A method for evaluating driving state of super-long highway tunnel based on human-vehicle cooperative information

Abstract

The application provides a long highway tunnel driving state evaluation method based on human-vehicle cooperative information. It includes: constructing a human-vehicle information index system for long highway tunnel driving state evaluation; collecting long highway tunnel driving state evaluation index data and preprocessing; using Hilbert-Huang algorithm to extract characteristic quantities reflecting state abnormalities in driver physiological information and vehicle driving information; constructing a Bayesian network structure based on human and vehicle state information; using similarity aggregation method for network parameter learning, and using contrast test method to verify the Bayesian network model. The application realizes fine processing of non-stationary physiological signals and vehicle driving signals in complex scenarios, extracts abnormal information, fully utilizes prior information of drivers and vehicle operation to make real-time judgment and probability update of driving state. It can evaluate the driving state in long highway tunnel environment and reduce the probability of traffic accidents in the tunnel.

Description

Technical Field

[0001] This invention relates to the field of vehicle safety assistance driving, and in particular to a method for assessing the driving status of vehicles in long highway tunnels based on human-vehicle collaborative information. Background Technology

[0002] Highway tunnels are relatively enclosed traffic structures. Compared to open roads, their internal driving environment is monotonous, cramped, poorly lit, and relatively poorly ventilated. In this environment, drivers require high concentration, are highly vigilant, and experience relative mental stress, making them prone to fatigue and operational errors. Currently, with the development of traffic engineering, the number and length of highway tunnels have reached unprecedented levels. Generally, an increase in tunnel length leads to a linear increase in travel time within the tunnel; for example, the Qinling Zhongnanshan Highway Tunnel is 18.02 km long, with an average travel time exceeding 15 minutes. Prolonged driving in the oppressive, dimly lit tunnel environment can cause traffic accidents, and the enclosed environment significantly increases the difficulty of accident rescue. Therefore, accurately assessing driving conditions inside exceptionally long highway tunnels is crucial for preventing traffic accidents within tunnels.

[0003] Driving status involves two dimensions: the driver and the vehicle. Specifically, it includes the driver's physiological state and the vehicle's operational state. Currently, driving status evaluation mainly focuses on the driver's physiological information to determine whether the driver is driving while fatigued, as exemplified by invention patents CN202210651936.5 and CN201811576142.7. However, these existing technologies still have significant shortcomings and limitations, mainly in the following aspects: (1) the driving scenarios involved are mainly open highway sections, and none of them involve the closed tunnel driving environment; (2) they usually rely on a single signal type to evaluate driving status, which may introduce detection errors; (3) they mostly use simple signal analysis methods, ignoring the dual uncertainties of physiological signals and driving status signals, especially the nonlinear characteristics of physiological signals in complex environments. In view of this, the present invention provides a method for evaluating the driving status of extra-long highway tunnels to improve the operational risk of extra-long highway tunnels. Summary of the Invention

[0004] Purpose of the invention: To address the limitations and shortcomings of the existing technologies, this invention proposes a method for assessing the driving status of drivers in long highway tunnels based on human-vehicle collaborative information. Based on the electrocardiogram, electroencephalogram, and electromyogram signals of the driver as they pass through the long highway tunnel, as well as the vehicle's driving status signals, the method utilizes Hilbert-Huang transform to detect the driving status in complex environments such as the single environment, poor lighting conditions, enclosed space, and poor ventilation of long highway tunnels.

[0005] Technical Solution: This invention discloses a method for assessing the driving status of long highway tunnels based on human-vehicle collaborative information, comprising the following steps:

[0006] Step 1: Construct a driver and vehicle information indicator system for assessing driving status in long highway tunnels;

[0007] Step 2: Collect driving status assessment index data for long highway tunnels. The data comes from actual vehicle testing and opinions from multiple experts, and the data is preprocessed. The driving status assessment index for long highway tunnels includes two dimensions: driver status and vehicle driving status. The driver status includes three secondary indicators: driver's electroencephalogram (EEG) signal, electrocardiogram (ECG) signal, and electromyogram (EMG) signal. The vehicle driving status includes two secondary indicators: real-time steering wheel angle and real-time vehicle speed.

