Distributed driving intention recognition method for vehicle based on vehicle stable state

Abstract

The application is suitable for the technical field of vehicle stability control, and provides a distributed drive vehicle human-like driving intention recognition method based on vehicle stability state, obtains vehicle longitudinal speed, lateral acceleration and front wheel angle signal through a sensor, establishes a system equation based on a two-degree-of-freedom vehicle model and a magic tire model, estimates a mass center side slip angle and a yaw angular velocity in real time by using an unscented Kalman filter, calculates a mass center side slip angle change rate, constructs a phase plane in combination with a six-line method, judges the vehicle stability state according to a region where a current state point is located, decides an expected motion strength according to the vehicle stability state, and finally outputs the obtained expected motion strength to an actuator control system, so as to realize collaborative control of each wheel drive and steering of the distributed drive vehicle, thereby realizing human-like driving intention tracking under the premise of ensuring stability.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of vehicle stability control, and particularly relates to a distributed drive vehicle human-like driving intention recognition method based on vehicle stable state. BACKGROUND

[0002] The distributed drive vehicle has the characteristics that the torque and steering of each wheel can be independently controlled, and has significant advantages in improving vehicle handling, stability and energy efficiency. However, how to accurately and timely recognize the driver's manipulation intention in such a complex system with multiple actuators working together, and on this basis to realize the human-like vehicle motion control, has become a research hotspot in the field of intelligent chassis and stability control.

[0003] In the traditional centralized drive vehicle, the driver's intention is usually inferred from the driver's operation signals such as steering wheel angle, accelerator pedal opening and brake pedal opening. However, in the distributed drive vehicle, since the torque of each drive wheel can be independently distributed, the coupling relationship between the vehicle motion state and the driver's manipulation input is more complex, and it is difficult to accurately reflect the real intention of the driver in the extreme working condition or unstable state by simply relying on the traditional input signal. Especially when the vehicle is in a critical stable or unstable state, if the expected motion strength of the driver cannot be timely recognized and reasonably corrected, the vehicle is easy to lose control, which seriously affects the driving safety.

[0004] In the prior art, some researches have tried to estimate the key state parameters of the vehicle based on the vehicle dynamics model and state observation method, and on this basis to carry out stability control. However, these methods mostly focus on the intervention of the control layer, lack of in-depth understanding and dynamic recognition of the driver's manipulation intention, which leads to the response lag of the control system or the inconsistency with the driver's expectation, affecting the driving experience and manipulation quality. SUMMARY

[0005] The purpose of the embodiment of the present application is to provide a distributed drive vehicle human-like driving intention recognition method based on vehicle stable state, which aims to solve the problems proposed in the background.

[0006] The embodiment of the present application is implemented as follows: the distributed drive vehicle human-like driving intention recognition method based on vehicle stable state comprises the following steps:

[0007] Step 1: based on the real-time measurement signals obtained by the vehicle sensor system, the vehicle's center of mass side slip angle and yaw rate are estimated by the unscented Kalman filter UKF algorithm;

[0008] Step 2: based on the estimated center of mass side slip angle, the center of mass side slip angle change rate is calculated, and the six-line method-based The phase-plane stability judgment method assesses the vehicle's stability state. This method employs a six-line approach. The phase plane is divided into three regions: the stable region, the unstable region, and the critical stable region, which correspond to the stable state, critical stable state, and unstable state of the vehicle, respectively.

[0009] Step 3: Based on the current vehicle stability determined in the above steps, different driving intention recognition methods are adopted for different vehicle stability states to obtain the driver's desired motion intensity.

[0010] When the vehicle is in a stable state, the expected motion intensity is calculated directly based on the vehicle's real-time parameters.

[0011] When the vehicle is in a critically stable or unstable state, the desired motion intensity is corrected by adjusting the sideslip angle and the rate of change of the sideslip angle.

[0012] Step 4: The desired motion intensity obtained from the final decision is used as a decision command and output to the actuator control system of the distributed drive vehicle; the actuator control system distributes and controls the torque / speed of the motor of each independent drive wheel and the rotation angle of each independent steering wheel according to the received desired longitudinal acceleration and desired lateral acceleration.

[0013] In a further technical solution, step 1 includes the following specific steps:

[0014] Longitudinal vehicle speed is obtained through vehicle sensor systems. Lateral acceleration and front wheel steering angle Based on the two-degree-of-freedom vehicle model and the magic tire model, the system's state vector, observation vector, and control input vector are defined.

