An automatic driving human-computer interaction dynamic regulation method and system

Abstract

This invention relates to the field of intelligent driving technology and provides a method and system for dynamic control of human-machine interaction in autonomous driving. The method includes: acquiring vehicle driving environment data, surrounding vehicle interaction data, traffic flow state data, environmental disturbance data, and multimodal state data before driver takeover; identifying the type of vehicle interaction scenario in which the vehicle is located; calculating a real-time comprehensive traffic flow risk value and a driver takeover readiness score; calculating a control authority transfer margin value; correcting the risk level index and driver state level index; determining a target matrix unit in a preset human-machine interaction strategy matrix; determining a target strategy parameter set from the target matrix unit; generating a continuous authority transfer curve based on the target strategy parameter set; and executing takeover reminders and control authority transfer according to the continuous authority transfer curve. This invention achieves a smooth and continuous transition between autonomous driving control and manual driving control.

Description

Technical Field

[0001] This invention relates to the field of intelligent driving technology, specifically to a method and system for dynamic control of human-machine interaction in autonomous driving. Background Technology

[0002] With the development of Level 3 autonomous vehicles, autonomous driving systems are able to undertake vehicle perception, decision-making, and control tasks under specific operating conditions. When the vehicle is in complex traffic flow, nearing the system's operating boundary, or experiencing sudden changes in the driving environment, it is necessary to issue a takeover warning to the driver and transfer vehicle control from the autonomous driving system to the driver.

[0003] Existing autonomous driving takeover technologies typically determine whether to trigger a takeover request based on factors such as environmental risk, collision time, driver status, or takeover capability, and provide alerts through visual, auditory, or tactile means. While these methods can achieve basic takeover warnings, they focus primarily on risk level classification or driver takeover capability assessment, rarely considering the continuous changes in manual and autonomous driving control permissions during the handover period. In continuous traffic flow scenarios, vehicle interactions such as following, lane changing, merging, or weaving cause takeover risks to change dynamically over time. If the permission transfer is too rapid, it may cause fluctuations in vehicle status; if the transfer is too slow, it may lead to a takeover delay.

[0004] Furthermore, existing solutions typically lack a closed-loop correction mechanism based on changes in traffic flow risk before and after takeover, vehicle status recovery information after takeover, and the execution results of permission handover. This makes it difficult to adaptively optimize takeover reminder parameters, permission handover duration, manual control permission opening slope, and autonomous driving control permission retention ratio in subsequent similar scenarios. Therefore, a dynamic control method for autonomous driving human-machine interaction is needed to address these issues. Summary of the Invention

[0005] In view of the shortcomings of the existing technology, the purpose of this invention is to provide a dynamic control method and system for human-machine interaction in autonomous driving, so as to solve the problems existing in the above-mentioned background technology.

[0006] This invention is implemented as follows: a dynamic control method for human-machine interaction in autonomous driving, the method comprising: Acquire vehicle driving environment data, surrounding vehicle interaction data, traffic flow status data, environmental disturbance data, and multimodal status data before driver takeover, and identify the type of vehicle interaction scenario in which the vehicle is located; Based on vehicle interaction scenario type, vehicle driving environment data, surrounding vehicle interaction data, traffic flow status data, and environmental disturbance data, calculate the real-time comprehensive traffic flow risk value; Calculate the driver takeover readiness score based on multimodal state data; The control authority transfer margin value is calculated based on the real-time integrated traffic flow risk value, the driver takeover readiness score, the rate of change of the real-time integrated traffic flow risk value, and the rate of change of the driver takeover readiness score. The risk level index and driver status index are corrected based on the control authority transfer margin value. Based on the corrected risk level index and driver status index, the target matrix unit is determined in the preset human-machine interaction strategy matrix. The target strategy parameter set is determined by the target matrix unit. The target strategy parameter set includes at least one of the following: takeover warning advance, warning intensity parameter, control authority handover start time, control authority handover duration, manual control authority opening slope, autonomous driving control authority retention ratio, and control authority handover termination condition. A continuous permission handover curve is generated based on the target strategy parameter group. Takeover reminders and control permission handover are executed according to the continuous permission handover curve, so that the manual driving control permission and the autonomous driving control permission change continuously within the preset handover period. After the transfer of control authority is completed, based on the changes in the real-time comprehensive traffic flow risk value before and after the takeover, the vehicle status recovery information after the takeover, and the execution results of the continuous authority transfer curve, at least one of the target strategy parameter group, target matrix unit parameter boundary, traffic flow risk calculation weight, and driver takeover readiness score weight in the human-machine interaction strategy matrix is ​​corrected in a closed loop.

[0007] As a further aspect of the present invention, the method is applied to the control authority handover process of L3-level autonomous vehicles in continuous traffic flow vehicle interaction scenarios. The continuous traffic flow vehicle interaction scenario is a vehicle-side control authority handover scenario in a non-signal-controlled continuous road segment where there is a continuous relative motion relationship between the vehicle and surrounding target interaction vehicles. The vehicle interaction scenario types include car-following scenarios, lane-changing scenarios, lane-merging scenarios, and weaving scenarios. The real-time comprehensive traffic flow risk value is used to characterize the degree of influence of the current vehicle interaction scenario on the continuous handover process of L3-level autonomous vehicle control authority, and serves as the input for the control authority handover margin value, the target matrix unit of the human-machine interaction strategy matrix, and the continuous authority handover curve parameters.

[0008] As a further aspect of the present invention, the step of calculating the real-time integrated traffic flow risk value specifically includes: Based on the type of vehicle interaction scenario, an adaptive weight for the scenario is determined between the corrected collision time MTTC and the post-intrusion time PET. When the vehicle interaction scenario type is a car-following scenario, the contribution weight of MTTC to micro vehicle interaction risk is set as the first dominant weight, and the contribution weight of PET to micro vehicle interaction risk is set as the first auxiliary weight. When the vehicle interaction scenario type is a lane changing scenario, a lane merging scenario, or a weaving scenario, the contribution weight of PET to micro vehicle interaction risk is set as the second dominant weight, and the contribution weight of MTTC to micro vehicle interaction risk is set as the second auxiliary weight. Based on the first dominant weight, the first auxiliary weight, the second dominant weight, and the second auxiliary weight, the risk mapping results of MTTC and PET are fused to obtain the micro-vehicle interaction risk. The micro-level vehicle interaction risk is incorporated as a component of the real-time comprehensive traffic flow risk value and used in the calculation of control authority handover margin, determination of target matrix units, and generation of continuous authority handover curve parameters.

[0009] As a further aspect of the present invention, the real-time integrated traffic flow risk value is obtained in the following manner: Based on the real-time traffic flow and dynamic capacity of a single lane in the target road segment, a dynamic saturation index is calculated; based on the dynamic saturation index, a macro-level traffic loss stability risk is calculated. By integrating macro-level traffic loss stabilization risk, micro-level vehicle interaction risk, and environmental disturbance correction terms, a real-time comprehensive traffic flow risk value is obtained. Among them, the weights of the micro-vehicle interaction risk, macro-traffic loss stability risk and environmental disturbance correction items are adjusted in a closed loop based on the changes in the real-time comprehensive traffic flow risk value before and after takeover, the vehicle status recovery information after takeover, and the execution results of the continuous authority transfer curve. This is used to adjust the control authority transfer margin value and continuous authority transfer curve parameters in subsequent similar vehicle interaction scenarios.

