System for measuring credibility of drivers with different temperament types to automatic driving

Abstract

The invention discloses a system for measuring the credibility of drivers with different temperament types for automatic driving. The system comprises a driver temperament type distinguishing module, a credibility measuring module, a credibility grade grading module, a credibility grade judging module and a comprehensive credibility evaluation module. According to the method, the difference of the credibility of drivers with different temperament types on automatic driving can be analyzed, and different coping strategies are adopted for the drivers with different temperament types; the credibility is measured by combining subjective factors and objective factors, and the measurement accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving, and in particular to a system for measuring the degree of trust in autonomous driving by drivers of different temperament types. Background Art

[0002] With the rapid development of science and technology, autonomous driving has become a hot topic in the field of road traffic. There are still many problems to be solved, including the trust between users and autonomous vehicles. The driver's trust in the vehicle will largely determine the efficiency of autonomous driving. Excessive trust may lead to traffic accidents, while lack of trust will also reduce the utilization rate of autonomous driving.

[0003] At present, the research on trust between drivers and autonomous vehicles mainly has the following problems:

[0004] (1) It does not take into account the differences between drivers themselves. Drivers with different temperament characteristics have different coping strategies for autonomous vehicles while driving, and drivers with different temperament characteristics also have different levels of trust in autonomous vehicles.

[0005] (2) When measuring drivers’ trust in autonomous vehicles, the influence of drivers’ subjective awareness was ignored, and subjective and objective factors were not combined for analysis during the measurement process;

[0006] (3) When measuring the driver's trust in the autonomous vehicle in real time, it is mainly achieved through the driver's behavior and eye movements. There are fewer measurements based on the driver's physiological indicators, which will lead to inaccurate trust measurement results. Summary of the invention

[0007] Purpose of the invention: In view of the above problems, the purpose of the present invention is to provide a system for measuring the trust of drivers of different temperament types in autonomous driving, based on physiological indicator measurements and combining subjective and objective factors to quantify the driver's trust in autonomous driving.

[0008] Technical solution: A system for measuring the trust of drivers of different temperament types in automatic driving of the present invention comprises a driver temperament type differentiation module, a trust measurement module, a trust level classification module, a trust level judgment module and a comprehensive trust evaluation module;

[0009] A driver temperament type differentiation module is used to differentiate the driver's temperament type, where the temperament types include choleric, sanguine, phlegmatic and melancholic;

[0010] The trust measurement module measures the driver's trust from both a subjective and objective perspective;

[0011] The trust level grading module uses the results of the trust measurement module to grade subjective trust and objective trust respectively;

[0012] A trust level judgment module is used to judge whether the subjective trust level classification is consistent with the objective trust level classification;

[0013] The comprehensive trust evaluation module is used to calculate the sum of the subjective trust value and the objective trust value, and perform intervention processing according to the calculation result.

[0014] Further, the trust measurement module includes a subjective trust measurement unit and an objective trust measurement unit;

[0015] In the subjective trust measurement unit, the driver's trust is quantified using the autonomous driving situational trust scale to obtain each driver's subjective trust value T 1 ;

[0016] In the objective trust measurement unit, the driver's physiological indicators and hand-foot interaction information are measured and quantified.

[0017] Furthermore, the driver's physiological indicators and hand-foot interaction information are measured and quantified, including:

[0018] Measure the driver's heart rate, blood pressure, blood oxygen saturation and brain wave information, and quantize each measurement value to [0, 1], denoted as P 1 , P 2 , P 3 and P 4 , and then converted into the corresponding physiological index frequency, the calculation formula is:

[0019]

[0020] In the formula, f h Indicates the frequency of a physiological index of the driver at this moment, P h is the value of a physiological index of the driver at this moment, h=1,2,3,4, P hmin It is the minimum value of a physiological index of the driver during driving. hmax It is the maximum value of a physiological index of the driver during driving;

[0021] Measuring the interaction between the driver's hands and feet through an inertial measurement unit 1 and i 2 , when the driver's right foot is in a relaxed state 1 =0, when the driver's right foot is in a tense state 1 =1, when the driver’s hands are in a relaxed state 2 =0, when the driver’s hands are in a tense state 2=1.