[0015] The state vector is: In the formula, for The state vector at time t, for The sideslip angle of the center of mass at any given moment. for The yaw rate at any given moment;

[0016] The observation vector is: In the formula, for The observation vector at time t, for Lateral acceleration at time;

[0017] The input vector is: In the formula, for The input vector at time step, For the front wheel angle at the moment;

[0018] Based on this, the nonlinear state equation of the system is established:

[0019] ;

[0020] And the nonlinear observation equation:

[0021] ;

[0022] where, and are process noise and observation noise, respectively; is a nonlinear function determined by the differential equation of the vehicle dynamics; is also a nonlinear function;

[0023] According to the state space equation, the Unscented Kalman Filter (UKF) algorithm is used to recursively and real-time estimate the state vector, i.e., the center side slip angle and yaw rate.

[0024] Further technical solutions, in the step 1, the estimation of the state vector by the Unscented Kalman Filter (UKF) algorithm, the specific steps are as follows:

[0025] First, set the initial value of the state vector and the initial value of the error covariance matrix ;

[0026] Secondly, for each estimation moment , according to the state estimation value and the error covariance matrix of the last moment, the Unscented transformation method is used to generate a set of Sigma points with weights ;

[0027] Then, the Sigma point set is propagated through the nonlinear state equation to obtain the predicted Sigma point set , and the prior predictive value and the prior error covariance matrix of the state vector are calculated according to the predicted Sigma point set; at the same time, the predicted Sigma point set is propagated through the nonlinear observation equation to obtain the observation predicted Sigma point set , and the predicted value of the observation vector is calculated according to the observation predicted Sigma point set;

[0028] After obtaining the observation vector at the current moment, the observation predicted covariance matrix and the cross-covariance matrix between the state and the observation and the Kalman gain is calculated , and the formula is ; the priori prediction value is fused with the current actual observation value, and the posteriori estimation value of the state vector is updated , and the update formula is ; the posteriori error covariance matrix is updated ;

[0029] Finally, the posteriori estimation value is output as the estimation result of the current time.

[0030] Further technical solutions, in the step 2, the specific range of the three regions is represented by the following formula:

[0031] The stable region:

[0032] ;

[0033] The critical stable region:

[0034] ;

[0035] The unstable region:

[0036] ;

[0037] In the above formula, is the coefficient of the four boundary straight lines of the stable region; is the correlation coefficient of the two straight lines corresponding to the stable region boundary divided by the double line method; the stable region boundary parameter is determined by the value of the road adhesion coefficient .

[0038] Further technical solutions, in the step 3, when the vehicle is in a stable state, the calculation formula of the expected motion intensity is:

[0039] ;

[0040] ;

[0041] In the formula, is the expected longitudinal acceleration; is the expected transverse acceleration; is the current center of mass speed of the vehicle, which is measured by a sensor; is the estimated current center of mass side slip angle of the vehicle; is the current center of mass side slip angle change rate of the vehicle; is the estimated current yaw rate of the vehicle;

[0042] When the vehicle is in critical stable state and unstable state, the correction process of expected motion intensity is as follows:

[0043] In On the phase diagram, the point corresponding to the current vehicle mass center side slip angle and mass center side slip angle rate of change and the origin line is a straight line, the formula is:

[0044] ;

[0045] This straight line intersects with one of the four stable region boundary lines corresponding to the straight line, here the stable region boundary line corresponding to the straight line is unified as:

[0046] ;

[0047] In the formula, and are the slope and the vertical intercept of the straight line, respectively, according to the intersected line, take or or or , take or or or ;

[0048] Solve the two straight lines together to get the point after correction :

[0049] ;

[0050] In the formula, is the corrected mass center side slip angle, is the corrected mass center side slip angle rate of change;

[0051] The motion intensity calculated according to the corrected mass center side slip angle and mass center side slip angle rate of change is the corrected expected motion intensity, the formula is:

[0052] ;

[0053] ;

[0054] In the formula, is the corrected expected longitudinal acceleration; is the corrected expected lateral acceleration.

