological function limitation parameters" into the subsequent assessment algorithm (such as the driving function fault tree model); after one assessment is completed, all switch states can be cleared through system commands or the reset button on the model, ready for the next use.

[0180] This testing module setup enables dynamic functional testing for further evaluation. For example, strength tests can be conducted by providing the applicant with a grip strength tester (integrated or external) to measure hand grip strength, or a traction tester for foot dorsiflexion strength testing; the data is transmitted to the control system module via wired or wireless means. Similarly, range of motion tests involve attaching motion sensors (such as inertial sensors) to the applicant's joints being evaluated (e.g., wrist, ankle) and allowing them to perform movements within a specified range; the sensor-recorded movement trajectory parameters are transmitted to the control system module. Furthermore, VR driving simulations can be performed by guiding the applicant to use a connected VR driving simulator for a simulated driving session, with their driving behavior data recorded synchronously. This dynamic test data is also aggregated by the control system module and can be combined with static constraint data for output, forming a more comprehensive evaluation profile.

[0181] This device transforms the traditional, abstract, and cumbersome form-filling process into an intuitive and efficient physical model interaction and electronic data collection process. It significantly improves the accuracy, efficiency, and user experience of the physiological data acquisition stage in driver qualification assessment, demonstrating good practical value and promising prospects for widespread application. Specific beneficial effects are as follows.

[0182] 1. Intuitive and easy to use, simple operation. Using a physical human body model for operation makes the assessment of key areas clear at a glance, avoiding ambiguity and misunderstandings that may arise from traditional textual descriptions. Assessors or users only need to press the corresponding switch on the model to complete the recording, making it especially suitable for assessment scenarios with limited education, language barriers, or those requiring rapid operation, thus lowering the operational threshold.

[0183] 2. Highly efficient and instant feedback. Adopting a one-button trigger design, the on / off action and indicator light status are synchronized in real time, providing clear visual feedback to confirm successful data acquisition. Compared to manually filling out forms, this significantly shortens data acquisition time and effectively improves the overall efficiency of the evaluation process.

[0184] 3. Accurate and standardized data, stored electronically. This avoids problems such as errors in selection, omissions, or illegible handwriting that may occur with manual entry. All operations are directly converted into electronic signals, processed and packaged by the control system module. The data format is uniform and standardized, and can be directly transmitted to the evaluation system through data interfaces (such as wired or wireless), ensuring the accuracy and consistency of the data source and facilitating subsequent storage, analysis, and retrieval.

[0185] 4. Integrated Functions, Comprehensive Assessment. Building upon basic data acquisition, the model can be further integrated with modules for strength testing, activity trajectory sensing, and VR driving simulation. This allows the model to not only record static physiological limitations but also dynamically test the strength, range of motion, and simulated driving behavior of key body parts, providing more comprehensive and multi-dimensional objective data support for driver qualification assessment and enhancing the scientific rigor and professionalism of the evaluation.

[0186] 5. Highly specialized and well-integrated with the process. Specifically designed for the recording and testing of physical function limitations in driver qualification assessments, the device closely integrates the assessment of key areas with driving operation requirements (such as turning the steering wheel, pressing the pedals, and observing the rearview mirrors). This high degree of integration with the assessment process enhances its standardization, normalization, and professionalism. It should be noted that all calculation formulas in this application employ, but are not limited to, regression analysis using machine learning algorithms to deeply analyze the collected parameters and identify their natural trends and interrelationships. Specialized software, such as Python's Scikit-learn library or R language, is used to automatically generate mathematical models that match the data. Then, cross-validation and other methods are used to objectively evaluate the model performance, combined with continuous feedback and optimization, to ensure that the created formulas truly reflect the inherent laws of the data, thereby guaranteeing their effectiveness and accuracy. In all calculation formulas in this application, the parameters in each formula undergo dimensionless processing within a consistent range to ensure that different physical quantities are compared on the same scale; dimensionless techniques include, but are not limited to, Min-Max Normalization and Z-Score standardization.

[0187] The technical solution of this invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments of this invention.

[0188] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0189] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for determining driving qualifications based on multi-point physiological feature input, characterized in that, The specific steps include: S1. Based on at least one static physiological function limitation parameter of the driver, a first risk factor is determined through a preset driving function fault tree model to characterize the theoretical failure risk in a preset driving task in a virtual reality environment. S2. Obtain driving behavior data generated by the driver when performing a preset driving task in the VR environment, and analyze the driving behavior data through a compensation effectiveness analysis model to obtain an adaptive strategy effectiveness index, so as to determine the second effectiveness factor used to characterize the driver's actual compensation ability. S3. Based on the first risk factor and the second efficiency factor, calculate the compensation efficiency load index, which characterizes the degree of compensatory effort required by the driver to achieve the observed driving performance. S4. When the compensation efficiency load index meets the preset disturbance triggering conditions, a disturbance command is generated to inject secondary disturbances into the VR environment. S5. In response to the generation of disturbance instructions, trigger the analysis of micro-instability characteristics of driving behavior data, and adjust the dynamic safety rating associated with the driving qualification level based on the analysis results of micro-instability characteristics.

