A method and device for controlling authority security of an artificial intelligence vehicle

Abstract

This application discloses a method and device for controlling the security of permissions for an artificial intelligence vehicle, relating to the field of automotive electronics technology. The technical solution is as follows: obtaining physiological monitoring data, driving behavior data, and environmental perception data within the current cycle; calculating the physiological abnormality index, operational deviation, and environmental risk coefficient for the current cycle; inputting the physiological abnormality index, operational deviation, and environmental risk coefficient into a risk prediction model that incorporates an attention mechanism to obtain a dynamic risk value for the next cycle; correcting the dynamic risk value based on a risk time decay factor, and matching the target risk level with the corrected dynamic risk value; determining the dynamic permission adjustment strategy corresponding to the target risk level using risk-permission linkage rules; performing real-time verification of the dynamic permission adjustment strategy based on the vehicle's driving scenario, generating a list of controllable functions, and executing permission authorization operations within the next cycle. The main purpose is to achieve precise permission adaptation for vehicle functions.

Description

Technical Field

[0001] The present application relates to the field of automotive electronics technology, and in particular to a method and device for controlling authority security of an artificial intelligence vehicle. Background Art

[0002] With the rapid development of artificial intelligence and autonomous driving technologies, smart cars are gradually evolving from single-function driver assistance to a "co-driving" model of human-machine collaboration. In this scenario, dynamically balancing driver control and vehicle autonomous decision-making has become a core challenge in ensuring driving safety. This is especially true in high-risk scenarios such as sudden illness, fatigued driving, and complex road conditions, which can lead to driver error or delayed system response, potentially causing accidents.

[0003] Currently, existing technologies for vehicle access security control mostly rely on fixed access rules based on a single-dimensional data source. For example, driver status is detected solely by steering wheel grip, or lane departure is detected solely by cameras. However, while this approach can achieve partial functional takeover, it lacks collaborative analysis of physiological indicators, operational behavior, and environmental risks, making it difficult to fully quantify dynamic risks. As a result, the vehicle's access management mechanism is unable to perceive multi-dimensional risks in real time and flexibly adjust vehicle function permissions. This severely restricts the security redundancy and scenario adaptability of vehicle access control, affecting the collaborative safety and control flexibility of vehicle driving. Summary of the Invention

[0004] In view of the above problems, this application provides a method and device for authority security control of artificial intelligence vehicles. The main purpose is to predict the comprehensive risk value of future cycles through multimodal data of physiology, behavior and environment, and to achieve precise authority adaptation of vehicle functions based on dynamic risk levels, thereby improving the collaborative safety and control flexibility of vehicle driving.

[0005] To solve the above technical problems, this application proposes the following solutions:

[0006] In a first aspect, the present application provides a method for controlling the security of permissions of an artificial intelligence vehicle, the method comprising:

[0007] Obtain the user's physiological monitoring data, driving behavior data, and vehicle environmental perception data within the current cycle;

[0008] Performing spatiotemporal alignment and noise filtering on the physiological monitoring data, the driving behavior data, and the environmental perception data, and respectively calculating the physiological abnormality index, the operation deviation, and the environmental risk factor of the current cycle;

[0009] Inputting the physiological abnormality index, operation deviation, and environmental risk coefficient of the current cycle into a risk prediction model fused with an attention mechanism to obtain a dynamic risk value for the next cycle, wherein the attention mechanism is used to dynamically assign global attention weights to the physiological abnormality index, the operation deviation, and the environmental risk coefficient based on the risk evolution relationship of historical cycles;

[0010] Correcting the dynamic risk value of the next period according to a risk time decay factor, and matching the corrected dynamic risk value with a target risk level within a preset risk level, wherein the risk time decay factor is determined by combining the dynamic risk value of the historical period and the real-time status parameters of the vehicle;

[0011] Determine a dynamic permission adjustment strategy corresponding to the target risk level using risk-permission linkage rules, wherein the dynamic permission adjustment strategy includes hierarchical authorization thresholds and real-time response rules for functional permissions;

[0012] The dynamic permission adjustment strategy is verified in real time based on the driving scenario of the vehicle, a list of controllable functions that are finally authorized is generated, and the permission authorization operation is executed in the next cycle to facilitate user security control.

[0013] In a second aspect, the present application provides an AI vehicle authority security control device, which is applied to the AI vehicle authority security control method of the first aspect, and the device includes:

[0014] The first acquisition unit is used to acquire the user's physiological monitoring data, driving behavior data and vehicle's environmental perception data in the current cycle;

[0015] a first calculation unit, configured to perform spatiotemporal alignment and noise filtering on the physiological monitoring data, the driving behavior data, and the environmental perception data, and respectively calculate a physiological abnormality index, an operation deviation, and an environmental risk factor of a current period;

[0016] a prediction unit, configured to input the physiological abnormality index, operational deviation, and environmental risk coefficient of the current cycle into a risk prediction model fused with an attention mechanism to obtain a dynamic risk value for the next cycle, wherein the attention mechanism is configured to dynamically assign global attention weights to the physiological abnormality index, operational deviation, and environmental risk coefficient based on risk evolution relationships over historical cycles;

[0017] a processing unit configured to modify the dynamic risk value of the next period according to a risk time decay factor, and match the modified dynamic risk value with a target risk level within a preset risk level, wherein the risk time decay factor is determined based on the dynamic risk value of the historical period and the real-time status parameters of the vehicle;

[0018] a determination unit, configured to determine a dynamic permission adjustment strategy corresponding to the target risk level by using a risk-permission linkage rule, wherein the dynamic permission adjustment strategy includes a hierarchical authorization threshold for functional permissions and a real-time response rule;

[0019] The authorization unit is used to verify the dynamic permission adjustment strategy in real time based on the driving scenario of the vehicle, generate a final authorized list of controllable functions, and perform permission authorization operations in the next cycle to facilitate user security control.

[0020] In order to achieve the above-mentioned purpose, according to the third aspect of the present application, a storage medium is provided, which includes a stored program, wherein when the program is running, the device where the storage medium is located is controlled to execute the permission security control method of the artificial intelligence car of the first aspect mentioned above.

[0021] In order to achieve the above-mentioned purpose, according to the fourth aspect of the present application, a processor is provided, which is used to run a program, wherein the program executes the permission security control method of the artificial intelligence car of the first aspect mentioned above when running.

[0022] By means of the above technical solution, the present application provides an artificial intelligence vehicle authority security control method and device, which first obtains the user's physiological monitoring data, driving behavior data and vehicle environmental perception data in the current cycle, and then performs spatiotemporal alignment and noise filtering on the physiological monitoring data, driving behavior data and environmental perception data, and respectively calculates the physiological abnormality index, operation deviation and environmental risk coefficient of the current cycle, and then inputs the physiological abnormality index, operation deviation and environmental risk coefficient of the current cycle into the risk prediction model of the fusion attention mechanism to obtain the dynamic risk value of the next cycle. The attention mechanism is used to dynamically allocate the physiological abnormality index, operation deviation and environmental risk coefficient according to the risk evolution relationship of the historical cycle. And the global attention weight of the environmental hazard factor, and then correct the dynamic risk value of the next cycle according to the risk time decay factor, and the corrected dynamic risk value matches the target risk level in the preset risk level. The risk time decay factor is determined by combining the dynamic risk value of the historical period and the real-time status parameters of the vehicle. Then, the risk-authorization linkage rule is used to determine the dynamic authority adjustment strategy corresponding to the target risk level. The dynamic authority adjustment strategy includes the hierarchical authorization threshold and real-time response rules of functional authorities. Finally, the dynamic authority adjustment strategy is verified in real time based on the driving scenario of the vehicle, and the final authorized controllable function list is generated. The authority authorization operation is executed in the next cycle to facilitate user safety control.The technical solution provided by the present application breaks through the limitations of single-dimensional risk assessment by synchronously collecting physiological monitoring data, driving behavior data and environmental perception data of the current cycle and calculating the physiological abnormality index, operation deviation and environmental risk factor respectively, realizing the "human-vehicle-environment" full-factor risk coupling analysis, avoiding the misjudgment caused by isolated judgment of traditional solutions, and improving the comprehensiveness and accuracy of subsequent risk prediction. The physiological abnormality index, operation deviation and environmental risk factor of the current cycle are input into the risk prediction model that dynamically allocates the global attention weight of the physiological abnormality index, operation deviation and environmental risk factor according to the risk evolution relationship of the historical cycle, and outputs the dynamic risk value of the next cycle. It can capture the risk evolution trend, realize risk pre-perception, make the authority adjustment ahead of the risk outbreak, correct the dynamic risk value based on the risk time decay factor, and match the preset risk level after correction, taking into account the impact of risk events at any time The characteristic of attenuation over time makes the matching of risk levels more accurate. Dynamic permission adjustment strategies are generated through risk-authorization linkage rules, and the dynamic permission adjustment strategies are verified in real time based on the driving scenario of the vehicle to generate a final authorized list of controllable functions, which can ensure the compatibility between the controllable function list and the driving scenario, thereby breaking the rigid restrictions of static permission rules and realizing elastic scaling of permissions. While ensuring core safety functions, it avoids excessive restrictions on the driver's reasonable operations, balances safety and driving freedom, and continuously updates the "human-vehicle-environment" data, dynamic risk values and controllable function lists on a periodic basis to form a "perception-prediction-decision-execution" closed loop, which can adapt to the continuous changes in long-term driving scenarios, thereby realizing real-time perception of multi-dimensional risks and flexible adjustment of vehicle function permissions, improving the safety redundancy and scenario adaptability of vehicle permission control, and thus ensuring the collaborative safety and control flexibility of vehicle driving.