[0010] As a further aspect of the present invention, the control authority handover margin value is used to determine the initiation mode and continuous authority handover curve parameters for continuous control authority handover. When the real-time integrated traffic flow risk value reaches the dynamic handover trigger threshold and the control authority handover margin value meets the first handover condition, a first continuous authority handover curve is generated, and takeover reminder and control authority handover are executed according to the first continuous authority handover curve; When the real-time comprehensive traffic flow risk value does not reach the dynamic handover trigger threshold, but the control authority handover margin value meets the second handover condition, or the rate of change of the real-time comprehensive traffic flow risk value is greater than the preset risk growth threshold, the start time of control authority handover is determined in advance, and at least one of the following is adjusted: the takeover reminder advance, the manual control authority opening slope, and the autonomous driving control authority retention ratio. The dynamic handover trigger threshold is adjusted based on the change in real-time comprehensive traffic flow risk value before and after takeover, vehicle status recovery information after takeover, and the execution results of the continuous permission handover curve.

[0011] As a further aspect of the present invention, the driver takeover readiness score is obtained by weighted fusion calculation based on driver physiological characteristic data, operation response data, and human-vehicle-environment interaction perception data. The driver physiological characteristic data includes the frequency of eye closure, the rate of change in pupil diameter, the change in heart rate, and / or the frequency of yawning. The operation response data includes the brake pedal hovering state, pedal response delay, steering wheel grip state, steering wheel angle change, and / or steering angular velocity smoothness. The human-vehicle-environment interaction perception data includes the proportion of effective forward-looking time, the duration of gaze deviation, the gaze return time, and / or the state of attention distribution. The weights of various data in the driver takeover readiness score are dynamically adjusted according to the vehicle interaction scenario type, the real-time comprehensive traffic flow risk value, the change in the real-time comprehensive traffic flow risk value before and after takeover, the vehicle state recovery information after takeover, and the execution results of the continuous permission transfer curve. Among them, when the execution result of the continuous permission handover curve is marked as a handover lag sample, the weights of the line-of-sight return time, line-of-sight deviation duration and / or pedal response delay are increased; When the execution result of the continuous permission handover curve is marked as a sample with insufficient risk release or a sample with too fast handover, increase the weight of operation response data and human-vehicle-environment interaction perception data in the driver takeover readiness score, and adjust at least one of the following in subsequent similar scenarios: the slope of manual control permission opening, the proportion of autonomous driving control permission retention, and the duration of control permission handover.

[0012] As a further aspect of the present invention, the control authority transfer margin value is used to dynamically adjust the target strategy parameter group; When the control authority handover margin value indicates an increased risk of takeover, increase the takeover warning advance and / or the intensity of multimodal warnings, reduce the rate of opening manual control authority and / or increase the proportion of autonomous driving control authority retention; When the control authority handover margin value indicates an improvement in the driver's readiness to take over, the rate of opening manual control authority and / or the proportion of autonomous driving control authority retention should be increased; A continuous permission handover curve is generated based on the adjusted target strategy parameter group, so that the manual driving control permission and the autonomous driving control permission change continuously within the preset handover time period. After the control authority is transferred, the target strategy parameter group and / or human-machine interaction strategy matrix are closed-loop modified based on the real-time changes in comprehensive traffic flow risk before and after the takeover, the vehicle status recovery information after the takeover, and the execution results of the continuous authority transfer curve, so as to update the control authority transfer strategy in subsequent similar vehicle interaction scenarios.

[0013] As a further aspect of the present invention, the human-computer interaction strategy matrix is ​​updated in a closed loop based on the takeover process samples; The takeover process sample includes vehicle interaction scenario type, real-time comprehensive traffic flow risk value, driver takeover readiness score, control authority handover margin value, target strategy parameter group, continuous authority handover curve, vehicle status recovery information after takeover, and changes in real-time comprehensive traffic flow risk value before and after takeover. Based on the changes in real-time comprehensive traffic flow risk values ​​before and after takeover, vehicle status recovery information after takeover, and the execution results of continuous permission handover curves, the target strategy parameter group is labeled with its effects to obtain samples of sufficient risk release, insufficient risk release, excessively fast handover, or delayed handover. When the target strategy parameter group is marked as a fully released risk sample, the calling priority of the target strategy parameter group is increased under the same or similar vehicle interaction scenarios, risk level index and driver status level index. When the target strategy parameter group is labeled as a sample with insufficient risk release, a sample with too fast handover, or a sample with delayed handover, at least one of the following parameters in the target strategy parameter group is modified: takeover warning advance amount, visual cue intensity, auditory cue intensity, tactile cue intensity, control authority handover duration, manual control authority opening slope, and autonomous driving control authority retention ratio. The modified strategy parameter set is written into the corresponding matrix unit in the human-computer interaction strategy matrix, or used as a candidate strategy parameter set for the corresponding matrix unit to determine the target strategy parameter set in subsequent similar takeover scenarios.

[0014] As a further aspect of the present invention, the vehicle state recovery information after takeover includes at least one of the following: vehicle lateral offset, lateral acceleration, yaw rate, longitudinal acceleration, vehicle speed fluctuation, lane keeping deviation, and change in safety margin with the target vehicle. The change in real-time comprehensive traffic flow risk value before and after the takeover is determined based on the real-time comprehensive traffic flow risk value at the start of the takeover and the real-time comprehensive traffic flow risk value within the preset evaluation time window after the takeover is completed. When the change in the real-time comprehensive traffic flow risk value indicates a decrease in risk, and the vehicle status recovery information after the takeover meets the preset stable recovery conditions, the corresponding continuous permission transfer curve is marked as a valid transfer curve. When the change in the real-time comprehensive traffic flow risk value indicates that the risk has not decreased, or when the vehicle status recovery information after the takeover does not meet the preset stable recovery conditions, the corresponding continuous authority transfer curve will be marked as a transfer curve to be corrected. For the handover curve to be corrected, at least one of the following should be corrected based on the reasons for the failure of risk reduction and the abnormal recovery type of vehicle status: the lead time for takeover warning, the duration of control authority handover, the slope of manual control authority opening, and the proportion of autonomous driving control authority retention.

[0015] As a further aspect of the present invention, the closed-loop update of the human-computer interaction strategy matrix includes a parameter boundary correction process, the specific steps of which are as follows: Based on multiple takeover process samples, the execution results of continuous permission transfer curves under the same or similar vehicle interaction scenarios, risk level index and driver status level index were statistically analyzed. Based on the execution results of the continuous permission handover curve, determine the takeover warning advance boundary, control permission handover duration boundary, manual control permission opening slope boundary, and autonomous driving control permission retention ratio boundary in the corresponding target matrix unit. When multiple consecutive takeover process samples are marked as samples with insufficient risk release, the lower limit of the takeover warning advance should be increased, the upper limit of the control authority transfer duration should be extended, or the lower limit of the autonomous driving control authority retention ratio should be increased. When multiple consecutive takeover process samples are marked as samples of too rapid handover, the upper limit of the slope for opening manual control permissions should be reduced, or the lower limit of the retention ratio of autonomous driving control permissions should be increased. When multiple consecutive takeover process samples are marked as delayed handover samples, increase the lower limit of the opening slope of manual control permissions, or decrease the upper limit of the retention ratio of autonomous driving control permissions. The corrected parameter boundaries are used as constraints on the corresponding target matrix units in the human-computer interaction strategy matrix to limit the range of values ​​for the target strategy parameter group in subsequent similar takeover scenarios.

