Real vehicle test and humanoid trajectory tracking method based on subjective handling stability evaluation

Abstract

The present application relates to the field of vehicle handling stability control, and proposes a real vehicle test and human-like trajectory tracking method based on subjective handling stability evaluation, which comprises the following steps: a real vehicle test scheme for driver subjective handling stability boundary calibration, a subjective handling stability boundary semi-empirical model design, a human-like trajectory tracking method based on subjective handling stability evaluation, etc. The present application establishes a mapping relationship between subjective handling stability perception and vehicle state through real vehicle testing, and forms a unified mathematical expression of the safety zone, transition zone and danger zone varying with vehicle speed by using a subjective handling stability boundary semi-empirical model that combines the kinetic mechanism and expert experience, thereby providing a quantitative basis for subjective handling stability evaluation. The human-like trajectory tracking control method based on subjective handling stability evaluation uses the semi-empirical model to determine the vehicle handling stability state in real time, adaptively adjusts the dynamic control requirements and the coordination weight of multiple actuators, and improves the handling stability and ride acceptance of unmanned vehicles.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of vehicle handling stability control, and particularly relates to a real vehicle test and human-like trajectory tracking method based on subjective handling stability evaluation. BACKGROUND

[0002] Existing vehicle stability evaluation is mostly based on vehicle dynamics theory, and the stability degree is determined by comparing the vehicle operating state with the theoretical stability boundary. However, there is a significant difference between the theoretical model and the subjective feeling of the driver, and the driver's acceptable stability range is usually smaller than the theoretical boundary. The objective evaluation index ignoring the subjective feeling may cause the active control system to deviate from the driver's expectation in the intervention time and control strength, affecting the driving comfort and safety.

[0003] The prior art has proposed a subjective and objective consistent vehicle stability evaluation method, which can establish the correlation between subjective and objective, but still has problems such as unsystematic test process and data processing, data missing in extreme working conditions, and large test quantity and high cost in normal working conditions, and it is urgent to establish a stability evaluation and trajectory tracking control method considering the prediction ability and test efficiency. SUMMARY

[0004] In order to solve the above technical problems, the present application provides a real vehicle test and human-like trajectory tracking method based on subjective handling stability evaluation, comprising the following steps:

[0005] Step 1, real vehicle test scheme for driver subjective handling stability boundary calibration;

[0006] The test scheme includes determination of road adhesion coefficient, determination of vehicle key parameters, and driver subjective evaluation. The determination of vehicle key parameters is to collect the lateral response data of the vehicle in the small perturbation linear region through sweep frequency test, and to identify the front and rear axle cornering stiffness at different vehicle speeds through least squares method. The driver's subjective evaluation is to determine the subjective boundaries of the vehicle in the stable, transition and dangerous regions under different road adhesion and vehicle speeds through the actual feeling of the driver under different steering inputs. On the basis of determining the test purpose and test conditions, a systematic test process is developed, and test data acquisition and data post-processing are carried out.

[0007] Step 2, design of subjective handling stability boundary semi-empirical model;

[0008] Based on the analysis of vehicle stability domain range by vehicle dynamics theory, a subjective handling stability boundary semi-empirical model is constructed by combining vehicle dynamics theory analysis and subjective evaluation expert experience knowledge, and a unified mathematical expression of the boundaries of stable region, transition region and dangerous region varying with vehicle speed is established. According to the relative position relationship between the real-time vehicle state point coordinates in the front and rear wheel cornering angle phase plane and the subjective handling stability boundary of the driver, a driver subjective handling stability evaluation index is designed.

[0009] Step three, human-like trajectory tracking method based on subjective handling stability evaluation;

[0010] According to the unmanned vehicle human-like trajectory tracking framework based on the subjective handling stability boundary semi-empirical model, combined with specific control algorithm, control target and control variable, the dynamic control demand weight adaptive adjustment is designed by using the driver's subjective handling stability evaluation index, and the safety dynamic demand is adjusted with the position of the real-time vehicle state in the subjective handling stability partition; the hyperbolic tangent function is used to design the multi-actuator coordinated control weight adaptive adjustment function, the vehicle prediction model for trajectory tracking MPC controller is constructed, the comprehensive cost function is designed, and the unmanned vehicle human-like trajectory tracking control method based on subjective handling stability evaluation is formed.

[0011] Further, in step one, the road adhesion coefficient determination test scheme measures the vehicle forward speed, longitudinal acceleration and lateral acceleration, and calculates the average friction coefficient of the high and low adhesion road; the vehicle key parameter determination test scheme measures the vehicle forward speed, steering wheel angle, front wheel steering angle, lateral acceleration, yaw rate and mass center side slip angle, and calculates the front and rear axle side stiffness; the driver's subjective evaluation test scheme measures the vehicle forward speed, steering wheel angle, front wheel steering angle, lateral acceleration, yaw rate and mass center side slip angle, and outputs the dynamic value and average value of the measurement parameters of the sine amplitude + sine constant amplitude test and slow step holding test corresponding to the subjective handling stability partition boundary respectively under different road adhesion and vehicle speed.

[0012] Further, the step of analyzing the vehicle stability domain range based on vehicle dynamics theory in step two comprises:

[0013] The rear wheel side slip angle at the saddle point position is in the descending area, indicating that the lateral motion reaches the maximum steady-state lateral acceleration state, and the yaw rate corresponding to the two saddle points is expressed as:

[0014]

[0015]

[0016] wherein, is the vehicle yaw rate, and are the yaw rate stability boundaries corresponding to saddle point 1 and saddle point 2 respectively; is the road adhesion coefficient, is the gravity acceleration, is the vehicle longitudinal speed;

[0017] The front wheel side slip angle corresponding to the saddle point is in the rising area and close to the saturated side slip angle, and the mass center side slip angle corresponding to the saddle point is expressed as:

[0018]

[0019]

[0020] where, is the vehicle mass center side slip angle, and are the mass center side slip angle stability boundaries corresponding to saddle point 1 and saddle point 2, respectively; is the front wheel side slip angle at maximum lateral force, ; is the vehicle mass, is the front wheel side slip stiffness, is the front wheel steering angle, and are the distances from the mass center to the front and rear wheels, respectively, is the wheelbase;

[0021] The front and rear wheel side slip angles and are calculated as follows:

[0022]

[0023] According to the yaw rate stability boundaries and the mass center side slip angle stability boundaries corresponding to the saddle point positions, the saddle point position coordinates in the phase plane and where, and are the front and rear wheel side slip angles corresponding to saddle point 1, respectively, and are the front and rear wheel side slip angles corresponding to saddle point 2, respectively:

[0024]

[0025]

[0026]

[0027]

[0028] The vehicle stability domain range is designed with the origin as the center and the average distance from the saddle point to the origin as the radius:

[0029]

[0030]

[0031]

[0032] where, and is the average position of the saddle points, is the average distance of the two saddle points to the origin of the phase plane.

[0033] Further, the step two of establishing the unified mathematical expression of the boundaries of the stable region, the transition region and the dangerous region varying with the vehicle speed comprises:

[0034] To reflect the overall trend of converging from low speed wide margin to high speed, the subjective handling stability boundary semi-empirical model adopts the Logistic function, and the mathematical form of the Logistic function in the subjective handling stability boundary semi-empirical model is .