[0022] Further, the trust level grading module includes a subjective trust grading unit and an objective trust grading unit;

[0023] In the subjective trust grading unit, the subjective trust is divided into high trust group, relative trust group and low trust group by K-modes clustering algorithm;

[0024] In the objective trust grading unit, the frequency of physiological indicators and interaction information are input into the dynamic Bayesian network to obtain the objective trust value, which is then divided into a high trust group, a relative trust group, and a low trust group.

[0025] Furthermore, the subjective trust is divided into a high trust group, a relative trust group and a low trust group by using the K-modes clustering algorithm, including the following steps:

[0026] Step 101, organize the subjective trust values ​​of all drivers into a discrete data set, set the number of groups K=3, the three groups correspond to the high trust group, the relatively trust group and the low trust group respectively, and randomly select an initial point for each group as the center of the group. The center represents the mode of each group, and one sample corresponds to the subjective trust value of one driver;

[0027] Step 102, calculate the Hamming distance between each sample and each group center, and assign the sample to the group represented by the group center closest to it; the calculation formula of the Hamming distance is:

[0028]

[0029] In the formula, and The samples are and samples The value of the mth attribute, is an indicator function used to determine whether the values ​​of two samples on the mth attribute are different, where n represents the total number of attributes;

[0030] Step 103, for each group, based on all samples currently assigned to it, update the group center so that the group center becomes the most common value of all samples on each attribute;

[0031] Step 104, repeating steps 102 to 103 until the group center no longer changes or the iteration is terminated when the preset number of iterations is reached, and determining whether each sample belongs to a high trust group, a relatively trust group or a low trust group according to the group to which it is assigned.

[0032] Furthermore, the physiological indicator frequency and interaction information are input into the dynamic Bayesian network to obtain the objective trust value, which includes the following steps:

[0033] Step 201, define a dynamic Bayesian network structure, including physiological indicator nodes: heart rate frequency f 1 , blood pressure frequency f 2 , blood oxygen saturation frequency f 3 , brain wave information frequency f 4 , the state space is a continuous value between 0 and 1; state node: right foot state i 1 、Both hands status i 2 , the state space is a binary state, i.e. 0 or 1;

[0034] Step 202, setting the transition probability of the dynamic Bayesian network, for each physiological indicator frequency f h Define the transition probability as P(f h , t+1 |f h , t );

[0035] Step 203, estimate the conditional probability table CPTs: for right foot state i 1 The probability distribution P(i 1 , t |f 1 , t ,f 2 , t ,f 3 , t ,f 4 , t ), indicating that in physiological indicators (f 1 ,f 2 ,f 3 ,f 4 ) Given the right foot state i 1 The probability of 2 The probability distribution P(i 2 , t |f 1 , t ,f 2 , t ,f 3 , t ,f 4 , t ) indicates that in physiological indicators (f 1 ,f 2 ,f 3 ,f 4 ) Given the state of both hands i 2 probability;

[0036] Step 204, mapping the actual physiological index data to the state space of the dynamic Bayesian network, filling the network with the preprocessed data, and updating the conditional probability table and the state transition matrix;

[0037] Step 205, define the confidence index, use the likelihood function to measure the matching degree between the model prediction value and the actual observation value P(data|model), and the calculation formula is:

[0038]

[0039] Where x is the state variable of the dynamic Bayesian network, and P(x) is the prior distribution of the state variable;

[0040] Step 206, by using the training data set and the test data set to evaluate the network performance and verify it, and then through iterative optimization, finally obtain the trust level T 2 .

[0041] Further, the comprehensive trust evaluation module includes a comprehensive trust calculation unit and a trust evaluation unit;

[0042] In the comprehensive trust calculation unit, the total trust is calculated using the subjective trust value and the objective trust value. The calculation formula is:

[0043]

[0044] In the formula, T represents the total trust, T 1 represents the subjective trust value, T 2 represents the objective trust value, T max Represents the minimum trust value among all trust values, T min represents the minimum trust value among all trust values, and α represents the weight of subjective trust;

[0045] In the trust evaluation unit, the total trust of the current driver T is compared with the strategy threshold T 0 The size of 0 , take over and intervene in the current driver.

