Method for evaluating driving ability of aided driving automobile based on multi-dimensional behavior analysis

Abstract

The invention relates to a multi-dimensional behavior analysis-based driving ability evaluation method for an auxiliary driving vehicle. The method comprises the following steps of: obtaining historical driving data of a target vehicle in a specified time period; preprocessing the historical driving data, and dividing data fragments for activating an auxiliary driving function; calculating a preset evaluation index based on the data fragment; wherein the preset evaluation indexes comprise an auxiliary driving function abuse rate, an emergency response delay index, a man-machine conflict frequency and a driving concentration index; and according to a calculation result of the preset evaluation index, generating a comprehensive ability evaluation report of the driver. The method is completely based on conventional data sources such as the existing CAN bus of the vehicle and log data of the auxiliary driving system, does not need any additional hardware, comprehensively evaluates the capability performance of the driver when the driver uses the auxiliary driving function from multiple dimensions, and is more comprehensive, low in calculation complexity, high in evaluation efficiency and easy to deploy and apply on a large scale.

Description

Technical Field

[0001] This invention relates to the fields of intelligent connected vehicle technology and traffic safety technology, and in particular to a method for evaluating the driving ability of assisted driving vehicles based on multi-dimensional behavior analysis. Background Technology

[0002] With the widespread adoption of Advanced Driver Assistance Systems (ADAS) and intelligent vehicles, the impact of driver behavior on driving safety, traffic efficiency, and overall road traffic safety is becoming increasingly prominent. When driving an assisted vehicle, the addition of the driver assistance system alters the driver's tasks to some extent, affecting their physiological and psychological state and changing the demands on perception, decision-making, and operation during driving. Accident analyses involving assisted vehicles have revealed that driver factors are a key contributor to accidents. Therefore, assessing driver capabilities in assisted vehicles is crucial for improving operational safety, preventing road accidents, and training and assessing assisted vehicle drivers. However, existing technologies primarily focus on evaluating vehicle or system performance, lacking quantitative analysis methods for assessing individual driver capabilities, particularly a lack of an effective, convenient, and low-cost tool to evaluate a driver's true capabilities in a "human-machine co-driving" mode. This capability includes not only traditional driving skills but also the ability to interact with automated systems, monitor system operation, and take over when necessary.

[0003] Existing evaluation methods often suffer from the following shortcomings:

[0004] 1. Reliance on specific hardware: Many advanced evaluations need to be conducted on simulators or on actual vehicles equipped with specialized equipment such as eye trackers and bioelectric sensors, which is costly and difficult to apply on a large scale.

[0005] 2. High real-time requirements: Some methods focus on real-time monitoring and alarms, which are computationally complex and require high performance from the vehicle-mounted computing platform.

[0006] 3. Limited evaluation dimensions: Most assessments focus only on single indicators such as takeover time, failing to comprehensively reflect a driver's overall capabilities, such as understanding of the system, trust in the system, and tendency to abuse it.

[0007] 4. Privacy Controversy: The continuous monitoring by in-car cameras can easily raise user privacy concerns.

[0008] Therefore, there is an urgent need for an evaluation method that does not rely on specific hardware, has low computational complexity, and can utilize the vehicle's existing and widely recorded driving data to conduct in-depth offline post-event analysis and retrospective assessment of driver capabilities from multiple dimensions based on historical performance. This method would provide support for safety education, insurance customization, function permission settings, driver management of assisted driving vehicles, and even human factor analysis of accidents involving assisted driving vehicles. Summary of the Invention

[0009] The purpose of this invention is to propose a method for evaluating the driving ability of assisted driving vehicles based on multi-dimensional behavior analysis, in order to solve the problems existing in the prior art. This method is based entirely on conventional data sources such as the vehicle's existing CAN bus and assisted driving system log data, without requiring any additional hardware. It comprehensively evaluates the driver's performance when using assisted driving functions from multiple dimensions, making it more comprehensive, with low computational complexity, high evaluation efficiency, and easy to deploy and apply on a large scale.

[0010] To achieve the above objectives, the present invention provides the following solution:

[0011] A method for assessing the driving ability of assisted driving vehicles based on multidimensional behavior analysis, comprising:

[0012] Acquire historical driving data of the target vehicle within a specified time period; wherein, the historical driving data includes: historical CAN data and ADAS log data;

[0013] The historical driving data is preprocessed, and data segments indicating the activation of driver assistance functions are identified.