[0035]

[0036] wherein, represents the critical vehicle speed perceived by the driver, is the longitudinal speed of the vehicle, represents the function difference between the low speed and the high speed interval, i.e. the difference between the low speed and the high speed , and the parameter determines the steepness of the boundary convergence;

[0037] To describe the "subjective optimal feeling" at the medium speed, the subjective handling stability boundary semi-empirical model introduces a Gaussian convex item on the Logistic baseline, and the mathematical form of the Gaussian convex item in the subjective handling stability boundary semi-empirical model is .

[0038]

[0039] wherein, represents the peak amplitude, is the optimal vehicle speed of the subjective feeling, represents the width of the medium speed range interval;

[0040] The subjective handling stability boundary semi-empirical model introduces a constant item , corresponding to the minimum stability margin of the vehicle in the high vehicle speed area, representing the minimum stability dynamic lower limit that the vehicle can still maintain in the limit state, providing a stability baseline for the model in the high vehicle speed area; the subjective handling stability boundary composite model is:

[0041] .

[0042] Further, the driver subjective handling stability evaluation index design method of step two comprises the following steps:

[0043] According to the front and rear wheel side slip angles Real-time vehicle state point coordinate in phase plane The relative position relationship between the driver's subjective handling stability boundary and the driver's subjective handling stability evaluation index ξ is designed as follows:

[0044]

[0045] wherein, , , are the driver's subjective handling stability evaluation index values corresponding to the safety zone, the transition zone and the danger zone boundaries, is the distance from the real-time vehicle state point to the origin of the phase plane:

[0046]

[0047] , and are the safety zone, the transition zone and the danger zone boundaries in the driver's subjective handling stability boundary:

[0048]

[0049]

[0050]

[0051] wherein, , , are the safety zone boundary , and corresponding subjective handling stability boundary composite models, , , , , , and are the parameters of the subjective handling stability boundary composite models at the safety zone, the transition zone and the danger zone boundaries, .

[0052] Further, in step three, for the dynamic control requirement, the safety dynamic requirement is adjusted according to the position of the subjective handling stability partition where the real-time vehicle state is located, and the subjective handling stability partition is applied to the self-driving vehicle as a condition for switching the dynamic safety requirement;

[0053] In combination with the driver's subjective handling stability evaluation index , an adaptive weight adjustment of multiple control targets is designed to realize the following dynamic switching of safety requirements:

[0054] When the vehicle state is in the safe region R1 of the subjective handling stability evaluation, , the dynamic safety requirement is trajectory tracking, the longitudinal speed tracking, the lateral position tracking and the yaw rate tracking in the vehicle state multi-objective function have large weight factors, and the torque vectoring control in the coordinated control architecture has a large weight factor;

[0055] When the vehicle state is in the transition region R2 of the subjective handling stability evaluation, , the dynamic safety requirement is weakened trajectory tracking, stability control is enhanced, the weight factor of the stability control corresponding to the vehicle state multi-objective function center side slip angle tracking increases with the increase of the driver's subjective handling stability evaluation index value, the weight factor of the active rear wheel steering in the coordinated control architecture gradually increases, and the weight factor of the torque vectoring control gradually decreases;

[0056] When the vehicle state is in the dangerous region R3 of the subjective handling stability evaluation, , the dynamic safety requirement is stability control as the main, trajectory tracking as the auxiliary, the center side slip angle tracking and the active rear wheel steering have large weight factors;

[0057] wherein, , , is the driver's subjective handling stability evaluation index value in the safe region, the transition region and the dangerous region respectively;

[0058] The control target weight adaptive adjustment function is as follows:

[0059]

[0060]

[0061]

[0062]

[0063] wherein, is the control target weight factor, , and are function adjustment parameters, , represents the lateral position, is the vehicle longitudinal speed, is the vehicle yaw rate, is the vehicle center side slip angle.

[0064] Further, in step three, the hyperbolic tangent function is used to design the multi-actuator coordinated control weight adaptive adjustment function as follows:

[0065]

[0066]

[0067]

[0068]

[0069] wherein, is a multi-actuator coordination control weight factor, , , , and are function adjustment parameters, , , and respectively represent the longitudinal acceleration, the front wheel steering angle, the rear wheel steering angle and the additional yaw moment;

[0070] Relying on the control architecture of the model predictive control algorithm, a control-oriented dynamics model of the unmanned vehicle is built, and a comprehensive cost function including multiple control objectives and multiple actuators is constructed, the steps being as follows:

[0071] Based on the force balance relationship of the vehicle, a three-degree-of-freedom single-track vehicle dynamics model including longitudinal, lateral and yaw motion degrees of freedom is established:

[0072]

[0073] wherein, is the vehicle mass, is the vehicle longitudinal velocity, is the vehicle lateral velocity, is the moment of inertia around the z-axis; is the tire force, the subscript respectively represent the longitudinal, lateral and vertical tire forces, the subscript respectively represent the front axle tire force and the rear axle tire force; and are respectively the distance from the mass center to the front wheel and the rear wheel, is the vehicle yaw angular velocity;

[0074] The front and rear axle lateral tire forces are obtained using the improved Magic Formula (MF) tire model:

[0075]

[0076] wherein, and are respectively the longitudinal and lateral tire forces, is a derating factor for the longitudinal tire force to provide the ability of the lateral tire force, ; is a shape factor, is a stiffness factor, is a peak, is a curvature value, is a vertical tire force; is a tire slip angle, front wheel slip angle , rear wheel slip angle ;

[0077] The lateral force is modeled by repeatedly linearizing the tire force model around the current slip angle at each time step to obtain an affine function of the slip angle to improve real-time computation performance:

[0078]

[0079]

[0080] wherein, and are the equivalent cornering stiffness of the front and rear axles, at each sampling time, the change of cornering stiffness should be updated according to the tire slip angle;

[0081] On the basis of a three-degree-of-freedom vehicle dynamics model, the vehicle position in the fixed ground coordinate system and the yaw angle between the reference ground axis and the vehicle longitudinal axis are added according to the geometric relationship, and the expression is:

[0082]

[0083]

[0084]

[0085] wherein, and are the longitudinal and lateral positions of the vehicle mass center in the ground coordinate system, is the yaw angle;

[0086] According to the small angle assumption, the vehicle prediction model for the trajectory tracking MPC controller is obtained as:

[0087]

[0088] Then, for the vehicle prediction model of the MPC controller built, by defining as the state vector, as the input variable, as the output of the system, the state space expression is obtained:

[0089]

[0090]

[0091] in, , and These are the state matrix, input matrix, and output matrix, respectively.

[0092] By using the current time reference point of the prediction model around the state vector and input variables Linearization is achieved through a first-order Taylor expansion, which speeds up model computation.

[0093]

[0094] Introducing error feedback into the linear prediction model: based on the model prediction values ​​obtained from the linear prediction model... Compared with the measured value of the current state quantity Calculate the error of the prediction model :

[0095]

[0096] Among them, the prediction model error ;

[0097] Substitute the current reference point The prediction model was then updated as follows:

[0098]

[0099] in, , , , .

[0100] By employing the forward Euler method, the prediction model update formula revolves around the sampling time. Discretization yields the discrete-time state-space equations:

[0101]

[0102] in, and This represents the state vectors that differ by one sampling step size after discretization. and These represent the discretized input vector and the prediction model error, respectively. It is the identity matrix. Indicates the current moment;

[0103] A comprehensive cost function is constructed to meet the dynamic control requirements of the controller and the smoothness of the actuator's action. The comprehensive cost function is expressed as follows:

[0104]

[0105] wherein, represents a comprehensive cost function, , respectively represent the prediction output and the reference output of a single sampling point, represents a control increment of a single sampling step; represents a dynamic control demand, , represents a multi-actuator coordinated control, as a weight matrix; , respectively represent a relaxation factor weight coefficient and a relaxation factor, and a relaxation term is introduced to solve the case that the optimization problem has no solution; is a prediction time range, is a control time range, represents a rolling step;

[0106] The optimization problem is re-expressed as a quadratic programming for solving, and the first element of the optimal control sequence is applied to the vehicle system; then the prediction range is pushed forward by one time interval, and the optimization problem is solved again by using the new process measurement value.