[0008] Step 3: Use the Hilbert-Huang algorithm to extract features of abnormal reaction states from the driver's physiological information and vehicle driving information;

[0009] Step 4: Construct a Bayesian network structure for driving status in long highway tunnels based on human and vehicle status information;

[0010] Step 5: Perform Bayesian network parameter learning and model validation for driving status in long highway tunnels; use similarity aggregation algorithm to train the prior probability and conditional probability distribution table of driving status evaluation index for long highway tunnels, and obtain a driving status evaluation model for long highway tunnels based on human-vehicle collaborative information.

[0011] Step 6: Based on prior information obtained in real time about the driver and vehicle, perform probability calculations and state assessments of the driving status in long highway tunnels.

[0012] Furthermore, in Step 2, the driving status index data for the long highway tunnel comes from driver and vehicle signals collected during actual vehicle testing, as well as opinions from multiple experts. Data preprocessing includes handling missing values ​​in the data and verifying the validity of the data, and converting the semantic descriptions of the experts into triangular fuzzy numbers.

[0013] Furthermore, in Step 3, the Hilbert-Huang algorithm is used to extract features of abnormal reaction states from the physiological information of drivers and vehicle driving information in long highway tunnels, including the following steps:

[0014] Step 3.1: Input the preprocessed time-travel signal data of people and vehicles from step S2;

[0015] Step 3.2: Select the Hilbert-Huang algorithm as the time-spectrum analysis algorithm for the signal;

[0016] Step 3.3: Extract the characteristic quantities of the abnormal reaction signal based on the Hilbert spectral analysis results;

[0017] Step 3.4: Set the Hilbert spectral thresholds for various signal anomalies;

[0018] Step 3.5: Output the characteristic quantities and range of the abnormal physiological information of the driver and the vehicle driving status in the long highway tunnel.

[0019] Furthermore, the Hilbert-Huang algorithm described in Step 3.2 includes the following steps:

[0020] (1) Calculate the mean value of the signal envelope, obtain the maximum and minimum values ​​of the signal x(t), obtain the upper and lower envelopes of the signal through cubic spline fitting, and calculate the mean value m1(t) of the envelope.

[0021] (2) Calculation of the intrinsic modal components of the signal: According to c1(t)=x(t)-m1(t), calculate the first intrinsic modal component of x(t), check that the number of maximum points and zero points of c1(t) does not differ by more than one, and the mean of the upper and lower envelopes of c1(t) is always 0. If not satisfied, repeat the above operation until the first intrinsic modal component c1(t) that satisfies the modal function condition is obtained.

[0022] (3) Empirical Mode Decomposition (EMD): Subtracting the first intrinsic mode component c1(t) from the original signal x(t) yields the residual signal r1(t) = x(t) - c1(t). This residual signal is used as the new original signal. The above operation is repeated until the signal residual r(t) is less than a preset value, at which point the EMD ends. The original signal is decomposed into n empirical mode components and 1 residual signal, i.e. Where c i (t) represents the i-th intrinsic mode component, and r(t) represents the residual signal of x(t) after empirical mode decomposition.

[0023] (4) Hilbert spectral analysis of intrinsic modal components, each intrinsic modal component is represented as: Where a i (t) is c i (t) Amplitude after real transformation It is C i (t) Frequency after real transform; Hilbert-Huang transform is performed on each intrinsic modal component to obtain its Hilbert time spectrum. Integrating the Hilbert time spectrum over the time window T yields the Hilbert marginal spectrum, which reflects the time-frequency characteristics of the signal.

[0024] (5) Determine the frequency range of the signal x(t) corresponding to the abnormal driver state, calculate its Hilbert marginal spectrum range, and output the characteristic quantity and characteristic quantity range of the abnormal signal state.

[0025] Furthermore, in Step 4, the Bayesian network structure for the driving state of the long highway tunnel includes a three-layer node structure: leaf nodes represent the driving state (T); intermediate nodes represent the driver state (D) and vehicle driving state (V); and the root nodes represent the Hilbert spectral features of the driver's electroencephalogram (B), electrocardiogram (H), and electromyogram (M), as well as the Hilbert spectral features of the vehicle's real-time steering wheel angle (A) and real-time speed (S).

[0026] Furthermore, in Step 5, the Bayesian network parameter learning employs a similarity aggregation algorithm to train the prior probability and conditional probability distribution table of the driving status evaluation index for long highway tunnels, including the following steps:

[0027] Step 5.1: Input the preprocessed expert opinions from step S2, where the expert opinions are represented by triangular fuzzy numbers;

[0028] Step 5.2: Use a similarity aggregation algorithm to fuse opinions from different experts;

[0029] Step 5.3: The aggregated expert opinions are then defuzzified to obtain the precise fault probability of the node;

[0030] Step 5.4: Repeat the above algorithm for each node unit to output the prior probability and conditional probability distribution table of all node units of the Bayesian network.