[0014] Based on the data fragment, preset evaluation indicators are calculated; wherein, the preset evaluation indicators include: abuse rate of driver assistance functions, emergency response delay index, human-machine conflict frequency, and driving attention index;

[0015] Based on the calculation results of the preset evaluation indicators, a comprehensive assessment report of the driver's capabilities is generated.

[0016] Optionally, the historical driving data includes: timestamp, vehicle speed, acceleration, steering wheel angle, turn signal status, ADAS status, ADAS mode, TJA / ICA activation status, accelerator pedal opening, brake pedal status, brake pressure, door status, seat belt status, forward distance warning information, and FCW / AEB warning records.

[0017] Optionally, the method for segmenting the data fragments in which the driver assistance function is activated includes:

[0018] Data slices are performed with each activation cycle of the driver assistance function as an analysis unit.

[0019] Optionally, the abuse rate of the driver assistance function is used to characterize the proportion of drivers who forcibly activate or use the driver assistance function in scenarios where it is clearly inappropriate to use it.

[0020] The criteria for determining the abuse rate of the driver assistance function include: activation under severe weather conditions, activation under complex road conditions, or activation under system restrictions.

[0021] The method for calculating the abuse rate of the aforementioned driver assistance function includes:

[0022] AAAR=(N_abuse / N_total)*100%;

[0023] Wherein, N_total is the total number of times the driver assistance function is activated within the statistical period, N_abuse is the number of times the activation is judged as abuse within the statistical period, and AAAR is the abuse rate of the driver assistance function.

[0024] Optionally, the emergency response delay index is used to measure the average delay time between the system issuing a takeover request or the occurrence of an emergency and the driver beginning effective operation;

[0025] The calculation method for the emergency response delay index includes:

[0026] Identify emergency event points; wherein, the emergency event points include at least one of the following: system takeover request, automatic emergency braking activation, or forward collision warning activation;

[0027] Determine the response start point; wherein, the response start point is the moment when the driver first makes a valid response operation;

[0028] Calculate the delay of a single event based on the emergency event point and the response start point;

[0029] The average value of the delays of all valid events within the statistical period is taken as the emergency response delay index.

[0030] Optionally, making an effective response operation includes: applying at least one of the following: applying braking greater than a preset braking threshold, applying steering greater than a preset steering threshold, or applying throttle greater than a preset throttle threshold.

[0031] Optionally, the human-machine conflict frequency is used to measure the frequency of inconsistencies between the driver's and the system's intentions to control the vehicle during the activation of the driver assistance function;

[0032] The criteria for determining the frequency of human-machine conflict include at least one of the following: steering conflict, acceleration conflict, or deceleration conflict.

[0033] The method for calculating the human-machine conflict frequency is as follows:

[0034] HCF = N_conflict / T_total;

[0035] Wherein, HCF is the human-machine conflict frequency, N_conflict is the number of conflicts that occur within the statistical period, and T_total is the total working time of the driver assistance system within the statistical period.

[0036] Optionally, the driving focus is used to measure the driver's sustained attention to the driving task during stable operation of the driver assistance system;

[0037] The method for calculating driving focus is as follows:

[0038] DAI=1-(T_distracted / T_total)*100%;

[0039] Where DAI is driving focus level, T_total is the total time the driver assistance system operates stably within the statistical period, and T_distracted is the cumulative time within the T_total duration during which the driver is judged to be in a distracted state.

[0040] Optionally, the method for determining the distraction state includes:

[0041] Preprocess the steering wheel torque signal;

[0042] Set up a sliding analysis window;

[0043] Calculate the standard deviation of the steering wheel torque signal within the window;

[0044] The standard deviation is compared with a preset extremely low threshold.

[0045] If the standard deviation is consistently lower than the preset extremely low threshold, it is determined that the driver has not effectively contacted the steering wheel during the corresponding time period, and this time period is accumulated into the distraction state.

[0046] The beneficial effects of this invention are as follows:

[0047] 1. Zero hardware dependency: It fully utilizes existing vehicle data without adding any hardware costs, making it easy to deploy and apply on a large scale.

[0048] 2. High computational efficiency: All indicator calculations are based on simple threshold judgment, time accumulation, statistical averaging and other operations. The algorithm complexity is low and can quickly process massive amounts of historical data on ordinary servers. It can be used to retrospectively analyze driving performance at any time in the past.