[0016] Another objective of this invention is to provide an autonomous driving human-machine interaction dynamic control system, the system comprising: The vehicle environment perception unit is used to collect information on traffic participants around the vehicle, road environment information, vehicle interaction information, and traffic flow status information, and to provide basic data support for subsequent risk calculation. The driver status acquisition unit is used to collect the driver's physiological characteristics, operational response data and visual behavior data before taking over, in order to characterize the driver's current takeover readiness status. The risk and readiness assessment unit is used to calculate the real-time comprehensive traffic flow risk value and the driver takeover readiness score based on the data from the vehicle environment perception unit and the driver status acquisition unit, respectively, to characterize the traffic environment risk level and the driver's response capability. The control authority transfer margin calculation unit is used to combine the above risk values ​​and readiness scores to calculate the safety margin index during the control transfer process, which is used to characterize whether the current system meets the control transfer conditions. The strategy matrix decision unit is used to correct the risk level index and driver status level index according to the control authority transfer margin value, and to determine the corresponding target strategy parameter group in the human-machine interaction strategy matrix. The continuous authority handover curve generation unit is used to generate a continuous change curve of control authority in the time dimension based on the target policy parameter group, so as to achieve a smooth transition between autonomous driving control authority and manual driving control authority. The human-computer interaction execution unit is used to execute visual cues, auditory cues, and tactile feedback human-computer interaction behaviors according to the target strategy parameter set; The vehicle control execution unit is used to dynamically switch vehicle control between the autonomous driving system and the driver based on the continuous authority transfer curve; The closed-loop correction unit is used to update the parameter combination and control handover strategy in the strategy matrix based on the execution results of the takeover process, so as to improve the control performance and safety in subsequent similar scenarios.

[0017] Another objective of this invention is to provide an electronic device, including a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, it implements the aforementioned dynamic control method for human-machine interaction in autonomous driving.

[0018] Another objective of this invention is to provide a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the aforementioned dynamic control method for human-machine interaction in autonomous driving is implemented.

[0019] Compared with the prior art, the beneficial effects of the present invention are: This invention achieves a smooth and continuous transition between autonomous driving control and manual driving control by constructing a continuous control transfer mechanism based on the coupling of traffic flow risk and driver takeover capability, effectively avoiding the control abruptness problem caused by traditional threshold-based takeover. It improves the comprehensiveness and reliability of driver state judgment through multimodal driver takeover readiness assessment, and realizes dynamic matching of risk and driver capability through a control authority transfer margin mechanism. In addition, it introduces a closed-loop correction mechanism based on takeover result feedback, enabling the system to have continuous optimization and adaptive adjustment capabilities, thereby improving the safety and stability of the overall system in complex traffic environments. Attached Figure Description

[0020] Figure 1 This is a flowchart of a dynamic control method for human-machine interaction in autonomous driving.

[0021] Figure 2 This is a schematic diagram of a continuous traffic flow vehicle interaction scenario in a dynamic control method for human-machine interaction in autonomous driving.

[0022] Figure 3 This is a flowchart for determining the real-time comprehensive traffic flow risk value in a dynamic control method for human-machine interaction in autonomous driving. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0024] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0025] like Figure 1 As shown in the figure, this embodiment of the invention provides a dynamic control method for human-machine interaction in autonomous driving, the method comprising the following steps: S1 acquires vehicle driving environment data, surrounding vehicle interaction data, traffic flow status data, environmental disturbance data, and multimodal status data before driver takeover, identifying the type of vehicle interaction scenario the vehicle is in. Step S1 is used to acquire multi-source status information during vehicle operation and to identify and classify the traffic interaction scenario in which the vehicle is located. Specifically, the external environment perception unit collects information on traffic participants around the vehicle, including the positions of vehicles in front and adjacent lanes, relative speeds, lane structure information, and traffic flow density information; the onboard communication module acquires road traffic flow status data and road environmental disturbance data.

[0026] In this embodiment of the invention, the vehicle's speed ranges from 60 to 90 km / h, the longitudinal distance between vehicles ranges from 20 to 50 m, and the relative speed ranges from -5 to 10 m / s. Simultaneously, a driver state acquisition unit acquires multimodal driver state information, including physiological characteristic data, operational response data, and visual behavior data. Physiological characteristics include heart rate variability (HRV) and eye-closing frequency; operational responses include steering wheel control smoothness and braking response delay; and visual behavior includes duration of gaze deviation and return-to-center time. Furthermore, a scene recognition model categorizes vehicle intersection scenarios into at least one of the following, lane-changing, lane-merging, or weaving scenarios.

[0027] S2 calculates the real-time comprehensive traffic flow risk value based on vehicle interaction scenario type, vehicle driving environment data, surrounding vehicle interaction data, traffic flow status data, and environmental disturbance data. The real-time comprehensive traffic flow risk value is composed of micro-level vehicle interaction risk, macro-level traffic flow stability risk, and an environmental disturbance correction term, expressed as follows: ,in For micro-level vehicle interaction risks, For macro traffic flow risks, This is an environmental disturbance correction term. These are weighting coefficients. .

[0028] S3. Based on multimodal state data, including physiological characteristic data, operational response data, and visual attention data, calculate the driver takeover readiness score.

[0029] S4 calculates the control authority transfer margin value based on the real-time comprehensive traffic flow risk value, driver takeover readiness score, rate of change of the real-time comprehensive traffic flow risk value, and rate of change of the driver takeover readiness score.

[0030] S5, correct the risk level index and driver status level index according to the control authority transfer margin value, and determine the target matrix unit in the preset human-machine interaction strategy matrix based on the corrected risk level index and driver status level index.

[0031] S6, the target strategy parameter set is determined by the target matrix unit. The target strategy parameter set includes at least one of the following: takeover warning advance amount, warning intensity parameter, control authority handover start time, control authority handover duration, manual control authority opening slope, autonomous driving control authority retention ratio, and control authority handover termination condition.

[0032] S7 generates a continuous permission handover curve based on the target strategy parameter group, and performs takeover reminders and control permission handover according to the continuous permission handover curve, so that the manual driving control permission and the automatic driving control permission change continuously within the preset handover period.

[0033] S8, after the control authority transfer is completed, based on the real-time changes in the comprehensive traffic flow risk value before and after the takeover, the vehicle status recovery information after the takeover, and the execution results of the continuous authority transfer curve, at least one of the following in the human-machine interaction strategy matrix—the target strategy parameter group, the target matrix unit parameter boundary, the traffic flow risk calculation weight, and the driver takeover readiness score weight—is closed-loop corrected. The closed-loop update of the human-machine interaction strategy matrix is ​​achieved through at least one of the following: a reinforcement learning model, a supervised learning model, a Bayesian optimization model, or a rule-based parameter correction model.

[0034] like Figure 2 As shown in the embodiment of the present invention, the method is applied to the control authority handover process of L3 autonomous vehicles in continuous traffic flow vehicle interaction scenarios. The continuous traffic flow vehicle interaction scenario is a vehicle-side control authority handover scenario in a non-signal-controlled continuous road segment where there is a continuous relative motion relationship between the vehicle and surrounding target interaction vehicles. The vehicle interaction scenario types include car-following scenario, lane-changing scenario, lane-merging scenario, and weaving scenario. The real-time comprehensive traffic flow risk value is used to characterize the degree of influence of the current vehicle interaction scenario on the continuous handover process of L3 autonomous vehicle control authority, and serves as the input for the control authority handover margin value, the target matrix unit of the human-machine interaction strategy matrix, and the continuous authority handover curve parameters.