[0046] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages:

[0047] 1. The present invention can analyze the differences in the trust of drivers with different temperament types in autonomous driving, and adopt different coping strategies for drivers with different temperament types during the autonomous driving process;

[0048] 2. The trust measurement of the present invention combines subjective and objective factors, and adds a trust level judgment module, so that the result reduces the influence of accidental factors and driver's subjective factors;

[0049] 3. The present invention is based on physiological information in the measurement of objective trust, so that the accuracy of real-time measurement is greatly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 A framework diagram of a system for measuring the trust of drivers with different temperament types in autonomous driving;

[0051] Figure 2 It is a flowchart of K-modes clustering algorithm;

[0052] Figure 3 Flowchart of the objective trust grading unit. DETAILED DESCRIPTION

[0053] In order to make the objectives, technical solutions and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments.

[0054] like Figure 1 As shown, a system for measuring the trust of drivers of different temperament types in automatic driving according to this embodiment includes a driver temperament type distinguishing module 100, a trust measurement module 200, a trust level grading module 300, a trust level judging module 400 and a comprehensive trust evaluation module 500;

[0055] A driver temperament type differentiation module is used to differentiate the driver's temperament type, where the temperament types include choleric, sanguine, phlegmatic and melancholic;

[0056] The trust measurement module measures the driver's trust from both a subjective and objective perspective;

[0057] The trust level grading module uses the results of the trust measurement module to grade subjective trust and objective trust respectively;

[0058] A trust level judgment module is used to judge whether the subjective trust level classification is consistent with the objective trust level classification;

[0059] The comprehensive trust evaluation module is used to calculate the sum of the subjective trust value and the objective trust value, and perform intervention processing according to the calculation result.

[0060] This embodiment distinguishes the temperament type of the test subject driver through the driver temperament type distinction module, so that different coping strategies can be adopted for the trust issue of drivers with different temperament types. After the test subject driver determines the temperament type, all test subjects respectively measure the subjective trust and objective trust of the test subject drivers with different temperament types through the trust measurement module. In order to prevent the occurrence of accidental factors or result errors caused by subjective factors of the test subject during the measurement process, a trust level classification module and a trust level judgment module are added to this example. The subjective and objective trust are graded and grouped respectively through the K-modes clustering algorithm and the DBN model. When the subjective and objective grading results of the same test subject are the same, the comprehensive trust evaluation module is performed, and the final total trust value is obtained through the weighted algorithm; then, the takeover intervention process is performed for drivers of different temperament types at different trust thresholds.

[0061] In the driver temperament type differentiation module, drivers are tested using the Chen Huichang Temperament Scale to classify their temperaments. The Chen Huichang Temperament Scale contains 60 questions, using a 5-point Likert scale, with options ranging from very consistent (+2) to completely inconsistent (-2), and the scores of each type of questions are accumulated. When a certain type scores the highest, it is considered that the temperament type of the subject during driving corresponds to that type. Temperament is divided into four categories: choleric, sanguine, phlegmatic, and melancholic.

[0062] In one example, the Chen Huichang Temperament Scale was used for testing. The questionnaire consisted of 60 questions using a 5-point Likert scale, with options ranging from very consistent (+2) to completely inconsistent (-2). Temperament is divided into four categories: choleric temperament corresponds to questions 2, 6, 9, 14, 17, 21, 27, 31, 36, 38, 42, 48, 50, 54, 58 in questions 1-59; sanguine temperament corresponds to questions 4, 8, 11, 16, 19, 23, 25, 29, 34, 40, 44, 46, 52, 56, 60; phlegmatic temperament corresponds to questions 1, 7, 10, 13, 18, 22, 26, 30, 33, 39, 43, 45, 49, 55, 57; melancholic temperament corresponds to questions 3, 5, 12, 15, 20, 24, 28, 32, 35, 37, 41, 47, 51, 53, 59. Add up the scores of the corresponding questions for each type. When the scores of a certain type exceed the scores of the other three by 4 points or more, it is considered to be that type; if the scores of two types are similar and exceed the scores of the other two types by 4 points or more, it is considered to be a mixture of the two types; if the scores of three types are similar and exceed the scores of another type by 4 points or more, it is considered to be a mixture of the three types. For the sake of clarity in the subsequent description, in this example, when the scores of a certain type exceed the scores of the other three, it is considered that the driver is of that type during driving, and the driver's temperament type can be quickly obtained through the Chen Huichang Temperament Scale.