[0049] 3. Comprehensive assessment: The assessment comprehensively depicts driver capabilities from four dimensions: abuse, response, conflict, and focus, resulting in a more scientific and multi-dimensional evaluation.

[0050] 4. Privacy Protection: The system analyzes vehicle control data, not the driver's facial or biometric features, fundamentally avoiding the risk of privacy breaches.

[0051] 5. Wide range of applications: It can be used in vehicle insurance UBI models, fleet safety management, personalized driver settings (adapted to their ability level), vehicle manufacturer assessment of the rationality of function design and after-sales service, targeted customization of driver assistance vehicle ability assessment and training education, and driver factor analysis in road traffic accidents, etc. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 This is a schematic diagram of a method for evaluating the driving ability of an assisted driving vehicle based on multidimensional behavior analysis, according to an embodiment of the present invention. Detailed Implementation

[0054] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0056] like Figure 1 As shown in the figure, this embodiment proposes a method for evaluating the driving ability of assisted driving vehicles based on multidimensional behavior analysis, including:

[0057] Acquire historical driving data of the target vehicle within a specified time period; wherein, the historical driving data includes: historical CAN data and ADAS log data;

[0058] The historical driving data is preprocessed, and data segments indicating the activation of driver assistance functions are identified.

[0059] Based on the data fragment, preset evaluation indicators are calculated; wherein, the preset evaluation indicators include: abuse rate of driver assistance functions, emergency response delay index, human-machine conflict frequency, and driving attention index;

[0060] Based on the calculation results of the preset evaluation indicators, a comprehensive assessment report of the driver's capabilities is generated.

[0061] Specifically, in this embodiment, data acquisition involves exporting historical CAN data and ADAS log data of a specified vehicle over a certain period of time from a vehicle telematics terminal (T-Box) or data center.

[0062] Data preprocessing: Clean the data, align the timestamps, and parse out the required signals (such as ACC status, steering wheel torque, brake pedal opening, alarm signals, etc.).

[0063] More specifically, firstly, historical driving data for a specified period is obtained from the vehicle data recording platform, including but not limited to: timestamps, vehicle speed, acceleration, steering wheel angle, turn signal status, ADAS status (whether activated), ADAS mode, TJA / ICA activation status, accelerator pedal opening, brake pedal status, brake pressure, door status, seat belt status, forward distance warning information, FCW / AEB warning records, etc.

[0064] The data was cleaned and aligned, and then sliced ​​into data segments with each "activation cycle of driver assistance function" as a basic unit of analysis.

[0065] Specifically, in this embodiment, the calculation of the four preset evaluation indicators includes:

[0066] 1. Driver Assistance System Abuse Rate (AAAR):

[0067] (1) Definition: The proportion of drivers who forcibly activate or use driver assistance functions in situations where it is clearly inappropriate to use them.

[0068] (2) Calculation method: AAAR = (N_abuse / N_total) * 100%;

[0069] N_total: The total number of times the driver assistance function was activated within the statistical period; N_abuse: The number of times the function was deemed "abused" within the statistical period.

[0070] (3) Criteria for determining “abuse” (meeting any one of these criteria constitutes one instance of abuse):

[0071] Activation in severe weather: The driver assistance system is activated when the windshield wipers are in a medium to high speed setting (e.g., for more than 30 seconds).

[0072] Activation on complex roads: Activate assisted driving on complex roads (such as sharp curves, ramps, and construction sections without clear lane markings) identified in high-precision map data or GPS data.

[0073] Activation under system restrictions: Attempt to activate when the vehicle speed is below or above the permitted operating range of the driver assistance system (e.g., 0 km / h or >140 km / h).

[0074] (4) Basis for parameter determination: The above judgment conditions are all based on the inapplicable scenarios clearly indicated in the user manuals and safety warnings of mainstream driver assistance systems. This method transforms safety specifications into quantifiable data judgment logic without the need for additional hardware.

[0075] 2. Emergency Response Delay Index (ERDI):

[0076] (1) Definition: Measures the average delay time between the system issuing a takeover request or an emergency event and the driver initiating effective action (such as applying torque or applying the brakes).

