A fuzzy neural network H∞ yaw moment controller based on phase plane analysis

Abstract

The application discloses a fuzzy neural network H∞ yaw moment controller based on phase plane analysis and belongs to the technical field of drive-by-wire chassis and vehicle stability control. The application quantitatively describes the real-time stability boundary in the vehicle driving process by the phase plane analysis technology, realizes accurate evaluation of the overall stability state of the system, designs the yaw moment controller by using the H∞ robust control theory, avoids the system chattering problem from the design level, guarantees the robustness of the control system, simultaneously introduces the fuzzy neural network online approximation controller parameter and the nonlinear mapping relationship between the vehicle driving state, realizes the autonomous optimization of the controller parameters under different working conditions, and thus the driving stability and adaptability of the distributed drive vehicle under the complex driving environment are significantly improved.

Description

Technical Field

[0001] This invention belongs to the field of drive-by-wire chassis and vehicle stability control technology, specifically relating to a yaw moment control method for distributed drive vehicles, and more particularly to a stability control strategy that integrates phase plane analysis, fuzzy neural networks and H∞ robust control. Background Technology

[0002] Distributed drive vehicles, with their unique advantage of independently controllable driving torque for all four wheels, offer the possibility of achieving breakthroughs in vehicle dynamics performance. However, under complex and variable driving conditions, such as high-speed driving, emergency steering, strong crosswinds, or sudden changes in road surface adhesion coefficient, the vehicle's body posture is prone to instability, leading to dangerous situations such as lateral slippage or even fishtailing. Therefore, designing a control method that can improve vehicle stability and adaptability under extreme conditions has become a research hotspot in the field of drive-by-wire chassis. Existing technologies for vehicle stability control mainly include the following implementation schemes:

[0003] 1. Stability control method based on tire nonlinear critical constraints The core principle of this method is to strictly limit the tire's operating state within the linear range of its mechanical properties. By monitoring tire force in real time, when it is determined to be close to the saturation point, the controller intervenes to adjust the driving or braking force to prevent the tire force from entering the nonlinear region and causing vehicle instability. The effectiveness of this method highly depends on the accuracy of the established vehicle dynamics model and tire model, striving to accurately predict the critical state of tire force through the model.

[0004] 2. Stability control method based on active rear-wheel steering This type of approach calculates an additional rear wheel steering angle by designing a feedback control law. This additional steering angle is designed to enable the vehicle's actual state (such as sideslip angle and yaw rate) to track the state output by a pre-defined ideal reference model. By actively controlling the rear wheels to generate additional lateral force, this method attempts to compensate for the vehicle's understeer or oversteer tendencies during steering, thereby pulling the vehicle back to an ideal stable trajectory.

[0005] 3. Yaw moment control method using sliding mode variable structure control theory This type of method has a similar objective to active rear-wheel steering control, namely, to directly intervene in the vehicle's yaw motion by calculating an additional yaw moment. Sliding mode control, as a typical nonlinear robust control method, focuses on designing a sliding surface and using a high-frequency switching control law to force system state variables (such as yaw rate error) to converge to zero along this sliding surface. This method exhibits strong invariance to system parameter perturbations and external disturbances.

[0006] However, the above-mentioned existing technical solutions still have the following shortcomings in practical applications: 1. The drawback of the existing scheme 1 (tire nonlinear critical constraint control) is that its control performance is overly dependent on the accuracy of the vehicle and tire models. When the vehicle's driving environment changes, such as a sudden change in the road adhesion coefficient or encountering crosswind interference, the critical saturation point of the tire force will change accordingly, causing the predefined constraint conditions to fail, resulting in distorted judgments and an inability to effectively prevent instability. In addition, this method focuses on monitoring the working state of a single tire and lacks a quantitative description of the overall stability of the vehicle system, making it difficult to accurately determine the global instability risk of the vehicle.