[0063] The trust measurement module 200 includes a subjective trust measurement unit 210 and an objective trust measurement unit 220;

[0064] In the subjective trust measurement unit 210, the driver's trust is quantified using the autonomous driving situation trust scale to obtain the subjective trust value T of each driver. 1 ;

[0065] In the objective trust measurement unit 220 , the driver's physiological indicators and hand-foot interaction information are measured and quantified.

[0066] Specifically, in the subjective trust measurement unit, subjective trust is measured by using the Situational Trust Scale for Autonomous Driving (STS-AD), a short questionnaire used to evaluate different aspects of situational trust. The STS-AD is divided into 6 subscales, each of which evaluates different aspects of situational trust, namely: system reliability, system capability, system transparency, system security, system adaptability, and system trust. The Likert seven-point scale is used for scoring. This method reduces the influence of unnecessary subjective factors and minimizes the impact of excessive trust and lack of trust. Based on the STS-AD, the trust after the experiment is quantified, and the quantification method is the numerical average of the options on the subjective trust scale of each subject. The scale is scored using the Likert seven-point scale, with a scoring range of 1 to 7, representing "strongly disagree" to "strongly agree", and the final average score is recorded as the subjective trust value T of the subject. 1 .

[0067] Measuring the driver's physiological indicators and hand-foot interaction information, and quantifying them including:

[0068] Measure the driver's heart rate, blood pressure, blood oxygen saturation and brain wave information, and quantize each measurement value to [0, 1], denoted as P 1 , P 2 , P 3 and P 4 , and then converted into the corresponding physiological index frequency, the calculation formula is:

[0069]

[0070] In the formula, f h Indicates the frequency of a physiological index of the driver at this moment, P h is the value of a physiological index of the driver at this moment, h=1,2,3,4, P hmin It is the minimum value of a physiological index of the driver during driving. hmax It is the maximum value of a physiological index of the driver during driving;

[0071] Measuring the interaction between the driver's hands and feet through an inertial measurement unit 1 and i 2 , when the driver's right foot is in a relaxed state 1 =0, when the driver's right foot is in a tense state 1 =1, when the driver’s hands are in a relaxed state 2 =0, when the driver’s hands are in a tense state 2 =1.

[0072] In the objective trust measurement unit, the subjects' physiological index data were collected through different measuring devices: Apple Watch Ultra to measure heart rate, Withings BPM Connect to measure blood pressure, Oxitone 1000i to measure blood oxygen saturation, and Muse 2 to measure brain wave information. The interaction information of the driver's hands and feet was measured through the inertial measurement unit (IMU). 1 and i 2 Among them, the interactive information i 1 ∈{0, 1}, when the driver’s right foot is in a relaxed state i 1 = 0, when the driver's right foot is in a tense state, that is, the driver's right foot makes an emergency braking action i 1 =1; interactive information i 2 ∈{0, 1}, when the driver’s hands are in a relaxed state i 2 = 0, when the driver's hands are in a tense state, that is, the driver's hands make the action of controlling the steering wheel i 2 =1.

[0073] The trust level grading module 300 includes a subjective trust grading unit 310 and an objective trust grading unit 320;

[0074] In the subjective trust grading unit, the subjective trust is divided into high trust group, relative trust group and low trust group by K-modes clustering algorithm;

[0075] In the objective trust grading unit, the frequency of physiological indicators and interaction information are input into the dynamic Bayesian network (DBN) to obtain the objective trust value, which is then divided into a high trust group, a relative trust group, and a low trust group.