[0023] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiment below. The accompanying drawings are for illustration purposes only and are not to be considered as limiting the present application. The same reference symbols are used throughout the drawings to represent the same components. In the drawings:

[0025] Figure 1 A flowchart of a method for controlling authority security of an artificial intelligence vehicle provided by an embodiment of the present application is shown;

[0026] Figure 2 A flowchart of another method for controlling the authority security of an artificial intelligence vehicle provided by an embodiment of the present application is shown;

[0027] Figure 3 A block diagram showing the composition of an authority security control device for an artificial intelligence vehicle provided by an embodiment of the present application is shown;

[0028] Figure 4 A block diagram of another artificial intelligence vehicle authority security control device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0029] The following describes exemplary embodiments of the present application in more detail with reference to the accompanying drawings. Although exemplary embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present application and to fully convey the scope of the present application to those skilled in the art.

[0030] Currently, existing technologies for vehicle access security control mostly rely on fixed access rules based on a single-dimensional data source. For example, driver status is detected solely by steering wheel grip, or lane departure is detected solely by cameras. However, while this approach can achieve partial functional takeover, it lacks collaborative analysis of physiological indicators, operational behavior, and environmental risks, making it difficult to fully quantify dynamic risks. As a result, the vehicle's access management mechanism is unable to perceive multi-dimensional risks in real time and flexibly adjust vehicle function permissions. This severely restricts the security redundancy and scenario adaptability of vehicle access control, affecting the collaborative safety and control flexibility of vehicle driving.

[0031] To this end, the embodiment of the present application provides a method for controlling the authority security of an artificial intelligence vehicle. This method can predict the comprehensive risk value of the future cycle through multimodal data of physiology, behavior and environment, and realize accurate authority adaptation of vehicle functions based on dynamic risk levels, thereby improving the coordinated safety and control flexibility of vehicle driving. The specific execution steps are as follows: Figure 1 Shown, including:

[0032] 101. Obtain the user's physiological monitoring data, driving behavior data, and vehicle environmental perception data within the current cycle.

[0033] In this step, the current cycle can be determined based on the collection requirements of physiological monitoring data, driving behavior data, and vehicle environmental perception data and the real-time requirements of risk prediction, for example, 30 seconds, 1 minute, 5 minutes, etc. Physiological monitoring data includes but is not limited to the user's breathing rate, micro-tremor amplitude, sitting stability, and the proportion of hand contact time with the steering wheel. Driving behavior data includes but is not limited to the accelerator and brake pedal pressure gradient, steering angular velocity, and vehicle acceleration data. Environmental perception data includes but is not limited to obstacle information, lane line status information, traffic flow information, and traffic event information.

[0034] Specifically, the detailed execution process of obtaining the user's physiological monitoring data, driving behavior data and vehicle's environmental perception data in the current cycle is as follows: collecting the user's breathing rate, micro-tremor amplitude, sitting posture stability and the proportion of time the hand is in contact with the steering wheel in the current cycle to obtain the user's physiological monitoring data in the current cycle; collecting the accelerator pedal pressure gradient, brake pedal pressure gradient, steering angular velocity and vehicle acceleration in the current cycle to obtain the user's driving behavior data in the current cycle; obtaining obstacle information, lane line status information, traffic flow information and traffic event information in the radar point cloud data of the vehicle's surrounding environment in the current cycle to obtain the vehicle's environmental perception data in the current cycle.

[0035] It should be noted that in the process of collecting physiological monitoring data, for the user's breathing frequency and micro-tremor amplitude, a non-contact millimeter-wave radar can be installed on the top of the car to collect the chest rise and fall frequency in real time to calculate the user's breathing frequency, and to calculate the micro-tremor amplitude through high-frequency signal analysis of small vibrations of the hands or head. For sitting posture stability, the smart seat pressure sensor can be embedded in the seat to collect the user's sitting posture stability, sitting posture pressure balance, sitting center of gravity offset distance, etc. in real time to calculate the sitting posture stability. As for the proportion of the time the hand is in contact with the steering wheel, the capacitance change of the steering wheel capacitance sensor can be used to determine whether the hand has left the steering wheel, and the proportion of the hand contact time can be determined based on the proportion of the time the hand is in contact with the steering wheel per unit time.

[0036] When collecting driving behavior data, high-precision pressure sensors can be built into the accelerator and brake pedals to determine the pressure gradients by calculating the rate of change of pressure per unit time. For steering angular velocity, a steering angle sensor can be used to collect the steering wheel's steering angular velocity (unit: ° / s). For vehicle acceleration, a chassis-mounted acceleration sensor can be used to collect longitudinal and lateral acceleration in real time.

[0037] During the environmental perception data collection process, obstacle information can be generated by collecting radar point cloud data from onboard radars and cameras. This information includes the number of obstacles, distance to obstacles, relative speed to obstacles, and obstacle type. Lane status information can be obtained by calculating lane integrity based on camera image recognition (such as intact, slightly damaged, or severely damaged) and curve coordinate system conversion. Traffic flow information can be obtained by using onboard cameras and vehicle-to-everything (V2X) communication to obtain information such as forward traffic density, road occupancy, and the difference between average vehicle speed and the speed limit. Traffic event information, such as accidents, construction, and sudden weather changes, can be received in real time through V2X communication.

[0038] 102. Perform spatiotemporal alignment and noise filtering on physiological monitoring data, driving behavior data, and environmental perception data, and calculate the physiological abnormality index, operation deviation, and environmental risk factor of the current cycle respectively.

[0039] In this step, after obtaining the aforementioned physiological monitoring data, driving behavior data, and environmental perception data, all sensor data are aligned to a unified time base (with an error of <10ms) using GPS timestamps to achieve time synchronization. The radar point cloud data and camera images are then mapped to the vehicle coordinate system to achieve coordinate system conversion. The dynamic time warping (DTW) algorithm is used to align different modal data (such as respiratory rate and steering angular velocity) along the time axis to achieve event correlation, thereby completing spatial alignment and obtaining spatiotemporally aligned data. The spatiotemporally aligned data is then subjected to noise filtering, including Kalman filtering and outlier detection. Kalman filtering is used to smooth continuous data such as acceleration and steering angle, while outlier detection uses the 3σ criterion to eliminate outliers (such as sudden acceleration changes) in the sensor data to ensure data accuracy.

[0040] The physiological abnormality index is calculated based on the user's breathing rate, micro-tremor amplitude, sitting posture stability, and the proportion of time the hand is in contact with the steering wheel. The operational deviation is calculated based on the accelerator and brake pedal pressure gradients, steering angular velocity, and vehicle acceleration. The environmental risk factor is calculated based on obstacle information, lane status information, traffic flow information, and traffic event information.