[0107] Advantages of the present application:

[0108] (1) The present application proposes a real vehicle test scheme for subjective handling stability partition boundary calibration, constructs a systematic test process and data processing method, establishes a mapping relationship between subjective handling stability perception and vehicle state, provides data support for semi-empirical model fitting, and reflects the differences in driving style and vehicle characteristics.

[0109] (2) The present application establishes a subjective handling stability boundary semi-empirical model that integrates the kinetic mechanism and expert experience, forms a unified mathematical expression of the safety zone, the transition zone and the danger zone varying with vehicle speed, realizes index prediction and test simplification, and provides a quantitative basis for subjective handling stability evaluation.

[0110] (3) The present application proposes a human-like trajectory tracking control method based on subjective handling stability evaluation, uses the semi-empirical model to determine the vehicle handling stability state in real time, adaptively adjusts the dynamic control demand and the multi-actuator coordination weight, and improves the handling stability and ride acceptance of the unmanned vehicle. BRIEF DESCRIPTION OF DRAWINGS

[0111] Figure 1 is a schematic diagram of the overall process of the present application;

[0112] Figure 2 is a schematic diagram of the real vehicle test process of the subjective handling stability evaluation of the driver;

[0113] Figure 3This is a schematic diagram of a human-like trajectory tracking method for autonomous vehicles based on subjective manipulation stability evaluation. Detailed Implementation

[0114] like Figure 1 As shown, this invention proposes a real-vehicle testing and humanoid trajectory tracking method based on subjective handling stability evaluation, including the following steps:

[0115] Step 1: Real-vehicle test plan for driver subjective control stability boundary calibration;

[0116] like Figure 2 As shown, the test plan includes determining the road surface adhesion coefficient, determining key vehicle parameters, and driver subjective evaluation. Based on the determined test objectives and conditions, a systematic test process is developed, including test data collection and post-processing, accumulating real-vehicle data and calibration experience to provide a basis for test sample reduction and model extrapolation. The specific steps are as follows:

[0117] 1. Test plan for determining the road surface friction coefficient

[0118] 1.1 Experimental Objective:

[0119] Confirming the road surface adhesion coefficient of the current test site provides a basic parameter for subsequent vehicle stability evaluation.

[0120] 1.2 Test conditions:

[0121] Electronic systems: ABS on, TCS and ESC off; Initial speed: 50 km / h; Test surfaces: High adhesion (dry asphalt) and low adhesion (compacted snow) surfaces.

[0122] 1.3 Test Methods:

[0123] 1.3.1) Accelerate the vehicle to 50 km / h and maintain a constant speed while driving in a straight line;

[0124] 1.3.2) Within the designated test area, quickly and smoothly depress the brake pedal to its maximum travel to perform straight-line braking with maximum braking intensity;

[0125] 1.3.3) Keep the direction stable and do not correct the steering wheel to avoid generating lateral tire forces until the vehicle comes to a complete stop;

[0126] 1.3.4) Record the maximum longitudinal deceleration during braking. (Unit: g)

[0127] 1.4 Experimental Data Acquisition and Processing:

[0128] 1.4.1) Testing instruments:

[0129] Vehicle speedometer; longitudinal acceleration sensor; lateral acceleration sensor.

[0130] 1.4.2) Measurement parameters (sampling frequency ≥ 100 Hz):

[0131] Vehicle forward speed (km / h); longitudinal acceleration (m / s2); lateral acceleration (m / s2).

[0132] 1.4.3) Each road surface repeated valid test not less than 3 times, the average value of each vehicle state.

[0133] 1.4.4) The repeatability error of the effective test results under the same vehicle speed is ≤±5%.

[0134] 1.4.5) Record ABS intervention, maximum deceleration and braking distance.

[0135] 1.4.6) The maximum lateral acceleration is not more than m / s².

[0136] 1.4.7) Calculate and output the road friction coefficient:

[0137]

[0138] Note: According to the ABS related standard GB / T 21670-2008, the utilization rate of ABS system attachment should not be less than 0.75. Since the maximum longitudinal deceleration value after ABS intervention is slightly lower than the limit value determined by the road attachment, the actual maximum longitudinal deceleration measured needs to be divided by 0.8 to obtain the true road friction coefficient.

[0139] 1.5 Output results:

[0140] Output the average friction coefficient of high and low adhesion road surface as the basic input parameter for subsequent test.

[0141] 2. Vehicle key parameter determination test scheme

[0142] 2.1 Test purpose:

[0143] Collect the lateral response data of the vehicle in the small perturbation linear region through sweep frequency test, and identify the front and rear axle cornering stiffness at different vehicle speeds through least squares method and , provide input parameters for subsequent stability evaluation model.

[0144] 2.2 Test conditions:

[0145] Road surface: dry, flat, high adhesion asphalt road; Initial state: drive, brake, steering and suspension system are normal, tire is well worn and tire pressure meets the manufacturer's standard; Electronic system: ABS is on, TCS and ESC are off; Load: driver + tester + data acquisition equipment; Environmental requirements: crosswind < 3 m / s, road slope < 2%.

[0146] 2.3 Test method:

[0147] 2.3.1) Accelerate to the target speed and maintain a straight line at a constant speed with a speed fluctuation of < ± 2 km / h.

[0148] 2.3.2) Under the premise of maintaining a constant speed, continuously perform the following requirements for 10 s:

[0149] 2.3.2.1) Sweep frequency range: gradually increase from 0.2 Hz to 2.0 Hz (about from every 5 s to every 0.5 s) ;

[0150] 2.3.2.2) Steering wheel angle peak: starting at ± 30°, linearly decreasing to ± 15° as the frequency increases (ensure that the maximum lateral acceleration does not exceed 0.4 g) ;

[0151] 2.3.2.3) The input trajectory should be smooth and continuous, with a steering wheel angle change approximating a sine wave, and no sudden or pause;

[0152] 2.3.2.4) Keep the throttle stable throughout the process, do not step on the brake or manually correct the vehicle attitude.

[0153] 2.3.3) After 10 s, the steering wheel is returned to the straight line and the vehicle is decelerated slowly to exit the test area.

[0154] 2.3.4) If the vehicle has a slight side slip or direction instability at high frequency stage (> 1.5 Hz), immediately stop the test and record the abnormality.

[0155] 2.3.5) Target speed: 50-120 km / h, with an interval of 10 km / h; The driver repeats the above 2.3.1) -2.3.5) operation instructions at least 3 times at each target speed to ensure data consistency.

[0156] 2.4 Test data acquisition and processing:

[0157] 2.4.1) Test instrument:

[0158] Speedometer; Steering wheel angle measuring instrument; Wheel angle measuring instrument; Lateral acceleration sensor; Yaw rate sensor; Mass center side slip angle sensor RT3000.

[0159] 2.4.2) Measurement parameters (sampling frequency ≥ 100 Hz):

[0160] Vehicle forward speed (km / h); steering wheel angle (°); front wheel steering angle (°); lateral acceleration (m / s2); yaw rate (° / s); vehicle center of gravity side slip angle (°).

