An evaluation method for the human-computer interaction interface outside the autonomous driving vehicle

Abstract

The present invention relates to the technical field of human-computer interaction between autonomous driving vehicles and road users, and specifically to a method for evaluating an off-vehicle human-computer interaction interface of an autonomous driving vehicle. During the evaluation, the availability of off-vehicle human-computer interaction is determined based on the three factors of effectiveness, efficiency and satisfaction from the aspects of dynamic behavior, interaction mode and interaction scenario factors of the autonomous driving vehicle, and corresponding evaluation indicators are proposed for the three factors in combination with the interaction characteristics between the autonomous driving vehicle and the road user, as well as a quantification method for each evaluation indicator. In particular, EEG data, blood oxygen data, skin electricity data, eye movement data and the like are introduced to objectively evaluate the off-vehicle human-computer interaction interface, and the optimal off-vehicle human-computer interaction interface is designed based on the hierarchical calculation evaluation results, and necessary information is transmitted to the road user through the off-vehicle human-computer interaction interface, and then based on the recognition and understanding of the interface information, traffic behavior decisions are made efficiently and accurately to ensure interaction safety.

Description

Technical Field

[0001] The present invention relates to the technical field of human-computer interaction between an autonomous driving vehicle and road users, and in particular to a method for evaluating an external human-computer interaction interface of an autonomous driving vehicle. Background Art

[0002] With the rapid development of new-generation information technologies such as artificial intelligence, mobile Internet, and big data, the new-generation intelligent transportation system with autonomous driving as its main feature will become a breakthrough in solving traffic problems. Autonomous driving technology has become a strategic direction for the development of the global automotive industry. It is an inevitable trend for road traffic to gradually develop from the era of traditional automobiles to the era of autonomous driving.

[0003] In an autonomous vehicle, the driver's main task will no longer be driving-related tasks, but non-driving tasks such as office work and entertainment. At this time, the driver no longer pays attention to the road conditions or there is no driver at all, resulting in the lack of communication between the driver's eyes, gestures, etc. with other road users during traditional driving. The human-computer interaction outside the autonomous vehicle for road users is one of the key methods to solve the lack of "non-verbal interaction". The road users interacting with autonomous vehicles mainly include traditional cars, ordinary bicycles, electric bicycles, motorcycles and pedestrians. Among them, human drivers will interact with autonomous vehicles through vehicle-to-vehicle interconnection technology, while relatively weak road users cyclists and pedestrians still need to judge the dynamic behavior of autonomous vehicles through human-computer interaction outside the vehicle. The unreasonable human-computer interaction interface of autonomous vehicles makes it impossible for road users to accurately and efficiently judge the dynamic behavior of autonomous vehicles, thereby inducing interaction risks. Therefore, it is necessary to design an evaluation method for the human-computer interaction interface outside the autonomous vehicle. Summary of the invention

[0004] The problem solved by the present invention is to provide a method for evaluating the human-computer interaction interface outside the vehicle of an autonomous driving vehicle, through which necessary information is transmitted to road users, and then based on the recognition and understanding of the interface information, traffic behavior decisions are made efficiently and accurately to ensure interaction safety.

[0005] In order to achieve the above object, the present invention adopts the following technical solution.

[0006] A method for evaluating an external human-machine interaction interface of an autonomous driving vehicle, the evaluation steps comprising:

[0007] S1: Construction of evaluation index system;

[0008] Set three first-level evaluation indicators, including effectiveness, efficiency and satisfaction;

[0009] Effectiveness includes two secondary indicators: information cognition and behavioral decision-making;

[0010] Among them, information cognition is measured by a three-level subjective indicator, information cognition correctness;

[0011] Efficiency includes two secondary indicators: perceived efficiency and decision-making efficiency;

[0012] Among them, perception efficiency is comprehensively measured by two three-level objective indicators, fixation time and number of fixations, and decision efficiency is measured by one three-level objective indicator, decision time;

[0013] Satisfaction includes three secondary indicators: emotional load, cognitive load and experience index. Emotional load is comprehensively measured by four third-level objective indicators: blinking frequency, breathing rate, real-time heart rate and skin conductance level.

[0014] Cognitive load is comprehensively measured by three three-level objective indicators, EEG power spectrum ratio, prefrontal rSO2, gaze entropy, and one three-level subjective indicator, subjective cognitive load;

[0015] The experience index is comprehensively measured by four three-level subjective indicators: aesthetic experience, functional experience, interactive experience, and trust experience;

[0016] S2: Measure rating indicators based on evaluation indicators

[0017] 1) Information cognition is measured by information cognition correctness. Information cognition correctness refers to whether the road user correctly understands the behavior intention of the autonomous vehicle represented by the human-machine interaction interface outside the vehicle. It is measured through open-ended questions after the experiment. Correct understanding is assigned 5 points and incorrect understanding is assigned 1 point.

[0018] 2) Behavioral decision-making is measured by behavioral decision correctness. Behavioral decision correctness refers to whether the traffic behavior decision made by the road user after reading the human-machine interaction interface outside the vehicle is correct. The correctness of the behavioral decision of the road user is observed during the experiment. A correct decision is assigned 5 points, and an incorrect decision is assigned 1 point.

[0019] 3) Perception efficiency refers to the efficiency of road users from discovering the autonomous vehicle to perceiving the vehicle's behavioral intention. It is comprehensively measured by gaze time and number of gaze points, and eye movement data is collected using a lightweight wearable eye tracker.

[0020] 31) Gaze time refers to the time interval from the beginning to the end of a road user's gaze when he or she is reading the human-machine interaction interface outside the vehicle. The longer the road user's gaze time on the human-machine interaction interface outside the vehicle, the lower the efficiency of visual information search, acquisition and understanding, and the lower the perception efficiency.

[0021] 32) The number of fixation points refers to the number of fixation points that a road user places on the interface when reading the human-machine interaction interface outside the vehicle. The more fixation points a road user places on the human-machine interaction interface outside the vehicle, the lower the visual search efficiency, and the lower the perception efficiency.