[0041] Specifically, the physiological monitoring data, driving behavior data and environmental perception data are subjected to spatiotemporal alignment and noise filtering, and the physiological abnormality index, operation deviation and environmental risk factor of the current cycle are calculated respectively. The detailed execution process is as follows: the physiological monitoring data, driving behavior data and environmental perception data are subjected to spatiotemporal alignment and noise filtering to obtain the processed physiological monitoring data, driving behavior data and environmental perception data; based on the processed physiological monitoring data, the respiratory rate abnormality score, the micro-tremor amplitude abnormality score, the sitting posture stability abnormality score and the contact time abnormality score are calculated respectively, and the respiratory rate abnormality score, the micro-tremor amplitude abnormality score, the sitting posture stability abnormality score and the contact time abnormality score are calculated respectively, and the physiological abnormality index, the operation deviation degree and the environmental risk factor of the current cycle are calculated respectively. The physiological abnormality index is determined based on the abnormal scores of the number of touches, sitting posture stability and contact time. The throttle pressure gradient deviation, brake pressure gradient deviation, steering angular velocity deviation and vehicle acceleration deviation are calculated based on the processed driving behavior data, and the operation deviation is determined based on the throttle pressure gradient deviation, brake pressure gradient deviation, steering angular velocity deviation and vehicle acceleration deviation. The obstacle hazard degree, lane line integrity, traffic flow hazard degree and event urgency are calculated based on the processed environmental perception data, and the environmental hazard coefficient is determined based on the obstacle hazard degree, lane line integrity, traffic flow hazard degree and event urgency.

[0042] It should be noted that the specific calculation process and expression for the physiological abnormality index of the current cycle are as follows:

[0043] The physiological indicators of the current cycle, such as respiratory rate, micro-tremor amplitude, sitting posture stability, and the proportion of hand contact time with the steering wheel, are converted into standardized scores (0-1) to indicate the degree of deviation from the normal range:

[0044] Abnormal respiratory rate score R:

[0045] ;

[0046] Among them, if R>1, then R=1 (maximum outlier).

[0047] Abnormal microtremor amplitude score T:

[0048] ;

[0049] Among them, if T<0, then T=0 (normal range); if T>1, then T=1.

[0050] Abnormal sitting stability score S:

[0051] Calculate the standard deviation of seat front / back / side-to-side pressure:

[0052] ;

[0053] in, is the pressure value of each pressure sensor, is the mean value, and N is the total number of pressure sensors embedded inside the seat, which is used to characterize the uniformity of seat pressure distribution.

[0054] ;

[0055] Among them, if S>1, then S=1, which means that the comprehensive situation of the center of gravity deviation and pressure balance of the vehicle driver and passengers has exceeded the safety threshold and is in a dangerous state that requires attention. In the subsequent applications such as permission control and safety warning based on S, as long as S=1, the corresponding high-risk handling process can be triggered.

[0056] Hand contact duration abnormality score H:

[0057] ;

[0058] If the contact time is less than 80%, the H value will increase.

[0059] ;

[0060] Among them, if H>1, then H=1 (completely disengaged), which means that the hand contact time has seriously failed to meet the safety requirements and is in dangerous situations such as "completely disengaged". In automobile driving safety management, once H>1, an emergency safety response can be triggered to ensure the safety and standardization of personnel operations during driving.

[0061] According to the impact of the above physiological indicators on the user's safe driving, the respiratory rate abnormality score, the micro-tremor amplitude abnormality score, the sitting posture stability abnormality score and the hand contact time abnormality score are set with corresponding weights in advance, for example, respiratory rate (0.3), micro-tremor amplitude (0.25), sitting posture stability (0.25), hand contact time (0.3). The current physiological abnormality index X is:

[0062] .

[0063] For the operation deviation of the current cycle, the specific calculation process and expression are as follows:

[0064] Accelerator pedal pressure gradient deviation and brake pedal pressure gradient deviation :

[0065] Baseline value: Set safety thresholds based on the vehicle model (e.g., the emergency acceleration threshold is an accelerator pedal pressure gradient of 5 m / s², and the emergency braking deceleration threshold is a pedal pressure gradient of 20 m / s²).

[0066]

[0067] If the deviation exceeds 100%, it is taken as 100% (maximum abnormality).

[0068] Steering angular velocity deviation :

[0069] Baseline value: Set a safety threshold based on the road type (e.g., highway / city) (e.g., the steering angle speed threshold on city roads is 300° / second).

[0070] ;

[0071] If the measured value exceeds the safety threshold, it is set to 1.0 (maximum anomaly).

[0072] Vehicle acceleration deviation :

[0073] Baseline value: Sets a safe acceleration range based on road speed limits and driving scenarios (e.g., a longitudinal acceleration safety range of ±3m / s²).

[0074] ;

[0075] If the deviation exceeds 3 times the standard deviation, the value corresponding to 3 times the standard deviation (such as 3.0) is taken.

[0076] According to the influence of the above deviation indicators on the user's safe driving, the corresponding weights are set for the accelerator pedal pressure gradient deviation, brake pedal pressure gradient deviation, steering angular velocity deviation and vehicle acceleration deviation in advance, for example, accelerator pedal (0.3), brake pedal (0.4), steering angular velocity (0.2), vehicle acceleration (0.1). The current operation deviation Y is:

[0077] .

[0078] For the environmental risk factor of the current cycle, the specific calculation process and expression are as follows:

[0079] Obstacle hazard level :

[0080] ;

[0081] Where N is the number of obstacles detected; is the distance to obstacle i; is the relative speed to the vehicle ; is the obstacle type weight (e.g. pedestrian = 1.2, vehicle = 0.8, etc.).

[0082] Lane status :

[0083] .

[0084] Traffic flow hazard :

[0085] ;

[0086] in, It is the road speed limit; is the average speed; Occupancy is the proportion of the road occupied by vehicles (usually expressed as a percentage), including the occupancy of lanes or road sections; is the traffic density; if <20% ,but Weighted doubling (congestion state).

[0087] V2X event impact weighting:

[0088] Incident urgency :

[0089] Set weights based on V2X event types (e.g., accident = 1.5, congestion = 0.8, construction = 0.6)

[0090] ;

[0091] in, is the event type, is the straight-line distance from the vehicle's current location to the location where the V2X event occurred, The event impact distance threshold (such as 100 meters).

[0092] Current environmental risk factor Z:

[0093] .

[0094] Based on the detailed implementation methods described above, the physiological abnormality index can be used to convert a driver's physiological state (such as respiratory rate, micro-tremors, sitting posture stability, etc.) into a numerical risk indicator, reflecting the degree of fatigue, distraction, or health abnormalities. The operational deviation can be used to analyze driving behavior data (such as sudden acceleration, sudden braking, sudden steering angle changes, etc.), quantify the degree of deviation between the driver's operation and the "normal driving template", and identify dangerous driving behaviors. The environmental risk coefficient can be used to integrate environmental perception data (such as obstacle distance, lane damage, traffic congestion, etc.) to assess the potential danger level of the surrounding environment. This achieves a fully coupled risk analysis of the "human-vehicle-environment" system, avoiding the misjudgments caused by traditional solutions based on isolated judgments from a single-dimensional data source, and effectively improving the comprehensiveness and accuracy of subsequent risk predictions.

[0095] 103. Input the physiological abnormality index, operation deviation and environmental risk coefficient of the current cycle into the risk prediction model integrated with the attention mechanism to obtain the dynamic risk value of the next cycle.

[0096] Among them, the attention mechanism is used to dynamically allocate the global attention weights of physiological abnormality index, operation deviation and environmental risk coefficient according to the risk evolution relationship of historical cycles.