[0076] The K-modes clustering algorithm is a clustering algorithm suitable for discrete data. It can be used to divide the driver's subjective trust in the autonomous driving system into different levels. K-modes is a method used in data mining to cluster categorical attribute data. Its algorithm idea is relatively simple and its time complexity is relatively low. Figure 2As shown, the K-modes clustering algorithm is used to divide the subjective trust into a high trust group, a relative trust group, and a low trust group, including the following steps:

[0077] Step 101, organize the subjective trust values ​​of all drivers into a discrete data set, set the number of groups K=3, the three groups correspond to the high trust group, the relatively trust group and the low trust group respectively, and randomly select an initial point for each group as the center of the group. The center represents the mode of each group, and one sample corresponds to the subjective trust value of one driver;

[0078] Step 102, calculate the Hamming distance between each sample and each group center, and assign the sample to the group represented by the group center closest to it; the calculation formula of the Hamming distance is:

[0079]

[0080] In the formula, and The samples are and samples The value of the mth attribute, is an indicator function used to determine whether the values ​​of two samples on the mth attribute are different, where n represents the total number of attributes;

[0081] Step 103, for each group, based on all samples currently assigned to it, update the group center so that the group center becomes the most common value of all samples on each attribute;

[0082] Step 104, repeating steps 102 to 103 until the group center no longer changes or the iteration is terminated when the preset number of iterations is reached, and determining whether each sample belongs to a high trust group, a relatively trust group or a low trust group according to the group to which it is assigned.

[0083] like Figure 3 As shown, the physiological indicator frequency and interaction information are input into the dynamic Bayesian network to obtain the objective trust value, which includes the following steps:

[0084] Step 201, define a dynamic Bayesian network structure, including physiological indicator nodes: heart rate frequency f 1 , blood pressure frequency f 2 , blood oxygen saturation frequency f 3 , brain wave information frequency f 4 , the state space is a continuous value between 0 and 1; state node: right foot state i 1 、Both hands status i 2 , the state space is a binary state, i.e. 0 or 1;

[0085] Step 202, setting the transition probability P of the dynamic Bayesian network, for each physiological indicator frequency f h Define the transition probability as P(f h , t+1 |f h , t );

[0086] Step 203, estimate the conditional probability table CPTs: for right foot state i 1 The probability distribution P(i 1 , t |f 1 , t ,f 2 , t ,f 3 , t ,f 4 , t ), indicating that in physiological indicators (f 1 ,f 2 ,f 3 ,f 4 ) Given the right foot state i 1 The probability of 2 The probability distribution P(i 2 , t |f 1 , t ,f 2 , t ,f 3 , t ,f 4 , t ) indicates that in physiological indicators (f 1 ,f 2 ,f 3 ,f 4 ) Given the state of both hands i 2 probability;

[0087] Step 204, mapping the actual physiological index data to the state space of the dynamic Bayesian network, using the data obtained in steps 201 to 203 to fill the dynamic Bayesian network, and updating the conditional probability table and the state transition matrix;

[0088] Step 205, define the confidence index, use the likelihood function to measure the matching degree between the model prediction value and the actual observation value P(data|model), and the calculation formula is:

[0089]

[0090] Where x is the state variable of the dynamic Bayesian network, and P(x) is the prior distribution of the state variable;

[0091] Step 206, by using the training data set and the test data set to evaluate the network performance and verify it, and then through iterative optimization, finally obtain the trust level T 2 , and T 2 In the range of 0 to 1.

[0092] Cross-validation can also be used for verification: that is, by dividing the data set into several subsets, training and testing the model on different subsets, and calculating the average error of the model. After the verification model is completed, it is iteratively optimized to finally obtain an accurate confidence level T. 2 The data set consists of 1,000 sets of physiological indicator frequencies and interaction information of test drivers, which are divided into training set and test set in a ratio of 7:3.

[0093] In the objective trust grading unit, the trust level T 2 Then the trust levels are graded, and [0, 1 / 3) is defined as the low trust group, [1 / 3, 2 / 3) as the relative trust group, and [2 / 3, 1] as the high trust group.