[0022] 4) Decision efficiency is measured by decision time, which refers to the time interval from seeing the autonomous vehicle to making a traffic behavior decision. The shorter the decision time, the higher the decision efficiency. High-definition cameras, timers and other equipment are used for data collection;

[0023] 5) Emotional load refers to the emotional tension of road users when reading the human-machine interaction interface outside the vehicle, which is comprehensively measured by blinking frequency, breathing rate, real-time heart rate and skin conductance level;

[0024] 51) There is a linear correlation between blinking frequency and emotional load. When a road user reads the human-machine interaction interface outside the vehicle, the higher the blinking frequency, the higher the emotional load. The wearable eye tracker is used to collect data. The blinking frequency calculation formula is as follows:

[0025]

[0026] Among them, BF represents the blinking frequency during the interaction, EBF represents the total number of blinks during the interaction time, and t represents the interaction time;

[0027] 52) When performing complex tasks or being under stress, there will be strong breathing or over-breathing. When road users read the human-machine interaction interface outside the vehicle, the higher the breathing rate, the higher the emotional load. The breathing patch sensor is used to collect data. The breathing frequency calculation formula is as follows:

[0028]

[0029] Among them, RR represents the respiratory rate during the interaction, ERR represents the total number of breaths during the interaction time, and t represents the interaction time;

[0030] 53) Real-time heart rate. The higher the real-time heart rate, the greater the emotional load. When the road user reads the human-machine interaction interface outside the vehicle, the higher the real-time heart rate average, the higher the emotional load. A multi-channel physiological instrument is used for data collection;

[0031] 54) The strength of skin conductance is positively correlated with the amount of sweat gland secretion. The secretion of sweat glands is affected by two main factors: temperature and physiological and psychological activities. In particular, physiological and psychological reactive sweat secretion is affected by the sympathetic nerves. When the nerve activity is enhanced, the amount of sweat gland secretion increases and the skin conductance level increases. A multi-conductor instrument is used for data collection. When road users read the human-machine interaction interface outside the vehicle, the higher the skin conductance level, the higher the emotional load.

[0032] 6) Cognitive load refers to the cognitive pressure of road users in the process of reading the human-machine interaction interface outside the car, which is comprehensively measured by EEG power spectrum ratio, prefrontal rSO2, gaze entropy and subjective cognitive load;

[0033] 61) The EEG power spectrum ratio can evaluate the cognitive load of road users in the process of reading the human-machine interaction interface outside the vehicle. It is mainly evaluated by the power spectrum ratio of θ / α in the prefrontal region of the scalp. The θ wave refers to the band with a frequency of 4-7Hz in the EEG signal, and the α wave refers to the band with a frequency of 8-13Hz in the EEG signal. By performing spectral analysis on the EEG signal, the average power of the θ wave and the α wave under different tasks and benchmarks is obtained. The higher the power spectrum ratio of θ / α, the greater the cognitive load generated by the test task. The EEG data is collected using an electroencephalometer. The calculation formula of the EEG power spectrum ratio is as follows:

[0034]

[0035] Among them, EER represents the EEG power spectrum ratio, Pθ represents the power spectrum value of theta wave, and Pα represents the power spectrum value of alpha wave;

[0036] 62) Prefrontal rSO2 is the local blood oxygen saturation of the prefrontal lobe, which has a high correlation with the blood oxygen saturation of the jugular bulb and can be used to quantify and analyze cognitive load. When road users read the human-machine interaction interface outside the vehicle, the greater the prefrontal rSO2 value, the greater the cognitive load. Functional near-infrared spectroscopy technology is used to collect data;

[0037] 63) Gaze entropy represents the uncertainty of gaze position within a given viewing time. Compared with other eye movement indicators, the measurement value of gaze entropy has the advantage of not being affected by external factors. Data is collected through a wearable eye tracker. When road users read the human-computer interaction interface outside the car, the greater the gaze entropy, the greater the cognitive load. The calculation formula of gaze entropy is as follows:

[0038]

[0039] Where E is the gaze entropy; v is the number of gaze areas; Pi is the probability that the subject gazes at area u, u = 1, 2, ·…·v;

[0040] 7) Subjective cognitive load The NASA-TLX scale was used to score subjective cognitive load, which consists of 6 indicators: mental demand, physical demand, time limit, self-performance, effort, and frustration.

[0041] 8) Experience index refers to the subjective experience evaluation of road users on the human-computer interaction interface outside the vehicle. The experience index is characterized by four dimensions: aesthetics, function, interaction, and trust. The aesthetic experience scale is used to measure the aesthetic experience index, the functional experience scale is used to measure the functional experience index, the interactive experience scale is used to measure the interactive experience index, and the trust scale is used to measure the trust index.

[0042] S3: Evaluation calculation through rating index measurement:

[0043] Q1: Indicator weight

[0044] The weights of the first-level indicators are:

[0045]

[0046] where w i represents the weight of the i-th first-level indicator, and n represents the number of first-level indicators;

[0047] The weights of the secondary indicators are:

[0048]

[0049] where w ij represents the weight of the jth secondary indicator under the ith first-level indicator, and m represents the number of secondary indicators under the ith first-level indicator;

[0050] The weights of the three-level indicators are:

[0051]

[0052] where w ijk represents the weight of the kth third-level indicator under the jth second-level indicator under the i-th first-level indicator, and q represents the number of third-level indicators under the jth second-level indicator under the i-th first-level indicator;

[0053] The relative importance of each indicator is scored using the priority relationship quantitative scaling method, with a quantitative scale of 0.1 to 0.9, that is, comparing two indicators:

[0054] When the scale value is 0.5, it indicates that the two indicators are equally important;

[0055] When the scale value is 0.6, it indicates that indicator 1 is slightly more important than indicator 2; if the value is 0.4, the opposite is true;

[0056] When the scale value is 0.7, it indicates that indicator 1 is significantly more important than indicator 2; if the value is 0.3, the opposite is true;

[0057] When the scale value is 0.8, it means that indicator 1 is much more important than indicator 2; if the value is 0.2, the opposite is true;

[0058] When the scale value is 0.9, it means that indicator 1 is extremely important than indicator 2; if the value is 0.1, the opposite is true;