[0097] In this step, a spatiotemporal convolutional network is used to process the spatiotemporal dependencies of the physiological abnormality index, operational deviation, and environmental risk factor, and the global attention weights of the physiological abnormality index, operational deviation, and environmental risk factor are dynamically allocated according to the attention mechanism. For example, when the event type is an emergency, the weight of the environmental risk factor is increased. The risk prediction model is specifically an LSTM-Transformer hybrid network model. The LSTM layer is used to capture the risk evolution relationship between the physiological abnormality index, operational deviation, and environmental risk factor of the historical period and the risk level, and the Transformer layer is used to calculate the global attention weights of the physiological abnormality index, operational deviation, and environmental risk factor based on the risk evolution relationship. The risk prediction model is used to output the dynamic risk value of the next period after inputting the physiological abnormality index, operational deviation, and environmental risk factor of the current period, that is, its output layer is the dynamic risk value. ∈[0,1], where: <0.3: low risk; 0.3≤ <0.7: medium risk; ≥0.7: high risk.

[0098] Training data is constructed based on a historical dataset. This dataset can specifically be historical driving data from the past one or three years. The same processing as in steps 101-102 is used: the historical driving data is divided into historical periods according to the time span of the current period, and the physiological abnormality index, environmental risk factor, and operational deviation corresponding to each historical period are calculated. The physiological abnormality index, environmental risk factor, and operational deviation are used as training data and, combined with real risk event records, are labeled with risk values to form training samples. These real risk event records include at least the event type (e.g., collision, sudden braking, lane departure), occurrence timestamp, and severity (e.g., high, medium, low). A dynamic time window can be pre-set for each event type. Based on the time window and occurrence timestamp, risk events are matched and associated. For example, the window length for emergency events (collision) is 1-5 seconds. If a risk event occurs within the time window, the risk value is assigned based on its severity. If multiple risk events occur within the time window, the highest severity among the multiple risk events is assigned the risk value.

[0099] A LSTM-Transformer hybrid network was constructed as a risk prediction model incorporating an attention mechanism. The LSTM layer was used to capture the risk evolution relationship between the physiological abnormality index, operational deviation, and environmental risk factor over historical periods and the risk level. The Transformer layer was used to calculate the global attention weights for the physiological abnormality index, operational deviation, and environmental risk factor based on this risk evolution relationship. The training samples were divided into a training set (70%), a validation set (15%), and a test set (15%). The LSTM-Transformer hybrid network model constructed above was trained using the training set. The accuracy of the trained LSTM-Transformer hybrid network model was verified using the validation set. The performance of the verified LSTM-Transformer hybrid network model was then tested using the test set to ensure that it met the pre-defined accuracy standards for the LSTM-Transformer hybrid network model. This resulted in a risk prediction model incorporating an attention mechanism.

[0100] By inputting the physiological abnormality index, operation deviation and environmental risk factor into the risk prediction model of the fusion attention mechanism, the dynamic risk value of the next cycle can be output.

[0101] 104. The dynamic risk value of the next period is revised based on the risk time decay factor, and the revised dynamic risk value matches the target risk level in the preset risk level.

[0102] Among them, the risk time decay factor is determined by combining the dynamic risk value of the historical period and the real-time status parameters of the vehicle.

[0103] Specifically, the detailed execution process of determining the risk time decay factor is as follows: perform a quantitative analysis of the time series trend of the dynamic risk value of the historical period, perform a quantitative assessment of the real-time status parameters of the vehicle, and obtain the risk time decay factor; based on the risk time decay factor, correct the dynamic risk value to obtain the corrected dynamic risk value. In this step, the dynamic risk values of each historical period are collected in advance. Statistical methods or machine learning models are used to analyze the trend changes of historical dynamic risk values, including but not limited to the moving average method, exponential smoothing method, and survival analysis. Among them, the moving average method: calculates the moving average under different window sizes to smooth data fluctuations and identify long-term trends; the exponential smoothing method: assigns higher weights to recent data to capture short-term trend changes; survival analysis: such as the Kaplan-Meier estimate or the Cox proportional hazard model, is used to evaluate the "survival" probability of a specific risk factor over time, and then infer the decay rate of its impact.

[0104] According to the above trend analysis results, determine the risk time decay factor , which represents the degree of attenuation of historical risk value over time. The specific expression is:

[0105]

[0106] in, As the basic exponential decay, the core decay term ensures that the historical dynamic risk value decreases over time. is the basic attenuation, t is the time difference from the current time; It is a trend enhancement factor that dynamically adjusts the decay rate according to the trend of historical risk value (increasing or decreasing). When the risk trend increases (such as the risk value increases for many consecutive days), the enhancement factor It may be positive, which reduces the decay factor (i.e., decays more slowly). When the risk trend decreases, the factor is enhanced. is negative, which increases the attenuation factor (i.e., accelerates decay). is the trend sensitivity coefficient, and Trend(t) is the trend slope calculated by moving average or linear regression; is the covariate adjustment factor, which adjusts the decay rate according to the historical volatility (such as standard deviation) of the historical dynamic risk value; g(S) is the weighted correction term of the vehicle real-time status parameter, and n represents the number of vehicle real-time status parameters involved in the weighted correction. is the weight coefficient of the real-time state parameter of the i-th vehicle, Normalize to [0,1], that is, the original vehicle real-time state parameters Converting to the range of [0,1] allows parameters of different types and dimensions to be calculated and analyzed on a unified scale. Center the parameter value around 0.5 (e.g. battery level =0.3, (0.3-0.5)=-0.2, reducing the decay factor).

[0107] In the process of calculating the risk time decay factor, exponential decay, trend analysis and covariate adjustment are integrated, combining the classic method of time decay (exponential decay) and cutting-edge technology of trend analysis (such as the Cox model). Through the trend enhancement factor, it ensures sensitivity to the rising trend of risk and avoids ignoring potential risks due to excessive decay. It can more flexibly reflect the decay law of historical dynamic risk values and effectively improve the accuracy and real-time performance of risk prediction.

[0108] The original dynamic risk value is combined with the risk time decay factor to obtain the revised dynamic risk value. The specific expression is:

[0109] ;

[0110] in, represents the original dynamic risk value, The corresponding risk time decay factor, g(S), is a weighted correction term for the vehicle's real-time status parameters.

[0111] It should be noted that to ensure that the correction process is as expected and does not introduce significant deviations, the above expression can be applied to the dynamic risk value of each historical period to generate a new series of corrected dynamic risk values. The dynamic risk values before and after the correction can be visualized as a comparison chart to intuitively compare the changes before and after the correction.

[0112] Accordingly, a corresponding dynamic risk value interval can be pre-set for each of the preset risk levels. The preset risk levels include low risk, medium risk, and high risk. Once the corrected dynamic risk value is obtained, it can be matched with the risk value interval corresponding to the preset risk level to determine the risk level for the next cycle, i.e., the target risk level.

[0113] Through the detailed implementation described above, by calculating a weighted correction term for the vehicle's real-time status parameters and introducing a risk decay factor, the revised dynamic risk value simultaneously accounts for the ongoing impact of historical trends and the decaying effect of real-time status. This dynamic risk value correction logic better aligns with the requirements of "historical trend analysis + real-time status perception," improving the accuracy and scenario adaptability of risk level matching. Matching the target risk level based on the revised dynamic risk value significantly enhances the accuracy, real-time nature, and robustness of risk assessment.

[0114] 105. Use risk-authority linkage rules to determine the dynamic authority adjustment strategy corresponding to the target risk level.

[0115] Among them, the dynamic permission adjustment strategy includes hierarchical authorization thresholds and real-time response rules for functional permissions.

[0116] In this step, a mapping relationship is established between each risk level and predefined vehicle permissions, namely the risk-permission linkage rule. This risk-permission linkage rule can be specifically combined with the permissions of all controllable functions in the vehicle as a permissions database. Within this permissions database, hierarchical authorization thresholds and real-time response rules are set for each risk level. The hierarchical authorization thresholds specifically refer to the scope of functions allowed / restricted and parameter thresholds. These are typically categorized into three levels: low risk, medium risk, and high risk. These thresholds can be dynamically adjusted based on safety indicators, traffic regulations, user habits, and other factors. For example, low risk allows automatic lane changes (if the vehicle behind is >50 meters away), with a cruising speed of ≤120 km / h; medium risk allows lane changes only in adjacent lanes (if the vehicle behind is >100 meters away), with a cruising speed of ≤100 km / h; and high risk prohibits lane changes, with a cruising speed of ≤80 km / h. Real-time response rules include response conditions, execution actions, and parameter restrictions. For example, conditions: The camera / radar detects a pedestrian less than 20 meters away from the vehicle and at an unprotected intersection; action: Immediately trigger automatic braking and disable the entertainment system sound output; parameters: Braking deceleration ≥ 6m / s², response time ≤ 0.3 seconds, until the pedestrian leaves the danger zone.