[0031] Furthermore, the similarity aggregation algorithm in Step 5.2 employs the following steps:

[0032] (1) Calculate the triangular fuzzy opinion similarity between any two experts and the average consistency measure of each expert;

[0033] (2) Calculate the relative consistency measure for each expert and obtain the consistency coefficient;

[0034] (3) Aggregate the triangular fuzzy opinions of all experts using the consistency coefficient of each expert as the weight;

[0035] (4) The aggregated triangular fuzzy opinions are defuzzified to obtain the precise fault probability of the nodes.

[0036] Furthermore, in Step 5, a comparative verification method is used to test the Bayesian network model for assessing driving conditions in long highway tunnels, including the following steps:

[0037] (1) Input the preprocessed time-history signal data of people and vehicles in step S2;

[0038] (2) Use a trained Bayesian network model to assess driving status;

[0039] (3) Compare the evaluation results and output the verification results.

[0040] Furthermore, the main steps in Step 6 are as follows:

[0041] Step 6.1: Obtain real-time signals of driver and vehicle status in long highway tunnels;

[0042] Step 6.2: Based on steps S1 to S5, establish a driving status assessment model for long highway tunnels based on human-vehicle collaborative information;

[0043] Step 6.3: Input the real-time signals of the driver and vehicle status, and use the above model to calculate the probability of the driving status being normal, slightly abnormal, or severely abnormal.

[0044] Step 6.4: Determine the driving status of the long highway tunnel according to the principle of maximum membership, and finally output the evaluated driving status.

[0045] Beneficial effects:

[0046] 1. This invention utilizes the electrocardiogram, electroencephalogram, and electromyogram signals of drivers passing through long highway tunnels to propose a driving state detection method based on Hilbert-Huang transform for long highway tunnels. This method enables the detection of driving state in complex environments such as single environment, poor lighting conditions, enclosed space, and poor ventilation in long highway tunnels. The Hilbert-Huang transform method can achieve refined processing of non-stationary physiological signals and vehicle driving signals in complex scenarios, and extract abnormal information from them.

[0047] 2. The Bayesian network structure established based on this invention can fully integrate human and vehicle information, and fully utilize prior information about driver and vehicle operation to perform real-time judgment and probability updates of driving status. This invention improves the accuracy of driving status detection in long highway tunnel environments, especially significantly improving detection accuracy when the driver is in a severely abnormal state and conventional indicators are fluctuating greatly. Attached Figure Description

[0048] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art are briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0049] Figure 1 This is a flowchart illustrating the driving status detection in an extra-long highway tunnel according to an embodiment of the present invention;

[0050] Figure 2 This is a schematic diagram of the Hilbert-Huang transform process for signals according to an embodiment of the present invention;

[0051] Figure 3 This is a schematic diagram of a Bayesian network structure according to an embodiment of the present invention. Detailed Implementation

[0052] The embodiments of the technical solution of the present invention will now be described in detail with reference to the accompanying drawings. These embodiments are merely illustrative of the technical solution of the present invention and are therefore intended to limit the scope of protection of the present invention.

[0053] It should be noted that, unless otherwise stated, the technical or scientific terms used in this application should have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0054] like Figure 1 As shown, this embodiment provides a method for evaluating driving status in long highway tunnels based on human-vehicle collaborative information. This method fully considers the dual uncertainties of driver physiological information and vehicle driving status in the environment of long highway tunnels. It establishes an evaluation index system for driving status in long highway tunnels from both human and vehicle information dimensions, and uses a Bayesian network to fully fuse multi-source information, thereby providing objective driving status evaluation results in the environment of long highway tunnels. The method includes the following steps:

[0055] Step 1: Construct a system of human and vehicle information indicators for assessing driving status in long highway tunnels.

[0056] The driving status assessment index system for extra-long highway tunnels is determined based on literature and tunnel driving accident reports. The assessment index includes two dimensions: driver status and vehicle driving status. Driver status includes three secondary indicators: electroencephalogram (EEG), electrocardiogram (ECG), and electromyogram (EMG). Vehicle driving status includes two secondary indicators: real-time steering wheel angle and real-time vehicle speed.