[0055] The present invention provides a distributed driving intention recognition method for vehicles based on vehicle stability. This method uses the unscented Kalman filter (UKF) algorithm to estimate key state parameters characterizing the vehicle's dynamic characteristics in real time, namely the centroid sideslip angle and yaw rate, and introduces a six-line method. Phase plane analysis is used to accurately determine the vehicle's stability state. For different stability states, this method does not always use the driver's input as the final control target. When the vehicle is in a stable state, the system tries to preserve the driver's original desired motion intensity; when the vehicle is in a critical or unstable state, the system actively intervenes. On the phase plane, the target point within the stabilization boundary is corrected based on the current unstable state point, and the corrected expected motion intensity is calculated accordingly. This allows the system to not only preserve the driver's original intention as much as possible, but also effectively prevent dangers caused by driver error or vehicle instability, thus improving active safety. This method has a clear flow, low computational load, does not rely on overly complex or difficult-to-obtain sensor signals, and is easy to implement on an onboard computing platform. It provides a practical technical solution for achieving high-performance stability control and driver assistance functions in distributed drive vehicles. Attached Figure Description

[0056] Figure 1 A flowchart of a distributed driving intention recognition method for vehicles based on vehicle stability states provided in an embodiment of the present invention. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0058] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0059] like Figure 1 As shown, a distributed driving intention recognition method for vehicles based on stable vehicle states, provided by an embodiment of the present invention, includes the following steps:

[0060] Step 1: Estimation of the sideslip angle and yaw rate;

[0061] Longitudinal vehicle speed can be obtained through vehicle sensor systems. Lateral acceleration Front wheel steering angle Real-time measurement signals are obtained, and the system's state vector, observation vector, and control input vector are defined based on a two-degree-of-freedom vehicle model and a magic tire model.

[0062] The state vector is: where, is the state vector at time t, is the side slip angle at time t, is the yaw rate at time t.

[0063] The observation vector is: where, is the observation vector at time t, is the lateral acceleration at time t.

[0064] The input vector is: where, is the input vector at time t, is the front wheel steering angle at time t.

[0065] Based on this, the nonlinear state equation of the system is established:

[0066]

[0067] and the nonlinear observation equation is:

[0068]

[0069] where, and are the process noise and observation noise, respectively; is a nonlinear function, which is determined by the differential equation of the dynamics of the vehicle; is also a nonlinear function, for example, the lateral acceleration is related to the side slip angle and the yaw rate as .

[0070] According to the state space equation, the unscented Kalman filter (UKF) algorithm is used to recursively and real-time estimate the state vector, i.e., the side slip angle and the yaw rate.

[0071] The estimation of the state vector by the unscented Kalman filter (UKF) algorithm has five steps:

[0072] First, set the initial value of the state vector and the initial value of the error covariance matrix .

[0073] Second, for each estimation time , according to the state estimation value at the last time​​​​​​​​​ and error covariance matrix , a set of Sigma points with weights is generated by using unscented transformation method .

[0074] Then, the set of Sigma points is propagated through the nonlinear state equation to obtain a set of predicted Sigma points , the prior predictive value of state vector and the prior error covariance matrix are calculated according to the set of predicted Sigma points ; meanwhile, the set of predicted Sigma points is propagated through the nonlinear observation equation to obtain a set of observation predicted Sigma points , the predictive value of observation vector is calculated according to the set of observation predicted Sigma points .

[0075] After the observation vector at the current time (i.e. ) is obtained, the observation predictive covariance matrix and the cross-covariance matrix between state and observation are calculated, and the Kalman gain is calculated, the formula of which is . The prior predictive value is fused with the actual observation value to update the posterior estimation value of state vector , the update formula of which is ; meanwhile, the posterior error covariance matrix is updated.

[0076] Finally, the posterior estimation value is output as the estimation result at the current time.

[0077] Step 2: Based on the estimated center of mass side slip angle, the center of mass side slip angle rate can be calculated, and the phase plane stable state judgment method based on six-line method is used to judge the stable state of the vehicle. This judgment method divides the phase plane into three regions, i.e. stable region, unstable region and critical stable region, which respectively correspond to the stable state, critical stable state and unstable state of the vehicle, and the specific ranges of the three regions are represented by the following formulas:

[0078] Stable region:

[0079] ;

[0080] In the formula, is the coefficient of the four boundary lines of the stable region.

[0081] Critical stable region:

[0082] ;

[0083] wherein, is the correlation coefficient of the two straight lines corresponding to the stable region boundary divided by the two-line method.

[0084] Unstable region:

[0085] ;

[0086] Since the boundary parameters of the six-line method are greatly affected by the road adhesion coefficient , the following Table 1 gives a phase plane stable boundary parameter.

[0087] Table 1

[0088]

[0089] Step 3: The above step estimates the current vehicle mass center side slip angle and the mass center side slip angle change rate to determine the current stable state of the vehicle, and different driving intention recognition methods are adopted for different vehicle stable states, so as to obtain the expected motion intensity of the driver.