[0035] like Figure 3 As shown in the preferred embodiment of the present invention, the step of calculating the real-time integrated traffic flow risk value specifically includes: determining a scenario-adaptive weight between the Modified Collision Time (MTTC) and the Post-Intrusion Time (PET) based on the vehicle interaction scenario type; when the vehicle interaction scenario type is a car-following scenario, setting the contribution weight of MTTC to the micro-vehicle interaction risk as the first dominant weight and the contribution weight of PET to the micro-vehicle interaction risk as the first auxiliary weight; when the vehicle interaction scenario type is a lane-changing scenario, a lane-merging scenario, or a weaving scenario, setting the contribution weight of PET to the micro-vehicle interaction risk as the second dominant weight and the contribution weight of MTTC to the micro-vehicle interaction risk as the second auxiliary weight; based on the first dominant weight, the first auxiliary weight, the second dominant weight, and the second auxiliary weight, fusing the risk mapping results of MTTC and PET to obtain the micro-vehicle interaction risk; using the micro-vehicle interaction risk as a component of the real-time integrated traffic flow risk value, and using it to participate in the calculation of the control authority handover margin value, the determination of the target matrix unit, and the generation of continuous authority handover curve parameters. For example, the weights for a car-following scenario are MTTC weight 0.7 and PET weight 0.3. Lane changing / merging / weaving scenarios are weighted as follows: MTTC weight 0.4, PET weight 0.6. In one embodiment, when MTTC≤2s, the microscopic risk increases rapidly; when PET≤2.5s, the lateral conflict risk increases significantly.

[0036] In a preferred embodiment of the present invention, the real-time comprehensive traffic flow risk value is obtained as follows: a dynamic saturation index is calculated based on the real-time traffic flow and dynamic capacity of a single lane in the target road segment; a macro-level traffic loss stability risk is calculated based on the dynamic saturation index; and the macro-level traffic loss stability risk, micro-level vehicle interaction risk, and environmental disturbance correction term are fused to obtain the real-time comprehensive traffic flow risk value. The weights of the micro-level vehicle interaction risk, macro-level traffic loss stability risk, and environmental disturbance correction term are adjusted in a closed-loop manner based on the change in the real-time comprehensive traffic flow risk value before and after takeover, vehicle status recovery information after takeover, and the execution results of the continuous authority transfer curve. This adjustment is used to adjust the control authority transfer margin value and continuous authority transfer curve parameters in subsequent similar vehicle interaction scenarios.

[0037] In this embodiment of the invention, the macro-risk weight is set under different traffic conditions, such as smooth traffic: Moderate congestion: High congestion: The environmental disturbance term is used to correct for the impact of the external environment on traffic risk, including changes in weather conditions, lighting conditions, and road surface adhesion coefficient. Environmental weights under different weather conditions. Different settings, such as sunny day Light rain Moderate rain / Low visibility at night Heavy rain / water accumulation / low-adhesion road surface When the road surface adhesion coefficient decreases, the weight of the environmental disturbance term is significantly increased to enhance the system's ability to respond to external uncertainties.

[0038] In this embodiment of the invention, when the vehicle is in the following typical scenario: in a car-following scenario, the vehicle speed is... For moderate traffic flow and light rain, the weighting is [value missing]. (Microscopic) (Macro) (Environment), thus obtaining the comprehensive traffic flow risk value. .

[0039] In a preferred embodiment of the present invention, the control authority handover margin value is used to determine the initiation mode and continuous control authority handover curve parameters for continuous handover. When the real-time comprehensive traffic flow risk value reaches the dynamic handover trigger threshold and the control authority handover margin value meets the first handover condition, a first continuous control authority handover curve is generated, and takeover reminder and control authority handover are executed according to the first continuous control authority handover curve. When the real-time comprehensive traffic flow risk value does not reach the dynamic handover trigger threshold, but the control authority handover margin value meets the second handover condition, or the rate of change of the real-time comprehensive traffic flow risk value is greater than the preset risk growth threshold, the start time of control authority handover is determined in advance, and at least one of the takeover reminder advance amount, the manual control authority opening slope, and the autonomous driving control authority retention ratio is adjusted. The dynamic handover trigger threshold is corrected based on the change in the real-time comprehensive traffic flow risk value before and after takeover, the vehicle status recovery information after takeover, and the execution result of the continuous control authority handover curve.

[0040] In a preferred embodiment of the present invention, the driver takeover readiness score is obtained by weighted fusion calculation based on driver physiological characteristic data, operation response data, and human-vehicle-environment interaction perception data. The driver physiological characteristic data includes eye closure frequency, pupil diameter change rate, heart rate change, and / or yawning frequency. The operation response data includes brake pedal hovering state, pedal response delay, steering wheel grip state, steering wheel angle change, and / or steering angular velocity smoothness. The human-vehicle-environment interaction perception data includes the percentage of effective forward-looking time, duration of gaze deviation, gaze return time, and / or attention distribution state. The weights of each data type in the driver takeover readiness score are determined according to the vehicle interaction scenario type and real-time comprehensive... The system dynamically corrects the risk values ​​of combined traffic flow, the changes in real-time combined traffic flow risk values ​​before and after takeover, the vehicle status recovery information after takeover, and the execution results of continuous authority handover curves. Specifically, when the execution results of continuous authority handover curves are marked as samples with delayed handover, the weights of gaze return time, gaze deviation duration, and / or pedal response delay are increased. When the execution results of continuous authority handover curves are marked as samples with insufficient risk release or excessively rapid handover, the weights of operation response data and human-vehicle-environment interaction perception data in the driver takeover readiness score are increased, and at least one of the following is adjusted in subsequent similar scenarios: the slope of manual control permission opening, the proportion of autonomous driving control permission retention, and the duration of control permission handover.

[0041] In this embodiment of the invention, the driver takeover readiness score is... This is used to characterize the driver's responsiveness to vehicle takeover tasks in the current traffic environment. Specifically, multimodal driver state data is acquired through a driver state acquisition unit. Physiological characteristic data includes heart rate variability (HRV), eye closure frequency, and fatigue index, with HRV ranging from 20 to 80 ms. Operational response data includes steering wheel correction frequency, brake pedal response delay, and steering angular velocity, with response delay ranging from 0.3 to 1.5 s. Visual attention data includes duration of gaze deviation and return time, with gaze deviation ranging from 0.5 to 4 s.

[0042] In this embodiment of the invention, the driver state vector is constructed after normalizing various features. The driver takeover readiness score was calculated based on a weighted fusion model. Among them, weight The score is dynamically adjusted based on driver fatigue level and traffic risk level. When the driver is in a state of high attention (view return time <1.5s), the takeover readiness score is usually greater than 0.6. When the driver is fatigued or distracted, the score drops to 0.2–0.4.

