Human-Machine Shared Steering and Collision Avoidance Control System and Method Based on Extension Decision

By constructing a two-dimensional extension set based on extension decision-making and integrating driver operation behavior and physiological fatigue characteristics, an adaptive and smooth allocation of control between the driver and the autonomous driving system in human-machine co-driving steering and collision avoidance control is achieved. This solves the problem of incomplete risk assessment in existing technologies and improves collision avoidance safety and human-machine interaction experience.

CN123104704APending Publication Date: 2026-11-17HEFEI UNIV
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN202611622016.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-20
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies in human-machine co-driving steering and collision avoidance control fail to effectively integrate the driver's steering behavior characteristics and EEG physiological fatigue characteristics, resulting in incomplete risk assessment. Furthermore, the allocation of control power lacks quantification and smoothness, affecting collision avoidance safety and human-machine interaction experience.

Method used

By constructing a two-dimensional extension set based on extension decision, and using driver steering and collision avoidance maneuvering errors and fatigue levels as features, the correlation function value is calculated to achieve adaptive and smooth allocation of control rights between the driver and the autonomous driving system. Combined with a first-order low-pass filter for dynamic smoothing processing, the continuous transition of control rights is ensured.

Benefits of technology

It enables multi-dimensional quantitative assessment of driver steering and collision avoidance safety risks, ensures the mathematical theoretical basis and real-time quantitative calculation capability of control allocation, improves collision avoidance safety and human-machine interaction comfort, avoids abrupt changes in control switching, and improves driver operating comfort and system acceptance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123104704A_ABST
    Figure CN123104704A_ABST
Patent Text Reader

Abstract

This invention provides a human-machine shared steering collision avoidance control system and method based on extension decision-making, relating to the field of intelligent vehicle control technology. The invention acquires driver steering wheel operation parameters, EEG parameters, and road parameters to calculate driver steering collision avoidance operation errors, driver fatigue level, and the first target front wheel angle. A two-dimensional extension set is constructed with operation error as the primary feature and fatigue level as the secondary feature, divided into classical, extension, and non-domain regions. The correlation function value is calculated based on the extension distance between the feature state points and the optimal point. The driver's control weight is determined according to the domain to which the correlation function value belongs. Finally, the front wheel angle of the driver-controlled vehicle and the first target front wheel angle are weighted and summed according to the control weight to output the second target front wheel angle. This invention achieves multi-dimensional fusion evaluation of driver steering operation behavior characteristics and EEG fatigue characteristics, as well as smooth adaptive allocation of control weight, balancing collision avoidance safety and human-machine interaction friendliness.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of intelligent vehicle control technology, specifically to a human-machine shared steering and collision avoidance control system and method based on extension decision-making. Background Technology

[0002] As intelligent driving technology evolves towards Level 2-3 human-machine co-driving stages, human-machine shared control in steering and collision avoidance scenarios has become a research hotspot in the field of intelligent vehicles. In emergency collision avoidance situations, the autonomous driving system and the driver jointly control the vehicle's steering. The timing and proportion of control allocation directly determine the vehicle's collision avoidance safety and the friendliness of human-machine interaction: premature system intervention can interfere with the driver's normal operation and reduce the driving experience; delayed intervention may miss the collision avoidance opportunity and cause a collision. Therefore, how to accurately assess the driver's steering and collision avoidance ability and achieve adaptive and smooth transition of control is a key problem that urgently needs to be solved in human-machine shared steering and collision avoidance control.

[0003] In the prior art, CN108732921B discloses a lateral extensional anti-aiming switching control method for autonomous vehicles. This method constructs an extensional set based on the vehicle's lateral position deviation and the curvature of the road ahead, dividing it into classical, extensional, and non-domain domains and corresponding to different control strategies, effectively improving lateral tracking stability in pure autonomous driving mode. However, this scheme only uses the vehicle's own motion state as the extensional feature input, completely neglecting the driver's maneuvering behavior and physiological state. Therefore, it is unsuitable for human-machine shared driving scenarios and cannot dynamically adjust control permissions based on the driver's real-time capabilities. To balance driver status and vehicle control performance, existing technologies have begun to introduce driver status assessment mechanisms to dynamically allocate permissions. For example, Chinese patent CN122143947A discloses a human-machine co-driving and shared control method that monitors driver status by collecting EEG signals and dynamically allocates driving permissions based on human-machine conflict index and safety benefits, thus improving the safety of human-machine co-driving to some extent. Another example is Chinese patent CN116449815A, which discloses a human-machine collaborative path tracking control method that considers driver status. It uses deep learning to identify driver status and adjust collaborative weights, achieving a balance between path tracking accuracy and driver experience. However, in these solutions, permission allocation is mostly based on conflict index, deep learning models, or fuzzy rules, lacking quantitative and multi-dimensional fusion assessment of driver steering and collision avoidance risks. In particular, it fails to integrate and model the driver's steering behavior characteristics with physiological fatigue characteristics, resulting in insufficient accuracy and smoothness in risk assessment, easily leading to abrupt control switching and poor human-machine interaction experience.

[0004] In summary, the existing technology lacks a human-machine shared steering and collision avoidance control scheme that can simultaneously integrate driver steering behavior characteristics and EEG fatigue characteristics, quantitatively assess steering and collision avoidance safety risks based on extension decision-making, and achieve adaptive and smooth allocation of control rights.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide a human-machine shared steering and collision avoidance control system and method based on extension decision-making, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: The human-machine shared steering and collision avoidance control system and method based on extension decision-making includes the following steps: The system acquires the driver's steering wheel operation parameters, EEG parameters, and road parameters at the current sampling time; calculates the driver's steering collision avoidance operation error based on the time-domain variation characteristics of the steering wheel operation parameters, determines the driver's fatigue level based on the driver's EEG parameter values, and determines the first target front wheel steering angle required for the autonomous driving system to control the vehicle based on the road parameters. A two-dimensional extension set is constructed with the driver's steering and collision avoidance operation error as the main feature and the driver's fatigue level as the secondary feature. The set is divided into classical domain, extension domain and non-domain. The ideal driving state is taken as the optimal point in the two-dimensional extension set. The correlation function value is calculated based on the positional relationship between the feature state point formed by the current main feature and secondary feature and the optimal point. The control weight of the driver to control the vehicle is determined according to the domain to which the correlation function value belongs. When the correlation function value belongs to the classical domain, the control weight corresponds to the driver having complete control of the vehicle. When the correlation function value belongs to the non-domain, the control weight corresponds to the autonomous driving system having complete control of the vehicle. When the correlation function value belongs to the extension domain, the control weight changes adaptively and continuously with the correlation function value, and transitions continuously at the boundaries between the classical domain and the extension domain, and between the extension domain and the non-domain. Based on the control weight, the driver-controlled front wheel angle is weighted and summed with the first target front wheel angle to obtain the second target front wheel angle under human-machine shared control, and then the result is output for execution.