[0090] When the vehicle is in a stable state, there is no need to correct the expected motion intensity, and the expected motion intensity can be obtained according to the real-time parameters of the vehicle, and the formula is:

[0091] ;

[0092] ;

[0093] wherein, is the expected longitudinal acceleration; is the expected transverse acceleration; is the current mass center speed of the vehicle, which is measured by a sensor; is the estimated current mass center side slip angle of the vehicle; is the current mass center side slip angle change rate of the vehicle; is the estimated current yaw rate of the vehicle.

[0094] When the vehicle is in a critical stable state and an unstable state, the expected motion intensity needs to be corrected so as to be in a stable state, and the correction process is as follows:

[0095] On the phase diagram, a straight line is obtained by connecting the point corresponding to the current mass center side slip angle and the mass center side slip angle change rate of the vehicle and the origin, and the formula is:

[0096] ;

[0097] This straight line intersects with one of the four straight lines corresponding to the stable region boundary at a point, where the straight lines corresponding to the stable region boundary are collectively referred to as:

[0098] ;

[0099] In the formula, and are the slope and the vertical intercept of the straight line, respectively, depending on the intersecting line, take or or or , take or or or .

[0100] Solving the two straight lines together gives the point after correction :

[0101] ;

[0102] In the formula, is the corrected centroid lateral angle, is the corrected centroid lateral angle rate of change.

[0103] The motion intensity calculated according to the corrected centroid lateral angle and the centroid lateral angle rate of change is the corrected expected motion intensity, and the formula is:

[0104] ;

[0105] ;

[0106] In the formula, is the corrected expected longitudinal acceleration; is the corrected expected lateral acceleration.

[0107] Step 4: The expected motion intensity obtained by the final decision is taken as the decision instruction and output to the actuator control system of the distributed drive vehicle; the actuator control system distributes and controls the torque / speed of the motor of each independent drive wheel and the rotation angle of each independent steering wheel according to the received expected longitudinal acceleration and expected lateral acceleration, and finally realizes the humanized intelligent driving that meets the driver's intention and ensures the stability of the vehicle.

[0108] In the embodiment of the present application, the method acquires the vehicle longitudinal speed, lateral acceleration and front wheel angle signals through sensors, establishes system equations based on a two-degree-of-freedom vehicle model and a magic tire model, and estimates the mass side slip angle and yaw rate in real time by using an unscented Kalman filter; further calculates the mass side slip angle rate, and constructs a six-line method based on the mass side slip angle rate and the mass side slip angle The phase plane is constructed according to the region where the current state point is located, and it is judged that the vehicle is in a stable, critical stable or unstable state; on this basis, the expected motion strength is decided according to the stable state of the vehicle: the expected longitudinal and lateral accelerations are directly calculated in the stable state, the target state is corrected through the intersection point of the phase plane boundary in the unstable and critical states, and the expected accelerations after correction are calculated; finally, the obtained expected motion strength is output to an actuator control system, the collaborative control of the distributed driving vehicle wheels is realized, and thus the human-like driving intention tracking is realized under the premise of ensuring stability.