[0043] In a preferred embodiment of the present invention, the control authority handover margin value is used to dynamically adjust the target strategy parameter set: when the control authority handover margin value indicates an increased current takeover risk, the takeover warning advance amount and / or multimodal warning intensity are increased, the manual control authority opening rate is reduced, and / or the autonomous driving control authority retention ratio is increased; when the control authority handover margin value indicates improved driver takeover readiness, the manual control authority opening rate is increased, and / or the autonomous driving control authority retention ratio is reduced; a continuous authority handover curve is generated based on the adjusted target strategy parameter set, so that the manual driving control authority and the autonomous driving control authority change continuously within a preset handover time period; and after the control authority handover is completed, the target strategy parameter set and / or human-machine interaction strategy matrix are closed-loop corrected based on the real-time comprehensive traffic flow risk changes before and after takeover, the vehicle status recovery information after takeover, and the execution results of the continuous authority handover curve, so as to update the control authority handover strategy in subsequent similar vehicle interaction scenarios.

[0044] In this embodiment of the invention, the control authority transfer margin value This is used to characterize the dynamic safety redundancy relationship between traffic flow risk status and driver takeover capability. The control handover margin value is based on the real-time comprehensive traffic flow risk value. Driver takeover readiness score It is calculated together with its changing trend, and its expression is as follows: ;in For real-time comprehensive traffic flow risk values, Rate the driver's readiness to take over. For the change in risk, This represents the change in the driver's state. These are the weighting coefficients; the range of values ​​for the weighting parameters is as follows: ; ; ; Among them, when traffic risk is in a high-growth state ( ), increase parameters The weighting is adjusted to enhance the risk-driven role. Typical ranges for control transfer margin values ​​under different operating conditions are as follows: Normal driving state: Warning and takeover status: High-risk takeover status: .when When the threshold is lower than the preset threshold, the system enters the rapid transfer of control phase and triggers the continuous permission transfer curve generation mechanism.

[0045] In this embodiment of the invention, based on the control authority handover margin value The risk level index and driver status level index are revised, and target strategy units are determined in the strategy matrix. Specifically, this is based on the real-time integrated traffic flow risk value. Driver takeover readiness rating The system state is divided into multiple risk level regions based on the combination of relationships, and a human-computer interaction strategy matrix is ​​constructed as follows: The risk level can be divided into three levels: low risk, medium risk, and high risk; the driver's readiness level can be divided into three levels: high readiness, medium readiness, and low readiness, thus forming... Strategy matrix structure.

[0046] In this embodiment of the invention, when and When entering a high-risk, low-preparation zone, the corresponding strategy unit should prioritize the enhanced takeover alert strategy; when and When entering a low-risk, high-preparation zone, the corresponding strategy unit reduces the intensity of takeover intervention. Control transfer margin value. Used to dynamically adjust the policy matrix index, when When the risk level is reduced, the risk level index is automatically adjusted upwards, thus moving the system into a higher priority takeover strategy zone.

[0047] In a preferred embodiment of the present invention, the human-computer interaction strategy matrix is ​​updated in a closed loop based on takeover process samples. The takeover process samples include vehicle interaction scenario type, real-time comprehensive traffic flow risk value, driver takeover readiness score, control authority handover margin value, target strategy parameter group, continuous authority handover curve, vehicle status recovery information after takeover, and changes in real-time comprehensive traffic flow risk value before and after takeover. Based on the execution results of the changes in real-time comprehensive traffic flow risk value before and after takeover, vehicle status recovery information after takeover, and continuous authority handover curve, the target strategy parameter group is labeled with its effectiveness, resulting in samples with sufficient risk release, insufficient risk release, excessively fast handover, or delayed handover. When the target strategy parameter group is labeled as having sufficient risk release... When sufficient samples are available, the invocation priority of the target strategy parameter group is increased under the same or similar vehicle interaction scenarios, risk level index, and driver status level index. When the target strategy parameter group is marked as a sample with insufficient risk release, a sample with too fast handover, or a sample with delayed handover, at least one of the following parameters in the target strategy parameter group is modified: takeover warning advance, visual cue intensity, auditory cue intensity, tactile cue intensity, control authority handover duration, manual control authority opening slope, and autonomous driving control authority retention ratio. The modified strategy parameter group is written into the corresponding matrix unit in the human-machine interaction strategy matrix, or used as a candidate strategy parameter group for the corresponding matrix unit for determining the target strategy parameter group in subsequent similar takeover scenarios.

[0048] In a preferred embodiment of the present invention, the target strategy parameter set is used to control takeover behavior during human-computer interaction. The target strategy parameter set includes takeover warning advance, visual cue intensity, auditory cue intensity, tactile cue intensity, control authority handover start time, control authority handover duration, manual control authority opening rate, and autonomous driving control authority retention ratio. The visual cue intensity, auditory cue intensity, and tactile cue intensity in the target strategy parameter set are configured in a hierarchical combination manner; specifically, the visual cue intensity can be divided into three levels: weak cue, medium cue, and strong cue; the auditory cue intensity can be divided into three levels: cue tone, warning tone, and continuous alarm; and the tactile cue intensity can be divided into three levels: light vibration, medium vibration, and strong vibration.

[0049] In this embodiment of the invention, the three types of alert intensity are based on the real-time comprehensive traffic risk value. Control authority transfer margin value Perform a joint mapping, the relationship of which is as follows: when and When using weak visual cues + low-frequency audible cues + light tactile feedback; when When using moderate visual cues + intermittent warning sounds + moderate tactile vibrations; or At that time, it uses strong visual bright prompts + continuous alarm sound + strong vibration tactile feedback.

[0050] In this embodiment of the invention, visual, auditory, and tactile cues have a synergistic enhancement relationship. When the intensity of any one modal cue increases, the intensity of the other modal cues increases synchronously to improve the efficiency of driver takeover response. When the vehicle is in a high-risk, rapid takeover scenario, the system prioritizes increasing the response speed of auditory cues; when the driver's gaze deviates for more than 2.5 seconds, the weight of visual cues is increased; and when the driver's operation response delay exceeds 1 second, the intensity of tactile feedback is increased. The lead time for takeover alerts ranges from 2 to 6 seconds and is dynamically adjusted according to the risk level. When the lead time is increased to 5-6 seconds; when When the lead time is reduced to 2-3 seconds, the duration of control handover is 3-8 seconds. In high-risk scenarios, the handover time is 3-4 seconds to speed up the takeover response. In low-risk scenarios, the handover time is 6-8 seconds to achieve a smooth transition.

[0051] In this embodiment of the invention, the rate at which permissions are granted is manually controlled. Ratio of autonomous driving control rights retained Satisfying complementary relationships, such as ;in This represents the percentage of human control authority. When the system is in a high-risk state, the intensity of visual and auditory cues is increased, while the rate at which human control authority is granted is reduced. When the system is in a low-risk state and the driver is highly prepared, the intensity of cues is reduced and the rate at which human control authority is granted is increased, in order to achieve a smooth and continuous transfer of control authority.

[0052] In a preferred embodiment of the present invention, the vehicle state recovery information after takeover includes at least one of the following: vehicle lateral offset, lateral acceleration, yaw rate, longitudinal acceleration, vehicle speed fluctuation, lane keeping deviation, and change in safety margin with the target vehicle. The change in real-time comprehensive traffic flow risk value before and after takeover is determined based on the real-time comprehensive traffic flow risk value at the start of takeover and the real-time comprehensive traffic flow risk value within a preset evaluation time window after takeover. When the change in real-time comprehensive traffic flow risk value indicates a decrease in risk, and the vehicle state recovery information after takeover meets the preset stable recovery conditions, the corresponding continuous permission transfer curve is marked as a valid transfer curve. When the change in real-time comprehensive traffic flow risk value indicates no decrease in risk, or the vehicle state recovery information after takeover does not meet the preset stable recovery conditions, the corresponding continuous permission transfer curve is marked as a transfer curve to be corrected. For the transfer curve to be corrected, at least one of the following is corrected based on the reason for the lack of risk reduction and the abnormal type of vehicle state recovery: takeover warning advance, control permission transfer duration, manual control permission opening slope, and autonomous driving control permission retention ratio.