[0161] 2.4.3) If the lateral acceleration exceeds 0.4 g, the data for this section is invalid.

[0162] 2.4.4) Repeatability error (sweep frequency range and steering wheel angle peak) ≤ ± 5% for at least three test results at the same vehicle speed;

[0163] 2.4.5) Least squares fitting is performed on each target vehicle speed data to calculate the front and rear axle cornering stiffness based on a two-degree-of-freedom vehicle dynamics model and .

[0164] 2.5 Output results:

[0165] 2.5.1) No less than 3 sets of valid data for each vehicle speed;

[0166] 2.5.2) Draw and the curve with vehicle speed, establish a look-up table database as input for subsequent stability evaluation.

[0167] 3. Driver subjective evaluation test scheme

[0168] 3.1 Test purpose:

[0169] Through the actual feeling of the driver under different steering inputs, determine the subjective boundaries of the stability zone, transition zone and danger zone of the vehicle under different road adhesion and vehicle speed, and provide calibration basis for subjective and objective consistent stability evaluation model.

[0170] 3.2 Test conditions:

[0171] Road surface: high adhesion road surface (dry asphalt) and low adhesion road surface (compacted snow); vehicle speed: 60-120 km / h, with an interval of 10 km / h; driver: no less than 3 subjective evaluators; system: ABS on, TCS and ESC off; environment: crosswind ≤ 3 m / s, road slope ≤ 2%; vehicle load: driver + tester + instrument equipment.

[0172] 3.3 Test method:

[0173] 3.3.1) Sine amplitude increase + sine constant amplitude (calibration of subjective boundaries)

[0174] 3.3.1.1) Accelerate the vehicle to the target speed, maintain a constant speed and straight line driving for 10 s, with a speed fluctuation of no more than ± 2 km / h.

[0175] 3.3.1.2) After the vehicle speed is stable, start to continuously turn the steering wheel at a frequency of 0.5-1.0 Hz (about 1 time per second back and forth).

[0176] 3.3.1.3) The initial steering wheel angle amplitude is ± 10°.

[0177] 3.3.1.4) Increase the amplitude by 5° every 2 s, keep the frequency unchanged, until the vehicle shows obvious oscillation or needs to be corrected.

[0178] 3.3.1.5) When it is felt that the vehicle direction response is developing from the linear zone (driver's subjective feeling: the vehicle responds linearly to the steering wheel, with no obvious delay or amplification response) to the transition zone, mark the subjective steering stability partition boundary "R1", record the steering wheel angle, and conduct a sine constant amplitude steering test with the steering wheel angle as the sine wave amplitude for more than 10 s.

[0179] 3.3.1.6) Continue to increase the steering angle, when the vehicle direction response is developing from the transition zone (driver's subjective feeling: the vehicle has a weak tendency to lose stability, with a slight lag or amplification response, and the vehicle shows slight side slip or correction demand) to the dangerous zone, mark the subjective steering stability partition boundary "R2", record the steering wheel angle, and conduct a sine constant amplitude steering test with the steering wheel angle as the sine wave amplitude for more than 10 s.

[0180] 3.3.1.7) Increase the steering angle again, when the vehicle direction response is close to the boundary of the dangerous zone (driver's subjective feeling: the vehicle is difficult to maintain direction or shows a tendency to spin, and feels that the vehicle is about to lose control, and increasing any actuator value will cause the vehicle to lose stability), mark the subjective steering stability partition boundary "R3", record the steering wheel angle, and conduct a sine constant amplitude steering test with the steering wheel angle as the sine wave amplitude for more than 10 s.

[0181] 3.3.1.8) Perform 3.3.1.1) - 3.3.1.7) at least 3 times at each vehicle speed, and take the average value.

[0182] 3.3.1.9) After each round, decelerate and stop, check the tire temperature and wear condition, and after the temperature returns, proceed to the next round.

[0183] 3.3.2) Slow step holding test (record steady state data)

[0184] 3.3.2.1) The steering wheel amplitude corresponding to the subjective partition boundary R1, R2, R3 determined in the previous step 3.3.1) is determined by the sinusoidal amplitude steering test. Refer to GB / T 6323-2014 steering angle step test for retesting.

[0185] 3.3.2.2) Accelerate to the target speed and drive straight at a constant speed.

[0186] 3.3.2.3) Smoothly reach the target steering wheel angle in 2-5 seconds (e.g. = 0.85, = 80 km / h, R2 corresponds to ±35°).

[0187] 3.3.2.4) The steering wheel remains unchanged for more than 3 seconds, with a speed fluctuation of ≤±2%.

[0188] 3.3.2.5) Repeat at least 3 times for each speed in the left and right directions.

[0189] 3.3.3) Vehicle speed: 60 km / h-120 km / h, with 10 km / h intervals Repeat the above.

[0190] 3.4 Test data acquisition and processing:

[0191] 3.4.1) Test instruments:

[0192] Speedometer; steering wheel angle measuring instrument; wheel angle measuring instrument; lateral acceleration sensor; yaw rate sensor; center of mass side slip angle sensor RT3000.

[0193] 3.4.2) Measured parameters (sampling frequency ≥100 Hz):

[0194] Vehicle forward speed (km / h); steering wheel angle (°); front wheel steering angle (°); lateral acceleration (m / s²); yaw rate (° / s); center of mass side slip angle (°).

[0195] 3.4.3) If a clear spin occurs, immediately terminate the test, and the data for this section is invalid.

[0196] 3.4.4) At each speed, repeat the test at least 3 times in each direction, and take the average of the calibration results to reduce test errors caused by other factors.

[0197] 3.4.5) Effective test is defined as the following: 3.3.1) the fluctuation of vehicle speed and steering wheel angle within 10 seconds is less than 5%, the difference of average steering angle between 3 groups of limited tests is less than 5%; 3.3.2) the fluctuation of vehicle speed and steering wheel angle within 3 seconds is less than 2%, the difference of left and right steering angle within each limited test is less than 2%, the difference of average steering wheel angle between 3 groups of limited tests is less than 2%.

[0198] 3.4.6) Record the measured parameter data corresponding to the subjective partition boundaries R1, R2 and R3 respectively shown in 3.4.2).

[0199] 3.5 Output results:

[0200] Output the measured parameter dynamic values and average values of 3 groups of effective tests corresponding to the subjective partition boundaries R1, R2 and R3 respectively under different road adhesion and vehicle speed, which are obtained by 3.3.1) sinusoidal amplitude increase + sinusoidal constant amplitude test and 3.3.2) slow step holding test.