[0109] The above merely describes the preferred embodiments of the present application and is not intended to limit the present application, and any modification, equivalent replacement and improvement made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A distributed driving intention recognition method for vehicles based on stable vehicle states, characterized in that, Includes the following steps: Step 1: Based on the real-time measurement signals acquired by the vehicle sensor system, estimate the vehicle's center of gravity sideslip angle and yaw rate using the unscented Kalman filter (UKF) algorithm; Step 2: Based on the estimated centroid sideslip angle, calculate the rate of change of the centroid sideslip angle using the six-line method. The phase-plane stability judgment method assesses the vehicle's stability state. This method employs a six-line approach. The phase plane is divided into three regions: the stable region, the unstable region, and the critical stable region, which correspond to the stable state, critical stable state, and unstable state of the vehicle, respectively. Step 3: Based on the current vehicle stability determined in the above steps, different driving intention recognition methods are adopted for different vehicle stability states to obtain the driver's desired motion intensity. When the vehicle is in a stable state, the expected motion intensity is calculated directly based on the vehicle's real-time parameters. When the vehicle is in a critically stable or unstable state, the desired motion intensity is corrected by adjusting the sideslip angle and the rate of change of the sideslip angle. Step 4: The desired motion intensity obtained from the final decision is used as a decision command and output to the actuator control system of the distributed drive vehicle; the actuator control system distributes and controls the torque / speed of the motor of each independent drive wheel and the rotation angle of each independent steering wheel according to the received desired longitudinal acceleration and desired lateral acceleration. In step 3, when the vehicle is in a stable state, the formula for calculating the desired motion intensity is: ； ； In the formula, The desired longitudinal acceleration; The desired lateral acceleration; The current center-of-gravity velocity of the vehicle is measured by sensors. To estimate the current sideslip angle of the vehicle's center of gravity; This represents the current rate of change of the vehicle's center of gravity sideslip angle. To estimate the current yaw rate of the vehicle; In step 3, the correction process for the desired motion intensity when the vehicle is in a critically stable state and an unstable state is as follows: exist On the phase diagram, the points corresponding to the vehicle's current sideslip angle and the rate of change of the sideslip angle are... A straight line is formed by connecting the origin to the origin, and the formula is: ； This straight line intersects at a point with one of the four lines corresponding to the boundary of the stable region. Here, the lines corresponding to the boundary of the stable region are unified as follows: ； In the formula, and These are the slope and y-intercept of the line, respectively. Solve the system of equations for the corrected points by combining the two lines. : ； In the formula, This is the corrected centroid sideslip angle. This is the corrected rate of change of the centroid sideslip angle; The motion intensity calculated based on the corrected sideslip angle and the rate of change of the sideslip angle is the corrected expected motion intensity, and the formula is: ； ； In the formula, This is the corrected desired longitudinal acceleration; This is the corrected desired lateral acceleration.

2. The distributed driving intention recognition method for vehicles based on stable vehicle states according to claim 1, characterized in that, Step 1 includes the following specific steps: Longitudinal vehicle speed is obtained through vehicle sensor systems. Lateral acceleration and front wheel steering angle Based on the two-degree-of-freedom vehicle model and the magic tire model, the system's state vector, observation vector, and control input vector are defined. The state vector is: In the formula, for The state vector at time t, for The sideslip angle of the center of mass at any given moment. for The yaw rate at any given moment; The observation vector is: In the formula, for The observation vector at time t, for Lateral acceleration at time; The input vector is: In the formula, for The input vector at time t, for The front wheel angle at that moment; Based on this, the nonlinear state equations of the system are established: ； And nonlinear observation equations: ； In the formula, and These are process noise and observation noise, respectively. It is a nonlinear function, determined by the vehicle's dynamic differential equation; It is also a nonlinear function; Based on the state-space equations, the Unscented Kalman Filter (UKF) algorithm is used to recursively and in real-time estimate the state vectors, namely the centroid sideslip angle and yaw rate.

3. The distributed driving intention recognition method for vehicles based on vehicle stability state according to claim 2, characterized in that, In step 1, the Unscented Kalman Filter (UKF) algorithm estimates the state vector, and the specific steps are as follows: First, set the initial values ​​of the state vector. Initial values ​​of the error covariance matrix ; Secondly, for each estimated time point Based on the state estimate from the previous moment And error covariance matrix A set of weighted Sigma points is generated using the unscented transformation method. ; Subsequently, the Sigma point set is propagated through a nonlinear state equation to obtain the predicted Sigma point set. Calculate the prior predicted value of the state vector based on the predicted Sigma point set. and prior error covariance matrix Simultaneously, the predicted Sigma point set is propagated through a nonlinear observation equation to obtain the observed predicted Sigma point set. The predicted value of the observation vector is calculated based on the observed prediction Sigma point set. ; Next, obtain the observation vector at the current time. Then, the observation-prediction covariance matrix is ​​calculated. and the cross-covariance matrix between state and observation And calculate the Kalman gain. Its calculation formula is The prior predictions are fused with the current actual observations to update the posterior estimate of the state vector. The updated formula is as follows Simultaneously update the posterior error covariance matrix. ; Finally, the posterior estimate is... Output the estimated result at the current moment.

4. The distributed driving intention recognition method for vehicles based on vehicle stability state according to claim 3, characterized in that, In step 2, the specific ranges of the three regions are represented by the following formula: Stable region: ； Critical stability region: ； Unstable region: ； In the above formula, These are the coefficients of the lines corresponding to the four boundaries of the stable region; The correlation coefficient between the two straight lines corresponding to the boundary of the stable zone using the double-line method; the boundary parameters of the stable zone are determined by the road surface adhesion coefficient. The value to be determined.

Images