[0057] Step 2: Collect driving status assessment index data for long highway tunnels and perform preprocessing.

[0058] The driving condition assessment index data for long highway tunnels includes driver physiological signals and vehicle status signals collected from actual vehicle tests, as well as expert opinions obtained through questionnaires.

[0059] The driver's physiological signals include electroencephalogram (EEG), electrocardiogram (ECG), and electromyogram (EMG) signals, all of which are collected using non-invasive wearable devices. The vehicle's status signals during operation include steering wheel angle and real-time vehicle speed, both of which can be acquired through onboard sensors. Tunnels are generally straight, and lane changes are typically prohibited during driving; therefore, the steering wheel angle and real-time vehicle speed of a vehicle normally traveling in a tunnel remain stable. Sudden changes in these test values ​​indicate an abnormal driving condition.

[0060] The driving condition assessment index data for long highway tunnels also includes expert opinions collected through questionnaires. Generally, when actual data is lacking and the probability of each event state cannot be accurately obtained, it is often necessary to obtain this information through expert-guided methods, leveraging expert knowledge and practical experience. To establish the correspondence between expert linguistic variables and fuzzy numbers, seven linguistic variables were introduced: very low, low, slightly low, moderate, slightly high, high, and very high (denoted by VL, L, ML, M, MH, H, and VH, respectively). In this embodiment, five expert questionnaires were distributed to industry experts.

[0061] Data preprocessing includes handling missing values ​​and validating the data, as well as converting expert semantic descriptions into triangular fuzzy numbers, as shown in Table 1.

[0062] Table 1. Linguistic variables and corresponding triangular fuzzy numbers for the probability of events occurring.

[0063]

[0064] Step 3: Use the Hilbert-Huang algorithm to extract features of abnormal reaction states from driver physiological information and vehicle driving information. This includes the following steps:

[0065] (1) Calculation of signal envelope mean. Obtain the maximum and minimum values ​​of signal x(t), obtain the upper and lower envelopes of the signal through cubic spline fitting, and calculate the mean m1(t) of the envelope;

[0066] (2) Calculation of the intrinsic modal components of the signal. Calculate the first intrinsic modal component of x(t) according to c1(t) = x(t) - m1(t). Check that the number of maxima and zero-crossings of c1(t) differs by no more than one, and that the mean of the upper and lower envelopes of c1(t) is always 0. If these conditions are not met, repeat the above operation until the first intrinsic modal component c1(t) that satisfies the modal function conditions is obtained.

[0067] (3) Empirical Mode Decomposition (EMD): Subtracting the first intrinsic mode component c1(t) from the original signal x(t) yields the residual signal r1(t) = x(t) - c1(t). This residual signal is used as the new original signal. The above operation is repeated until the signal residual r(t) is less than a preset value, at which point the EMD ends. The original signal is decomposed into n empirical mode components and 1 residual signal, i.e. Where c i (t) represents the i-th intrinsic mode component, and r(t) represents the residual signal of x(t) after empirical mode decomposition.

[0068] (4) Hilbert spectral analysis of intrinsic modal components, each intrinsic modal component is represented as: Where a i (t) is c i (t) Amplitude after real transformation It is C i (t) Frequency after real transform; Hilbert-Huang transform is performed on each intrinsic modal component to obtain its Hilbert time spectrum. Integrating the Hilbert time spectrum over the time window T yields the Hilbert marginal spectrum, which reflects the time-frequency characteristics of the signal.

[0069] (5) Determine the frequency range of the signal x(t) corresponding to the abnormal driver state, calculate its Hilbert marginal spectrum range, and output the characteristic quantity and characteristic quantity range of the abnormal signal state.

[0070] Following the steps above, the time-frequency domain interval of the EEG signal reflects the driver's driving state based on the β rhythm, whose Hilbert marginal spectral energy is... In the formula, H B (ω) represents the Hilbert marginal spectrum of the EEG signal, and β_UL and β_DL are the upper and lower limits of the β rhythm frequency range, respectively. The time-frequency domain range of the ECG signal reflects the driver's driving state based on the LF rhythm, whose Hilbert marginal spectrum energy is... In the formula, H H (ω) represents the Hilbert marginal spectrum of the electrocardiogram (ECG) signal, where LF_UL and LF_DL are the upper and lower limits of the LF rhythm frequency range, respectively. The time-frequency domain interval of the electromyography (EMG) signal reflects the driver's state based on the MF rhythm (median frequency band), and the energy of the Hilbert marginal spectrum is... In the formula H M (ω) represents the Hilbert marginal spectrum of the electrocardiogram signal, and MF_UL and MF_DL are the upper and lower limits of the MF rhythm frequency range, respectively.