[0007] 2. The drawback of the existing solution 2 (active rear-wheel steering control) is that under extreme driving conditions (such as extreme cornering or low-traction surfaces), the tires enter a highly nonlinear region, and the lateral forces they generate approach saturation, drastically reducing the available control margin. At this point, relying solely on the additional lateral forces generated by active rear-wheel steering is insufficient to guarantee the vehicle's stability margin, and the vehicle still faces the risk of slippage or fishtailing. Furthermore, this method is essentially a state-tracking control, which also lacks a quantitative description of the overall stability of the vehicle system.

[0008] 3. The drawback of the existing scheme 3 (sliding mode variable structure control) is that the inherent high-frequency switching characteristic of sliding mode control can cause "chattering" in the system. In vehicle stability control, which places stringent demands on actuator and system response, chattering introduces system uncertainty, exacerbates actuator load, and may excite unmodeled high-frequency dynamics of the vehicle, affecting control performance and ride comfort. Although some researchers have improved sliding mode control by introducing fuzzy logic or replacing symbolic functions, these methods usually only reduce chattering but cannot completely eliminate it, and they significantly increase the computational burden on the controller.

[0009] To address the aforementioned issues, this invention proposes a fuzzy neural network H∞ yaw moment controller based on phase plane analysis. Summary of the Invention

[0010] The purpose of this invention is to provide a fuzzy neural network H∞ yaw moment controller based on phase plane analysis to solve the following problems: (1) Traditional control methods (such as tire critical constraint control and state tracking control) lack quantitative description of the overall stability of the vehicle system, resulting in insufficient adaptability under complex and ever-changing driving conditions. (2) Some nonlinear control methods (such as sliding mode control) have inherent system chattering problems, which affect control quality and system stability.

[0011] To achieve the above objectives, the present invention adopts the following technical solution: A fuzzy neural network H∞ yaw moment controller based on phase plane analysis includes: The phase plane analysis module is configured to construct a phase plane diagram based on the vehicle's two-degree-of-freedom dynamics model and magic tire model, and to use the bi-line method to divide the stable and unstable regions of the phase plane in order to calculate the weight coefficients characterizing the stability of the vehicle's current driving state. The fuzzy neural network weight self-adjustment module is configured to take vehicle speed and road surface adhesion coefficient as inputs and use the fuzzy neural network to output the state variable weight parameters and control variable weight parameters of the H∞ controller. The H∞ yaw moment controller is configured to construct the system state space equation based on the deviation between the actual and ideal values ​​of the vehicle state variables. It then solves the state feedback control law and outputs the additional yaw moment by combining the state variable weight parameters and control variable weight parameters output by the fuzzy neural network weight self-adjustment module with the weight coefficients output by the phase plane analysis module.

[0012] Preferably, the phase plane analysis module is further configured as follows: A phase plane with the sideslip angle and sideslip angular velocity of the center of mass as coordinate axes is constructed based on the two-degree-of-freedom dynamics model of the vehicle and the magic tire model. The stability boundary of the phase plane is determined by the biline method, and the phase plane is divided into stable and unstable regions. Based on the vehicle's current sideslip angle and sideslip velocity on the phase plane, calculate its distance to the center of stability and the nearest distance to the stability boundary equation, and calculate the weighting coefficients based on the distances.

[0013] Preferably, the fuzzy neural network weight self-adjustment module includes a five-layer network structure: The first layer is the input layer, which is used to fuzzify the vehicle speed and road surface adhesion coefficient. The Gaussian membership function is used to calculate the membership degree corresponding to the input quantity. The second layer is the rule layer, which is used to calculate the trigger strength of each rule based on preset fuzzy rules; The third layer is the normalization layer, which is used to normalize the trigger strength calculation; The fourth layer is the consequent layer, which is used to calculate the output of each rule based on the normalized trigger strength and adjustable parameters; The fifth layer is the output layer, which is used to calculate the final output state variable weight parameters and control variable weight parameters using a weighted average method.

[0014] Preferably, the H∞ yaw moment controller is further configured as follows: The system state variables are defined as the difference between the actual center-of-mass sideslip angle and the ideal center-of-mass sideslip angle, and the difference between the actual yaw rate and the ideal yaw rate. Using the additional yaw moment as the control variable and the front and rear wheel steering angles as disturbance terms, the system state-space equations are constructed. The state feedback gain matrix is ​​obtained by solving a convex optimization problem that satisfies the linear matrix inequality constraint. The additional yaw moment is calculated based on the state feedback gain matrix and the system state variables.