[0059] Taking the calculation of the weight of the first-level indicator as an example, the expert scoring method is used to assign importance to the first-level indicators of the same level: effectiveness K1, efficiency K2 and satisfaction K3. aij represents Ki and Kj, where i, j = 1, 2, 3;

[0060] Construct a fuzzy complementary judgment matrix, the fuzzy complementary judgment matrix M is:

[0061]

[0062] Perform consistency check on the judgment matrix M, calculate the maximum characteristic root of the matrix, and calculate its consistency index:

[0063]

[0064] Query the average random consistency index RI (n=3, take RI=0.58), and calculate the consistency ratio CR:

[0065]

[0066] Judging consistency: When CR < 0.1, the consistency of the matrix can be considered acceptable;

[0067] The square root method is used to calculate the eigenvector W'. The row product of the judgment matrix is ​​calculated and then the nth square root is taken, as shown in the following formula:

[0068] W′=[w′1,w′2,…w′ i ],i=1,2,…n

[0069] in

[0070] Further obtain the weight value of the first-level indicator:

[0071]

[0072] Then we can get the weight value of the first-level index. Similarly, we can get the weights of the second-level index and the third-level index.

[0073] Q2: Raw data processing

[0074] The processing of raw data includes two aspects: subjective data processing and objective data processing. Subjective data is obtained by taking the average of the scale data of multiple subjects, and the data range is 1 to 5 points. The objective data uses the function mapping relationship f(x) to convert the raw data into a 5-point score consistent with the subjective scale. According to the large sample experimental data, the evaluation index data obeys the normal distribution, that is, X~N(μ,σ), so as to determine the function expression:

[0075]

[0076] Among them, x represents the actual value of objective data; μ and σ represent the mean and standard deviation of this type of objective data respectively;

[0077] Q3: Comprehensive evaluation

[0078]

[0079] Among them, S represents the comprehensive score of the human-machine interaction interface outside the autonomous driving vehicle; S ijk Indicates the 5-point score corresponding to the third-level indicators.

[0080] In S1, effectiveness means that when a road user interacts with an autonomous vehicle, the human-computer interaction interface outside the vehicle can guide the user to make correct information understanding and traffic behavior decisions. It includes two secondary indicators: information cognition and behavior decision. Information cognition is measured by a third-level subjective indicator, information cognition correctness. If the information cognition index is larger, the effectiveness index is larger. Behavior decision is measured by a third-level subjective indicator, behavior decision correctness. If the behavior decision index is larger, the effectiveness index is larger.

[0081] In S1, efficiency means that when a road user interacts with an autonomous vehicle, the human-computer interaction interface outside the vehicle can guide the road user to make correct information understanding and traffic behavior decisions more timely and efficiently. It includes perception efficiency and decision efficiency. The perception efficiency is comprehensively measured by two three-level objective indicators, gaze time and number of gaze points. If the perception efficiency index is larger, the efficiency index is larger. The decision efficiency is measured by a three-level objective indicator, decision time. If the decision efficiency index is larger, the efficiency index is larger.

[0082] In S1, satisfaction refers to the fact that when a road user interacts with an autonomous vehicle, under the guidance of human-computer interaction outside the vehicle, the psychological and cognitive pressure of the road user when interacting with the autonomous vehicle can be alleviated, and the acceptability and comfort of the interaction process can be improved. It includes emotional load, cognitive load and experience index. The emotional load is comprehensively measured by four three-level objective indicators: blinking frequency, breathing rate, real-time heart rate and skin conductance level. If the emotional load index is larger, the satisfaction is smaller. The cognitive load is comprehensively measured by three three-level objective indicators: EEG power spectrum ratio, prefrontal rSO2, gaze entropy and one three-level subjective indicator: subjective cognitive load. If the cognitive load index is larger, the satisfaction is smaller. The experience index is comprehensively measured by four three-level subjective indicators: aesthetic experience, functional experience, interactive experience and trust experience. If the experience index index is larger, the satisfaction is greater.

[0083] A method for extracting a human-computer interaction scene outside an autonomous driving vehicle, the extraction process comprising:

[0084] The first level: includes the initial set of dynamic behaviors of autonomous driving vehicles and the initial set of interaction modes, which are filtered by filter A. The filtering criteria are to eliminate the dynamic behaviors of vehicles that can be represented by turn signals, reverse lights, and brake lights in traditional ways;

[0085] Filtering is performed through filters B and C. The criterion of filter B is to eliminate those interaction modes that do not cause the risk of traffic conflict;

[0086] The criterion of filter C is to eliminate those interaction modes with similar or insignificant conflict characteristics; the merging interactions of road users from the left and right sides of the autonomous vehicle can be combined into one mode of side merging interactions;

[0087] The second level: includes the extracted combination of dynamic behaviors and interaction patterns of autonomous driving vehicles and the initial set of interaction scenario factors; among them, the extracted combination of dynamic behaviors and interaction patterns of autonomous driving vehicles is the combination after screening at the first level;

[0088] The initial set of interaction scenario factors is screened by filter D. The screening criterion is to eliminate factors that have similar or no significant differences in the requirements for human-machine interaction outside the vehicle;

[0089] The third level: the extracted combination of interaction scenario factors, autonomous driving vehicle dynamic behaviors, and interaction modes, which is the combination after the second level screening, is screened by filter E. The screening criterion is to eliminate interaction scenarios that do not conform to reality.

[0090] At the fourth level, through screening by filter A, filter B, filter C, filter D, and filter E, n interaction scenarios between autonomous driving vehicles and road users with significant feature differences are extracted, and actual tests and evaluations are carried out in different interaction scenarios.

[0091] In the first level, the dynamic behavior of the autonomous driving vehicle refers to all possible dynamic behaviors performed by the car, including driving direction: forward / reverse, turning, changing lanes, and driving speed: constant speed, acceleration, deceleration, thereby proposing an initial set of dynamic behaviors of the autonomous driving vehicle.

[0092] In the first level, the interaction mode refers to all possible interaction modes when the autonomous vehicle and the road user meet at the same location. Two factors that affect the interaction mode are considered: the relative position relationship between the road user and the autonomous vehicle and the moving direction, and thus an initial set of interaction modes is proposed;

[0093] In the second level, interaction scene factors refer to factors that affect the attributes of the interaction scene, such as road type, traffic environment, right of way, initial vehicle speed, etc., and an initial set of interaction scene factors is proposed.