[0008] Furthermore, the step of calculating the first target front wheel steering angle based on road parameters specifically involves: generating a desired lane-changing collision avoidance trajectory based on the longitudinal coordinates, lateral coordinates, and heading angle in the road parameters, and determining the desired lateral position on the desired lane-changing collision avoidance trajectory based on the vehicle state at the current sampling time. The difference between the desired lateral position and the actual lateral position of the vehicle is used as the lateral displacement deviation. A closed-loop correction is performed on the lateral displacement deviation to obtain the first target front wheel steering angle.

[0009] Furthermore, driver steering collision avoidance errors are determined based on the time-domain variation characteristics of the driver's steering wheel operation parameters, including: A sliding time window of a preset length is constructed with the current sampling time as the endpoint. The steering wheel angle and the differential of the steering wheel angle input by the driver within the sliding time window are extracted, as well as the first target front wheel angle and the differential of the first target front wheel angle at each sampling time within the same sliding time window. The steering wheel angle and the differential of the steering wheel angle are converted to the front wheel angle side according to the steering transmission ratio. The absolute value of the angle deviation between the converted steering wheel angle and the first target front wheel angle, as well as the absolute value of the differential of the angle deviation between the converted steering wheel angle and the first target front wheel angle, are calculated at each sampling time. The absolute values ​​of the angle deviation and the absolute values ​​of the differential of the angle are accumulated within the sliding time window to obtain the cumulative amount of angle deviation and the cumulative amount of differential of the angle. The cumulative differential deviation of the rotation angle is weighted according to a preset weight and combined with the cumulative deviation of the rotation angle. Then, it is mapped to a unified numerical range through a bounded monotonic compression function to obtain the original manipulation error at the current sampling time. The smoothing weight is adaptively determined based on the original control error at the current sampling time. The larger the original control error, the higher the proportion of the current sampling time value in the smoothing weight. The smaller the original control error, the higher the proportion of the historical value in the smoothing weight. The driver's steering and collision avoidance control error after smoothing at the previous sampling time and the original control error at the current sampling time are weighted and combined according to the smoothing weight to obtain the driver's steering and collision avoidance control error after smoothing at the current sampling time, which is used as the main feature quantity.

[0010] Furthermore, the driver's steering and collision avoidance maneuvering error was calculated. Then, the dynamic smoothing filter factor is determined based on the manipulation error at the current sampling time. A first-order low-pass filter is used for smoothing: in, : No. The driver's steering and collision avoidance error after smoothing at the sampling time is dimensionless and serves as the main feature of the extension decision model. For the first The dynamic smoothing filter factor at the sampling time is dimensionless and ranges from (0,1]. It is based on the steering and collision avoidance maneuver error at the current time. Adaptive determination; No. The original manipulation error assessment value at the sampling time is dimensionless; Smooth steering avoidance error As the principal characteristic of the two-dimensional extension set.

[0011] Furthermore, the calculation of driver fatigue based on EEG parameters specifically involves: based on EEG values... Fatigue levels are classified as follows: values ​​in [40, 100) indicate no fatigue, values ​​in [30, 40) indicate mild fatigue, and values ​​in (0, 30) indicate severe fatigue. The brainwave values ​​were normalized using the following formula to obtain the driver fatigue level with values ​​ranging from [0,1]. : in, For the first The driver fatigue level after normalization at the sampling time is dimensionless and ranges from 0 to 1; the larger the value, the higher the fatigue level, and it serves as a secondary feature of the extension decision model. For the first The numerical values ​​of EEG characteristics at the sampling time are dimensionless.

[0012] Furthermore, the division of the two-dimensional extension set is specifically as follows: a two-dimensional feature plane is established with the principal feature quantity as the horizontal axis and the secondary feature quantity as the vertical axis. The two-dimensional extension set is divided into classical domain, extension domain and non-domain in the two-dimensional feature plane with the optimal point as the reference. Connect the optimal point with the current feature state point to form a feature ray. The feature ray intersects the boundaries of the classical domain and the extension domain, respectively. Calculate the correlation function value based on the relative position of the current feature state point on the feature ray and each boundary intersection point: When the current feature state point is within the boundary of the classical domain, the correlation function value belongs to the classical domain; When the current feature state point is located between the boundary of the classical domain and the boundary of the extension domain, the correlation function value belongs to the extension domain; When the current feature state point is located outside the boundary of the extension domain, the correlation function value belongs to the non-domain.

[0013] Furthermore, the calculation of the correlation function value specifically involves: taking the origin (0,0) of the two-dimensional extension set as the optimal point of the characteristic state, for any characteristic state point... Connect the origin O and point The ray intersects the boundary between the classical domain and the extensional domain at... The boundaries of points, intersections, extended regions, and non-regions are at... Point; calculate the point using the following formula. to the interval Extendable distance: in, Characteristic state points to the interval Extendable distance, Let be the current feature state point in the two-dimensional extension feature space. The current feature state point on the one-dimensional number axis Boundary points of classical domain The distance between them (i.e., the absolute value of the difference between the coordinates of the two points on a one-dimensional number axis). For a ray in a two-dimensional feature space Direction, from the origin Time The interval corresponds to the range of the classical domain; The correlation function characterizing the driver's steering collision avoidance safety risk is determined based on the extended distance, and the reference formula is as follows: in, The larger the correlation function value, the lower the safety risk for the driver when steering to avoid a collision; the smaller the correlation function value, the higher the safety risk for the driver when steering to avoid a collision.

[0014] Furthermore, when the correlation function value belongs to the extensionable domain, the driver's control weights change continuously with the correlation function value, including: The driver control weight is determined by an S-shaped mapping: as the correlation function value changes from the classical domain side to the non-domain side, the driver control weight smoothly decreases from the end where the driver has complete control of the vehicle to the end where the autonomous driving system has complete control of the vehicle. The slope and center position of the S-shaped mapping are determined by a preset adjustment coefficient, and the sum of the driver control weights and the control weights of the autonomous driving system output by the mapping remains a fixed total.