[0015] Preferably, the ideal yaw rate and the ideal center-of-gravity sideslip angle are determined based on the dynamic relationship of the vehicle during steady-state steering and the road surface adhesion conditions.

[0016] Preferably, the H∞ yaw moment controller is further configured to use the weighting coefficients output by the phase plane analysis module as weighting factors to weight the state variables in the system state space equation.

[0017] Preferably, the vehicle speed ranges from 30 km / h to 120 km / h, and the road surface adhesion coefficient ranges from 0.3 to 0.8.

[0018] Preferably, the two-degree-of-freedom vehicle dynamics model is a linearized model based on Newton's second law, used to describe the vehicle's lateral and yaw motions.

[0019] Preferably, the magic tire model is used to calculate tire lateral forces, and its inputs include tire vertical load and tire slip angle.

[0020] The present invention further protects a vehicle yaw stability control system, including the fuzzy neural network H∞ yaw moment controller based on phase plane analysis as proposed above.

[0021] Compared with the prior art, the present invention has at least the following beneficial effects: (1) Due to the use of phase plane analysis, compared with traditional stability control strategies, such as tire nonlinear critical constraint stability control and active rear wheel steering stability control strategies, this invention realizes real-time quantitative description of the overall vehicle stability and improves the accuracy of the controller.

[0022] (2) The present invention designs a yaw stability control strategy based on the H∞ state feedback robust control algorithm, abandoning the traditional stability control method using sliding mode variable structure control theory. Therefore, the chattering problem caused by discontinuous switching of the system will not occur, thus improving the driving stability of distributed drive vehicles.

[0023] (3) The present invention uses a fuzzy neural network architecture to realize the adaptive adjustment of the H∞ controller parameters. Different driving conditions (vehicle speed, road adhesion coefficient) are used as inputs to the fuzzy neural network to solve for the optimized H∞ controller state weight parameters. Qand control quantity weight parameters R This improves the vehicle's adaptive capabilities in complex driving environments. However, traditional stability control strategies lack parameter adjustment methods and have poor algorithm robustness, thus failing to guarantee the vehicle's stability when external driving conditions change. Attached Figure Description

[0024] Figure 1 This refers to the two-degree-of-freedom vehicle dynamics model proposed in Embodiment 1 of the present invention; Figure 2 This is a flowchart of the fuzzy neural network proposed in Embodiment 1 of the present invention; Figure 3 This is a schematic diagram of the phase plane stable region division proposed in Embodiment 1 of the present invention; Figure 4 This is a flowchart of the yaw moment controller proposed in Embodiment 1 of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0026] This invention proposes an adaptive H∞ yaw moment controller based on phase plane analysis for distributed drive vehicles. The H∞ yaw moment controller calculates the yaw moment based on the vehicle's real-time stable state through phase plane analysis, thereby increasing vehicle stability. Traditional stability control strategies, such as tire nonlinear critical constraint-based stability control and active rear-wheel steering stability control strategies, lack quantitative evaluation of overall vehicle stability and cannot calculate system control quantities based on the vehicle's real-time state. Furthermore, yaw moment control using sliding mode variable structure control theory suffers from chattering issues that cannot be eliminated, easily increasing system uncertainty in vehicle stability control problems and reducing algorithm control accuracy.

[0027] Furthermore, to improve the adaptive capability of distributed drive vehicles to different driving environments, this invention employs a fuzzy neural network to adaptively adjust the H∞ controller parameters, thereby improving the accuracy of the control algorithm. Vehicle speed and road adhesion coefficient are used as inputs to the fuzzy neural network structure, and the H∞ controller state weights... Q Weights of control quantities RTo improve the vehicle's adaptability to different driving conditions, the parameters of the yaw moment controller are adaptively adjusted. Traditional stability control strategies, however, lack parameter adjustment mechanisms, making them ill-suited for complex and variable driving conditions and exhibiting low algorithm robustness.