[0094] The beneficial effects of the present invention are as follows: the present invention proposes a hierarchical screening method by demarcating the scene area between the autonomous driving vehicle and the cyclists and pedestrians, and extracts multiple interactive scenes with significant feature differences in accordance with the screening criteria from aspects such as the dynamic behavior of the autonomous driving vehicle, the interaction mode, and the interaction scene factors;

[0095] Regarding the usability of human-computer interaction outside the vehicle, the system usability is determined based on the three factors of effectiveness, efficiency and satisfaction. Combined with the interaction characteristics between autonomous vehicles and road users, corresponding evaluation indicators are proposed for the three factors, as well as the quantification method of each evaluation indicator; in particular, EEG data, blood oxygen data, skin electricity data, eye movement data, etc. are introduced to objectively evaluate the human-computer interaction interface outside the vehicle;

[0096] Based on the hierarchical calculation and evaluation results, the optimal vehicle-exterior human-computer interaction interface is designed, and the necessary information is transmitted to road users through the vehicle-exterior human-computer interaction interface. Then, based on the recognition and understanding of the interface information, traffic behavior decisions are made efficiently and accurately to ensure interactive safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0097] Figure 1 It is a schematic diagram of the method flow of the present invention;

[0098] Figure 2 A schematic diagram showing the subjectivity and objectivity of indicators in the evaluation method of the present invention;

[0099] Figure 3 It is a schematic diagram of the relationship between the evaluation index levels of the present invention. DETAILED DESCRIPTION

[0100] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0101] Specific examples are given below.

[0102] See also Figure 1-3 , a method for evaluating an autonomous driving vehicle's exterior human-machine interaction interface, the evaluation steps comprising:

[0103] S1: Construction of evaluation index system;

[0104] Set three first-level evaluation indicators, including effectiveness, efficiency and satisfaction;

[0105] Effectiveness includes two secondary indicators: information cognition and behavioral decision-making;

[0106] Among them, information cognition is measured by a three-level subjective indicator, information cognition correctness;

[0107] Effectiveness refers to the ability of the human-machine interface outside the vehicle to guide road users to make correct information understanding and traffic behavior decisions when interacting with autonomous vehicles. It includes two secondary indicators: information cognition and behavior decision. Information cognition is measured by a third-level subjective indicator, information cognition correctness. If the information cognition indicator is larger, the effectiveness indicator is larger. Behavior decision is measured by a third-level subjective indicator, behavior decision correctness. If the behavior decision indicator is larger, the effectiveness indicator is larger.

[0108] Efficiency includes two secondary indicators: perceived efficiency and decision-making efficiency;

[0109] Among them, perception efficiency is comprehensively measured by two three-level objective indicators, gaze time and number of gaze points, and decision efficiency is measured by one three-level objective indicator, decision time; efficiency means that when road users interact with autonomous vehicles, the human-machine interaction interface outside the vehicle can guide road users to make correct information understanding and traffic behavior decisions more timely and efficiently, which includes perception efficiency and decision efficiency. Perception efficiency is comprehensively measured by two three-level objective indicators, gaze time and number of gaze points. If the perception efficiency index is larger, the efficiency index is larger. Decision efficiency is measured by one three-level objective indicator, decision time. If the decision efficiency index is larger, the efficiency index is larger.

[0110] Satisfaction includes three secondary indicators: emotional load, cognitive load and experience index. Emotional load is comprehensively measured by four third-level objective indicators: blinking frequency, breathing rate, real-time heart rate and skin conductance level.

[0111] Cognitive load is comprehensively measured by three three-level objective indicators, EEG power spectrum ratio, prefrontal rSO2, gaze entropy, and one three-level subjective indicator, subjective cognitive load;

[0112] The experience index is comprehensively measured by four three-level subjective indicators: aesthetic experience, functional experience, interactive experience, and trust experience;

[0113] Satisfaction refers to the fact that when a road user interacts with an autonomous vehicle, under the guidance of human-computer interaction outside the vehicle, the psychological and cognitive pressure of the road user in the interaction with the autonomous vehicle can be alleviated, and the acceptability and comfort of the interaction process can be improved. It includes emotional load, cognitive load and experience index. The emotional load is comprehensively measured by four three-level objective indicators: blinking frequency, breathing rate, real-time heart rate and skin conductance level. If the emotional load index is larger, the satisfaction is smaller. The cognitive load is comprehensively measured by three three-level objective indicators: EEG power spectrum ratio, prefrontal rSO2, gaze entropy and one three-level subjective indicator: subjective cognitive load. If the cognitive load index is larger, the satisfaction is smaller. The experience index is comprehensively measured by four three-level subjective indicators: aesthetic experience, functional experience, interactive experience and trust experience. If the experience index index is larger, the satisfaction is greater.

[0114] S2: Measure rating indicators based on evaluation indicators

[0115] 1) Information cognition is measured by information cognition correctness. Information cognition correctness refers to whether the road user correctly understands the behavior intention of the autonomous vehicle represented by the human-machine interaction interface outside the vehicle. It is measured through open-ended questions after the experiment. Correct understanding is assigned 5 points and incorrect understanding is assigned 1 point.

[0116] 2) Behavioral decision-making is measured by behavioral decision correctness. Behavioral decision correctness refers to whether the traffic behavior decision made by the road user after reading the human-machine interaction interface outside the vehicle is correct. The correctness of the behavioral decision of the road user is observed during the experiment. A correct decision is assigned 5 points, and an incorrect decision is assigned 1 point.

[0117] 3) Perception efficiency refers to the efficiency of road users from discovering the autonomous vehicle to perceiving the vehicle's behavioral intention. It is comprehensively measured by gaze time and number of gaze points, and eye movement data is collected using a lightweight wearable eye tracker.

[0118] 31) Gaze time refers to the time interval from the beginning to the end of a road user's gaze when he or she is reading the human-machine interaction interface outside the vehicle. The longer the road user's gaze time on the human-machine interaction interface outside the vehicle, the lower the efficiency of visual information search, acquisition and understanding, and the lower the perception efficiency.