[0015] Furthermore, the driver's steering wheel angle is converted according to the steering transmission ratio to obtain the front wheel angle controlled by the driver; the driver's control weight is used as the weight of the driver's side front wheel angle, and the difference between the fixed total and the driver's control weight is used as the weight of the first target front wheel angle. The driver's side front wheel angle and the first target front wheel angle are weighted and summed to obtain the second target front wheel angle, which is then output to the vehicle steering actuator.

[0016] This application also includes a human-machine shared steering and collision avoidance control system based on extension decision-making, applicable to the above method, including: Parameter acquisition module: acquires the driver's steering wheel operation parameters, EEG parameters and road parameters at the current sampling time; calculates the driver's steering collision avoidance operation error based on the time domain variation characteristics of the steering wheel operation parameters; determines the driver's fatigue level based on the driver's EEG parameter values; and determines the first target front wheel steering angle required for the autonomous driving system to control the vehicle based on the road parameters. Feature evaluation module: Construct a two-dimensional extension set with the driver's steering and collision avoidance operation error as the main feature and the driver's fatigue level as the secondary feature, and divide it into classical domain, extension domain and non-domain. The ideal driving state is taken as the optimal point in the two-dimensional extension set. The correlation function value is calculated based on the positional relationship between the feature state point formed by the current main feature and secondary feature and the optimal point. Extensional decision module: Determines the control weight of the driver controlling the vehicle based on the domain to which the correlation function value belongs. When the correlation function value belongs to the classical domain, the control weight corresponds to the driver having complete control of the vehicle. When the correlation function value belongs to a non-domain, the control weight corresponds to the autonomous driving system having complete control of the vehicle. When the correlation function value belongs to the extensional domain, the control weight adaptively and continuously changes with the correlation function value, and continuously transitions at the boundaries between the classical domain and the extensional domain, and between the extensional domain and the non-domain. Control output module: Based on the control weights, the driver-controlled front wheel angle of the vehicle is weighted and summed with the first target front wheel angle to obtain the second target front wheel angle under human-machine shared control and output it for execution.

[0017] Compared with the prior art, the beneficial effects of the present invention are: This invention integrates driver steering behavior characteristics and EEG physiological fatigue characteristics for multidimensional assessment, overcoming the shortcomings of existing technologies that rely solely on a single vehicle or driver state, resulting in incomplete risk assessment. By constructing a two-dimensional extension set with driver steering and collision avoidance error as the primary characteristic and driver fatigue level as the secondary characteristic, it can simultaneously characterize the degree of driver operational deviation and physiological fatigue state within the same framework, thereby achieving a multidimensional and quantitative assessment of driver steering and collision avoidance safety risks. This invention, based on extension decision theory and taking the ideal driving state as the optimal point, calculates the correlation function value by using the extension distance from feature state points to the boundaries of the classical and extension domains. This quantitatively maps the driver's steering and collision avoidance safety risks to three regions: the classical domain, the extension domain, and a non-domain. Based on this, three control modes are defined: "full driver control," "human-machine shared control," and "full autonomous driving control." Compared to existing technologies that allocate control rights based on conflict indices or fuzzy rules, this invention's control allocation has a clear mathematical theoretical foundation and real-time quantitative calculation capability, strong interpretability, low computational complexity, and is easy to implement in engineering. This invention employs an S-shaped mapping within the extension domain to determine the driver's control weights. This allows the control weights to smoothly decrease as the correlation function value changes from the classical domain side to the non-domain side, avoiding abrupt changes in control weights during mode switching. Combined with a first-order low-pass filter, dynamic smoothing is applied to driver steering and collision avoidance maneuvers, effectively suppressing high-frequency noise and random fluctuations in the original evaluation values. This further ensures the smoothness of control weight transitions, improving driver comfort and system acceptability during human-machine co-driving. This invention achieves a second target front wheel steering angle output under human-machine shared control by weighting and summing the driver's controlled front wheel steering angle with the first target front wheel steering angle of the autonomous driving system according to control weights. When the driver's steering and collision avoidance maneuvers are minor and the driver is not fatigued, the vehicle is primarily controlled by the driver, fully respecting the driver's operational intentions. When the driver's steering and collision avoidance maneuvers are significant or the driver is excessively fatigued, the autonomous driving system gains primary control and promptly executes emergency collision avoidance maneuvers. This control strategy can adaptively adjust the degree of control participation of the driver and the autonomous driving system according to the driver's real-time state, maximizing the use of the driver's control ability while ensuring collision avoidance safety, thus achieving a balance between safety and user-friendly human-machine interaction. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the overall method flow of the present invention; Figure 2 This is a schematic diagram of the two-dimensional extension set of the present invention; Figure 3 A dual Y-axis graph showing the relationship between driver steering and collision avoidance maneuvering errors and the values ​​of the associated function. Figure 4 A dual Y-axis graph showing the relationship between driver fatigue level and the value of the correlation function; Figure 5 This is a schematic diagram of the overall system modules of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0020] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0021] Example: Please see Figure 1 The present invention provides a technical solution: The human-machine shared steering and collision avoidance control method based on extension decision-making includes the following steps: S1: Obtain the driver's steering wheel operation parameters, EEG parameters, and road parameters at the current sampling time; calculate the driver's steering collision avoidance operation error based on the time-domain variation characteristics of the steering wheel operation parameters, determine the driver's fatigue level based on the driver's EEG parameter values, and determine the first target front wheel steering angle required for the autonomous driving system to control the vehicle based on the road parameters. The desired lane-change collision avoidance trajectory is generated based on the longitudinal coordinates, lateral coordinates, and heading angle in the road parameters, and the desired lateral position on the desired lane-change collision avoidance trajectory is determined based on the vehicle state at the current sampling time. The difference between the desired lateral position and the actual lateral position of the vehicle is used as the lateral displacement deviation. A closed-loop correction is performed on the lateral displacement deviation to obtain the first target front wheel steering angle.

[0022] The desired lane-changing collision avoidance trajectory is a fifth-order polynomial trajectory, which at the current sampling time... The expected lateral displacement is: in, Indicates the first The expected lateral displacement of the lane-changing collision avoidance trajectory at the sampling time. At the current sampling time, For lane changing time, This refers to the lane width.