[0028] The following description, in conjunction with relevant accompanying drawings and specific examples, illustrates a fuzzy neural network H∞ yaw moment controller based on phase plane analysis proposed in this invention, specifically including the following content.

[0029] Example 1: This invention proposes a fuzzy neural network H∞ yaw moment controller based on phase plane analysis, comprising: (I) Two-degree-of-freedom dynamic model of the vehicle A two-degree-of-freedom dynamic model is used to describe the lateral and yaw dynamics of the vehicle. It assumes small wheel steering angles and ignores the difference in steering angles between the inner and outer wheels. The corresponding two-degree-of-freedom model is as follows: Figure 1 As shown.

[0030] Based on Newton's second law, the simplified two-degree-of-freedom dynamic equation is shown in equation (1). (1) In the formula, For the overall vehicle quality, and These are the front and rear wheelbases, and These are the lateral forces of the front and rear tires, For the moment of inertia of yaw rotation, and These are the vehicle's longitudinal and lateral speeds, respectively. Let yaw rate be angular velocity. When the tire operates within the linear region, the following relationship holds: (2) in, and These are the lateral stiffness of the front and rear tires, respectively. and Let these be the front and rear wheel slip angles when the vehicle is in motion, respectively. Figure 1 The angular relationship (2) can be expressed as: (3) Defined by the centroid sideslip angle Introducing additional yaw moment The two-degree-of-freedom differential equation for the vehicle is shown below: (4) When the vehicle maintains steady steering... and Considering the limitations of road surface adhesion conditions during actual driving, the ideal yaw rate and the center of gravity sideslip angle are as shown in equation (5): (5) The system state variables are defined as the difference between the actual centroid sideslip angle and the yaw rate and the corresponding ideal variables based on the phase plane weight allocation, as shown in Equation (6).

[0031] (6) The control quantity is the additional yaw rate. The interference item is the front and rear wheel steering angles. The system state-space equation is then shown in equation (7): (7) in: , , , , ,

[0032] (II) H∞ Yaw Stability Control Strategy The H∞ control algorithm is a robust control strategy suitable for controlling highly nonlinear systems such as vehicles. Its essence is to solve for the state feedback control quantity. This minimizes the H∞ norm of the closed-loop transfer function matrix of the system, effectively suppressing external disturbances. The closed-loop transfer function... , This indicates the level of disturbance suppression. When a symmetric positive definite matrix exists... And the system that satisfies equation (8) is asymptotically stable.

[0033] (8) Introducing feedback control quantity The above formula can be converted to: (9) To solve for the positive definite matrix With feedback gain ,make The above problem is transformed into a convex optimization problem with linear matrix inequality constraints (LMI), and then... The above equation is transformed by congruence as shown in equation (10).

[0034] (10) make , The above expression can then be equivalent to: (11) The transformation relationship from state equation to transfer function is as follows: Then the closed-loop transfer function can be expressed as ,because Then the transfer function is equivalent to .Will , and Replace with , , ,pass Transforming the above equation using congruences, we get: (12) The system control problem is transformed into solving the optimal solution. and The optimization problem is expressed as: (13) The expression for the final feedback control law of the controller is: (14) (III) Fuzzy Neural Networks The first layer of the fuzzy neural network architecture is the input layer, which fuzzifies the input variables. Each node in the first layer represents a logical linguistic value, and a Gaussian membership function is chosen to convert the precise value into the corresponding fuzzy set membership degree. (15) in, and These represent the number of input variables and the number of fuzzy variables, respectively. The membership function corresponding to the fuzzy set. and These represent the centrality and width of the membership function, respectively.

[0035] The second layer calculates the trigger strength based on the rules corresponding to each node: (16) In the above formula Indicates the first The trigger strength of the rule, and These are the first corresponding input quantities. The membership degree of each rule. The third layer completes the normalization calculation of the excitation intensity: (17) The fourth layer performs consequent value calculation, outputs each rule, and completes fuzzy inference: (18) In the above formula , as well as The first The adjustment parameters of the rule are as follows: the fifth layer uses a weighted average rule to control the output quantity as shown in equation (19). The process of the fuzzy neural network based on the five-layer architecture is as follows: Figure 2 As shown.