[0119] 32) The number of fixation points refers to the number of fixation points that a road user places on the interface when reading the human-machine interaction interface outside the vehicle. The more fixation points a road user places on the human-machine interaction interface outside the vehicle, the lower the visual search efficiency, and the lower the perception efficiency.

[0120] 4) Decision efficiency is measured by decision time, which refers to the time interval from seeing the autonomous vehicle to making a traffic behavior decision. The shorter the decision time, the higher the decision efficiency. High-definition cameras, timers and other equipment are used for data collection;

[0121] 5) Emotional load refers to the emotional tension of road users when reading the human-machine interaction interface outside the vehicle, which is comprehensively measured by blinking frequency, breathing rate, real-time heart rate and skin conductance level;

[0122] 51) There is a linear correlation between blinking frequency and emotional load. When a road user reads the human-machine interaction interface outside the vehicle, the higher the blinking frequency, the higher the emotional load. The wearable eye tracker is used to collect data. The blinking frequency calculation formula is as follows:

[0123]

[0124] Among them, BF represents the blinking frequency during the interaction, EBF represents the total number of blinks during the interaction time, and t represents the interaction time;

[0125] 52) When performing complex tasks or being under stress, there will be strong breathing or over-breathing. When road users read the human-machine interaction interface outside the vehicle, the higher the breathing rate, the higher the emotional load. The breathing patch sensor is used to collect data. The breathing frequency calculation formula is as follows:

[0126]

[0127] Among them, RR represents the respiratory rate during the interaction, ERR represents the total number of breaths during the interaction time, and t represents the interaction time;

[0128] 53) Real-time heart rate. The higher the real-time heart rate, the greater the emotional load. When the road user reads the human-machine interaction interface outside the vehicle, the higher the real-time heart rate average, the higher the emotional load. A multi-channel physiological instrument is used for data collection;

[0129] 54) The strength of skin conductance is positively correlated with the amount of sweat gland secretion. The secretion of sweat glands is affected by two main factors: temperature and physiological and psychological activities. In particular, physiological and psychological reactive sweat secretion is affected by the sympathetic nerves. When the nerve activity is enhanced, the amount of sweat gland secretion increases and the skin conductance level increases. A multi-conductor instrument is used for data collection. When road users read the human-machine interaction interface outside the vehicle, the higher the skin conductance level, the higher the emotional load.

[0130] 6) Cognitive load refers to the cognitive pressure of road users in the process of reading the human-machine interaction interface outside the car, which is comprehensively measured by EEG power spectrum ratio, prefrontal rSO2, gaze entropy and subjective cognitive load;

[0131] 61) The EEG power spectrum ratio can evaluate the cognitive load of road users in the process of reading the human-machine interaction interface outside the vehicle. It is mainly evaluated by the power spectrum ratio of θ / α in the prefrontal region of the scalp. The θ wave refers to the band with a frequency of 4-7Hz in the EEG signal, and the α wave refers to the band with a frequency of 8-13Hz in the EEG signal. By performing spectral analysis on the EEG signal, the average power of the θ wave and the α wave under different tasks and benchmarks is obtained. The higher the power spectrum ratio of θ / α, the greater the cognitive load generated by the test task. The EEG data is collected using an electroencephalometer. The calculation formula of the EEG power spectrum ratio is as follows:

[0132]

[0133] Among them, EER represents the EEG power spectrum ratio, Pθ represents the power spectrum value of theta wave, and Pα represents the power spectrum value of alpha wave;

[0134] 62) Prefrontal rSO2 is the local blood oxygen saturation of the prefrontal lobe, which has a high correlation with the blood oxygen saturation of the jugular bulb and can be used to quantify and analyze cognitive load. When road users read the human-machine interaction interface outside the vehicle, the greater the prefrontal rSO2 value, the greater the cognitive load. Functional near-infrared spectroscopy technology is used to collect data;

[0135] 63) Gaze entropy represents the uncertainty of gaze position within a given viewing time. Compared with other eye movement indicators, the measurement value of gaze entropy has the advantage of not being affected by external factors. Data is collected through a wearable eye tracker. When road users read the human-computer interaction interface outside the car, the greater the gaze entropy, the greater the cognitive load. The calculation formula of gaze entropy is as follows:

[0136]

[0137] Where E is the gaze entropy; v is the number of gaze areas; Pi is the probability that the subject gazes at area u, u = 1, 2, ·…·v;

[0138] 7) Subjective cognitive load The NASA-TLX scale was used to score subjective cognitive load, which consists of 6 indicators: mental demand, physical demand, time limit, self-performance, effort, and frustration.

[0139] 8) Experience index refers to the subjective experience evaluation of road users on the human-computer interaction interface outside the vehicle. The experience index is characterized by four dimensions: aesthetics, function, interaction, and trust. The aesthetic experience scale is used to measure the aesthetic experience index, the functional experience scale is used to measure the functional experience index, the interactive experience scale is used to measure the interactive experience index, and the trust scale is used to measure the trust index.

[0140] S3: Evaluation calculation through rating index measurement:

[0141] Q1: Indicator weight

[0142] The weights of the first-level indicators are:

[0143]

[0144] where w i represents the weight of the i-th first-level indicator, and n represents the number of first-level indicators;

[0145] The weights of the secondary indicators are:

[0146]

[0147] where w ij represents the weight of the jth secondary indicator under the ith first-level indicator, and m represents the number of secondary indicators under the ith first-level indicator;

[0148] The weights of the three-level indicators are:

[0149]

[0150] where w ijk represents the weight of the kth third-level indicator under the jth second-level indicator under the i-th first-level indicator, and q represents the number of third-level indicators under the jth second-level indicator under the i-th first-level indicator;

[0151] The relative importance of each indicator is scored using the priority relationship quantitative scaling method, with a quantitative scale of 0.1 to 0.9, that is, comparing two indicators:

[0152] When the scale value is 0.5, it indicates that the two indicators are equally important;

[0153] When the scale value is 0.6, it indicates that indicator 1 is slightly more important than indicator 2; if the value is 0.4, the opposite is true;