[0071] The time-frequency domain interval of the steering wheel angle signal reflecting abnormal vehicle driving status is based on the HF rhythm, whose Hilbert marginal spectral energy is... In the formula, H A(ω) represents the Hilbert marginal spectrum of the steering wheel angle signal, and HF_UL and HF_DL are the upper and lower limits of the HF rhythm frequency range, respectively. The time-frequency domain range of the vehicle's real-time speed signal reflects the driver's driving state based on the SF rhythm, whose Hilbert marginal spectrum energy is... In the formula, H S (ω) represents the Hilbert marginal spectrum of the vehicle's real-time speed signal, and SF_UL and SF_DL are the upper and lower limits of the SF rhythm frequency range, respectively.

[0072] Step 4: Construct a Bayesian network structure based on human and vehicle status information;

[0073] The driving status assessment network for long highway tunnels comprises a three-layer node structure: leaf nodes represent driving status (T); intermediate nodes represent driver status (D) and vehicle driving status (V); and the root nodes represent the Hilbert spectral features of the driver's electroencephalogram (B), electrocardiogram (H), and electromyogram (M), as well as the Hilbert spectral features of the vehicle's real-time steering wheel angle (A) and real-time speed (S). The network structure is as follows: Figure 2 As shown.

[0074] Step 5: Bayesian network parameter learning and model validation for driving status assessment in long highway tunnels.

[0075] The parameters of a Bayesian network are learned using a similarity aggregation method, following these steps:

[0076] (a) Obtaining expert language descriptions of basic events k (k=1,2,…,λ), and convert it into a standard fuzzy number.

[0077] (b) Calculate the opinions of the two experts and Similarity:

[0078]

[0079] In the formula, λ represents the number of experts surveyed. It is a similarity function. and From expert E u (u=1,2,…,λ) and E v The standard fuzzy number of the opinion obtained at (v=1,2,…,λ).

[0080] (c) Calculate the average consensus measure among experts:

[0081]

[0082] (d) Calculate the relative consistency measure for each expert:

[0083]

[0084] (e) Calculate the expert consensus coefficient:

[0085] CC(E u )=β·w(E u )+(1-β)·RA(E u (4)

[0086] In the formula, w(E) u ) represents the expert weights, and β is the weight of w(E). u For RA(E) u The relaxation coefficient.

[0087] (f) Calculate the aggregated results of the fuzzy opinions:

[0088]

[0089] (g) Converting fuzzy numbers into fuzzy possibility fractions:

[0090] S f =0.25g1 + 0.5g0 + 0.25g2 (6)

[0091] In the formula, g0, g1, and g2 are the starting point, midpoint, and ending point of the triangular fuzzy number, respectively.

[0092] The final probability table of the nodes is shown in Table 2 below:

[0093] Table 2 Prior Probability Table of Nodes

[0094]

[0095]

[0096] Specifically, model validation is performed according to the following steps:

[0097] (1) Input the preprocessed time-history signal data of people and vehicles in step S2.

[0098] (2) A Bayesian network model was used to evaluate the driving status.

[0099] (3) Based on the existing literature and the results of actual vehicle testing, the evaluation results are compared and the verification results are output as shown in Table 3.

[0100] Table 3 Comparison of Bayesian network model evaluation results for driving status in long highway tunnels

[0101]

[0102] Step 6: Based on the prior information of the driver and vehicle driving obtained in real time, perform probability calculation and state assessment on the driving status of the long highway tunnel.

[0103] Specifically, the probability calculation and condition assessment of driving conditions in extra-long highway tunnels are carried out according to the following steps:

[0104] (1) In the polymorphic B-Ns structure, the root node is x i The intermediate node is y. j The leaf node is T. and T q These represent the risk states of the root node, intermediate nodes, and leaf nodes, respectively, where a i =0,1,…,u i -1, b j =0,1,…,v j -1, q=0,1,…,r-1,u i v j , where r represents the number of states of a node.

[0105] (2) When the root node x i Status is The probabilities are respectively At that time, leaf node T is in a risky state. q The probability calculation formula is as follows:

[0106]

[0107] In the formula, π(T) is the set of parent nodes of leaf node T.