2. The method for determining driving qualifications based on multi-point physiological feature input according to claim 1, characterized in that: The first risk factor in S1 is the top event failure rate of the preset driving task, which is obtained by using the static physiological function limitation parameters as the basic failure events through the driving function fault tree model and performing probability propagation calculation. And / or, the compensation effectiveness analysis model in S2 is a preset model used to analyze time-series data of driving behavior; The second effectiveness factor includes the success rate of the observed task calculated based on driving behavior data, and the effectiveness index of the adaptive strategy output by the compensation effectiveness analysis model; And / or, the specific steps in S3 for calculating the compensation efficiency load index are: calculating the compensation efficiency load index that characterizes the functional relationship between the second efficiency factor and the first risk factor; And / or, the preset disturbance triggering condition in S4 is: the instantaneous value of the compensation efficiency load index in the current time window exceeds a preset multiple of the moving average value in the past time windows; And / or, the micro-instability features in S5 include: the increase in steering wheel vibration energy of the steering wheel angle signal within a preset frequency band, and the rate of change of the gaze entropy value of the driver's eye gaze point distribution.

3. The method for determining driving qualifications based on multi-point physiological feature input according to claim 2, characterized in that: The calculation steps for the first risk factor and the second effectiveness factor are as follows: Input the driver's static physiological function limitation parameters and the driving behavior data of the driver recorded by the VR scene engine during the execution of basic events to form a multi-dimensional time series data segment; Based on the input static physiological function limitation parameters, the corresponding baseline event failure rate is retrieved from the physiological limitation-benchmark failure rate lookup table. The VR scene engine calculates the scene complexity factor based on the current task configuration; The failure rate after dynamic modulation is calculated based on the baseline failure rate of the basic event and the scenario complexity factor. The failure rate after dynamic modulation is used as the probability of the basic event and input into the preset driving function fault tree model. The driving function fault tree model also includes basic events affected by different physiological limitations; the driving function fault tree model performs probability propagation calculations and finally outputs the top event failure rate, which is the first risk factor; The recorded complete driving behavior data is input into the encoder of the pre-trained Siamese neural network; Output the behavior sequence embedding vector; Calculate the distance between the behavior sequence embedding vector and the standard behavior center vector of the underlying event, and calculate the success rate of the observation task accordingly; The weight distribution is extracted from the self-attention layer of the network, and the attention weight distribution entropy is calculated. Based on the attention weight distribution entropy, the effectiveness index of the adaptive strategy is calculated. Output the first risk factor and the second effectiveness factor.

4. The method for determining driving qualifications based on multi-point physiological feature input according to claim 3, characterized in that: The calculation steps for the compensation efficiency load index are as follows: Input the first risk factor and the second effectiveness factor to observe the task success rate and adaptive strategy effectiveness index; The product of the success rate of the observation task and the effectiveness index of the adaptive strategy is multiplied and then the square root is taken to calculate the comprehensive compensation performance score. Obtain the nonlinear index of performance contribution; using the comprehensive compensation performance score as the base and the nonlinear index of performance contribution as the exponent, perform exponentiation to obtain the load numerator; Obtain the first risk factor; Subtracting the first risk factor from the constant 1 yields the median value representing the safety margin; Obtain a non-linear index for risk aversion; Using the safety margin as the base and the risk aversion nonlinear index as the exponent, an exponentiation operation is performed. Add a very small smoothing constant to the result; this gives the final load denominator. Divide the obtained load numerator by the obtained load denominator to obtain the preliminary compensation efficiency load index; Set the lower and upper limits of output; If the calculated value is less than the lower limit of the output, the final output will be the lower limit of the output. If it is greater than the output limit, then the output will be the output limit. Otherwise, output the original calculated value; the final output is the compensation efficiency load index after amplitude limiting.