[0023] Lane change collision avoidance trajectory at the current sampling time Desired longitudinal coordinates in the geodetic coordinate system and horizontal coordinates for: in, and The current sampling time The longitudinal and lateral coordinates of a road in a geodetic coordinate system. The current sampling time Road heading angle in geodetic coordinate system and The current sampling time The longitudinal and lateral coordinates of the desired lane-changing collision avoidance trajectory in the geodetic coordinate system.

[0024] Based on the current sampling time Longitudinal coordinates of the desired lane-changing collision avoidance trajectory and the vehicle's actual longitudinal coordinates Calculate the current sampling time The actual lateral displacement deviation value collected below : The autonomous driving system employs a lateral tracking control algorithm to reduce the lateral displacement deviation of the vehicle when tracking the desired lane-change collision avoidance trajectory. This algorithm takes the lateral displacement deviation and its historical changes as input, and outputs the first target front wheel steering angle control value required for the autonomous driving system to fully control the vehicle at the current sampling moment. As a feasible form of discrete control, its reference calculation formula is: In the formula, Sampling time The first target front wheel steering angle control amount required for the vehicle to be fully controlled by the autonomous driving system. , , These are the control gains for the deviation ratio term, deviation accumulation term, and deviation change term in the lateral tracking control algorithm, respectively, all of which are preset normal values; At the current sampling time, The actual window length is used for integration. Preset window length for integration. Sampling time The actual lateral displacement deviation value collected below. To backtrack from the current sampling time The cumulative sum of deviations within each sampling time point, when hour, ;when hour, , ; This is the index of the sampling time.

[0025] The calculation of driver steering and collision avoidance error based on steering wheel angle specifically involves: constructing a length of... N A sliding time window (with the same length as the preset integration window mentioned above) is used to extract the driver's steering wheel angle sequence within the window. Column, steering wheel angle differential sequence The sequence of front wheel steering angles for the autonomous driving system. Expected front wheel steering angle differential sequence ;when At that time, historical data was insufficient. There are 10 sampling points, at which point only existing data is used. Historical data are used in the calculation, and the sampling time is reached. Then, the full sliding time window is used for calculation.

[0026] The differential of the steering wheel angle input by the driver is obtained by performing a first-order numerical differential on the steering wheel angle, with units of . The calculation formula is as follows: The differential of the desired front wheel steering angle required to track the desired lane-change collision avoidance trajectory is obtained by performing a first-order numerical differential of the front wheel steering angle, with units of . The calculation formula is as follows: Calculate the sum of the rotation angle deviations within the sliding window and the sum of the differential rotation angle deviations using the following formula: in, For the first The sum of the absolute values ​​of the front wheel steering angle deviation within the sliding time window at the sampling time. For the first The sum of the absolute values ​​of the differential deviation of the front wheel steering angle within the sliding time window at the sampling time. N The length of the sliding time window (total number of sampling points). Here, represents the steering gear ratio of the steering system, and represents an inherent parameter of the vehicle. For the first The differential steering wheel angle input by the driver at the sampling time. For the first The derivative of the expected front wheel steering angle at the sampling moment. No. The steering wheel angle input by the driver at the sampling time. For the first The expected front wheel steering angle at the sampling time; Calculating driver steering and collision avoidance maneuvering errors using the hyperbolic tangent function: in, : No. The original driver steering and collision avoidance control error assessment value at the sampling time, with a value range of (0, 1); This is the front wheel steering angle deviation threshold, in degrees. The threshold values ​​for the differential deviation of the front wheel steering angle, measured in degrees per second, are used to normalize the cumulative steering angle deviation and the cumulative differential steering angle deviation within the sliding window to a uniform order of magnitude. Their values ​​are determined through offline simulation and driver-in-the-loop testing calibration: Under typical collision avoidance conditions, the distribution characteristics of the cumulative steering angle deviation and the cumulative differential steering angle deviation within the sliding window during normal driver operation are statistically analyzed. The 95th percentile (or the mean plus three standard deviations) of each value is taken as the initial calibration value for the corresponding threshold, and then fine-tuned based on the collision avoidance success rate and the driver's subjective acceptance. The weighting coefficient of the angle differential deviation term is used to adjust the degree of influence of driver input deviation on error assessment. It is dimensionless and is a preset positive constant. The value range is 0 to 1. When The larger the value, the higher the proportion of the differential deviation of the steering angle in the assessment of handling errors; when The smaller the value, the higher the proportion of the steering angle deviation in the assessment of handling errors. In the simulation calculation of this embodiment, a value of 0.4 is preferred.

[0027] This formula transforms the normalized angle deviation into an error index between 0 and 1 through nonlinear mapping, effectively avoiding evaluation distortion caused by extreme values. When the value approaches 1, it indicates that the driver's steering operation deviates significantly from the expected trajectory, posing a high risk of collision avoidance maneuvering error; conversely, if... A value close to 0 indicates that the driver's actions closely match the expected path, and the steering behavior is stable and accurate.

[0028] Driver fatigue levels are calculated based on electroencephalogram (EEG) parameters, specifically: based on EEG data... Fatigue levels are classified by value: values ​​in [40, 100) indicate no fatigue, values ​​in [30, 40) indicate mild fatigue, and values ​​in (0, 30) indicate excessive fatigue. The classification is based on the use of electroencephalogram (EEG) characteristics to determine driver fatigue levels. EEG signals can be measured using portable EEG measuring devices. The Karolinska Sleepiness Scale (KSS) is an internationally recognized reference for classifying fatigue levels, dividing fatigue into nine levels from very alert to very drowsy: levels 1-4 are from very alert to moderately alert; level 5 is neither alert nor drowsy; level 6 is slightly drowsy; level 7 is drowsy, staying alert is effortless; level 8 is drowsy, staying alert is somewhat difficult; and level 9 is very drowsy, requiring effort to stay alert. According to the EEG characteristic values ​​for fatigue levels in the appendix of the "Rules for Determining Fatigue Driving by Motor Vehicle Drivers" GA / T2372-2026, drivers with KSS levels 1-7 are considered not fatigued, KSS level 8 is mildly fatigued, and KSS level 9 is excessively fatigued. Therefore, at the current sampling time... The relationship between lower brain wave values ​​and fatigue levels can be expressed as follows: The brainwave values ​​were normalized using the following formula to obtain the driver fatigue level with values ​​ranging from [0,1]. : in, For the first The driver fatigue level after normalization at the sampling time is dimensionless and ranges from 0 to 1; the larger the value, the higher the fatigue level, and it serves as a secondary feature of the extension decision model. For the first The numerical values ​​of EEG characteristics at the sampling time are dimensionless.