[0077] (2) Calculation method:

[0078] First, identify the "emergency event point" (T_event). Event sources include:

[0079] System takeover request: The data shows clear signals such as "hands on the steering wheel" warning (steering_wheel_warning) or "takeover request" (please take over immediately).

[0080] Automatic Emergency Braking (AEB) Activation: The moment when the AEB system suddenly intervenes.

[0081] Forward Collision Warning (FCW) Activation: The moment when the FCW system issues a strong audible warning.

[0082] Then determine the "response start point" (T_response). This is defined as the moment when the driver first makes a valid response, and must include at least one of the following:

[0083] Apply significant braking (brake_pressure > threshold K_brake).

[0084] Apply a significant steering (steering_angle_change_rate > threshold K_steer).

[0085] I decisively stepped on the gas (throttle_pedal_open>threshold K_throttle).

[0086] Calculate the delay for a single event: Delay_i = T_response - T_event.

[0087] The average delay of all valid events within the statistical period is taken as ERDI: ERDI = Average(Delay_i).

[0088] (3) Parameter determination basis: The thresholds K_brake, K_steer, and K_throttle can be calibrated according to the vehicle model and data characteristics. They are usually set to 2-3 standard deviations beyond the normal cruise operation range. The system sensitivity and false triggering are also taken into account to filter out unintentional minor operations.

[0089] 3. Human-Machine Conflict Frequency (HCF):

[0090] (1) Definition: Measures the frequency of inconsistency between the driver's and the system's intention to control the vehicle during the activation of the driver assistance function, reflecting the mismatch in the understanding of vehicle control rights between the two parties and the smoothness of their cooperation.

[0091] (2) Calculation method: HCF = N_conflict / T_total

[0092] N_conflict (Number of Conflicts): The number of "conflicts" that occurred within the statistical period.

[0093] T_total: Total operating time (hours) of the driver assistance system during the statistical period.

[0094] 3) Conflict event determination: (A conflict is counted if any one of the following conditions is met):

[0095] Steering conflict: While the system is performing Lane Keeping Assist Steering (LKA), the driver applies a counteracting, continuous steering torque (steering_torque > threshold K_torque, and lasts > 0.5 seconds).

[0096] Acceleration / Deceleration Conflict: When adaptive cruise control (ACC) is decelerating while following another vehicle, the driver may instead press the accelerator; or when ACC is accelerating, the driver may instead press the brake. (ACC_decelerating AND throttle_pedal_open>K_throttle) OR (ACC_accelerating AND brake_pressure>K_brake).

[0097] The threshold is determined through power steering system characteristics and ergonomic experiments on the vehicle platform. It is obtained by learning from a large amount of smooth driving data to ensure that it can capture obviously intentional, anti-system operations and ignore unintentional shaking or touching.

[0098] 4. Driving Attention Index (DAI):

[0099] (1) Definition: Measures the driver's sustained attention to the driving task during the stable operation of the driver assistance system. This indicator indirectly determines whether the driver is in a "hands-off" or "distracted" state by analyzing the driver's micro-management behavior of the steering wheel.

[0100] (2) Calculation method: DAI = 1 - (T_distracted / T_total) * 100%

[0101] T_total: The total duration of stable operation of the driver assistance system within the statistical period. The criteria for stable operation are: system functions are activated (e.g., LKA / ACC status bit is True) and there are no system alarms (e.g., hands-on-steering warning, takeover request) or emergency events (e.g., AEB / FCW activation). These status signals can be obtained directly from the vehicle's CAN bus data.

[0102] T_distracted: The cumulative time within the duration of T_total during which the driver is considered to be in a distracted state.

[0103] (3) Scheme for determining the distracted state (T_distracted):

[0104] Preprocess the steering torque signal (e.g., filter and denoise).

[0105] Set a sliding analysis window (e.g., window length W = 60 seconds).

[0106] Calculate the standard deviation σ_window of the steering wheel torque signal within the window.

[0107] Compare σ_window with a preset very low threshold σ_low.

[0108] If σ_window remains below the threshold σ_low, it is determined that the driver has not effectively touched the steering wheel during this time period (i.e., is in a "hands-off" distracted state), and this time period is accumulated into T_distracted.

[0109] (4) Basis for parameter determination:

[0110] Core principle: A focused driver maintains "gentle supervision" of the driver assistance system, with their hands making frequent, low-amplitude micro-adjustments in sync with the system, resulting in continuous and minute fluctuations in the steering wheel torque signal. Conversely, completely releasing the hands leads to extremely low, almost zero, variance in the torque signal.