5. The method for determining driving qualifications based on multi-point physiological feature input according to claim 4, characterized in that: The calculation steps for dynamic security rating are as follows: At the beginning of each time step, input the compensation efficiency load index and continuously monitor the health status of the data streams for the first risk factor and the second efficiency factor; Check the latest timestamp of the first risk factor. If the difference between the current time and the timestamp exceeds the information source interruption timeout threshold, it is determined that the acquisition of the first risk factor has been interrupted, and the system immediately jumps to the first degraded mode. Check the latest timestamp of the data stream for the second performance factor; If the difference between the current time and the timestamp exceeds the source interruption timeout threshold, the acquisition of the second performance factor is determined to be interrupted, and the system immediately jumps to the second degradation mode. If both the first risk factor and the second efficiency factor are obtained normally, then continue; store the latest compensation efficiency load index into the short-term and long-term sliding windows; Calculate the long-term moving average and standard deviation of the compensation efficiency load index; Calculate the adaptive trigger ratio; If the instantaneous value of the compensation efficiency load index is greater than the long-term moving average multiplied by the adaptive triggering multiple, the triggering condition is determined to be met, and the next step is executed. Otherwise, end the current loop; Calculate the overshoot of the compensation efficiency load index, which is the instantaneous value minus the long-term moving average multiplied by the adaptive triggering factor; calculate the specific disturbance intensity value based on the overshoot and disturbance intensity mapping factor; generate a disturbance command containing the disturbance type, the calculated intensity, and the preset duration. The disturbance command is sent to the VR scene engine actuator in the technical environment; after the disturbance command is issued, the high-frequency data analysis mode is immediately entered; the steering wheel vibration energy increment and the rate of change of gaze entropy value are calculated in parallel; and the final attitude instability comprehensive score is calculated by fusion. Based on the comprehensive attitude instability score, a penalty deduction is applied to the rating; the penalty deduction is equal to the maximum dynamic safety rating multiplied by the comprehensive attitude instability score; the dynamic safety rating is obtained by subtracting the penalty deduction from the maximum dynamic safety rating; the current cycle ends. The first downgrade mode stops the calculation of S3 and S4, and directly calculates the downgraded dynamic security rating based on the second performance factor output by S2. The final report explicitly states that the evaluation results are based on performance data and do not include theoretical risk analysis, thus reducing the confidence level. The second downgrade mode is to stop the calculation of S3 and S4; based on the first risk factor output by S1, the dynamic security rating of the downgrade is directly calculated. The final report explicitly states that the assessment results are based on risk projections and do not include actual performance verification, thus reducing the confidence level. The final output of each loop is either an updated dynamic security rating or a suboptimal rating with explicit downgrade instructions.

6. The method for determining driving qualifications based on multi-point physiological feature input according to claim 5, characterized in that: Based on the dynamic safety rating, the eligible driver's license levels are determined as follows: If the dynamic safety rating is ≥0.85, then you can obtain an A1 driver's license for large passenger vehicles; If the dynamic safety rating is 0.82 or less and the rating is less than 0.85, then a driver's license for city buses (A3) can be obtained. If the dynamic safety rating is 0.80 or less and the rating is less than 0.82, then an A2 driver's license can be obtained for towing trailers. If the dynamic safety rating is 0.75 or less and the rating is less than 0.80, then a B1 driver's license for medium-sized passenger vehicles can be obtained. If the dynamic safety rating is 0.70 or less and the rating is less than 0.75, then a B2 driver's license for large trucks can be obtained. If the dynamic safety rating is 0.65 or less and the rating is less than 0.70, then a C1 or C2 driver's license for small cars can be obtained. If the dynamic safety rating is 0.60 ≤ Dynamic Safety Rating < 0.65, then a Class D driver's license for ordinary three-wheeled motorcycles can be obtained. If the dynamic safety rating is less than 0.60, no driver's license can be obtained.

7. A method for determining driving qualifications based on multi-point physiological feature input as described in any one of claims 1-6, characterized in that: The static physiological function limitation parameters are obtained through a human body model device capable of acquiring physical condition limitation parameters. The human body model device includes a human body model body with assessment key parts divided and set according to driving operation needs, switch components respectively set on each assessment key part, indicator lights set on each assessment key part and controlled by the corresponding switch components, an interactive module for user operation and interaction, and a control system module for overall control. The switch components, indicator lights, and interactive module are connected to the control system module for signal or data transmission. The user's static physical function limitation parameters are obtained by the user operating the switch components or the interactive module on the human body model body.

8. The method for determining driving qualifications based on multi-point physiological feature input according to claim 7, characterized in that: It also includes a testing module for testing the user's physical function, which is connected to the control system module for signal or data transmission.

9. The method for determining driving qualifications based on multi-point physiological feature input as described in claim 8, characterized in that, The testing module includes a strength testing module for testing the strength of the user's palms and feet to obtain strength parameters of key assessment areas for evaluation and judgment; And / or, including activity sensing devices worn on key assessment points of the user's body to obtain activity trajectory parameters for assessment and judgment; And / or, VR driving simulation devices used to allow users to simulate driving operations in order to obtain driving behavior data for evaluation and judgment.

10. A method for determining driving qualifications based on multi-point physiological feature input for physical condition level, as described in any one of claims 7-9, characterized in that... A light-transmitting housing is provided for the key assessment parts and the areas defined according to the driving operation needs, and the indicator lights are set inside the light-transmitting housing; And / or, the interaction module is set on a display screen device on the main body of the human body model or is a separate display screen device.

Images