[0201] Step two, design of subjective handling stability boundary semi-empirical model;

[0202] Based on vehicle dynamics theory, analyze the vehicle stability domain range; construct a subjective handling stability boundary semi-empirical model that integrates vehicle dynamics theory analysis and subjective evaluator's experience knowledge, establish a unified mathematical expression of the boundaries of stable zone, transition zone and dangerous zone changing with vehicle speed; design a driver subjective handling stability evaluation index according to the relative position relationship between the real-time vehicle state point coordinates in the front and rear wheel side slip angle phase plane and the driver subjective handling stability boundary; the model can explain the change law of the stability boundary from the physical mechanism, realize index prediction in the extreme working condition with insufficient data, realize test simplification and efficient fitting in the conventional working condition with sufficient data, and provide theoretical basis and quantitative standard for vehicle stability subjective evaluation. The specific steps are as follows:

[0203] Firstly, based on vehicle dynamics theory, analyze the vehicle stability domain range. The rear wheel side slip angle is located in the descending zone at the saddle point position, indicating that the lateral motion reaches the maximum steady-state lateral acceleration state, and the yaw rate corresponding to the two saddle points is expressed as:

[0204]

[0205]

[0206] wherein, is the vehicle yaw rate, and are the yaw rate stability boundaries corresponding to saddle point 1 and saddle point 2 respectively; is the road adhesion coefficient, is the gravitational acceleration, Vx is the vehicle longitudinal velocity;

[0207] The front wheel side slip angle corresponding to the saddle point is located in the rising region and is close to the saturation side slip angle, and the center of mass side slip angle corresponding to the saddle point is expressed as:

[0208]

[0209]

[0210] wherein, Vx is the vehicle longitudinal velocity; and are the center of mass side slip angle stability boundaries corresponding to saddle point 1 and saddle point 2, respectively; is the tire side slip angle at which the front wheel side force is maximum, ; is the vehicle mass, is the front wheel side stiffness, is the front wheel steering angle, and are the distances from the center of mass to the front wheel and the rear wheel, respectively, is the wheelbase;

[0211] The front and rear wheel side slip angles and are calculated as follows:

[0212]

[0213] According to the yaw rate stability boundary and the center of mass side slip angle stability boundary corresponding to the saddle point position, the coordinates of the saddle point position in the phase plane and .

[0214]

[0215]

[0216]

[0217]

[0218] wherein, and are the front wheel side slip angle and the rear wheel side slip angle corresponding to saddle point 1, respectively, and are the front wheel side slip angle and the rear wheel side slip angle corresponding to saddle point 2, respectively;

[0219] Therefore, the vehicle stability domain range is designed with the origin as the center and the average distance from the saddle point to the origin as the radius: ​

[0220]

[0221]

[0222]

[0223] where, and is the average position of the two saddle points, is the average distance of the two saddle points to the origin of the phase plane.

[0224] Then, in the subjective handling stability evaluation system, the driver divides the vehicle stability domain range into three typical partitions according to the response characteristics and controllability differences of the vehicle at different speeds: stable area, transition area and dangerous area. The three boundary curves respectively reflect the boundary conditions of the gradual evolution of the vehicle from a stable state to an unstable state, which not only is influenced by the constraints of vehicle dynamics, but also is dominated by the subjective handling stability perception of the driver. In order to mathematically depict the gradual change characteristics from "stable-transition-dangerous", and make the model reflect the coupling law between the driver's psychology and the dynamics law, the subjective handling stability boundary semi-empirical model proposed by the application adopts the composite structure of Logistic function term, Gaussian function term and constant term, which respectively describes the change law of the vehicle dynamics state evolution and the driver's subjective handling stability perception at different speed intervals. The structure is continuous in geometry and reasonable in physics, and can uniformly describe the change mechanism of the stability boundary from the aspects of driver perception, vehicle response and dynamic limit.

[0225] From the "overall trend", the three stability boundaries all present the overall downward law with the increase of the vehicle speed. The fundamental characteristic of the driver's subjective perception is that: when the vehicle speed is low, the lateral acceleration response and lateral stability margin of the vehicle are large, the driver can easily correct the attitude and maintain controllability, so the stability boundary can be appropriately relaxed; however, as the vehicle speed continuously increases, the psychological load of the driver in perceiving the vehicle sideslip and correcting the trajectory increases sharply, and the dynamic characteristics of the vehicle are closer to the nonlinear region - especially after the tire side slip characteristics enter the saturation section, the sensitivity of the vehicle to the steering input increases and the stability margin decreases, so that the driver feels obvious "tension" and "fear". Therefore, from the perspective of subjective handling stability, the higher the vehicle speed, the more conservative the driver tends to determine the "acceptable stable area", which is consistent with the contraction of the dynamic limit. In order to reflect this overall trend from low speed wide to high speed convergence, the model adopts the Logistic function, and its mathematical form is:

[0226]

[0227] where, represents the critical vehicle speed perceived by the driver, is the longitudinal speed of the vehicle, represents the function difference between the low-speed and high-speed intervals, parameter determines the steepness of the boundary convergence.

[0228] The monotone smooth descending characteristic of Logistic accurately simulates the "boundary convergence" law under the coupling of such subjective feelings and dynamic constraints: from the perspective of vehicle dynamics, as the vehicle speed increases, the tire side slip angle approaches the saturation zone, and the vehicle transitions from linear response to nonlinear response, with the stability margin gradually decreasing; from the subjective perspective of the driver, as the vehicle speed increases, the control difficulty increases, the visual load and psychological tension rise, and the driver will voluntarily reduce the acceptable stability boundary.

[0229] Therefore, when the vehicle speed is lower than , the boundary changes slowly and tends to be flat; when the vehicle speed approaches or exceeds , the controllability perceived by the driver rapidly decreases, and the stability boundary presents obvious convergence, thereby forming the impression of the driver's subjective "rapid narrowing of the stability zone".

[0230] From the "mid-section feature", the real vehicle test results of step one show that in the medium speed range, the driver usually considers that the vehicle's steering response is the most natural and controllable, and the subjective score also often reaches a peak in this interval, and there is a phenomenon that the vehicle's stability performance is enhanced in the medium speed range. At this time, the tire is in the linear-nonlinear transition zone, the steering response is sensitive but still predictable, and the vehicle has good balance. To describe this "subjective optimal feeling" at medium speed, a Gaussian convex term is introduced above the Logistic baseline, and its mathematical form is:

[0231]

[0232] wherein, represents the peak amplitude, is the optimal vehicle speed of subjective feeling, represents the width of the medium speed range interval.

[0233] This term represents the "maneuvering comfort zone" of the driver in a certain speed range: in the medium speed range (about 70~100 km / h), the tire side stiffness of the vehicle is well matched, the yaw response is smooth, the steering is sensitive and predictable. At this time, the driver can easily correct the posture and feel that the vehicle is "good in follow-up and natural in control", and subjectively considers that the vehicle has the highest stability. This medium speed range is also the most commonly used speed range by the driver in daily driving, with rich control experience and strong psychological adaptability.

[0234] Therefore, when When the vehicle stability, controllability and linear response work together, the subjective feeling of the driver is the best, and the boundary slightly expands outward; when the vehicle speed deviates from When the vehicle stability, controllability and linear response work together, the subjective feeling of the driver is the best, and the boundary slightly expands outward; when the vehicle speed deviates from

[0235] From the “reference level”, the constant term corresponds to the minimum stability margin of the vehicle in the high-speed region, that is, the subjective stability boundary of the driver in the “danger zone”. When the vehicle speed rises to the limit interval, the vehicle enters the nonlinear region of tire lateral force, the phase margin and gain margin of the system decrease, and the response of the yaw motion is unstable. In this region, the driver has obviously felt that the directional response is too fast, the attitude correction is difficult, the yaw and sideslip tend to be unpredictable, and the driver feels “fear” or “danger” subjectively. At this time, the boundary tends to converge, and the constant term represents the minimum stability dynamic lower limit that the vehicle can still maintain in the limit state, providing a stability baseline for the model in the high-speed region.

[0236] In summary, the subjective steering stability boundary composite model is obtained as follows:

[0237]

[0238] The subjective steering stability boundary composite model not only mathematically guarantees the continuity and derivability of the curve, but also explains the variation law of the subjective steering stability boundary of the driver from both physical and psychological angles.