[0029] S2: Construct a two-dimensional extension set with the driver's steering and collision avoidance operation error as the main feature and the driver's fatigue level as the secondary feature, and divide it into classical domain, extension domain and non-domain. Take the ideal driving state as the optimal point in the two-dimensional extension set, and calculate the correlation function value according to the positional relationship between the feature state point formed by the current main feature and secondary feature and the optimal point. Calculations show that the driver's steering and collision avoidance control errors were obtained. Then, the dynamic smoothing filter factor is determined based on the manipulation error at the current sampling time. A first-order low-pass filter is used for smoothing: in, : No. The driver's steering and collision avoidance error after smoothing at the sampling time is dimensionless and serves as the main feature of the extension decision model. For the first The dynamic smoothing filter factor at the sampling time is dimensionless and ranges from (0,1]. It is based on the steering and collision avoidance maneuver error at the current time. Adaptive determination; No. The original manipulation error assessment value at the sampling time is dimensionless; The dynamic smoothing filter factor The following formula is used to determine the original manipulation error at the current sampling time. The adaptive determination reference formula is as follows: Equivalent land: in, A preset positive constant is used to adjust the sensitivity of the smoothing weights to changes in manipulation errors; it is dimensionless. In this embodiment, The range of values ​​is Preferred value =0.5.

[0030] From the above formula, it can be seen that when When it increases, Reduce the weight of the original manipulation error at the current sampling time. Increase; when When decreasing, Increase the smoothed value at the previous sampling time. The weight of increases.

[0031] Therefore, when the driver's steering error increases sharply, the smoothing result can quickly track the changes in the original error and ensure that the system increases the control weight of the autonomous driving system in a timely manner. When the driver's operation is smooth and the error is small, the smoothing result is based on historical values, which can effectively suppress high-frequency noise and random fluctuations in the original evaluation value and avoid unnecessary jitter of the control weight.

[0032] Smooth steering avoidance error As the principal characteristic of the two-dimensional extension set; the driver's fatigue level These are secondary eigenvectors; A two-dimensional feature plane is established with the principal feature as the abscissa and the secondary feature as the ordinate. The two-dimensional extension set is divided into classical domain, extension domain and non-domain in the two-dimensional feature plane with the optimal point as the reference. In the classical domain, the driver has complete control over the vehicle and the automatic driving system does not participate in the control, so it will not interfere with the driver's normal operation. Connect the optimal point with the current feature state point to form a feature ray. The feature ray intersects the boundaries of the classical domain and the extension domain, respectively. Calculate the correlation function value based on the relative position of the current feature state point on the feature ray and each boundary intersection point: When the current feature state point is within the boundary of the classical domain, the correlation function value belongs to the classical domain; when the current feature state point is between the boundary of the classical domain and the boundary of the extended domain, the correlation function value belongs to the extended domain; when the current feature state point is outside the boundary of the extended domain, the correlation function value belongs to the non-domain. Please see Figure 2 In this embodiment, the following is selected: For the boundary between the classical domain and the extended domain, select To define the boundary between the extended field and the non-field, the original manipulation error evaluation function is used. The mapping has a theoretical range of [0,1), therefore... Defined as the upper limit of the characteristic value of driver steering and collision avoidance maneuvering error, this boundary corresponds to the limit state in which the driver loses the ability to perform effective collision avoidance maneuvers; it is determined through offline simulation and driver-in-the-loop testing. The collision avoidance success rate and driver subjective acceptance rating were statistically analyzed under different error levels. At this time, the driver can maintain the vehicle within the desired trajectory safety envelope by relying on their own operation, and system intervention will significantly reduce driver acceptance. The risk of a vehicle collision increases dramatically. Therefore, The threshold values ​​are calibrated as the critical thresholds between the classical and extended domains, thereby maximizing driver acceptance while ensuring collision avoidance safety. The EEG values ​​corresponding to the boundary between no fatigue and mild fatigue were selected as the boundary between the classical domain and the extensional domain. The EEG values ​​corresponding to the boundary between mild and excessive fatigue were selected as the boundary between the extended domain and the non-extended domain. According to EEG values correspond This is a state of non-fatigue; correspond This indicates a state of mild fatigue. correspond This indicates a state of excessive fatigue, leading to the conclusion that: The above calculation of the correlation function value is specifically as follows: taking the origin (0,0) of the two-dimensional extension set as the optimal point of the characteristic state, for any characteristic state point... Connect the origin O and point The ray intersects the boundary between the classical domain and the extensional domain at... The boundaries of points, intersections, extended regions, and non-regions are at... Point; calculate the point using the following formula. to the interval Extendable distance: in, Characteristic state points to the interval Extendable distance, Let be the current feature state point in the two-dimensional extension feature space. The current feature state point on the one-dimensional number axis Boundary points of classical domain The distance between them (i.e., the absolute value of the difference between the coordinates of the two points on a one-dimensional number axis). For a ray in a two-dimensional feature space Direction, from the origin Time The interval corresponds to the range of the classical domain; The correlation function characterizing the driver's steering collision avoidance safety risk is determined based on the extended distance, and the reference formula is as follows: in, The larger the correlation function value, the lower the safety risk for the driver when steering to avoid a collision; the smaller the correlation function value, the higher the safety risk for the driver when steering to avoid a collision.

[0033] S3: Determine the control weight of the driver controlling the vehicle based on the domain to which the correlation function value belongs. When the correlation function value belongs to the classical domain, the control weight corresponds to the driver having complete control of the vehicle. When the correlation function value belongs to the non-domain, the control weight corresponds to the autonomous driving system having complete control of the vehicle. When the correlation function value belongs to the extension domain, the control weight adapts and changes continuously with the correlation function value, and transitions continuously at the boundaries between the classical domain and the extension domain, and between the extension domain and the non-domain. As the correlation function value gradually decreases over time, the driver's control over the vehicle also gradually decreases, thus achieving a smooth transition of control, and a continuous transition at the boundaries between the classical and extended domains, and between the extended and non-domains. It should be noted that the correlation function value is not required to change continuously in a linear manner. Even if the correlation function value changes abruptly, the control weight will be updated adaptively in real time, dynamically adjusting the control weight based on the driver's steering and collision avoidance safety risks. This fully utilizes the collaborative capabilities between the driver and the autonomous driving system, improving collision avoidance safety and human-machine interface friendliness.