[0111] The method for determining the extremely low threshold σ_low: This threshold is obtained through statistical learning using a large amount of historical data. The specific method is as follows:

[0112] Collect a large amount of data on periods of time known as “focused” driving (e.g., smooth cruise data segments when the vehicle is in manual driving mode).

[0113] Calculate the distribution of the standard deviation of steering wheel torque for these data within the same sliding window (W = 60s).

[0114] Take the 5th percentile of this distribution as the threshold σ_low.

[0115] Rationale: This threshold represents the level of steering wheel fluctuation even during the most "smooth" and focused driving. Therefore, if the steering wheel fluctuation (standard deviation) consistently falls below this threshold over any given period, there is a very high statistical confidence that the driver is not touching the steering wheel (hands off), indicating a state of distraction.

[0116] After calculating the four indicators, normalization and scoring are performed: a scorecard is created for each indicator, mapping the calculated raw values ​​to standardized scores. Then, results are generated and visualized: a detailed evaluation report is produced, including the overall score, sub-scores, radar charts, historical trend analysis, and improvement suggestions.

[0117] The four calculated metrics (AAAR, ERDI, HCF, DAI) are compared with preset thresholds or a benchmark model built based on a large number of samples. Finally, an evaluation report is generated, which can:

[0118] A comprehensive capability score is given, which can be displayed in the form of a radar chart to intuitively reflect the strength of its "human-machine co-driving" ability, that is, the driving ability of the assisted driving vehicle.

[0119] Point out the driver's weaknesses (e.g., "The emergency response is delayed, and vigilance training needs to be strengthened");

[0120] Drivers are classified (e.g., excellent, qualified, high-risk) or profiled (e.g., aggressive, dependent, cautious).

[0121] The following example illustrates the implementation process of this embodiment, using a car manufacturer's desire to assess the capabilities of its user base when using the Intelligent Cruise Control (ICA) function:

[0122] (1) Data collection: Download all driving data segments of the target vehicle that have activated the ICA function in the past month from the cloud.

[0123] (1) Calculate AAAR: Analysis revealed that the user activated ICA multiple times during 3 heavy rains (when the wipers were running at high speed continuously), N_abuse=3, and the total number of activations N_total=58, so its AAAR=(3 / 58)*100%=5.17%.

[0124] (2) Calculation of ERDI: The system recorded two explicit "take over immediately" requests. The first time, the driver applied the brakes hard after 2.5 seconds; the second time, the driver took over after 4.1 seconds. Taking the 90th percentile, ERDI = 4.1s.

[0125] (3) Calculate HCF: During a total of 15 hours of ICA use, a total of 10 pedal operations opposite to ACC acceleration / deceleration were detected, HCF = 10 / 15 = 0.67 times / hour.

[0126] (4) Calculate DAI: In the above 15 hours, using 60-second windows, calculate the standard deviation of steering wheel torque within each window. Given that the attention threshold σ_low = 0.05 N·m, obtained through extensive manual driving data, statistically analyze the window times where the standard deviation consistently falls below 0.05 N·m, accumulating the total distraction time T_distracted = 3 hours. Calculate DAI = 1 - (3 / 15) * 100% = 80%. This value indicates that during stable system operation, the driver remains focused (hands on the steering wheel) 80% of the time and is in a hands-free, distracted state for 20% of the time.

[0127] (5) Comprehensive Assessment: Report generated: The driver exhibits some abuse of driver assistance features (AAAR slightly high), severely delayed emergency response (ERDI extremely high), normal frequency of human-machine conflict, and needs to improve focus (DAI 80%, with a 20% risk of hands-off driving). The driver is assessed as a high-risk driver. It is recommended that the driver receive education on the proper use of driver assistance features, specialized training in emergency takeover capabilities, and be specifically warned to avoid hands-off driving.

[0128] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for assessing the driving ability of assisted driving vehicles based on multidimensional behavior analysis, characterized in that, include: Acquire historical driving data of the target vehicle within a specified time period; wherein, the historical driving data includes: historical CAN data and ADAS log data; The historical driving data is preprocessed, and data segments indicating the activation of driver assistance functions are identified. Based on the data fragment, preset evaluation indicators are calculated; wherein, the preset evaluation indicators include: abuse rate of driver assistance functions, emergency response delay index, human-machine conflict frequency, and driving attention index; Based on the calculation results of the preset evaluation indicators, a comprehensive assessment report of the driver's capabilities is generated.