[0239] Logistic term: embodies the overall “boundary contraction” caused by the increase of the vehicle speed, the psychological tension of the driver and the decrease of the vehicle stability;

[0240] Gaussian term: depicts the local “boundary expansion” caused by the “best controllability” felt by the driver at medium speed;

[0241] Constant term: corresponds to the residual stability capability of the vehicle in the limit state.

[0242] Therefore, the shape of the curve is not empirically fitted, but is the result of coupling the vehicle dynamics characteristics and the subjective perception mechanism of the driver, which reflects the mapping relationship between the psychological safety margin of the driver and the physical stability boundary of the vehicle. This function form can be used universally in different vehicle models and test data to establish a quantitative description model of the speed-stability subjective feeling, and provides a theoretical basis for subsequent stability control strategies and driving experience optimization.

[0243] Finally, according to the front and rear wheel sideslip angle Real-time vehicle state point coordinates in the phase plane The relative position relationship between the real-time vehicle state point and the driver's subjective handling stability boundary is designed, and the driver's subjective handling stability evaluation index ξ is:

[0244]

[0245] Among them, , , The driver's subjective handling stability evaluation index value corresponding to the safety zone, transition zone and dangerous zone boundary, The distance from the real-time vehicle state point to the origin of the phase plane:

[0246]

[0247] , And The safety zone, transition zone and dangerous zone boundary in the driver's subjective handling stability boundary:

[0248]

[0249]

[0250]

[0251] Among them, , , The driver's subjective handling stability boundary composite model corresponding to the safety zone boundary , And , , , , , And The parameters of the subjective handling stability boundary composite model in the safety zone, transition zone and dangerous zone boundary, , And The average position of the saddle point, The average distance from the two saddle points to the origin of the phase plane.

[0252] According to the dynamic driving requirements of the driver for handling and stability, the driver's subjective handling stability evaluation index value is defined as the safety zone boundary , the transition zone boundary , and the dangerous zone boundary .​

[0253] Step three, human-like trajectory tracking method based on subjective handling stability evaluation;

[0254] As Figure 3 shown, the subjective handling stability boundary semi-empirical model and subjective handling stability evaluation method determined by step two are combined with specific control algorithm, control target and control variable to design dynamic control demand and weight adaptive adjustment of multi-actuator coordination control of unmanned vehicle trajectory tracking, forming an unmanned vehicle human-like trajectory tracking method based on subjective handling stability evaluation, and realizing human-like driving to improve the acceptance of unmanned driving technology by passengers. The specific steps are as follows:

[0255] For dynamic control demand, adjust safety dynamic demand according to the position of real-time vehicle state in subjective handling stability partition. Subjective handling stability partition represents different vehicle stability states under the subjective feeling of passengers, which can be used as a condition to switch dynamic safety demand in unmanned vehicle. Combined with the subjective handling stability evaluation index of the driver, design adaptive weight adjustment of multiple control targets to clearly realize the dynamic switching of safety demand shown in Table 1.

[0256] Table 1

[0257]

[0258] Referring to Table 1, the relationship between real-time vehicle state and dynamic control demand is described as follows: when the vehicle state is in the safety zone R1 of subjective handling stability evaluation, since the vehicle has sufficient stability margin, the dynamic safety demand only considers the trajectory tracking ability of the vehicle, so the longitudinal speed tracking, lateral position tracking and yaw rate tracking in the vehicle state multi-objective function have a larger weight factor; since the moment vector control is a direct control of yaw rate, it has a significant effect on improving vehicle handling performance, so the moment vector control has a larger weight factor in the coordination control architecture.

[0259] In the process of the vehicle state gradually moving away from the safety zone, the safety demand of lateral stability control is continuously enhanced, when the vehicle state is in the transition zone R2 of subjective handling stability evaluation, the weight factor of the corresponding centroid side slip angle tracking in the stability control of the vehicle state multi-objective function increases with the increase of the subjective evaluation index, while the priority of trajectory tracking decreases with the increase of the subjective evaluation index. Since active rear wheel steering can provide additional lateral force, it is beneficial to ensure the lateral stability of the vehicle, so the weight factor of active rear wheel steering gradually increases in the coordination control architecture, and the weight factor of moment vector control gradually decreases.

[0260] When the vehicle state is in the dangerous region R3 of the subjective handling stability evaluation, the driver has a strong subjective feeling of instability, and the dynamic safety demand is mainly to ensure the vehicle to run stably, so the ideal center side slip angle tracking and active rear wheel steering have a large weight factor.

[0261] The control target weight adaptive adjustment function is as follows:

[0262]

[0263]

[0264]

[0265]

[0266] wherein, is the control target weight factor, , and are function adjustment parameters, , represents the lateral position, is the vehicle longitudinal speed, is the vehicle yaw rate, is the vehicle center side slip angle.

[0267] In order to reduce the actuator chattering, the weight adaptive adjustment function of the actuator should be continuous and smooth, and the hyperbolic tangent function is used to design the multi-actuator coordinated control weight adaptive adjustment function as follows:

[0268]

[0269]

[0270]

[0271]

[0272] wherein, is the multi-actuator coordinated control weight factor, , , , and are function adjustment parameters, , , , and respectively represent the longitudinal acceleration, the front wheel steering angle, the rear wheel steering angle and the additional yaw moment.

[0273] Based on the control architecture of model predictive control algorithm, a control-oriented vehicle dynamics model is built, a comprehensive cost function including multiple control objectives (including longitudinal speed tracking, lateral position tracking, yaw rate tracking and center side slip angle tracking) and multiple actuators (including front wheel steering angle, rear wheel steering angle and additional yaw moment) is constructed, and the dynamic control requirements and multiple actuator coordinated control are realized by using the weight adaptive adjustment based on subjective handling stability evaluation to realize the comprehensive optimization of trajectory tracking accuracy, vehicle stability and actuator smoothness of human-like driving style.

[0274] Based on the force balance relationship of the vehicle, a three-degree-of-freedom single-track vehicle dynamics model including longitudinal, lateral and yaw motion degrees of freedom is established:

[0275]

[0276] Wherein, is the mass of the vehicle; is the lateral velocity of the vehicle; is the rotational inertia around the z-axis; is the tire force, and the subscripts respectively represent the longitudinal, lateral and vertical tire forces, and the subscripts respectively represent the front axle tire force and the rear axle tire force. The variable has a derivative, which is a basic mathematical expression, and the same applies below.

[0277] In the present application, the front and rear axle lateral tire forces are obtained by using the improved Magic Formula (MF) tire model, which can capture the decrease of the lateral tire force caused by the applied longitudinal tire force :

[0278]

[0279] Wherein, is a derating factor for the ability of the longitudinal tire force to the lateral tire force, ; is a shape factor, is a stiffness factor, is a peak value, is a curvature value, is a vertical tire force; is a tire side slip angle, the front wheel side slip angle , and the rear wheel side slip angle ;

[0280] By repeatedly linearizing the tire force model around the current side slip angle at each time step, the lateral force is modeled to obtain an affine function of the side slip angle to improve the real-time calculation performance:

[0281]

[0282]

[0283] where, and are the equivalent cornering stiffness of front and rear axles, at each sampling time, the change of cornering stiffness should be updated according to the tire cornering angle.

[0284] Based on the three-degree-of-freedom vehicle dynamics model, the vehicle position in the fixed ground coordinate system and the yaw angle between the reference ground axis and the vehicle longitudinal axis are added according to the geometric relationship, the expression is:

[0285]

[0286]

[0287]

[0288] where, and are the longitudinal and lateral positions of the vehicle mass center in the ground coordinate system, is the yaw angle.