[0034] Within the extension domain, the continuous nonlinear mapping between the driver's control weights and the associated function values ​​can take various forms, such as sigmoid functions, piecewise linear functions, and polynomial functions. As a preferred implementation, a sigmoid mapping is used: As the correlation function value changes from the classical domain side to the non-domain side, the driver control weight smoothly decreases from the end where the driver has complete control of the vehicle to the end where the autonomous driving system has complete control of the vehicle. The slope and center position of the S-shaped mapping are determined by a preset adjustment coefficient, and the sum of the driver control weights and the control weights of the autonomous driving system output by the mapping remains a fixed total.

[0035] Within the extended domain, the driver's control authority is calculated using the following formula: in, For the first The driver's control over the vehicle at the time of sampling. The preset adjustment coefficient of the weighting function. Used to adjust the transition steepness of the S-shaped mapping curve. Used to adjust the midpoint offset position of the S-shaped mapping curve. and The value of is determined jointly through boundary constraints and simulation calibration: First, to ensure that the control weights are at both ends of the extension domain ( and ) smoothly approximate 0 and 1 respectively, and at the point in the extensional domain ( At point ) equal distribution of human-machine control rights is achieved, and the derivation is as follows: and Secondly, in a high-fidelity vehicle dynamics simulation environment, with collision avoidance success rate, smooth control handover, and driver's subjective acceptance as comprehensive optimization objectives, [the following is observed / optimized / optimized / etc.]. and A parameter scan was performed, and the optimal value was finally determined. At this time, when K( P When )=1, Approaching full driver control; when K( P When )=0, It approaches full control of autonomous driving. This parameter setting allows the control weights to smoothly approach the saturation value at both ends of the extension domain, with a moderate transition slope in the middle, balancing smoothness and response speed; The correlation function value of the current feature state, when the correlation function value At that time, the characteristic state is in the classical domain, and the driver's control over the vehicle is... The vehicle is entirely under the driver's control, and the automatic driving system does not intervene; when At that time, the characteristic state is in the extension domain, and the driver has control over the vehicle. The value is dynamically adjusted according to the value of the correlation function, and the vehicle is controlled jointly by the driver and the autonomous driving system; when When the characteristic state is in a non-domain, the driver's control over the vehicle is... The vehicle is fully controlled by the autonomous driving system and performs emergency collision avoidance.

[0036] In one specific embodiment, the association function It is obtained by fusing two features: driver steering and collision avoidance maneuvering error and driver fatigue level. To analyze the correlation function... The time-domain variation pattern affected by a single feature was investigated using the control variable method. Data sets with correlation functions varying only with driver steering and collision avoidance errors and data sets with correlation functions varying only with driver fatigue were obtained. The original sampled data were downsampled, and the resulting data are shown in Table 1.

[0037] Table 1. Statistics on Safety Risks for Drivers Steering and Avoiding Collisions To visually demonstrate the time-domain variation of the correlation function value with a single feature quantity, simulations were performed based on the control variable method described in Table 1, and the resulting curves are shown below. Figure 3 , Figure 4 As shown; based on the data shown in Table 1, combined with Figure 3 It can be seen that as driver steering error increases, the correlation function value tends to decrease, indicating an increase in driver steering error safety risk. Conversely, as driver steering error decreases, the correlation function value tends to increase, indicating a decrease in driver steering error safety risk. Figure 3 The driver's steering and collision avoidance maneuvering error showed a trend of first increasing and then decreasing, while the correlation function value showed an opposite trend of first decreasing and then increasing. The opposite trends of the two verified the negative correlation between maneuvering error and safety risk.

[0038] Based on the data shown in Table 1, combined with Figure 4 It can be seen that as driver fatigue increases, the correlation function value tends to decrease, indicating an increased risk to the driver's steering and collision avoidance safety. When the driver fatigue level is 0.6, the correlation function value is 1, reaching the boundary value between the classical domain and the extension domain. Within the extension domain, the driver is mildly fatigued, and the risk to the driver's steering and collision avoidance safety is relatively high. When the driver fatigue level is 0.7, the correlation function value is 0, reaching the boundary value between the extension domain and the non-domain. Within the non-domain, the driver is excessively fatigued, and the risk to the driver's steering and collision avoidance safety is very high.

[0039] S4: Based on the control weight value, the front wheel angle controlled by the driver and the first target front wheel angle are weighted and summed to obtain the second target front wheel angle under human-machine shared control and output for execution.

[0040] The weighted summation yields the second target front wheel steering angle, specifically: The steering wheel angle input by the driver via steering ratio The conversion yields the front wheel steering angle corresponding to the driver's input. The first target front wheel steering angle calculated by the autonomous driving system is denoted as... The weighted sum of the two values, calculated using the following formula, yields the second target front wheel steering angle under human-machine shared control. And output to the vehicle steering actuator: in, For the first The second target, the front wheel steering angle, is the shared control output of the human-machine interface at the sampling moment. No. The driver's control authority at the sampling time. No. At the sampling time, the driver manipulated the corresponding front wheel steering angle. For the first The front wheel steering angle of the first target corresponding to the sampling time of the autonomous driving system.

[0041] The driver's control authority The value range is [0, 1]. When hour, The vehicle is entirely under the driver's control; the automatic driving system does not intervene in steering control. hour, The vehicle is fully controlled by the autonomous driving system, performing emergency collision avoidance maneuvers; when At this time, the vehicle is in a human-machine shared control state, where the driver and the autonomous driving system jointly determine the steering output. Because... The correlation function value of the extension decision is dynamically determined and continuously adjusted as the correlation function value changes between the classical domain, extension domain and non-domain. Therefore, the above weighted summation can achieve a smooth transition of control between the driver and the autonomous driving system, avoiding vehicle steering vibration or driver discomfort caused by sudden switching of control. The steering ratio The transmission ratio of the vehicle steering system can be determined by the structural parameters of the steering system or dynamically adjusted according to vehicle speed or steering conditions; whether a fixed transmission ratio or a variable transmission ratio is used, both are equivalent replacements of the present invention. The second target front wheel steering angle can be output to the steering motor of the steer-by-wire system or the power steering motor of the electric power steering system, and the steering actuator performs the steering collision avoidance control. In one specific embodiment, the driver can first manipulate the corresponding front wheel steering angle. The first target of the autonomous driving system is the front wheel steering angle. Amplitude limiting or filtering is performed, and then the front wheel angle of the second target is calculated according to the above weighted summation formula to further suppress the influence of abnormal signals on the control output. However, the above modified processing does not change the basic control architecture and weighted fusion method of the present invention.