[0094] In the trust level judgment module, it is judged whether the trust group obtained by the subjective trust grading unit and the trust group obtained by the objective trust grading unit for the same subject are consistent. If the trust groups are consistent, the total trust is calculated. If not, re-measurement is performed.

[0095] The comprehensive trust evaluation module 500 includes a comprehensive trust calculation unit 510 and a trust evaluation unit 520;

[0096] In the comprehensive trust calculation unit 510, the total trust is calculated using the subjective trust value and the objective trust value, and the calculation formula is:

[0097]

[0098] In the formula, T represents the total trust, T 1 represents the subjective trust value, T 2 represents the objective trust value, T max Represents the minimum trust value among all trust values, T min represents the minimum trust value among all trust values, and α represents the weight of subjective trust;

[0099] In the trust evaluation unit 520, the total trust T of the current driver is compared with the strategy threshold T 0 The size of 0 , take over and intervene in the current driver.

[0100] The total trust T obtained is in the range of 0 to 1. Drivers with different temperaments should adopt different coping strategies when their trust is different. Drivers with different temperaments should set different trust thresholds for takeover intervention, which is called the strategy threshold in this example.

[0101] Drivers with different temperaments adopt different strategies when driving, which in turn affects driving behavior. Drivers with phlegmatic and sanguine characteristics are relatively stable during driving, so the strategy threshold can be set to a general 0.5; choleric drivers tend to take risks. The more obvious the choleric characteristics are, the worse their safety and emotional control ability is. Therefore, different high strategy thresholds can be set according to different levels of choleric drivers; melancholic drivers are lonely and weird, slow in action, and have relatively poor driving skills. Therefore, different low strategy thresholds can be set according to different levels of melancholic drivers. When the driver's total trust T is greater than the corresponding strategy threshold, the current driver needs to be taken over and intervened.

Claims

1. A system for measuring the trust of drivers of different temperament types in autonomous driving, characterized in that: It includes a driver temperament type differentiation module, a trust measurement module, a trust level classification module, a trust level judgment module and a comprehensive trust evaluation module; A driver temperament type differentiation module is used to differentiate the driver's temperament type, where the temperament types include choleric, sanguine, phlegmatic and melancholic; The trust measurement module measures the driver's trust from both a subjective and objective perspective; The trust level grading module uses the results of the trust measurement module to grade subjective trust and objective trust respectively; A trust level judgment module is used to judge whether the subjective trust level classification is consistent with the objective trust level classification; The comprehensive trust evaluation module is used to calculate the sum of the subjective trust value and the objective trust value, and perform intervention processing according to the calculation result.

2. The system for measuring the trust of drivers of different temperament types in automatic driving according to claim 1, characterized in that: The trust measurement module includes a subjective trust measurement unit and an objective trust measurement unit; In the subjective trust measurement unit, the driver's trust is quantified using the autonomous driving situational trust scale to obtain each driver's subjective trust value T1; In the objective trust measurement unit, the driver's physiological indicators and hand-foot interaction information are measured and quantified.

3. The system for measuring the trust of drivers of different temperament types in automatic driving according to claim 2, characterized in that: Measuring the driver's physiological indicators and hand-foot interaction information, and quantifying them including: The driver's heart rate, blood pressure, blood oxygen saturation and brain wave information are measured, and each measurement value is quantized to [0, 1], recorded as P1, P2, P3 and P4 respectively, and then converted into the corresponding physiological index frequency. The calculation formula is: In the formula, f h Indicates the frequency of a physiological index of the driver at this moment, P h is the value of a physiological index of the driver at this moment, h=1,2,3,4, P hmin It is the minimum value of a physiological index of the driver during driving. hmax It is the maximum value of a physiological index of the driver during driving; The interaction information i1 and i2 between the driver's hands and feet are measured by the inertial measurement unit. When the driver's right foot is in a relaxed state, i1=0, and when the driver's right foot is in a tense state, i1=1. When the driver's hands are in a relaxed state, i2=0, and when the driver's hands are in a tense state, i2=1.