[0289] According to the small angle assumption, the vehicle prediction model for the trajectory tracking MPC controller is obtained by combining the above formulas:

[0290]

[0291] Then, for the vehicle prediction model of the MPC controller built, by defining as the state vector, as the input variable, as the output of the system, the state space expression is obtained:

[0292]

[0293]

[0294] where, , and are the state matrix, input matrix and output matrix respectively;

[0295] By performing first-order Taylor expansion around the current time reference point to realize linearization processing, the model calculation speed is accelerated:

[0296]

[0297] To compensate for the effect of high order terms neglected in the first order linearized nonlinear model, an error feedback is introduced in the linear prediction model: the model prediction value obtained from the linear prediction model is subtracted from the measured value of the current state variable to obtain the prediction model error

[0298]

[0299] where the prediction model error is defined as

[0300] Substituting the reference point at the current time , the prediction model is updated as

[0301]

[0302] where , , ,

[0303] Further, by using the forward Euler method, the above equation is discretized around the sampling time to obtain the discrete-time state-space equation:

[0304]

[0305] where and denote the state vector at the next sampling step, and denote the input vector and the prediction model error after discretization, respectively, is the identity matrix, denotes the current time;

[0306] To meet the dynamic control requirements of the controller and the smoothness of the actuator action, the integrated cost function is expressed as an optimization problem, aiming to minimize the error of vehicle trajectory tracking and stability maintenance, and to avoid the impact of severe actuator operation on comfort. Therefore, the integrated cost function is expressed as follows:

[0307]

[0308] where denotes the integrated cost function, , denote the prediction output and the reference output of a single sampling point, respectively, denotes the control increment of a single sampling step; denotes the dynamic control requirement, , ​​denotes a multi-actuator coordinated control, as a weight matrix; 、 respectively denote a relaxation factor weight coefficient and a relaxation factor, and a relaxation term is introduced to solve the case that the optimization problem has no solution; is a prediction time range, is a control time range, denotes a rolling step.

[0309] According to the above analysis, the optimization problem can be re-expressed as a quadratic programming problem for solving, and the first element of the optimal control sequence is applied to the vehicle system; then the prediction range is pushed forward by one time interval, and the optimization problem is solved again using the new process measurement value.

[0310] In summary, according to the subjective steering stability evaluation, the dynamic control demand and the control architecture are adaptively adjusted to realize the human-like trajectory tracking method of the unmanned vehicle.

Claims

1. A real-vehicle testing and humanoid trajectory tracking method based on subjective handling stability evaluation, characterized in that: Includes the following steps: Step 1: Determine the road surface adhesion coefficient, key vehicle parameters, and driver subjective evaluation through a real vehicle test plan. Key vehicle parameters are obtained by collecting lateral response data of the vehicle in the linear region of small disturbance through frequency sweep test, and the front and rear axle lateral stiffness at different vehicle speeds is identified by the least squares method. Driver subjective evaluation is determined by the driver's actual feeling under different steering inputs to determine the subjective boundaries of the vehicle's stability, transition, and danger zones under different road surface adhesion and vehicle speeds. Step 2: Analyze the vehicle stability domain based on vehicle dynamics theory; construct a semi-empirical model of subjective handling stability boundary that integrates vehicle dynamics theory analysis and subjective evaluation experience; establish a unified mathematical expression for the boundary of stable zone, transition zone, and danger zone as a function of vehicle speed; design driver subjective handling stability evaluation index based on the relative positional relationship between the real-time vehicle state point coordinates in the front and rear wheel slip angle phase plane and the driver's subjective handling stability boundary. Step two, which describes establishing a unified mathematical expression for the changes in the boundaries of the stable zone, transition zone, and danger zone with vehicle speed, includes: To reflect the overall trend of convergence from low-speed to high-speed, the semi-empirical model of subjective manipulation stability boundary adopts the Logistic function, whose mathematical form is... for: in, This indicates the critical speed perceived by the driver. For the longitudinal speed of the vehicle, The function represents the difference between the low-speed and high-speed ranges, with parameters... Determines the steepness of boundary convergence; To describe the "subjective optimal feeling" at mid-range speeds, the subjective manipulation stability boundary semi-empirical model introduces a Gaussian convex term on top of the Logistic baseline, its mathematical form being... for: in, Indicates peak amplitude. The optimal speed as perceived subjectively. Indicates the width of the medium speed range; Subjective manipulation stability boundary semi-empirical model introduces constant term This corresponds to the minimum stability margin of the vehicle in the high-speed region, representing the minimum dynamic lower limit of stability that the vehicle can maintain under extreme conditions, providing a stability baseline for the model in the high-speed region; the subjective handling stability boundary composite model is: ； The driver subjective handling stability evaluation index design method described in step two includes the following steps: Based on the front and rear wheel slip angles Real-time vehicle status point coordinates in phase plane The relative positional relationship between the driver's subjective control stability boundary and the driver's subjective control stability boundary is used to design the driver's subjective control stability evaluation index ξ as follows: in, These are the driver's subjective handling stability evaluation index values ​​corresponding to the boundaries of the safe zone, transition zone, and danger zone, respectively. The distance from the real-time vehicle status point to the origin of the phase plane: , and These are the boundaries of the safe zone, transition zone, and danger zone within the driver's subjective control stability boundary: in, These are the boundaries of the safe zone. The corresponding subjective manipulation stability boundary composite model, These represent the parameters of the subjective manipulation stability boundary composite model at the boundaries of the safe zone, transition zone, and danger zone, respectively. , and The average position of the saddle point. This represents the average distance from the two saddle points to the origin of the phase plane. Step 3: Based on the semi-empirical model of subjective maneuver stability boundary for autonomous vehicles, an adaptive adjustment of dynamic control demand weights is designed using the driver's subjective maneuver stability evaluation index. The safety dynamic demand is adjusted according to the position of the vehicle in the subjective maneuver stability zone in real time. A hyperbolic tangent function is used to design an adaptive adjustment function for multi-actuator coordinated control weights, a vehicle prediction model is constructed, and a comprehensive cost function is designed to form an autonomous vehicle human trajectory tracking control method based on subjective maneuver stability evaluation.

2. The method for real-vehicle testing and humanoid trajectory tracking based on subjective handling stability evaluation according to claim 1, characterized in that: In step one, the road surface adhesion coefficient determination test plan measures the vehicle's forward speed, longitudinal acceleration, and lateral acceleration, and calculates and outputs the average friction coefficient of high and low adhesion road surfaces; the vehicle key parameter determination test plan measures the vehicle's forward speed, steering wheel angle, front wheel steering angle, lateral acceleration, yaw rate, and center of gravity sideslip angle, and calculates the front and rear axle lateral stiffness; the driver subjective evaluation test plan measures the vehicle's forward speed, steering wheel angle, front wheel steering angle, lateral acceleration, yaw rate, and center of gravity sideslip angle, and outputs the dynamic values ​​and average values ​​of the measured parameters for the sine amplification + sine constant amplitude test and the slow step hold test corresponding to the subjective handling stability zone boundaries under different road surface adhesion and vehicle speeds.