[0042] Please see Figure 5 The present invention also provides a human-machine shared steering and collision avoidance control system based on extension decision-making, applicable to the above method, comprising: Parameter acquisition module: acquires the driver's steering wheel operation parameters, EEG parameters and road parameters at the current sampling time; calculates the driver's steering collision avoidance operation error based on the time domain variation characteristics of the steering wheel operation parameters; determines the driver's fatigue level based on the driver's EEG parameter values; and determines the first target front wheel steering angle required for the autonomous driving system to control the vehicle based on the road parameters. Feature evaluation module: Construct a two-dimensional extension set with the driver's steering and collision avoidance operation error as the main feature and the driver's fatigue level as the secondary feature, and divide it into classical domain, extension domain and non-domain. The ideal driving state is taken as the optimal point in the two-dimensional extension set. The correlation function value is calculated based on the positional relationship between the feature state point formed by the current main feature and secondary feature and the optimal point. Extensional decision module: Determines the control weight of the driver controlling the vehicle based on the domain to which the correlation function value belongs. When the correlation function value belongs to the classical domain, the control weight corresponds to the driver having complete control of the vehicle. When the correlation function value belongs to a non-domain, the control weight corresponds to the autonomous driving system having complete control of the vehicle. When the correlation function value belongs to the extensional domain, the control weight adaptively and continuously changes with the correlation function value, and continuously transitions at the boundaries between the classical domain and the extensional domain, and between the extensional domain and the non-domain. Control output module: Based on the control weights, the driver-controlled front wheel angle of the vehicle is weighted and summed with the first target front wheel angle to obtain the second target front wheel angle under human-machine shared control and output it for execution.

[0043] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0044] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0045] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0046] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A human-machine shared steering and collision avoidance control method based on extension decision-making, characterized in that, The specific steps include: The system acquires the driver's steering wheel operation parameters, EEG parameters, and road parameters at the current sampling time. Based on the temporal variation characteristics of the steering wheel operation parameters, it accumulates the angular deviation and differential deviation between the driver's steering wheel angle and the expected front wheel angle of the autonomous driving system within a sliding time window. The system calculates the driver's steering collision avoidance operation error through compression mapping and adaptive smoothing. It determines the degree of fatigue based on the EEG parameter values ​​and determines the first target front wheel angle required for the autonomous driving system to control the vehicle based on the road parameters. A two-dimensional extension set is constructed with driver steering and collision avoidance control error as the main feature and fatigue degree as the secondary feature. It is divided into classical domain, extension domain and non-domain. The ideal driving state is the optimal point. The correlation function value is calculated based on the positional relationship between the feature state point composed of the current main and secondary features and the optimal point. The driver's control weight is determined based on the domain to which the correlation function value belongs. In the classical domain, the driver has complete control of the vehicle. In the non-domain, the autonomous driving system has complete control of the vehicle. In the extension domain, the control weight adapts and changes continuously with the correlation function value, and transitions continuously at the boundaries of each domain. Based on the control weight, the driver-controlled front wheel angle is weighted and summed with the first target front wheel angle to obtain the second target front wheel angle under human-machine shared control, and then the result is output for execution.

2. The human-machine shared steering and collision avoidance control method based on extension decision-making according to claim 1, characterized in that, The calculation of the first target front wheel steering angle based on road parameters specifically involves: generating a desired lane-changing collision avoidance trajectory based on the longitudinal coordinates, lateral coordinates, and heading angle in the road parameters, and determining the desired lateral position on the desired lane-changing collision avoidance trajectory based on the vehicle state at the current sampling time; The difference between the desired lateral position and the actual lateral position of the vehicle is used as the lateral displacement deviation. A closed-loop correction is performed on the lateral displacement deviation to obtain the first target front wheel steering angle.

3. The human-machine shared steering and collision avoidance control method based on extension decision-making according to claim 1, characterized in that, Driver steering and collision avoidance errors are determined based on the time-domain variation characteristics of the driver's steering wheel operation parameters, including: A sliding time window of a preset length is constructed with the current sampling time as the endpoint. The steering wheel angle and the differential of the steering wheel angle input by the driver within the sliding time window are extracted, as well as the first target front wheel angle and the differential of the first target front wheel angle at each sampling time within the same sliding time window. The steering wheel angle and the differential of the steering wheel angle are converted to the front wheel angle side according to the steering transmission ratio. The absolute value of the angle deviation between the converted steering wheel angle and the first target front wheel angle, as well as the absolute value of the differential of the angle deviation between the converted steering wheel angle and the first target front wheel angle, are calculated at each sampling time. The absolute values ​​of the angle deviation and the absolute values ​​of the differential of the angle are accumulated within the sliding time window to obtain the cumulative amount of angle deviation and the cumulative amount of differential of the angle. The cumulative differential deviation of the rotation angle is weighted according to a preset weight and combined with the cumulative deviation of the rotation angle. Then, it is mapped to a unified numerical range through a bounded monotonic compression function to obtain the original manipulation error at the current sampling time. The smoothing weight is adaptively determined based on the original control error at the current sampling time. The larger the original control error, the higher the proportion of the current sampling time value in the smoothing weight. The smaller the original control error, the higher the proportion of the historical value in the smoothing weight. The driver's steering and collision avoidance control error after smoothing at the previous sampling time and the original control error at the current sampling time are weighted and combined according to the smoothing weight to obtain the driver's steering and collision avoidance control error after smoothing at the current sampling time, which is used as the main feature quantity.

4. The human-machine shared steering and collision avoidance control method based on extension decision-making according to claim 3, characterized in that, After calculating the driver's steering and collision avoidance error, the smoothing coefficient is adaptively determined based on the error at the current sampling time. A recursive smoothing method is used to fuse the manipulation error at the current sampling time with the smoothed manipulation error at the previous sampling time. The larger the manipulation error at the current sampling time, the smaller the smoothing coefficient, so as to increase the influence of the manipulation error at the current sampling time on the fusion result. The smaller the manipulation error at the current sampling moment, the larger the smoothing coefficient, so as to increase the influence of the smoothing result at the previous sampling moment on the fusion result; The smoothed driver steering and collision avoidance maneuvering error at the current sampling moment is obtained as the main feature quantity of the two-dimensional extension set.