2. The method for evaluating the driving ability of assisted driving vehicles based on multidimensional behavior analysis according to claim 1, characterized in that, The historical driving data includes: timestamp, vehicle speed, acceleration, steering wheel angle, turn signal status, ADAS status, ADAS mode, TJA / ICA activation status, accelerator pedal opening, brake pedal status, brake pressure, door status, seat belt status, forward distance warning information, and FCW / AEB warning records.

3. The method for evaluating the driving ability of assisted driving vehicles based on multidimensional behavior analysis according to claim 1, characterized in that, The method for segmenting the data fragments in which the driver assistance function is activated includes: Data slices are performed with each activation cycle of the driver assistance function as an analysis unit.

4. The method for evaluating the driving ability of assisted driving vehicles based on multidimensional behavior analysis according to claim 1, characterized in that, The abuse rate of the driver assistance function is used to characterize the proportion of drivers who forcibly activate or use the driver assistance function in scenarios where it is clearly inappropriate to use it. The criteria for determining the abuse rate of the driver assistance function include: activation under severe weather conditions, activation under complex road conditions, or activation under system restrictions. The method for calculating the abuse rate of the aforementioned driver assistance function includes: AAAR=(N_abuse / N_total)*100%; Wherein, N_total is the total number of times the driver assistance function is activated within the statistical period, N_abuse is the number of times the activation is judged as abuse within the statistical period, and AAAR is the abuse rate of the driver assistance function.

5. The method for evaluating the driving ability of assisted driving vehicles based on multidimensional behavior analysis according to claim 1, characterized in that, The emergency response delay index is used to measure the average delay time between the system issuing a takeover request or an emergency occurring and the driver beginning effective operation. The calculation method for the emergency response delay index includes: Identify emergency event points; wherein, the emergency event points include at least one of the following: system takeover request, automatic emergency braking activation, or forward collision warning activation; Determine the response start point; wherein, the response start point is the moment when the driver first makes a valid response operation; Calculate the delay of a single event based on the emergency event point and the response start point; The average value of the delays of all valid events within the statistical period is taken as the emergency response delay index.

6. The method for evaluating the driving ability of assisted driving vehicles based on multidimensional behavior analysis according to claim 5, characterized in that, Effective response actions include: applying at least one of the following: applying braking greater than a preset braking threshold, applying steering greater than a preset steering threshold, or applying throttle greater than a preset throttle threshold.

7. The method for evaluating the driving ability of assisted driving vehicles based on multidimensional behavior analysis according to claim 1, characterized in that, The human-machine conflict frequency is used to measure the frequency of inconsistencies between the driver's and the system's intentions to control the vehicle during the activation of the driver assistance function. The criteria for determining the frequency of human-machine conflict include at least one of the following: steering conflict, acceleration conflict, or deceleration conflict. The method for calculating the human-machine conflict frequency is as follows: HCF = N_conflict / T_total; Wherein, HCF is the human-machine conflict frequency, N_conflict is the number of conflicts that occur within the statistical period, and T_total is the total working time of the driver assistance system within the statistical period.

8. The method for evaluating the driving ability of assisted driving vehicles based on multidimensional behavior analysis according to claim 1, characterized in that, The driving focus is used to measure the driver's sustained attention to the driving task during stable operation of the driver assistance system; The method for calculating driving focus is as follows: DAI=1-(T_distracted / T_total)*100%; Where DAI is driving focus level, T_total is the total time the driver assistance system operates stably within the statistical period, and T_distracted is the cumulative time within the T_total duration during which the driver is judged to be in a distracted state.

9. The method for evaluating the driving ability of assisted driving vehicles based on multidimensional behavior analysis according to claim 8, characterized in that, The method for determining the distraction state includes: Preprocess the steering wheel torque signal; Set up a sliding analysis window; Calculate the standard deviation of the steering wheel torque signal within the window; The standard deviation is compared with a preset extremely low threshold. If the standard deviation is consistently lower than the preset extremely low threshold, it is determined that the driver has not effectively contacted the steering wheel during the corresponding time period, and this time period is accumulated into the distraction state.

Images