3. The method for real-vehicle testing and humanoid trajectory tracking based on subjective handling stability evaluation according to claim 1, characterized in that: Step two, which involves analyzing the vehicle stability domain based on vehicle dynamics theory, includes: At the saddle point, the rear wheel slip angle is in the decreasing region, indicating that the lateral motion has reached the maximum steady-state lateral acceleration state. The yaw rate corresponding to the two saddle points is expressed as: in, Let yaw rate be the vehicle's angular velocity. and These are the yaw rate stability boundaries corresponding to saddle point 1 and saddle point 2, respectively. The road surface adhesion coefficient, It is the acceleration due to gravity. The longitudinal speed of the vehicle; The front wheel slip angle corresponding to the saddle point is in the rising region and close to the saturation slip angle. The center-of-gravity slip angle corresponding to the saddle point is expressed as: in, The sideslip angle is the angle at the vehicle's center of gravity. and These are the centroid sideslip angle stability boundaries corresponding to saddle points 1 and 2, respectively. This is the tire slip angle when the lateral force on the front wheel is at its maximum. ; For vehicle quality, For the front wheel lateral stiffness, This refers to the front wheel steering angle. and These are the distances from the center of gravity to the front and rear wheels, respectively. Wheelbase; Front and rear wheel slip angles and The calculation formula is as follows: Based on the yaw rate stability boundary and the center of mass sideslip angle stability boundary corresponding to the saddle point position, we obtain... Saddle point position coordinates in phase plane and ,in, and These are the front wheel slip angle and rear wheel slip angle corresponding to saddle point 1, respectively. and These are the front wheel slip angle and rear wheel slip angle corresponding to saddle point 2, respectively: Design the vehicle stability domain with the origin as the center and the average distance from the saddle point to the origin as the radius: in, and The average position of the saddle point. This represents the average distance from the two saddle points to the origin of the phase plane.

4. The method for real-vehicle testing and humanoid trajectory tracking based on subjective handling stability evaluation according to claim 1, characterized in that: In step three, for dynamic control requirements, the safety dynamic requirements are adjusted according to the position of the subjective handling stability zone of the real-time vehicle status. The subjective handling stability zone is used as a condition for switching dynamic safety requirements in autonomous vehicles. Combined with driver's subjective control stability evaluation index An adaptive weight adjustment for multiple control targets is designed to achieve dynamic switching of the following safety requirements: When the vehicle's condition is within the safe zone R1 of the subjective handling stability evaluation... The dynamic safety requirement is trajectory tracking. In the multi-objective function of vehicle state, longitudinal velocity tracking, lateral position tracking and yaw rate tracking have large weight factors. In the coordinated control architecture, torque vector control has a large weight factor. When the vehicle is in the transition zone R2 of the subjective handling stability evaluation The dynamic safety requirements are reduced trajectory tracking and enhanced stability control. In the multi-objective function of vehicle state, the weight factor of the centroid sideslip angle tracking corresponding to stability control increases with the increase of the driver's subjective operation stability evaluation index value. In the coordinated control architecture, the weight factor of active rear wheel steering gradually increases, while the weight factor of torque vector control gradually decreases. When the vehicle's condition is within the danger zone R3 of the subjective handling stability evaluation... Dynamic safety requirements prioritize stability control, supplemented by trajectory tracking, with center of gravity sideslip angle tracking and active rear wheel steering having significant weighting factors. in, , , These are the evaluation index values ​​for the driver's subjective handling stability in the safe zone, transition zone, and danger zone, respectively. The adaptive adjustment function for the control target weights is as follows: in, To control the target weighting factor, , and Adjusting parameters for the function, , Indicates horizontal position. For the longitudinal speed of the vehicle, Let yaw rate be the vehicle's angular velocity. The vehicle's center of gravity sideslip angle.

5. The method for real-vehicle testing and humanoid trajectory tracking based on subjective handling stability evaluation according to claim 1, characterized in that: In step three, the hyperbolic tangent function is used to design the adaptive adjustment function for the weights of the multi-actuator coordinated control, as follows: in, Weighting factors for multi-actuator coordinated control. , , , and Adjusting parameters for the function, , and These represent longitudinal acceleration, front wheel steering angle, rear wheel steering angle, and additional yaw moment, respectively. Based on the control architecture of model predictive control algorithm, a control-oriented dynamic model of autonomous vehicle is built, and a comprehensive cost function including multiple control objectives and multiple actuators is constructed. The steps are as follows: Based on the force balance relationship of the vehicle, a three-degree-of-freedom monorail vehicle dynamics model including longitudinal, lateral, and yaw motion degrees of freedom is established: in, It's about vehicle quality. For the longitudinal speed of the vehicle, It is the vehicle's lateral speed. It is the moment of inertia about the z-axis; It refers to tire force, subscript. These represent the longitudinal, lateral, and vertical tire forces, respectively, with subscripts. These represent the front axle tire force and the rear axle tire force, respectively. These are the front and rear wheel steering angles, respectively. These are the distances from the center of gravity to the front and rear wheels, respectively. The vehicle's yaw rate; The front and rear axle lateral tire forces were obtained using the modified Magic Formula (MF) tire model: in, For lateral tire force, A depreciation factor that provides capability for longitudinal tire forces to lateral tire forces. ; For longitudinal tire force, It is the shape factor. It is the stiffness factor. It is the peak value. It is the curvature value. It is the vertical force on the tire; It refers to the tire slip angle and the front wheel slip angle. Rear wheel slip angle ; Lateral forces are modeled by iteratively linearizing the tire force model around the current slip angle at each time step, thus obtaining an affine function of the slip angle to improve real-time computational performance. in, It is the equivalent lateral stiffness of the front and rear axles. At each sampling time, the change in lateral stiffness should be updated according to the tire slip angle. Based on the three-degree-of-freedom vehicle dynamics model, the vehicle position in a fixed geodetic coordinate system and the yaw angle between the reference ground axis and the vehicle's longitudinal axis are added according to geometric relationships. The expression is: in, These refer to the longitudinal and lateral positions of the vehicle's center of gravity in the geodetic coordinate system, respectively. It is the yaw angle; Based on the small angle assumption, the vehicle prediction model for the trajectory tracking MPC controller is obtained as follows: Then, for the vehicle prediction model of the built MPC controller, by defining For state vectors, For input variables, The state-space expression for the system output is obtained as follows: in, These are the state matrix, input matrix, and output matrix, respectively. By analyzing the prediction model around the current reference point Linearization is achieved through a first-order Taylor expansion, which speeds up model computation. Introducing error feedback into the linear prediction model: based on the model prediction values ​​obtained from the linear prediction model... Compared with the measured value of the current state quantity Calculate the error of the prediction model : Among them, the prediction model error ; Substitute the current reference point The prediction model was then updated as follows: in, , , , ; By employing the forward Euler method, the prediction model update formula revolves around the sampling time. Discretization yields the discrete-time state-space equations: in, This represents the state vectors that differ by one sampling step size after discretization. These represent the discretized input vector and the prediction model error, respectively. It is the identity matrix. Indicates the current time; A comprehensive cost function is constructed to meet the dynamic control requirements of the controller and the smoothness of the actuator's action. The comprehensive cost function is expressed as follows: in, Represents the comprehensive cost function. These represent the predicted output and reference output for a single sampling point, respectively. This represents the control increment for a single sampling step. This indicates a need for dynamic control. , This indicates multi-actuator coordinated control. As a weight matrix; Let represent the relaxation factor weight coefficient and the relaxation factor, respectively. A relaxation term is introduced to address the case where the optimization problem has no solution. It is the predicted time range. It controls the time range. Indicates the rolling step size; The optimization problem is reformulated as a quadratic programming problem and solved by applying the first element of the optimal control sequence to the vehicle system. The prediction range is then advanced by one time interval, and the optimization problem is solved again using the new process measurements.

Images