[0053] In this embodiment of the invention, the continuous authority transfer curve is based on the proportion of manual driving control rights. It indicates that its form of expression is as follows: ,in For the transfer rate coefficient, At the moment when control is balanced, and the stated It is not a fixed parameter, but rather determined by real-time integrated traffic flow risk values. Control authority transfer margin value It is obtained by modulating the intensity of multimodal interaction prompts together, especially in high-risk states. Pick Under medium risk conditions Pick Under low-risk conditions Pick At the same time, the control balance moment Transfer margin based on control authority Perform dynamic correction when Hand over in advance, when The handover is delayed; furthermore, the intensity of visual cues, auditory cues, and tactile feedback constitutes a multimodal interaction intensity vector. When the driver's attention decreases or the gaze deviates for more than 2.5 seconds, the visual and auditory intensities are increased to accelerate the handover process, while the interaction intensity is reduced to achieve a smooth transition when the risk is low and the driver is highly prepared. This allows the control of manual driving and autonomous driving to change continuously within a preset time window and complete a gradual handover.

[0054] In a preferred embodiment of the present invention, the closed-loop update of the human-machine interaction strategy matrix includes a parameter boundary correction process. The specific steps are as follows: based on multiple takeover process samples, statistically analyze the execution results of continuous permission handover curves under the same or similar vehicle interaction scenarios, risk level index, and driver status level index; based on the continuous permission handover curve execution results, determine the takeover warning advance boundary, control permission handover duration boundary, manual control permission opening slope boundary, and autonomous driving control permission retention ratio boundary in the corresponding target matrix unit; when multiple consecutive takeover process samples are marked as insufficient risk release samples, increase the takeover warning advance... The lower limit is set by extending the upper limit of the control authority handover duration or increasing the lower limit of the autonomous driving control authority retention ratio. When multiple consecutive takeover process samples are marked as samples with excessively fast handover, the upper limit of the human control authority opening slope is reduced or the lower limit of the autonomous driving control authority retention ratio is increased. When multiple consecutive takeover process samples are marked as samples with delayed handover, the lower limit of the human control authority opening slope is increased or the upper limit of the autonomous driving control authority retention ratio is reduced. The corrected parameter boundaries are used as constraints on the corresponding target matrix units in the human-machine interaction strategy matrix to limit the value range of the target strategy parameter group in subsequent similar takeover scenarios.

[0055] In this embodiment of the invention, the human-computer interaction strategy matrix and continuous authority transfer curve parameters are corrected in a closed loop based on the takeover execution results to achieve the system's adaptive optimization capability. Specifically, after each control transfer is completed, the system acquires information on state changes before and after the takeover, including real-time changes in comprehensive traffic flow risk. The driver state recovery information after takeover and the execution deviation of the continuous authority transfer curve are included. The driver state recovery information includes vehicle lateral offset, lateral acceleration, yaw rate, lane keeping deviation, and vehicle speed fluctuation. A lateral offset range of 0.2-0.8m is used as the stable recovery reference interval. And when the vehicle's state recovers to a stable threshold, this execution is marked as a fully risk-released sample. If a vehicle exhibits significant fluctuations, it is marked as a sample with insufficient risk release or an abnormal sample to be transferred, and the strategy parameters are updated based on the sample markings.

[0056] In this embodiment of the invention, for samples with insufficient risk release or abnormal handover, the lead time for takeover alerts is increased, the duration of control authority handover is extended, or the curve slope parameter is reduced. To slow down the rate of change of control, for samples where the handover is too rapid and driver takeover is unstable, the proportion of autonomous driving control retained is reduced while the intensity of visual and auditory cues is increased. For samples where risks are fully released, the priority of their corresponding strategies in the strategy matrix is ​​increased, thus prioritizing their use in similar traffic scenarios and driver states. Furthermore, the update process is implemented through reinforcement learning or rule correction mechanisms, that is, the strategy matrix parameters are iteratively updated based on the historical sample set, so that the strategy matrix is ​​continuously optimized in the dimensions of risk level index and driver state level index. This gradually converges to the optimal control strategy distribution, thereby achieving adaptive evolution of the human-machine interaction control strategy.

[0057] This invention also provides an autonomous driving human-machine interaction dynamic control system. The system includes a vehicle environment perception unit, a driver state acquisition unit, a risk and readiness assessment unit, a control authority handover margin calculation unit, a strategy matrix decision unit, a continuous authority handover curve generation unit, a human-machine interaction execution unit, a vehicle control execution unit, and a closed-loop correction unit. Each unit works together to complete the dynamic allocation and continuous handover of control authority between the autonomous driving system and the driver.

[0058] In this embodiment of the invention, the vehicle environment perception unit is used to collect information on traffic participants around the vehicle, road environment information, vehicle interaction information, and traffic flow status information, providing basic data support for subsequent risk calculation. The driver status acquisition unit is used to collect the driver's physiological characteristic data, operational response data, and visual behavior data before takeover, characterizing the driver's current takeover readiness state. The risk and readiness assessment unit is used to calculate the real-time comprehensive traffic flow risk value and the driver takeover readiness score based on the data from the vehicle environment perception unit and the driver status acquisition unit, respectively, to characterize the traffic environment risk level and the driver's response capability. The control authority transfer margin calculation unit is used to calculate the safety margin index during the control authority transfer process by combining the above risk value and readiness score, characterizing whether the current system meets the control authority transfer conditions. The strategy matrix decision unit is used to correct the risk level index and driver status level index according to the control authority transfer margin value, and determine the corresponding target strategy parameter set in the human-machine interaction strategy matrix. The continuous authority transfer curve generation unit is used to generate a continuous change curve of control authority in the time dimension based on the target strategy parameter set, to achieve a smooth transition between autonomous driving control authority and manual driving control authority. The human-machine interaction execution unit is used to execute human-machine interaction behaviors such as visual cues, auditory cues, and tactile feedback according to the target strategy parameter set. The vehicle control execution unit is used to realize the dynamic switching of vehicle control between the autonomous driving system and the driver according to the continuous permission transfer curve. The closed-loop correction unit is used to update the parameter combination and control transfer strategy in the strategy matrix according to the execution result of the takeover process, so as to improve the control performance and safety in subsequent similar scenarios.

[0059] This invention also provides an L3-level autonomous driving vehicle, comprising: external perception sensors for collecting traffic flow information, road environment information, and vehicle interaction information around the vehicle; driver monitoring sensors for collecting driver takeover readiness status data; a human-machine interaction actuator for executing visual, auditory, and / or tactile cues; a vehicle controller for determining a target strategy parameter set based on a real-time comprehensive traffic flow risk value, a driver takeover readiness score, and a control authority handover margin value, and generating a continuous authority handover curve based on the target strategy parameter set; and a vehicle control actuator for executing a continuous handover between manual driving control authority and autonomous driving control authority based on the continuous authority handover curve; wherein, the vehicle controller is further configured to perform closed-loop correction on the strategy parameter set and / or parameter boundaries in the human-machine interaction strategy matrix based on the change in the real-time comprehensive traffic flow risk value before and after takeover, the vehicle state recovery information after takeover, and the execution result of the continuous authority handover curve.