[0117] After determining the target risk level, the dynamic permission adjustment strategy corresponding to the target risk level can be quickly determined through the risk-permission linkage rule.

[0118] 106. Based on the vehicle's driving scenario, the dynamic permission adjustment strategy is verified in real time, a list of finally authorized controllable functions is generated, and permission authorization operations are executed within the next cycle to facilitate user security control.

[0119] In this step, the corresponding controllable functions are determined based on the permissions corresponding to the dynamic permission adjustment policy determined in step 105. To better align the dynamic permission adjustment policy with the vehicle's driving scenario, adapt to the continuous changes in long-term driving scenarios, improve the security redundancy and scenario adaptability of vehicle permission control, and thus ensure the coordinated safety and control flexibility of vehicle driving, the dynamic permission adjustment policy can be verified in real time based on the vehicle's driving scenario, thereby finalizing the authorized controllable function list.

[0120] Specifically, the dynamic authority adjustment strategy is verified in real time based on the driving scenario of the vehicle, and the detailed execution process of generating the final authorized controllable function list is as follows: determine the driving scenario according to the road type, traffic environment and road conditions of the vehicle; verify the hierarchical authorization threshold and real-time response rules based on the key safety indicators preset in the driving scenario to obtain the verification result; if the verification result is passed, the controllable function list is generated according to the dynamic authority adjustment strategy; if the verification result is failed, the dynamic authority adjustment strategy is adaptively corrected, and the controllable function list is generated according to the corrected dynamic authority adjustment strategy.

[0121] In this step, GPS / high-precision maps are used to determine the current road type (freeway / urban road / tunnel). Cameras identify lane markings, traffic signs, traffic lights, pedestrians, and vehicles. LiDAR detects surrounding obstacle density, road boundaries, and terrain changes. Radar measures the distance to the vehicle ahead and its relative speed, thereby determining the traffic environment. Environmental sensors detect weather (rain, snow, fog, or haze), road slipperiness, and visibility. A convolutional neural network (CNN) is used to extract and classify these data based on driving scenarios. For example, on a freeway: speeds > 80 km / h, clear lane markings, and low traffic density. On a city road: speeds < 50 km / h, dense traffic lights, and numerous pedestrians and non-motorized vehicles. In a tunnel: weak GPS signal, low visibility, and a low speed limit. In extreme weather: rainfall > 5 mm / h and visibility < 100 m.

[0122] For example, key indicators are defined for each driving scenario, as shown in Table 1:

[0123] Table 1

[0124]

[0125] The hierarchical authorization thresholds and real-time response rules are compared and verified against the quantitative standards of their corresponding scenarios against key safety indicators to determine if there are any conflicts. If not, verification passes, meaning the scenario rules are consistent with the dynamic permission policy and meet safety indicators. A list of controllable functions is generated based on the permissions corresponding to the dynamic permission adjustment policy. If there are conflicts, verification fails, meaning the scenario rules are more restrictive or the dynamic policy violates safety indicators. If the scenario rules require stricter restrictions (e.g., prohibiting autonomous driving in tunnels), the permission policy can be dynamically adjusted. If safety indicators are not met (e.g., vehicle speeds > 120 km / h), the real-time response rules can be modified to trigger an emergency speed limit command. A list of controllable functions is generated based on the corrected permissions corresponding to the dynamic permission adjustment policy. Once the list of controllable functions is determined, permission control commands are sent to the ECU via the CAN bus, preserving the controllable functions for user use while disabling other functions. The current target risk level and a list of available functions, consisting of controllable functions, are displayed on the vehicle's display, allowing users to easily monitor the vehicle's current status. Voice notifications can also be used to provide user alerts. For example, if the authority is restricted, a voice prompt can be given: "Due to high risk, please take over driving immediately."

[0126] Based on the above Figure 1It can be seen from the implementation method that the present application provides a method for security control of permissions for artificial intelligence vehicles. The technical solution provided by the present application breaks through the limitations of single-dimensional risk assessment by synchronously collecting physiological monitoring data, driving behavior data and environmental perception data of the current period and calculating the physiological abnormality index, operation deviation and environmental risk factor respectively, and realizes the "human-vehicle-environment" full-factor risk coupling analysis, avoids the misjudgment caused by isolated judgment of traditional solutions, and improves the comprehensiveness and accuracy of subsequent risk prediction. The physiological abnormality index, operation deviation and environmental risk factor of the current period are input into the risk prediction model that dynamically allocates the global attention weight of the physiological abnormality index, operation deviation and environmental risk factor according to the risk evolution relationship of the historical period, and outputs the dynamic risk value of the next period. It can capture the risk evolution trend, realize risk pre-perception, make the permission adjustment ahead of the risk outbreak, correct the dynamic risk value based on the risk time decay factor, and match the preset The risk level takes into account the characteristic that the impact of risk events decays over time, making the matching of risk levels more accurate. Dynamic permission adjustment strategies are generated through risk-permission linkage rules, and the dynamic permission adjustment strategies are verified in real time based on the driving scenario of the vehicle to generate a final authorized list of controllable functions. This can ensure the compatibility between the controllable function list and the driving scenario, thereby breaking the rigid restrictions of static permission rules and achieving elastic scaling of permissions. While ensuring core safety functions, it avoids excessive restrictions on the driver's reasonable operations, balancing safety and driving freedom, and continuously updates the "human-vehicle-environment" data, dynamic risk values and controllable function lists on a periodic basis to form a "perception-prediction-decision-execution" closed loop, which can adapt to the continuous changes in long-term driving scenarios, thereby achieving real-time perception of multi-dimensional risks and flexible adjustment of vehicle function permissions, improving the security redundancy and scenario adaptability of vehicle permission control, and thus ensuring the collaborative safety and control flexibility of vehicle driving.

[0127] Furthermore, the preferred embodiment of the present application is in the above Figure 1 Based on this, a detailed description of the process of security control of permissions for artificial intelligence vehicles is given. The specific steps are as follows: Figure 2 Shown, including:

[0128] 201. Obtain physiological abnormality index, operation deviation and environmental risk factor of historical period.

[0129] In this step, historical driving data from the past one to three years is collected. This historical driving data specifically includes physiological monitoring data, driving behavior data, and environmental perception data from the historical period. Combined with the calculation of the physiological abnormality index, operating deviation, and environmental risk factor for the current period in steps 101-103, the physiological abnormality index, operating deviation, and environmental risk factor corresponding to each historical period are calculated based on the physiological monitoring data, driving behavior data, and environmental perception data from the historical period.

[0130] The three types of data are aligned in time and space, and outliers are processed before being stored in a database. Each record is guaranteed to contain the following fields: timestamp (accurate to the second), physiological abnormality index, operational deviation, and environmental risk factor. This facilitates subsequent retrieval.

[0131] 202. Physiological abnormality index, operation deviation and environmental risk factor are used as training data, and the training data are marked with risk levels based on real risk event records to obtain training samples.

[0132] Among them, the real risk event records include event type, occurrence timestamp and severity.

[0133] In this step, you can obtain records of actual accidents, congestion, construction, and other risk events from accident reports, traffic management departments, or insurance companies. Each record should include the event type (e.g., accident, congestion), occurrence timestamp, and severity (e.g., minor, moderate, severe).

[0134] Specifically, the training data is marked with risk levels based on real risk event records, and the detailed execution process of obtaining training samples is as follows: obtain the dynamic time windows corresponding to different event types; according to the dynamic time windows, match and associate the event types with the training data according to the occurrence timestamps; if the matched and associated event type in the training data is a single event, the training data is marked with risk values based on the severity corresponding to the single event; if the matched and associated event type in the training data is a compound event, the training data is marked with risk values based on the highest severity in the compound event.