[0108] (3) Given the root node x i In a state of risk Under the given conditions, leaf node T is in each risk state T q The posterior probability is:

[0109]

[0110] (4) Calculate the membership degree of each state according to the membership function formula of each risk state, and finally determine the state to which the leaf node belongs according to the principle of maximum membership degree, so as to determine the driver's driving state. In the formula, p is the probability value of each state of the leaf node.

[0111]

[0112] Specifically, in Step 6, this example selects the real-time signals of the driver and vehicle after the driving time in the long highway tunnel exceeds 500 seconds as input data and inputs them into the assessment model constructed above. After probability calculation, P{normal driving state} = 30%, P{mildly abnormal driving state} = 42%, and P{severely abnormal signal} = 28%. According to the membership function formula of each risk state, the membership degree of each state is calculated, and the driving state is rated as slightly abnormal, which is consistent with the actual driving state after the driving time in the long highway tunnel exceeds 500 seconds.

[0113] The above embodiments are only for illustrating the technical concept and features of the present invention, and are intended to enable those skilled in the art to understand the content of the present invention and implement it accordingly. They should not be construed as limiting the scope of protection of the present invention. All equivalent transformations or modifications made in accordance with the spirit and essence of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for assessing the driving status of long highway tunnels based on human-vehicle collaborative information, characterized in that, Includes the following steps: Step 1: Construct a driver and vehicle information indicator system for assessing driving status in long highway tunnels; Step 2: Collect driving condition assessment index data for long highway tunnels. The data comes from actual vehicle testing and opinions from multiple experts, and the data is preprocessed. The driving status assessment index for extra-long highway tunnels includes two dimensions: driver status and vehicle driving status. The driver status includes three secondary indicators: the driver's electroencephalogram (EEG) signal, electrocardiogram (ECG) signal, and electromyogram (EMG) signal. The vehicle driving status includes two secondary indicators: the real-time steering wheel angle and the real-time vehicle speed. Step 3: Use the Hilbert-Huang algorithm to extract features of abnormal reaction states from the driver's physiological information and vehicle driving information; Step 4: Construct a Bayesian network structure for driving status in long highway tunnels based on human and vehicle status information; Step 5: Perform Bayesian network parameter learning and model validation for driving status in long highway tunnels; A similarity aggregation algorithm was used to train the prior probability and conditional probability distribution tables of driving status assessment indicators for long highway tunnels, and a driving status assessment model for long highway tunnels based on human-vehicle collaborative information was obtained. Step 6: Based on prior information obtained in real time about the driver and vehicle, perform probability calculations and state assessments of the driving status in long highway tunnels.

2. The method for assessing the driving status of long highway tunnels based on human-vehicle collaborative information according to claim 1, characterized in that: In Step 2, the driving status index data for the long highway tunnel comes from driver and vehicle signals collected during actual vehicle testing, as well as opinions from multiple experts. Data preprocessing includes handling missing values ​​and verifying the validity of the data, and converting the experts' semantic descriptions into triangular fuzzy numbers.

3. The method for evaluating the driving status of long highway tunnels based on human-vehicle cooperative information according to claim 1, characterized in that: In Step 3, the Hilbert-Huang algorithm is used to extract features of abnormal reaction states from the physiological information of drivers and vehicle driving information in long highway tunnels, including the following steps: Step 3.1: Input the preprocessed time-travel signal data of people and vehicles from step S2; Step 3.2: Select the Hilbert-Huang algorithm as the time-spectrum analysis algorithm for the signal; Step 3.3: Extract the characteristic quantities of the abnormal reaction signal based on the Hilbert spectral analysis results; Step 3.4: Set the Hilbert spectral thresholds for various signal anomalies; Step 3.5: Output the characteristic quantities and range of the abnormal physiological information of the driver and the vehicle driving status in the long highway tunnel.