[0060] This invention also provides an electronic device, including a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, it implements the aforementioned dynamic control method for human-machine interaction in autonomous driving.

[0061] This invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the aforementioned dynamic control method for human-machine interaction in autonomous driving.

[0062] This invention also provides a computer program product, including a computer program or computer instructions, which, when executed by a processor, implement the aforementioned dynamic control method for human-machine interaction in autonomous driving.

[0063] It should be noted that the specific embodiments described in this invention are only used to explain the technical principles and implementation process of this invention, and do not constitute a limitation on the scope of protection of this invention. For those skilled in the art, without departing from the core idea of ​​this invention, equivalent substitutions or parameter adjustments can be made to the traffic flow risk modeling method, driver takeover readiness assessment method, control authority handover margin calculation method, and the functional form of continuous authority handover curves. For example, replacing MTTC or PET with other conflict time indicators, or resetting the weight allocation method of multimodal interaction intensity, all fall within the scope of protection of this invention.

[0064] Meanwhile, the human-computer interaction strategy matrix and closed-loop correction mechanism in this invention can be adaptively extended according to different vehicle platforms, different levels of autonomous driving and different traffic environments. Its core lies in realizing the continuity and adaptive optimization of control rights through the coupling relationship between traffic flow risk, driver status and control authority transfer process, rather than being limited to the parameter values ​​or implementation forms described in specific embodiments.

[0065] Furthermore, the present invention can be implemented in the form of software, hardware, or a combination of both. Its operating platform can be an in-vehicle controller, an embedded system, or a cloud-based collaborative computing platform, but none of these affect the implementation of the core method of the present invention.

[0066] It should be understood that although the steps in the flowcharts of the various embodiments of the present invention are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the various embodiments may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0067] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the disclosure in the specification and embodiments. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the claims.

Claims

1. A dynamic control method for human-machine interaction in autonomous driving, characterized in that, The method includes: Acquire vehicle driving environment data, surrounding vehicle interaction data, traffic flow status data, environmental disturbance data, and multimodal status data before driver takeover, and identify the type of vehicle interaction scenario in which the vehicle is located; Based on vehicle interaction scenario type, vehicle driving environment data, surrounding vehicle interaction data, traffic flow status data and environmental disturbance data, calculate real-time comprehensive traffic flow risk value; Calculate the driver takeover readiness score based on multimodal state data; The control authority transfer margin value is calculated based on the real-time integrated traffic flow risk value, the driver takeover readiness score, the rate of change of the real-time integrated traffic flow risk value, and the rate of change of the driver takeover readiness score. The risk level index and driver status index are corrected based on the control authority transfer margin value. Based on the corrected risk level index and driver status index, the target matrix unit is determined in the preset human-machine interaction strategy matrix. The target strategy parameter set is determined by the target matrix unit. The target strategy parameter set includes at least one of the following: takeover warning advance, warning intensity parameter, control authority handover start time, control authority handover duration, manual control authority opening slope, autonomous driving control authority retention ratio, and control authority handover termination condition. A continuous permission handover curve is generated based on the target strategy parameter group. Takeover reminders and control permission handover are executed according to the continuous permission handover curve, so that the manual driving control permission and the autonomous driving control permission change continuously within the preset handover period. After the transfer of control authority is completed, based on the changes in the real-time comprehensive traffic flow risk value before and after the takeover, the vehicle status recovery information after the takeover, and the execution results of the continuous authority transfer curve, at least one of the target strategy parameter group, target matrix unit parameter boundary, traffic flow risk calculation weight, and driver takeover readiness score weight in the human-machine interaction strategy matrix is ​​corrected in a closed loop.

2. The autonomous driving human-machine interaction dynamic control method according to claim 1, characterized in that, The method is applied to the control authority handover process of L3 autonomous vehicles in continuous traffic flow vehicle interaction driving scenarios. The continuous traffic flow vehicle interaction driving scenario is a vehicle-side control authority handover scenario in non-signal-controlled continuous road sections where there is a continuous relative motion relationship between the vehicle and surrounding target interaction vehicles. Vehicle interaction scenarios include following scenarios, lane changing scenarios, lane merging scenarios, and weaving scenarios; The real-time integrated traffic flow risk value is used to characterize the impact of the current vehicle interaction scenario on the continuous handover process of control authority of L3 autonomous vehicles, and serves as the input for the control authority handover margin value, the target matrix unit of the human-machine interaction strategy matrix, and the parameters of the continuous authority handover curve.

3. The autonomous driving human-machine interaction dynamic control method according to claim 1, characterized in that, The steps for calculating the real-time integrated traffic flow risk value specifically include: Based on the type of vehicle interaction scenario, an adaptive weight for the scenario is determined between the corrected collision time MTTC and the post-intrusion time PET. When the vehicle interaction scenario type is a car-following scenario, the contribution weight of MTTC to micro vehicle interaction risk is set as the first dominant weight, and the contribution weight of PET to micro vehicle interaction risk is set as the first auxiliary weight. When the vehicle interaction scenario type is a lane changing scenario, a lane merging scenario, or a weaving scenario, the contribution weight of PET to micro vehicle interaction risk is set as the second dominant weight, and the contribution weight of MTTC to micro vehicle interaction risk is set as the second auxiliary weight. Based on the first dominant weight, the first auxiliary weight, the second dominant weight, and the second auxiliary weight, the risk mapping results of MTTC and PET are fused to obtain the micro-vehicle interaction risk. The micro-level vehicle interaction risk is incorporated as a component of the real-time comprehensive traffic flow risk value and used in the calculation of control authority handover margin, determination of target matrix units, and generation of continuous authority handover curve parameters.

4. The autonomous driving human-machine interaction dynamic control method according to claim 3, characterized in that, The real-time integrated traffic flow risk value is obtained through the following methods: Based on the real-time traffic flow and dynamic capacity of a single lane in the target road segment, a dynamic saturation index is calculated; based on the dynamic saturation index, a macro-level traffic loss stability risk is calculated. By integrating macro-level traffic loss stabilization risk, micro-level vehicle interaction risk, and environmental disturbance correction terms, a real-time comprehensive traffic flow risk value is obtained. Among them, the weights of the micro-vehicle interaction risk, macro-traffic loss stability risk and environmental disturbance correction items are adjusted in a closed loop based on the changes in the real-time comprehensive traffic flow risk value before and after takeover, the vehicle status recovery information after takeover, and the execution results of the continuous authority transfer curve. This is used to adjust the control authority transfer margin value and continuous authority transfer curve parameters in subsequent similar vehicle interaction scenarios.

5. The dynamic control method for human-machine interaction in autonomous driving according to claim 1, characterized in that, The control authority handover margin value is used to determine the initiation method and continuous control authority handover curve parameters for continuous handover. When the real-time integrated traffic flow risk value reaches the dynamic handover trigger threshold and the control authority handover margin value meets the first handover condition, a first continuous authority handover curve is generated, and takeover reminder and control authority handover are executed according to the first continuous authority handover curve; When the real-time comprehensive traffic flow risk value does not reach the dynamic handover trigger threshold, but the control authority handover margin value meets the second handover condition, or the rate of change of the real-time comprehensive traffic flow risk value is greater than the preset risk growth threshold, the start time of control authority handover is determined in advance, and at least one of the following is adjusted: the takeover reminder advance, the manual control authority opening slope, and the autonomous driving control authority retention ratio. The dynamic handover trigger threshold is adjusted based on the change in real-time comprehensive traffic flow risk value before and after takeover, vehicle status recovery information after takeover, and the execution results of the continuous permission handover curve.