[0154] When the scale value is 0.7, it indicates that indicator 1 is significantly more important than indicator 2; if the value is 0.3, the opposite is true;

[0155] When the scale value is 0.8, it means that indicator 1 is much more important than indicator 2; if the value is 0.2, the opposite is true;

[0156] When the scale value is 0.9, it means that indicator 1 is extremely important than indicator 2; if the value is 0.1, the opposite is true;

[0157] Taking the calculation of the weight of the first-level indicator as an example, the expert scoring method is used to assign importance to the first-level indicators of the same level: effectiveness K1, efficiency K2 and satisfaction K3. aij represents Ki and Kj, where i, j = 1, 2, 3;

[0158] Construct a fuzzy complementary judgment matrix, the fuzzy complementary judgment matrix M is:

[0159]

[0160] Perform consistency check on the judgment matrix M, calculate the maximum characteristic root of the matrix, and calculate its consistency index:

[0161]

[0162] Query the average random consistency index RI (n=3, take RI=0.58), and calculate the consistency ratio CR:

[0163]

[0164] Judging consistency: When CR < 0.1, the consistency of the matrix can be considered acceptable;

[0165] The square root method is used to calculate the eigenvector W'. The row product of the judgment matrix is ​​calculated and then the nth square root is taken, as shown in the following formula:

[0166] W′=[w′1,w′2,…w′ i ],i=1,2,…n

[0167] in

[0168] Further obtain the weight value of the first-level indicator:

[0169]

[0170] Then we can get the weight value of the first-level index. Similarly, we can get the weights of the second-level index and the third-level index.

[0171] Q2: Raw data processing

[0172] The processing of raw data includes two aspects: subjective data processing and objective data processing. Subjective data is obtained by taking the average of the scale data of multiple subjects, and the data range is 1 to 5 points. The objective data uses the function mapping relationship f(x) to convert the raw data into a 5-point score consistent with the subjective scale. According to the large sample experimental data, the evaluation index data obeys the normal distribution, that is, X~N(μ,σ), so as to determine the function expression:

[0173]

[0174] Among them, x represents the actual value of objective data; μ and σ represent the mean and standard deviation of this type of objective data respectively;

[0175] Q3: Comprehensive evaluation

[0176]

[0177] Among them, S represents the comprehensive score of the human-machine interaction interface outside the autonomous driving vehicle; S ijk It indicates the 5-point score corresponding to the three-level indicators;

[0178] A method for extracting a human-computer interaction scene outside an autonomous driving vehicle, the extraction process comprising:

[0179] The first level: includes the initial set of dynamic behaviors of autonomous driving vehicles and the initial set of interaction modes, which are filtered by filter A. The filtering criteria are to eliminate the dynamic behaviors of vehicles that can be represented by turn signals, reverse lights, and brake lights in traditional ways;

[0180] Filtering is performed through filters B and C. The criterion of filter B is to eliminate those interaction modes that do not cause the risk of traffic conflict;

[0181] The criterion of filter C is to eliminate those interaction modes with similar or insignificant conflict characteristics; the merging interactions of road users from the left and right sides of the autonomous vehicle can be combined into one mode of side merging interactions;

[0182] The second level: includes the extracted combination of dynamic behaviors and interaction patterns of autonomous driving vehicles and the initial set of interaction scenario factors; among them, the extracted combination of dynamic behaviors and interaction patterns of autonomous driving vehicles is the combination after screening at the first level;

[0183] The initial set of interaction scenario factors is screened by filter D. The screening criterion is to eliminate factors that have similar or no significant differences in the requirements for human-machine interaction outside the vehicle;

[0184] The third level: the extracted combination of interaction scenario factors, autonomous driving vehicle dynamic behaviors, and interaction modes, which is the combination after the second level screening, is screened by filter E. The screening criterion is to eliminate interaction scenarios that do not conform to reality.

[0185] At the fourth level, through screening by filter A, filter B, filter C, filter D, and filter E, n interaction scenarios between autonomous driving vehicles and road users with significant feature differences are extracted, and actual tests and evaluations are carried out in different interaction scenarios.

[0186] In the first level, the dynamic behavior of the autonomous driving vehicle refers to all possible dynamic behaviors performed by the car, including driving direction: forward / reverse, turning, changing lanes, and driving speed: constant speed, acceleration, deceleration, thereby proposing an initial set of dynamic behaviors of the autonomous driving vehicle.

[0187] In the first level, the interaction mode refers to all possible interaction modes when the autonomous vehicle and the road user meet at the same location. Two factors that affect the interaction mode are considered: the relative position relationship between the road user and the autonomous vehicle and the moving direction, and thus an initial set of interaction modes is proposed;

[0188] In the second level, interaction scene factors refer to factors that affect the attributes of the interaction scene, such as road type, traffic environment, right of way, initial vehicle speed, etc., and an initial set of interaction scene factors is proposed.

[0189] The present invention proposes a hierarchical screening method by defining the scene area between the autonomous driving vehicle and the cyclists and pedestrians. From the aspects of the dynamic behavior of the autonomous driving vehicle, the interaction mode, the interaction scene factors, etc., multiple interaction scenes with significant feature differences are extracted hierarchically according to the screening criteria.

[0190] Regarding the usability of human-computer interaction outside the vehicle, the system usability is determined based on the three factors of effectiveness, efficiency and satisfaction. Combined with the interaction characteristics between autonomous vehicles and road users, corresponding evaluation indicators are proposed for the three factors, as well as the quantification method of each evaluation indicator; in particular, EEG data, blood oxygen data, skin electricity data, eye movement data, etc. are introduced to objectively evaluate the human-computer interaction interface outside the vehicle;

[0191] Based on the hierarchical calculation and evaluation results, the optimal vehicle-exterior human-computer interaction interface is designed, and the necessary information is transmitted to road users through the vehicle-exterior human-computer interaction interface. Based on the recognition and understanding of the interface information, traffic behavior decisions are made efficiently and accurately to ensure interactive safety.