4. The method for evaluating the driving status of long highway tunnels based on human-vehicle collaborative information according to claim 3, characterized in that: The Hilbert-Huang algorithm described in Step 3.2 includes the following steps: (1) Calculate the mean value of the signal envelope, obtain the maximum and minimum values ​​of the signal x(t), obtain the upper and lower envelopes of the signal through cubic spline fitting, and calculate the mean value m1(t) of the envelope. (2) Calculation of the intrinsic modal components of the signal: According to c1(t)=x(t)-m1(t), calculate the first intrinsic modal component of x(t), check that the number of maximum points and zero points of c1(t) does not differ by more than one, and the mean of the upper and lower envelopes of c1(t) is always 0. If not satisfied, repeat the above operation until the first intrinsic modal component c1(t) that satisfies the modal function condition is obtained. (3) Empirical Mode Decomposition (EMD): Subtracting the first intrinsic mode component c1(t) from the original signal x(t) yields the residual signal r1(t) = x(t) - c1(t). This residual signal is used as the new original signal. The above operation is repeated until the signal residual r(t) is less than a preset value, at which point the EMD ends. The original signal is decomposed into n empirical mode components and 1 residual signal, i.e. Where c i (t) represents the i-th intrinsic mode component, and r(t) represents the residual signal of x(t) after empirical mode decomposition. (4) Hilbert spectral analysis of intrinsic modal components, each intrinsic modal component is represented as: Where a i (t) is c i (t) Amplitude after real transformation It is C i (t) Frequency after real transform; Hilbert-Huang transform is performed on each intrinsic modal component to obtain its Hilbert time spectrum. Integrating the Hilbert time spectrum over the time window T yields the Hilbert marginal spectrum, which reflects the time-frequency characteristics of the signal. (5) Determine the frequency range of the signal x(t) corresponding to the abnormal driver state, calculate its Hilbert marginal spectrum range, and output the characteristic quantity and characteristic quantity range of the abnormal signal state.

5. The method for evaluating the driving status of long highway tunnels based on human-vehicle collaborative information according to claim 1, characterized in that: In Step 4, the Bayesian network structure for the driving state of the long highway tunnel includes a three-layer node structure: leaf nodes represent the driving state T; and intermediate nodes represent the driver state D and the vehicle driving state V. The root nodes are the Hilbert spectral features of the driver's brain electroencephalogram (B), electrocardiogram (H), and electromyogram (M), as well as the real-time steering wheel angle (A) and real-time speed (S) of the vehicle.

6. The method for evaluating the driving status of long highway tunnels based on human-vehicle collaborative information according to claim 5, characterized in that: In Step 5, Bayesian network parameter learning employs a similarity aggregation algorithm to train the prior probability and conditional probability distribution table of driving status evaluation indicators for long highway tunnels, including the following steps: Step 5.1: Input the preprocessed expert opinions from step S2, where the expert opinions are represented by triangular fuzzy numbers; Step 5.2: Use a similarity aggregation algorithm to fuse opinions from different experts; Step 5.3: The aggregated expert opinions are then defuzzified to obtain the precise fault probability of the node; Step 5.4: Repeat the above algorithm for each node unit to output the prior probability and conditional probability distribution table of all node units of the Bayesian network.

7. The method for assessing the driving status of long highway tunnels based on human-vehicle collaborative information according to claim 6, characterized in that: The similarity aggregation algorithm in Step 5.2 uses the following steps: (1) Calculate the triangular fuzzy opinion similarity between any two experts and the average consistency measure of each expert; (2) Calculate the relative consistency measure for each expert and obtain the consistency coefficient; (3) Aggregate the triangular fuzzy opinions of all experts using the consistency coefficient of each expert as the weight; (4) The aggregated triangular fuzzy opinions are defuzzified to obtain the precise fault probability of the nodes.

8. The method for assessing the driving status of long highway tunnels based on human-vehicle collaborative information according to claim 1, characterized in that: In Step 5, a comparative verification method is also used to test the Bayesian network model for assessing driving conditions in long highway tunnels, including the following steps: (1) Input the preprocessed time-history signal data of people and vehicles in step S2; (2) Use a trained Bayesian network model to assess driving status; (3) Compare the evaluation results and output the verification results.

9. The method for assessing the driving status of long highway tunnels based on human-vehicle cooperative information according to claim 1, characterized in that: The main steps in Step 6 are as follows: Step 6.1: Obtain real-time signals of driver and vehicle status in long highway tunnels; Step 6.2: Based on steps S1 to S5, establish a driving status assessment model for long highway tunnels based on human-vehicle collaborative information; Step 6.3: Input the real-time signals of the driver and vehicle status, and use the above model to calculate the probability of the driving status being normal, slightly abnormal, or severely abnormal. Step 6.4: Determine the driving status of the long highway tunnel according to the principle of maximum membership, and finally output the evaluated driving status.

Images