[0135] It's important to note that different types of risk events have varying durations of impact on driving behavior and risk factors. For example, an accident typically causes significant changes in driver behavior (e.g., sudden braking, jerking the steering wheel) and physiological state (e.g., accelerated heart rate) within a few minutes before and after the accident. Congestion can have impacts lasting up to ten or even dozens of minutes, as vehicles gradually slow down, stop, and then restart. Construction can also have even longer-lasting effects, as it not only alters normal driving routes but can also cause changes in traffic flow. Therefore, given the varying timeliness of different event types, different dynamic time windows can be pre-defined based on event type to better capture these differences. This allows for more accurate correlation between historical physiological abnormality indices, historical operational deviations, and historical environmental risk factors and actual risk events, thereby improving the accuracy of risk prediction models. For example, for an accident, data within 5 minutes before and after the event is considered relevant; for congestion, data within 15 minutes before and after the event is considered relevant; and for construction, data within 30 minutes before and after the event is considered relevant.

[0136] After obtaining the dynamic time windows corresponding to different time types, for each piece of training data (i.e., the historical physiological abnormality index, historical operational deviation, and historical environmental risk factor at each time point), all event records within the corresponding dynamic time window are searched. If a single event exists, the severity of the event is directly used as the risk marker. If a complex event exists, the most serious event is selected as the risk marker. For example, low risk (e.g., no event or minor incident); medium risk (e.g., general congestion); and high risk (e.g., major accident or severe congestion).

[0137] Furthermore, to further improve the accuracy of linking training data to event types, dynamic time windows allow for precise matching of historical periods with physiological abnormality indices, operational deviations, and environmental risk factors based on the location of the event. This eliminates data points unrelated to specific risk events, reduces unnecessary interference, and helps improve the quality of model training, allowing it to focus more on truly meaningful information.

[0138] 203. Construct an LSTM-Transformer hybrid network model as a risk prediction model integrating attention mechanism.

[0139] Among them, the LSTM layer is used to capture the risk evolution relationship between the physiological abnormality index, operation deviation, environmental risk factor and risk level in the historical period, and the Transformer layer is used to calculate the global attention weight of the physiological abnormality index, operation deviation and environmental risk factor based on the risk evolution relationship.

[0140] In this step, the LSTM layer captures risk evolution. LSTM (Long Short-Term Memory) networks are used to model long-term dependencies in time series, capturing the temporal evolution of physiological, operational, and environmental data. By inputting multivariate time series data consisting of three indicators from historical periods (physiological abnormality index, operational deviation, and environmental risk factor), the model outputs a hidden state for each time step, representing the risk evolution characteristics of the current period. The Transformer layer calculates global attention weights. The Transformer's self-attention mechanism dynamically calculates weights for each time step, assigning higher attention to key periods (e.g., abnormal behavior before a high-risk event). By inputting the hidden state sequence output by the LSTM layer, the model outputs a weighted context vector, representing the global risk characteristics. LSTM parameters include: number of hidden layer units: 64-256 (adjusted based on data complexity); bidirectional LSTM: enhances the model's ability to capture contextual features; Transformer parameters: number of multi-head attention heads: 8-16; positional encoding: embeds time step information using sine / cosine functions. The output layer can be used for classification tasks: using the Softmax activation function to output probability distribution (such as P(low)=0.1, P(medium)=0.5, P(high)=0.7); regression tasks: output continuous risk values (such as a dynamic risk index between 0 and 1). The loss function of this risk prediction model is .in, =0.7, used to ensure classification accuracy; =0.3, used to ensure timing continuity.

[0141] 204. Use training samples to train a risk prediction model that integrates attention mechanism.

[0142] This step is combined with the description of the risk prediction model training process in step 103 of the above method, and the same content will not be repeated here.

[0143] 205. Obtain the user's physiological monitoring data, driving behavior data, and vehicle's environmental perception data within the current cycle.

[0144] This step is combined with the description of step 101 of the above method, and the same contents will not be repeated here.

[0145] 206. Perform spatiotemporal alignment and noise filtering on physiological monitoring data, driving behavior data, and environmental perception data, and calculate the physiological abnormality index, operation deviation, and environmental risk factor of the current cycle respectively.

[0146] This step is combined with the description of step 102 of the above method, and the same contents are not repeated here.

[0147] 207. Input the physiological abnormality index, operation deviation and environmental risk coefficient of the current cycle into the risk prediction model integrated with the attention mechanism to obtain the dynamic risk value of the next cycle.

[0148] This step is combined with the description of step 103 of the above method, and the same contents are not repeated here.

[0149] 208. The dynamic risk value of the next period is corrected based on the risk time decay factor, and the corrected dynamic risk value matches the target risk level in the preset risk level.

[0150] This step is combined with the description of step 104 of the above method, and the same contents are not repeated here.

[0151] 209. Use risk-authority linkage rules to determine the dynamic authority adjustment strategy corresponding to the target risk level.

[0152] This step is combined with the description of step 105 of the above method, and the same contents are not repeated here.

[0153] 210. Based on the vehicle's driving scenario, the dynamic permission adjustment strategy is verified in real time, a list of finally authorized controllable functions is generated, and permission authorization operations are executed within the next cycle to facilitate user security control.

[0154] This step is combined with the description of step 106 of the above method, and the same contents are not repeated here.

[0155] Furthermore, to ensure accurate adaptation of dynamic risk levels to vehicle functions, thereby improving coordinated safety and control flexibility during vehicle driving, in addition to steps 201-210, the following steps are further included: monitoring user satisfaction with the list of controllable functions in the next cycle, and calculating the risk event incidence rate of the vehicle in the next cycle; calculating the safety compatibility between the dynamic permission adjustment strategy and the target risk level based on the user satisfaction and risk event incidence rate; and triggering an update prompt for the risk-permission linkage rule if the safety compatibility is lower than a preset threshold.

[0156] In this step, the user's satisfaction can be determined through direct feedback, behavioral data, and sentiment analysis. Direct feedback can collect user ratings of controllable functions (such as 1-5 star ratings) and comment texts (such as "smooth operation" and "response delay"); behavioral data can be analyzed through logs to analyze the user's frequency of use of functions, length of stay, number of incorrect operations, etc.; sentiment analysis can use natural language processing (NLP) technology to analyze the sentiment polarity of user comments (such as positive / neutral / negative) and extract negative feedback keywords (such as "stuttering" and "insufficient authority"). Specifically, a scoring system can be set up, which includes quantitative scores corresponding to the indicators involved in the above three methods, and the scores of each method are calculated separately. Then, they are weighted with the preset weights to calculate the usage satisfaction score, which ranges from 0 to 1.

[0157] The vehicle's risk event rate can be calculated by recording high-risk events (such as operational errors and environmental hazards) that occur within the next cycle through sensors, logs, or user reports. Alternatively, the dynamic risk value correction results from steps 203-209 can be combined to calculate the actual frequency of risk events. Specifically, the current cycle can be used as a benchmark, and the number and type of risk events (such as accidents, congestion, and operational errors) in the next cycle can be counted to calculate the risk event rate.

[0158] After obtaining user satisfaction and risk event rates, the safety fit can be calculated by combining them with their respective preset importance factors. A pre-set fit threshold (e.g., 0.8) is set and the safety fit is compared to this threshold. If the safety fit is equal to or above the threshold, the risk-permission linkage rules do not need to be adjusted, ensuring coordinated safety and control flexibility. If the safety fit is below the threshold, the risk-permission linkage rules need to be adjusted to further improve coordinated safety and control flexibility, triggering a risk-permission linkage rule update prompt. This update prompt includes: Suggestions: Automatically generate recommendations for risk-permission linkage rules (e.g., "Change high-risk permissions from 'Allow' to 'Restrict'"), provide attribution analysis of risk event rates (e.g., "Operational error rate increased by 20%"), etc. Notifications can be displayed on the vehicle's display or sent via email, SMS, or other means, prompting users to promptly update risk-permission linkage rules to ensure coordinated safety and control flexibility.