4. The system for measuring the trust of drivers of different temperament types in automatic driving according to claim 3, characterized in that: The trust level grading module includes a subjective trust level grading unit and an objective trust level grading unit; In the subjective trust grading unit, the subjective trust is divided into high trust group, relative trust group and low trust group by K-modes clustering algorithm; In the objective trust grading unit, the frequency of physiological indicators and interaction information are input into the dynamic Bayesian network to obtain the objective trust value, which is then divided into a high trust group, a relative trust group, and a low trust group.

5. The system for measuring the trust of drivers of different temperament types in automatic driving according to claim 4, characterized in that: The K-modes clustering algorithm is used to divide the subjective trust into high trust group, relative trust group and low trust group, which includes the following steps: Step 101, organize the subjective trust values ​​of all drivers into a discrete data set, set the number of groups K=3, the three groups correspond to the high trust group, the relatively trust group and the low trust group respectively, and randomly select an initial point for each group as the center of the group. The center represents the mode of each group, and one sample corresponds to the subjective trust value of one driver; Step 102, calculate the Hamming distance between each sample and each group center, and assign the sample to the group represented by the group center closest to it; the calculation formula of the Hamming distance is: In the formula, and The samples are and samples The value of the mth attribute, is an indicator function used to determine whether the values ​​of two samples on the mth attribute are different, where n represents the total number of attributes; Step 103, for each group, based on all samples currently assigned to it, update the group center so that the group center becomes the most common value of all samples on each attribute; Step 104, repeating steps 102 to 103 until the group center no longer changes or the iteration is terminated when the preset number of iterations is reached, and determining whether each sample belongs to a high trust group, a relatively trust group or a low trust group according to the group to which it is assigned.

6. The system for measuring the trust of drivers of different temperament types in automatic driving according to claim 5, characterized in that: Inputting the physiological indicator frequency and interaction information into the dynamic Bayesian network to obtain the objective trust value includes the following steps: Step 201, define a dynamic Bayesian network structure, including physiological index nodes: heart rate frequency f1, blood pressure frequency f2, blood oxygen saturation frequency f3, brain wave information frequency f4, the state space is a continuous value between 0 and 1; state nodes: right foot state i1, both hands state i2, the state space is a binary state, that is, 0 or 1; Step 202, setting the transition probability P of the dynamic Bayesian network, for each physiological indicator frequency f h Define the transition probability as P(f h , t+1 |f h , t ); Step 203, estimate the conditional probability table CPTs: the probability distribution P(i1, t |f1, t ,f2, t ,f3, t ,f4, t ), which represents the probability of right foot state i1 given the physiological indicators (f1, f2, f3, f4); for the probability distribution of both hands state i2 P(i2, t |f1, t ,f2, t ,f3, t ,f4, t ) represents the probability of the two-hand state i2 when the physiological indicators (f1, f2, f3, f4) are given; Step 204, mapping the actual physiological index data to the state space of the dynamic Bayesian network, filling the network with the preprocessed data, and updating the conditional probability table and the state transition matrix; Step 205, define the confidence index, use the likelihood function to measure the matching degree between the model prediction value and the actual observation value P (data | model), and the calculation formula is: Where x is the state variable of the dynamic Bayesian network, and P(x) is the prior distribution of the state variable; Step 206, the network performance is evaluated and verified by using the training data set and the test data set, and then after iterative optimization, the trust level T2 is finally obtained.

7. The system for measuring the trust of drivers of different temperament types in automatic driving according to claim 6, characterized in that: The comprehensive trust evaluation module includes a comprehensive trust calculation unit and a trust evaluation unit; In the comprehensive trust calculation unit, the total trust is calculated using the subjective trust value and the objective trust value. The calculation formula is: In the formula, T represents the total trust, T1 represents the subjective trust value, T2 represents the objective trust value, and T max Represents the minimum trust value among all trust values, T min represents the minimum trust value among all trust values, and α represents the weight of subjective trust; In the trust evaluation unit, the total trust T of the current driver is compared with the strategy threshold T0. When T>T0, the current driver is taken over and intervened.

Images