5. The human-machine shared steering and collision avoidance control method based on extension decision-making according to claim 1, characterized in that, The driver's fatigue level is calculated based on EEG parameters. Specifically, the driver's fatigue level is classified according to the EEG feature values ​​at the current sampling time and a preset threshold. When the EEG feature values ​​are above 40 and below 100, the driver is considered not fatigued. When the values ​​are above 30 and below 40, the driver is considered mildly fatigued. When the values ​​are above 0 and below 30, the driver is considered excessively fatigued. The brainwave feature values ​​are subjected to inverse linear normalization, which maps the brainwave feature values ​​to a fatigue level index between 0 and 1. The larger the brainwave feature value, the lower the corresponding fatigue level, and the smaller the brainwave feature value, the higher the corresponding fatigue level. The normalized driver fatigue level is used as a secondary feature of the extension decision model.

6. The human-machine shared steering and collision avoidance control method based on extension decision-making according to claim 1, characterized in that, The specific division of the two-dimensional extension set is as follows: a two-dimensional feature plane is established with the principal feature quantity as the horizontal axis and the secondary feature quantity as the vertical axis. The two-dimensional extension set is divided into classical domain, extension domain and non-domain in the two-dimensional feature plane with the optimal point as the reference. Connect the optimal point with the current feature state point to form a feature ray. The feature ray intersects the boundaries of the classical domain and the extension domain, respectively. Calculate the correlation function value based on the relative position of the current feature state point on the feature ray and each boundary intersection point: When the current feature state point is within the boundary of the classical domain, the correlation function value belongs to the classical domain; When the current feature state point is located between the boundary of the classical domain and the boundary of the extension domain, the correlation function value belongs to the extension domain; When the current feature state point is located outside the boundary of the extension domain, the correlation function value belongs to the non-domain.

7. The human-machine shared steering and collision avoidance control method based on extension decision-making according to claim 1, characterized in that, The specific steps for calculating the correlation function value are as follows: Using the origin in the two-dimensional extension set as the optimal point of the characteristic state, for any characteristic state point, the intersection point of the classical domain boundary and the extension domain boundary in the direction of the line connecting the characteristic state point and the origin is determined, and the positional relationship of the characteristic state point relative to the classical domain and the extension domain is obtained accordingly. The distance relationship between the feature state point and the classical domain interval is defined as the extension distance. When the feature state point is within the classical domain, the extension distance takes a negative value to characterize the degree to which the feature state point is within the classical domain and close to the optimal state. When the feature state point is located between the boundary of the classical domain and the boundary of the extension domain, the extension distance takes a positive value to characterize the degree to which the feature state point deviates from the classical domain but is still within the extension domain. Based on the extension distance, a correlation function is constructed to characterize the driver's steering and collision avoidance safety risk. The correlation function value can reflect the degree of proximity of the current feature state to the classical domain and the degree of deviation from the boundary of the extension domain. The larger the correlation function value, the lower the driver's steering and collision avoidance safety risk, and the smaller the correlation function value, the higher the driver's steering and collision avoidance safety risk.

8. The human-machine shared steering and collision avoidance control method based on extension decision-making according to claim 7, characterized in that, When the correlation function value belongs to the extension domain, the driver's control weights change continuously with the correlation function value, including: The driver control weight is determined by an S-shaped mapping: as the correlation function value changes from the classical domain side to the non-domain side, the driver control weight smoothly decreases from the end where the driver has complete control of the vehicle to the end where the autonomous driving system has complete control of the vehicle. The slope and center position of the S-shaped mapping are determined by a preset adjustment coefficient, and the sum of the driver control weights and the control weights of the autonomous driving system output by the mapping remains a fixed total.

9. The human-machine shared steering and collision avoidance control method based on extension decision-making according to claim 1, characterized in that, The driver's steering wheel angle is converted according to the steering transmission ratio, and the driver controls the front wheel angle of the vehicle. The driver's control weight is used as the weight of the driver's side front wheel angle, and the difference between the fixed total and the driver's control weight is used as the weight of the first target front wheel angle. The driver's side front wheel angle and the first target front wheel angle are weighted and summed to obtain the second target front wheel angle, which is then output to the vehicle steering actuator.

10. A human-machine shared steering and collision avoidance control system based on extension decision, applicable to the method of any one of claims 1-9, characterized in that, include: Parameter acquisition module: acquires the driver's steering wheel operation parameters, EEG parameters and road parameters at the current sampling time; calculates the driver's steering collision avoidance operation error based on the time domain variation characteristics of the steering wheel operation parameters; determines the driver's fatigue level based on the driver's EEG parameter values; and determines the first target front wheel steering angle required for the autonomous driving system to control the vehicle based on the road parameters. Feature evaluation module: Construct a two-dimensional extension set with the driver's steering and collision avoidance operation error as the main feature and the driver's fatigue level as the secondary feature, and divide it into classical domain, extension domain and non-domain. The ideal driving state is taken as the optimal point in the two-dimensional extension set. The correlation function value is calculated based on the positional relationship between the feature state point formed by the current main feature and secondary feature and the optimal point. Extensional decision module: Determines the control weight of the driver controlling the vehicle based on the domain to which the correlation function value belongs. When the correlation function value belongs to the classical domain, the control weight corresponds to the driver having complete control of the vehicle. When the correlation function value belongs to a non-domain, the control weight corresponds to the autonomous driving system having complete control of the vehicle. When the correlation function value belongs to the extensional domain, the control weight adaptively and continuously changes with the correlation function value, and continuously transitions at the boundaries between the classical domain and the extensional domain, and between the extensional domain and the non-domain. Control output module: Based on the control weights, the driver-controlled front wheel angle of the vehicle is weighted and summed with the first target front wheel angle to obtain the second target front wheel angle under human-machine shared control and output it for execution.

Citation Information

Patent Citations

  • A Laterally Extendable Preview Switching Control Method for Autonomous Vehicles

    CN108732921B

  • Man-machine cooperation path tracking control method considering driving state of driver

    CN116449815A

  • Man-machine co-driving sharing control method and related equipment

    CN122143947A