[0192] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for evaluating the human-machine interaction interface outside an autonomous driving vehicle, characterized in that: The evaluation steps include: S1: Construction of evaluation index system; Set three first-level evaluation indicators, including effectiveness, efficiency and satisfaction; Effectiveness includes two secondary indicators: information cognition and behavioral decision-making; Among them, information cognition is measured by a three-level subjective indicator, information cognition correctness; Efficiency includes two secondary indicators: perceived efficiency and decision-making efficiency; Among them, perception efficiency is comprehensively measured by two three-level objective indicators, fixation time and number of fixations, and decision efficiency is measured by one three-level objective indicator, decision time; Satisfaction includes three secondary indicators: emotional load, cognitive load and experience index. Emotional load is comprehensively measured by four third-level objective indicators: blinking frequency, breathing rate, real-time heart rate and skin conductance level. Cognitive load is comprehensively measured by three three-level objective indicators, EEG power spectrum ratio, prefrontal rSO2, gaze entropy, and one three-level subjective indicator, subjective cognitive load; The experience index is comprehensively measured by four three-level subjective indicators: aesthetic experience, functional experience, interactive experience, and trust experience; S2: Measure rating indicators based on evaluation indicators 1) Information cognition is measured by information cognition correctness. Information cognition correctness refers to whether the road user correctly understands the behavior intention of the autonomous vehicle represented by the human-machine interaction interface outside the vehicle. It is measured through open-ended questions after the experiment. Correct understanding is assigned 5 points and incorrect understanding is assigned 1 point. 2) Behavioral decision-making is measured by behavioral decision correctness. Behavioral decision correctness refers to whether the traffic behavior decision made by the road user after reading the human-machine interaction interface outside the vehicle is correct. The correctness of the behavioral decision of the road user is observed during the experiment. A correct decision is assigned 5 points, and an incorrect decision is assigned 1 point. 3) Perception efficiency refers to the efficiency of road users from discovering the autonomous vehicle to perceiving the vehicle's behavioral intention. It is comprehensively measured by gaze time and number of gaze points, and eye movement data is collected using a lightweight wearable eye tracker. 31) Gaze time refers to the time interval from the beginning to the end of a road user's gaze when reading the human-machine interaction interface outside the vehicle. The longer the road user's gaze time on the human-machine interaction interface outside the vehicle, the lower the efficiency of visual information search, information acquisition, and information understanding, and the lower the perception efficiency; 32) The number of fixation points refers to the number of fixation points that a road user places on the interface when reading the human-machine interaction interface outside the vehicle. The more fixation points a road user places on the human-machine interaction interface outside the vehicle, the lower the visual search efficiency, and the lower the perception efficiency. 4) Decision efficiency is measured by decision time, which refers to the time interval from seeing the autonomous vehicle to making a traffic behavior decision. The shorter the decision time, the higher the decision efficiency. High-definition cameras, timers and other equipment are used for data collection; 5) Emotional load refers to the emotional tension of road users when reading the human-machine interaction interface outside the vehicle, which is comprehensively measured by blinking frequency, breathing rate, real-time heart rate and skin conductance level; 51) There is a linear correlation between blinking frequency and emotional load. When a road user reads the human-machine interaction interface outside the vehicle, the higher the blinking frequency, the higher the emotional load. The wearable eye tracker is used to collect data. The blinking frequency calculation formula is as follows: Among them, BF represents the blinking frequency during the interaction, EBF represents the total number of blinks during the interaction time, and t represents the interaction time; 52) When performing complex tasks or being under stress, there will be strong breathing or over-breathing. When road users read the human-machine interaction interface outside the vehicle, the higher the breathing rate, the higher the emotional load. The breathing patch sensor is used to collect data. The breathing frequency calculation formula is as follows: Among them, RR represents the respiratory rate during the interaction, ERR represents the total number of breaths during the interaction time, and t represents the interaction time; 53) Real-time heart rate. The higher the real-time heart rate, the greater the emotional load. When the road user reads the human-machine interaction interface outside the vehicle, the higher the real-time heart rate average, the higher the emotional load. A multi-channel physiological instrument is used for data collection; 54) The strength of skin conductance is positively correlated with the amount of sweat gland secretion. The secretion of sweat glands is affected by two main factors: temperature and physiological and psychological activities. In particular, physiological and psychological reactive sweat secretion is affected by the sympathetic nerves. When the nerve activity is enhanced, the amount of sweat gland secretion increases and the skin conductance level increases. A multi-conductor instrument is used for data collection. When road users read the human-machine interaction interface outside the vehicle, the higher the skin conductance level, the higher the emotional load. 6) Cognitive load refers to the cognitive pressure of road users in the process of reading the human-machine interaction interface outside the car, which is comprehensively measured by EEG power spectrum ratio, prefrontal rSO2, gaze entropy and subjective cognitive load; 61) The EEG power spectrum ratio can evaluate the cognitive load of road users in the process of reading the human-machine interaction interface outside the vehicle. It is mainly evaluated by the power spectrum ratio of θ / α in the prefrontal region of the scalp. The θ wave refers to the band with a frequency of 4-7Hz in the EEG signal, and the α wave refers to the band with a frequency of 8-13Hz in the EEG signal. By performing spectral analysis on the EEG signal, the average power of the θ wave and the α wave under different tasks and benchmarks is obtained. The higher the power spectrum ratio of θ / α, the greater the cognitive load generated by the test task. The EEG data is collected using an electroencephalometer. The calculation formula of the EEG power spectrum ratio is as follows: Among them, EER represents the EEG power spectrum ratio, P θ Represents the power spectrum value of theta wave, P α Indicates the power spectrum value of α wave; 62) Prefrontal rSO2 is the local blood oxygen saturation of the prefrontal lobe, which has a high correlation with the blood oxygen saturation of the jugular bulb and can be used to quantify and analyze cognitive load. When road users read the human-machine interaction interface outside the vehicle, the greater the prefrontal rSO2 value, the greater the cognitive load. Functional near-infrared spectroscopy technology is used to collect data; 63) Gaze entropy represents the uncertainty of gaze position within a given viewing time. Compared with other eye movement indicators, the measurement value of gaze entropy has the advantage of not being affected by external factors. Data is collected through a wearable eye tracker. When road users read the human-computer interaction interface outside the car, the greater the gaze entropy, the greater the cognitive load. The calculation formula of gaze entropy is as follows: Where E is the gaze entropy; v is the number of gaze areas; P u is the probability that the subject fixates on area u, u=1,2,···,v; 7) Subjective cognitive load: The NASA-TLX scale was used to score subjective cognitive load, which includes 6 indicators: mental demand, physical demand, time limit, self-performance, effort, and frustration; 8) Experience index refers to the subjective experience evaluation of road users on the human-machine interaction interface outside the vehicle. The experience index is characterized by four dimensions: aesthetics, function, interaction, and trust. The aesthetic experience scale is used to measure the aesthetic experience index, the functional experience scale is used to measure the functional experience index, the interactive experience scale is used to measure the interactive experience index, and the trust scale is used to measure the trust index; S3: Evaluation calculation through rating index measurement: Q1: Indicator weight The weights of the first-level indicators are: where w i represents the weight of the i-th first-level indicator, and n represents the number of first-level indicators; The weights of the secondary indicators are: where w ij represents the weight of the jth secondary indicator under the ith first-level indicator, and m represents the number of secondary indicators under the ith first-level indicator; The weights of the three-level indicators are: where w ijk represents the weight of the kth third-level indicator under the jth second-level indicator under the i-th first-level indicator, and q represents the number of third-level indicators under the jth second-level indicator under the i-th first-level indicator; The relative importance of each indicator is scored using the priority relationship quantitative scaling method, with a quantitative scale of 0.1 to 0.9, that is, comparing two indicators: When the scale value is 0.5, it indicates that the two indicators are equally important; When the scale value is 0.6, it indicates that indicator 1 is slightly more important than indicator 2; if the value is 0.4, the opposite is true; When the scale value is 0.7, it indicates that indicator 1 is significantly more important than indicator 2; if the value is 0.3, the opposite is true; When the scale value is 0.8, it means that indicator 1 is much more important than indicator 2; if the value is 0.2, the opposite is true; When the scale value is 0.9, it means that indicator 1 is extremely important than indicator 2; if the value is 0.1, the opposite is true; Taking the calculation of the weight of the first-level indicator as an example, the expert scoring method is used to assign importance to the first-level indicators of the same level: effectiveness K1, efficiency K2 and satisfaction K3. ij Represents Ki and Kj, where i, j = 1, 2, 3; Construct a fuzzy complementary judgment matrix, the fuzzy complementary judgment matrix M is: Perform consistency check on the judgment matrix M, calculate the maximum characteristic root of the matrix, and calculate its consistency index: Query the average random consistency index RI (n=3, take RI=0.58), and calculate the consistency ratio CR: Judging consistency: When CR < 0.1, the consistency of the matrix can be considered acceptable; The square root method is used to calculate the eigenvector W'. The row product of the judgment matrix is ​​calculated and then the nth square root is taken, as shown in the following formula: W′=[w′1,w2,…w′ i ],i=1,2,…n in Further obtain the weight value of the first-level indicator: Then we can get the weight value of the first-level index. Similarly, we can get the weights of the second-level index and the third-level index. Q2: Raw data processing The processing of raw data includes two aspects: subjective data processing and objective data processing. Subjective data is obtained by taking the average of the scale data of multiple subjects, and the data range is 1 to 5 points. The objective data uses the function mapping relationship f(x) to convert the raw data into a 5-point score consistent with the subjective scale. According to the large sample experimental data, the evaluation index data obeys the normal distribution, that is, X~N(μ,σ), so as to determine the function expression: Among them, x represents the actual value of objective data; μ and σ represent the mean and standard deviation of this type of objective data respectively; Q3: Comprehensive evaluation Among them, S represents the comprehensive score of the human-machine interaction interface outside the autonomous driving vehicle; S ijk Indicates the 5-point score corresponding to the third-level indicators.