[0159] Furthermore, as a response to the above Figure 1-Figure 2The implementation of the method embodiment shown, the embodiment of the present application provides an authority security control device for an artificial intelligence car, which is used to predict the comprehensive risk value of the future period through multimodal data of physiology, behavior and environment, and to achieve accurate authority adaptation of vehicle functions based on dynamic risk levels, thereby improving the collaborative safety and control flexibility of car driving. The embodiment of this device corresponds to the aforementioned method embodiment. For ease of reading, this embodiment will no longer repeat the details of the aforementioned method embodiment one by one, but it should be clear that the device in this embodiment can correspond to all the contents of the aforementioned method embodiment. Specifically, Figure 3 As shown, the device includes:

[0160] The first acquisition unit 31 is used to acquire the user's physiological monitoring data, driving behavior data and vehicle's environmental perception data in the current cycle;

[0161] a first calculation unit 32 for performing spatiotemporal alignment and noise filtering on the physiological monitoring data, the driving behavior data, and the environmental perception data, and respectively calculating a physiological abnormality index, an operation deviation, and an environmental risk factor of a current period;

[0162] A prediction unit 33 is configured to input the physiological abnormality index, operational deviation, and environmental risk factor of the current cycle into a risk prediction model that incorporates an attention mechanism to obtain a dynamic risk value for the next cycle. The attention mechanism is configured to dynamically assign global attention weights to the physiological abnormality index, operational deviation, and environmental risk factor based on the risk evolution relationship of historical cycles.

[0163] a processing unit 34 configured to modify the dynamic risk value of the next period according to a risk time decay factor, and match the modified dynamic risk value to a target risk level within a preset risk level, wherein the risk time decay factor is determined by combining the dynamic risk value of the historical period and the real-time status parameters of the vehicle;

[0164] A determination unit 35 is configured to determine a dynamic permission adjustment strategy corresponding to the target risk level using a risk-permission linkage rule, wherein the dynamic permission adjustment strategy includes a hierarchical authorization threshold for functional permissions and a real-time response rule;

[0165] The authorization unit 36 is used to verify the dynamic permission adjustment strategy in real time based on the driving scenario of the vehicle, generate a final authorized controllable function list, and perform permission authorization operations in the next cycle to facilitate user security control.

[0166] Further, such as Figure 4 As shown, the first acquiring unit 301 includes:

[0167] The first collection module 3011 is used to collect the user's breathing rate, micro-tremor amplitude, sitting posture stability, and the proportion of time the hand is in contact with the steering wheel during the current cycle to obtain the user's physiological monitoring data during the current cycle;

[0168] The second collection module 3012 is used to collect the accelerator pedal pressure gradient, brake pedal pressure gradient, steering angular velocity, and vehicle acceleration in the current cycle to obtain the user's driving behavior data in the current cycle;

[0169] The first acquisition module 3013 is used to obtain obstacle information, lane line status information, traffic flow information and traffic event information from the radar point cloud data of the vehicle's surrounding environment in the current cycle, and obtain the vehicle's environmental perception data in the current cycle.

[0170] Further, such as Figure 4 As shown, the calculation unit 302 includes:

[0171] A preprocessing module 3021 is configured to perform spatiotemporal alignment and noise filtering on the physiological monitoring data, the driving behavior data, and the environmental perception data to obtain processed physiological monitoring data, driving behavior data, and environmental perception data;

[0172] a first determining module 3022, configured to calculate a respiratory rate abnormality score, a microtremor amplitude abnormality score, a sitting posture stability abnormality score, and a contact duration abnormality score based on the processed physiological monitoring data, and determine the physiological abnormality index according to the respiratory rate abnormality score, the microtremor amplitude abnormality score, the sitting posture stability abnormality score, and the contact duration abnormality score;

[0173] a second determining module 3023, configured to calculate, based on the processed driving behavior data, a throttle pressure gradient deviation, a brake pressure gradient deviation, a steering angular velocity deviation, and a vehicle acceleration deviation, and determine the operation deviation according to the throttle pressure gradient deviation, the brake pressure gradient deviation, the steering angular velocity deviation, and the vehicle acceleration deviation;

[0174] The third determination module 3024 is used to calculate the obstacle hazard, lane line integrity, traffic flow hazard and event urgency based on the processed environmental perception data, and determine the environmental hazard coefficient according to the obstacle hazard, lane line integrity, traffic flow hazard and event urgency.

[0175] Further, such as Figure 4 As shown, the device also includes:

[0176] The second acquisition unit 307 is configured to acquire the physiological abnormality index, operation deviation, and environmental risk coefficient of the historical period before inputting the physiological abnormality index, operation deviation, and environmental risk coefficient of the current period into the risk prediction model of the fusion attention mechanism to obtain the dynamic risk value of the next period;

[0177] a marking unit 308 for using the physiological abnormality index, operation deviation, and environmental risk factor of the historical period as training data, and marking the training data with risk values according to actual risk event records to obtain training samples;

[0178] A construction unit 309 is configured to construct an LSTM-Transformer hybrid network model as the risk prediction model integrating the attention mechanism, wherein the LSTM layer is configured to capture the risk evolution relationship between the physiological abnormality index, the operational deviation, and the environmental risk factor of the historical period and the risk level, and the Transformer layer is configured to calculate the global attention weights of the physiological abnormality index, the operational deviation, and the environmental risk factor based on the risk evolution relationship;

[0179] The training unit 310 is used to train the risk prediction model integrating the attention mechanism using the training samples.

[0180] Further, such as Figure 4 As shown, the real risk event record includes the event type, the event occurrence stamp and the severity; the marking unit 308 includes:

[0181] The second acquisition module 3081 is used to acquire dynamic time windows corresponding to different event types;

[0182] an association module 3082, configured to match and associate the event type with the training data according to the dynamic time window and the occurrence event stamp;

[0183] a marking module 3083 configured to mark the training data with a risk value based on the severity corresponding to the single event if the event type matched in the training data is a single event;

[0184] The marking module 3083 is further configured to mark the training data with a risk value according to the highest severity of the compound event if the event type matched and associated in the training data is a compound event.

[0185] Furthermore, the authorization unit 306 includes:

[0186] The fourth determining module 3061 is configured to determine the driving scenario based on the road type, traffic environment, and road conditions on which the vehicle is located;

[0187] A verification module 3062 is configured to verify the hierarchical authorization threshold and the real-time response rule based on the key safety indicators preset in the driving scenario, and obtain a verification result;

[0188] The generating module 3063 is configured to generate the controllable function list according to the dynamic permission adjustment policy if the verification result is passed;

[0189] The generating module 3063 is configured to adaptively correct the dynamic permission adjustment policy if the verification result is failure, and generate the controllable function list according to the corrected dynamic permission adjustment policy.

[0190] Further, such as Figure 4 As shown, the device also includes:

[0191] A monitoring and statistics unit 311 is used to monitor user satisfaction with the controllable function list in the next period, and to count the occurrence rate of risk events of the vehicle in the next period;

[0192] A second calculation unit 312 is configured to calculate a security compatibility between the dynamic permission adjustment policy and the target risk level based on the usage satisfaction and the risk event occurrence rate;

[0193] The triggering unit 313 is configured to trigger an update prompt message of the risk-authorization linkage rule if the security compliance is lower than a preset threshold.

[0194] Furthermore, the embodiment of the present application also provides a storage medium, which is used to store a computer program, wherein when the computer program is running, the device where the storage medium is located is controlled to execute the above Figure 1-Figure 2 The permission security control method of the artificial intelligence car described in.

[0195] Furthermore, the embodiment of the present application also provides a processor, which is used to run a program, wherein the program executes the above Figure 1-Figure 2 The permission security control method of the artificial intelligence car described in.

[0196] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0197] It is understood that the relevant features of the above methods and devices can be referenced to each other. In addition, the terms "first" and "second" in the above embodiments are used to distinguish between the embodiments, and do not represent the advantages and disadvantages of the embodiments.

[0198] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0199] The algorithms and displays provided herein are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems may also be used together with the teachings herein. Based on the above description, it is apparent that the structure required for constructing such systems is suitable. In addition, the present application is not directed to any specific programming language. It should be understood that various programming languages may be utilized to implement the present application described herein, and the description of the specific languages above is provided for the purpose of disclosing the preferred embodiment of the present application.