6. The method for dynamic control of human-machine interaction in autonomous driving according to claim 1, characterized in that, The driver takeover readiness score is calculated by weighted fusion of driver physiological characteristic data, operational response data, and human-vehicle-environment interaction perception data; driver Physiological data include eye closing frequency, pupil diameter change rate, heart rate change and / or yawning frequency; operational response data include brake pedal hovering status, pedal response delay, steering wheel grip status, steering wheel angle change and / or steering angular velocity smoothness. The human-vehicle-environment interaction perception data includes the effective forward-looking time ratio, the duration of line-of-sight deviation, the line-of-sight return time and / or the attention distribution status; the weights of various data in the driver takeover readiness score are dynamically adjusted based on the vehicle interaction scenario type, real-time comprehensive traffic flow risk value, the change in real-time comprehensive traffic flow risk value before and after takeover, vehicle status recovery information after takeover, and the execution results of the continuous permission transfer curve. Among them, when the execution result of the continuous permission handover curve is marked as a handover lag sample, the weights of the line-of-sight return time, line-of-sight deviation duration and / or pedal response delay are increased; When the execution result of the continuous permission handover curve is marked as a sample with insufficient risk release or a sample with too fast handover, increase the weight of operation response data and human-vehicle-environment interaction perception data in the driver takeover readiness score, and adjust at least one of the following in subsequent similar scenarios: the slope of manual control permission opening, the proportion of autonomous driving control permission retention, and the duration of control permission handover.

7. The method for dynamic control of human-machine interaction in autonomous driving according to claim 1, characterized in that, The control authority transfer margin value is used to dynamically adjust the target strategy parameter group; When the control authority handover margin value indicates an increased risk of takeover, increase the takeover warning advance and / or the intensity of multimodal warnings, reduce the rate of opening manual control authority and / or increase the proportion of autonomous driving control authority retention; When the control authority handover margin value indicates an improvement in the driver's readiness to take over, the rate of opening manual control authority and / or the proportion of autonomous driving control authority retention should be increased; A continuous permission handover curve is generated based on the adjusted target strategy parameter group, so that the manual driving control permission and the autonomous driving control permission change continuously within the preset handover time period. After the control authority is transferred, the target strategy parameter group and / or human-machine interaction strategy matrix are closed-loop modified based on the real-time changes in comprehensive traffic flow risk before and after the takeover, the vehicle status recovery information after the takeover, and the execution results of the continuous authority transfer curve, so as to update the control authority transfer strategy in subsequent similar vehicle interaction scenarios.

8. The method according to claim 7, characterized in that, The human-computer interaction strategy matrix is ​​updated in a closed loop based on the takeover process samples. The takeover process sample includes vehicle interaction scenario type, real-time comprehensive traffic flow risk value, driver takeover readiness score, control authority handover margin value, target strategy parameter group, continuous authority handover curve, vehicle status recovery information after takeover, and changes in real-time comprehensive traffic flow risk value before and after takeover. Based on the changes in real-time comprehensive traffic flow risk values ​​before and after takeover, vehicle status recovery information after takeover, and the execution results of continuous permission handover curves, the target strategy parameter group is labeled with its effects to obtain samples of sufficient risk release, insufficient risk release, excessively fast handover, or delayed handover. When the target strategy parameter group is marked as a fully released risk sample, the calling priority of the target strategy parameter group is increased under the same or similar vehicle interaction scenarios, risk level index and driver status level index. When the target strategy parameter group is labeled as a sample with insufficient risk release, a sample with too fast handover, or a sample with delayed handover, at least one of the following parameters in the target strategy parameter group is modified: takeover warning advance amount, visual cue intensity, auditory cue intensity, tactile cue intensity, control authority handover duration, manual control authority opening slope, and autonomous driving control authority retention ratio. The modified strategy parameter set is written into the corresponding matrix unit in the human-computer interaction strategy matrix, or used as a candidate strategy parameter set for the corresponding matrix unit to determine the target strategy parameter set in subsequent similar takeover scenarios.

9. The method according to claim 8, characterized in that, The vehicle status recovery information after takeover includes at least one of the following: vehicle lateral offset, lateral acceleration, yaw rate, longitudinal acceleration, vehicle speed fluctuation, lane keeping deviation, and change in safety margin with the target vehicle. The change in real-time comprehensive traffic flow risk value before and after the takeover is determined based on the real-time comprehensive traffic flow risk value at the start of the takeover and the real-time comprehensive traffic flow risk value within the preset evaluation time window after the takeover is completed. When the change in the real-time comprehensive traffic flow risk value indicates a decrease in risk, and the vehicle status recovery information after the takeover meets the preset stable recovery conditions, the corresponding continuous permission transfer curve is marked as a valid transfer curve. When the change in the real-time comprehensive traffic flow risk value indicates that the risk has not decreased, or when the vehicle status recovery information after the takeover does not meet the preset stable recovery conditions, the corresponding continuous authority transfer curve will be marked as a transfer curve to be corrected. For the handover curve to be corrected, at least one of the following should be corrected based on the reasons for the failure of risk reduction and the abnormal recovery type of vehicle status: the lead time for takeover warning, the duration of control authority handover, the slope of manual control authority opening, and the proportion of autonomous driving control authority retention.

10. An autonomous driving human-machine interaction dynamic control system, used to execute the autonomous driving human-machine interaction dynamic control method according to any one of claims 1 to 9, characterized in that, The system includes: The vehicle environment perception unit is used to collect information on traffic participants around the vehicle, road environment information, vehicle interaction information, and traffic flow status information, and to provide basic data support for subsequent risk calculation. The driver status acquisition unit is used to collect the driver's physiological characteristics, operational response data and visual behavior data before taking over, in order to characterize the driver's current takeover readiness status. The risk and readiness assessment unit is used to calculate the real-time comprehensive traffic flow risk value and the driver takeover readiness score based on the data from the vehicle environment perception unit and the driver status acquisition unit, respectively, to characterize the traffic environment risk level and the driver's response capability. The control authority transfer margin calculation unit is used to combine the above risk values ​​and readiness scores to calculate the safety margin index during the control transfer process, which is used to characterize whether the current system meets the control transfer conditions. The strategy matrix decision unit is used to correct the risk level index and driver status level index according to the control authority transfer margin value, and to determine the corresponding target strategy parameter group in the human-machine interaction strategy matrix. The continuous authority handover curve generation unit is used to generate a continuous change curve of control authority in the time dimension based on the target policy parameter group, so as to achieve a smooth transition between autonomous driving control authority and manual driving control authority. The human-computer interaction execution unit is used to execute visual cues, auditory cues, and tactile feedback human-computer interaction behaviors according to the target strategy parameter set; The vehicle control execution unit is used to dynamically switch vehicle control between the autonomous driving system and the driver based on the continuous authority transfer curve; The closed-loop correction unit is used to update the parameter combination and control handover strategy in the strategy matrix based on the execution results of the takeover process, so as to improve the control performance and safety in subsequent similar scenarios.

Images