2. The method for evaluating the human-machine interaction interface outside the autonomous driving vehicle according to claim 1, characterized in that: In S1, effectiveness means that when a road user interacts with an autonomous vehicle, the human-computer interaction interface outside the vehicle can guide the user to make correct information understanding and traffic behavior decisions. It includes two secondary indicators: information cognition and behavior decision. Information cognition is measured by a third-level subjective indicator, information cognition correctness. If the information cognition index is larger, the effectiveness index is larger. Behavior decision is measured by a third-level subjective indicator, behavior decision correctness. If the behavior decision index is larger, the effectiveness index is larger.

3. The method for evaluating the human-machine interaction interface outside the autonomous driving vehicle according to claim 1, characterized in that: In S1, efficiency means that when a road user interacts with an autonomous vehicle, the human-computer interaction interface outside the vehicle can guide the road user to make correct information understanding and traffic behavior decisions more timely and efficiently. It includes perception efficiency and decision efficiency. The perception efficiency is comprehensively measured by two three-level objective indicators, gaze time and number of gaze points. If the perception efficiency index is larger, the efficiency index is larger. The decision efficiency is measured by a three-level objective indicator, decision time. If the decision efficiency index is larger, the efficiency index is larger.

4. The method for evaluating the human-machine interaction interface outside the autonomous driving vehicle according to claim 1, characterized in that: In S1, satisfaction refers to the fact that when a road user interacts with an autonomous vehicle, under the guidance of human-computer interaction outside the vehicle, the psychological and cognitive pressure of the road user when interacting with the autonomous vehicle can be alleviated, and the acceptability and comfort of the interaction process can be improved. It includes emotional load, cognitive load and experience index. The emotional load is comprehensively measured by four three-level objective indicators: blinking frequency, breathing rate, real-time heart rate and skin conductance level. If the emotional load index is larger, the satisfaction is smaller. The cognitive load is comprehensively measured by three three-level objective indicators: EEG power spectrum ratio, prefrontal rSO2, gaze entropy and one three-level subjective indicator: subjective cognitive load. If the cognitive load index is larger, the satisfaction is smaller. The experience index is comprehensively measured by four three-level subjective indicators: aesthetic experience, functional experience, interactive experience and trust experience. If the experience index index is larger, the satisfaction is greater.

Images