[0200] In addition, the memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0201] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0202] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0203] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0204] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0205] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0206] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0207] Computer-readable media includes permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0208] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0209] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0210] The above are merely embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should all be included within the scope of the claims of the present application.

Claims

1. A method for controlling the authority security of an artificial intelligence vehicle, characterized in that: The method comprises: Obtain the user's physiological monitoring data, driving behavior data, and vehicle environmental perception data within the current cycle; Performing spatiotemporal alignment and noise filtering on the physiological monitoring data, the driving behavior data, and the environmental perception data, and respectively calculating the physiological abnormality index, the operation deviation, and the environmental risk factor of the current cycle; Inputting the physiological abnormality index, operation deviation, and environmental risk coefficient of the current cycle into a risk prediction model fused with an attention mechanism to obtain a dynamic risk value for the next cycle, wherein the attention mechanism is used to dynamically assign global attention weights to the physiological abnormality index, the operation deviation, and the environmental risk coefficient based on the risk evolution relationship of historical cycles; Correcting the dynamic risk value of the next period according to a risk time decay factor, and matching the corrected dynamic risk value with a target risk level within a preset risk level, wherein the risk time decay factor is determined by combining the dynamic risk value of the historical period and the real-time status parameters of the vehicle; Determine a dynamic permission adjustment strategy corresponding to the target risk level using risk-permission linkage rules, wherein the dynamic permission adjustment strategy includes hierarchical authorization thresholds and real-time response rules for functional permissions; The dynamic permission adjustment strategy is verified in real time based on the vehicle's driving scenario, a final list of authorized controllable functions is generated, and permission authorization operations are performed in the next cycle to facilitate user security control; Obtain the user's physiological monitoring data, driving behavior data, and vehicle environmental perception data for the current cycle, including: Collecting the user's breathing rate, micro-tremor amplitude, sitting posture stability, and the proportion of time the hand is in contact with the steering wheel during the current cycle to obtain the user's physiological monitoring data during the current cycle; Collecting the accelerator pedal pressure gradient, brake pedal pressure gradient, steering angular velocity, and vehicle acceleration during the current cycle to obtain driving behavior data of the user during the current cycle; Obstacle information, lane line status information, traffic flow information, and traffic event information are obtained from radar point cloud data of the vehicle's surrounding environment during the current cycle to obtain environmental perception data of the vehicle during the current cycle.

2. The method according to claim 1, characterized in that Performing spatiotemporal alignment and noise filtering on the physiological monitoring data, the driving behavior data, and the environmental perception data, and calculating the physiological abnormality index, operation deviation, and environmental risk factor of the current cycle, respectively, including: performing spatiotemporal alignment and noise filtering on the physiological monitoring data, the driving behavior data, and the environmental perception data to obtain processed physiological monitoring data, driving behavior data, and environmental perception data; Calculating a respiratory rate abnormality score, a microtremor amplitude abnormality score, a sitting posture stability abnormality score, and a contact time abnormality score based on the processed physiological monitoring data, and determining the physiological abnormality index according to the respiratory rate abnormality score, the microtremor amplitude abnormality score, the sitting posture stability abnormality score, and the contact time abnormality score; Calculating a throttle pressure gradient deviation, a brake pressure gradient deviation, a steering angular velocity deviation, and a vehicle acceleration deviation based on the processed driving behavior data, and determining the operation deviation according to the throttle pressure gradient deviation, the brake pressure gradient deviation, the steering angular velocity deviation, and the vehicle acceleration deviation; Based on the processed environmental perception data, the obstacle hazard, lane line integrity, traffic flow hazard and event urgency are calculated respectively, and the environmental hazard coefficient is determined according to the obstacle hazard, the lane line integrity, the traffic flow hazard and the event urgency.

3. The method according to claim 1, characterized in that Before inputting the physiological abnormality index, operation deviation, and environmental risk coefficient of the current cycle into the risk prediction model of the fusion attention mechanism to obtain the dynamic risk value of the next cycle, the method further includes: Acquire physiological abnormality index, operation deviation, and environmental risk factor of a historical period; use the physiological abnormality index, operation deviation, and environmental risk factor of the historical period as training data, and mark the training data with risk values based on actual risk event records to obtain training samples; An LSTM-Transformer hybrid network model is constructed as the risk prediction model of the fused attention mechanism, wherein the LSTM layer is used to capture the risk evolution relationship between the physiological abnormality index, the operational deviation, and the environmental risk factor of the historical period and the risk level, and the Transformer layer is used to calculate the global attention weights of the physiological abnormality index, the operational deviation, and the environmental risk factor based on the risk evolution relationship; The risk prediction model integrating the attention mechanism is trained using the training samples.

4. The method according to claim 3, characterized in that The actual risk event record includes the event type, event occurrence stamp and severity; The training data is labeled with risk values according to actual risk event records to obtain training samples, including: Obtaining dynamic time windows corresponding to different event types; According to the dynamic time window, matching and associating the event type with the training data according to the occurrence event stamp; If the event type matched in the training data is a single event, marking the training data with a risk value based on the severity corresponding to the single event; If the event type matched and associated in the training data is a compound event, the training data is marked with a risk value according to the highest severity of the compound event.

5. The method according to claim 1, wherein The dynamic permission adjustment strategy is verified in real time based on the vehicle's driving scenario to generate a final list of authorized controllable functions, including: Determining the driving scenario based on the road type, traffic environment, and road conditions of the vehicle; Verifying the hierarchical authorization threshold and the real-time response rule based on the key safety indicators preset in the driving scenario, respectively, to obtain verification results; If the verification result is passed, the controllable function list is generated according to the dynamic permission adjustment strategy; If the verification result is failure, the dynamic permission adjustment policy is adaptively corrected, and the controllable function list is generated according to the corrected dynamic permission adjustment policy.

6. The method according to claim 1, characterized in that The method further comprises: Monitoring user satisfaction with the list of controllable functions during the next period, and calculating the incidence rate of risk events for vehicles during the next period; Calculating the security compatibility of the dynamic permission adjustment strategy with the target risk level based on the usage satisfaction and the risk event occurrence rate; If the security compatibility is lower than a preset threshold, an update prompt message of the risk-authority linkage rule is triggered.

7. An authority security control device for an artificial intelligence vehicle, characterized in that: The method for controlling authority security of an artificial intelligence vehicle according to any one of claims 1 to 6, wherein the device comprises: The first acquisition unit is used to acquire the user's physiological monitoring data, driving behavior data and vehicle's environmental perception data in the current cycle; a first calculation unit, configured to perform spatiotemporal alignment and noise filtering on the physiological monitoring data, the driving behavior data, and the environmental perception data, and respectively calculate a physiological abnormality index, an operation deviation, and an environmental risk factor of a current period; a prediction unit, configured to input the physiological abnormality index, operational deviation, and environmental risk coefficient of the current cycle into a risk prediction model fused with an attention mechanism to obtain a dynamic risk value for the next cycle, wherein the attention mechanism is configured to dynamically assign global attention weights to the physiological abnormality index, operational deviation, and environmental risk coefficient based on risk evolution relationships over historical cycles; a processing unit configured to modify the dynamic risk value of the next period according to a risk time decay factor, and match the modified dynamic risk value with a target risk level within a preset risk level, wherein the risk time decay factor is determined based on the dynamic risk value of the historical period and the real-time status parameters of the vehicle; a determination unit, configured to determine a dynamic permission adjustment strategy corresponding to the target risk level by using a risk-permission linkage rule, wherein the dynamic permission adjustment strategy includes a hierarchical authorization threshold for functional permissions and a real-time response rule; The authorization unit is used to verify the dynamic permission adjustment strategy in real time based on the driving scenario of the vehicle, generate a final authorized list of controllable functions, and perform permission authorization operations in the next cycle to facilitate user security control.

8. A storage medium, characterized in that: The storage medium includes a stored program, wherein when the program is running, the device where the storage medium is located is controlled to execute the permission security control method for an artificial intelligence vehicle as described in any one of claims 1 to 6.

9. A processor, characterized in that: The processor is used to run a program, wherein the program, when running, executes the authority security control method for an artificial intelligence vehicle as described in any one of claims 1 to 6.

Images