[0036] (19) By setting different driving conditions, the final control quantity is calculated based on the linear matrix inequality constraint (LMI). The vehicle speed range was selected to be 30 km / h to 120 km / h, and the road surface adhesion coefficient range was 0.3 to 0.8. The neural network input was set to vehicle speed and road surface adhesion coefficient, and the output was the state variables and control weight parameters of the H∞ controller. This enables adaptive adjustment of controller parameters under different vehicle driving conditions.

[0037] (iv) Phase plane analysis For a highly nonlinear system like a vehicle, the dynamic characteristics of the tires and suspension, as well as external disturbances such as crosswinds and road adhesion conditions, all increase the system's uncertainty during operation. Traditional linear analysis methods cannot effectively capture the vehicle's dynamic behavior; therefore, phase plane analysis is introduced to quantitatively describe vehicle stability. The phase plane method is a graphical analytical method that does not require solving dynamic equations; the stability of the system can be intuitively reflected through the phase trajectory of the system state. Based on the bilinear method, system state points in the stable region converge to the origin, while state points in the unstable region diverge.

[0038] A second-order autonomous system is defined as shown in equation (20): (20) In the above formula At this point, a second-order autonomous system is equivalent to: (twenty one) According to the definition of phase plane theory, the direction of the tangent to the trajectory of the system's state points is the direction of the phase trajectory, expressed as: When the vehicle's nonlinear system simultaneously satisfies as well as When the equilibrium point is reached, the state point is either an equilibrium point or a saddle point; otherwise, it is a constant point. Within the stable region, the system state points converge to the equilibrium point. The accurate description of tire lateral force is achieved by combining the magic tire model, as shown in equation (22): (twenty two) in The fitting coefficients are given by the following formula: The change in vertical load on the tires when the vehicle is in motion is: (twenty three) Combining the two-degree-of-freedom dynamics model of the vehicle and equations (20) to (23), based on the bilinear method ( ) Delineate the stable boundary of the phase trajectory as follows Figure 3 As shown.

[0039] The stable and unstable regions of the phase plane are represented as follows: (twenty four) in and These are the road surface adhesion coefficient and the acceleration due to gravity, respectively. and These are the vehicle stability coefficient and wheelbase, respectively. The formula for calculating the system state variable weighting coefficients based on phase plane analysis is shown in equation (25). (25) In the above formula The distance from the system state point to the stability center of the phase plane. This represents the shortest distance from the system state point to the phase plane stability boundary equation. The flowchart of the fuzzy neural network H∞ yaw moment controller based on phase plane analysis is as follows: Figure 4 As shown.

[0040] In summary, this invention quantitatively describes the real-time stable state of the entire system by analyzing and fitting the stability boundary of vehicle driving using phase plane analysis. A yaw moment stability controller is designed using H∞ robust control theory to avoid system chattering. A fuzzy neural network is employed to approximate the nonlinear relationship between controller parameters and vehicle driving state, enabling controller parameter optimization under different operating conditions and thus improving the vehicle's adaptive capability.

[0041] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A fuzzy neural network H∞ yaw moment controller based on phase plane analysis, characterized in that, include: The phase plane analysis module is configured to construct a phase plane diagram based on the vehicle's two-degree-of-freedom dynamics model and magic tire model, and to use the bi-line method to divide the stable and unstable regions of the phase plane in order to calculate the weight coefficients characterizing the stability of the vehicle's current driving state. The fuzzy neural network weight self-adjustment module is configured to take vehicle speed and road surface adhesion coefficient as inputs and use the fuzzy neural network to output the state variable weight parameters and control variable weight parameters of the H∞ controller. The H∞ yaw moment controller is configured to construct the system state space equation based on the deviation between the actual and ideal values ​​of the vehicle state variables. It then solves the state feedback control law and outputs the additional yaw moment by combining the state variable weight parameters and control variable weight parameters output by the fuzzy neural network weight self-adjustment module with the weight coefficients output by the phase plane analysis module.

2. The fuzzy neural network H∞ yaw moment controller based on phase plane analysis according to claim 1, characterized in that, The phase plane analysis module is further configured as follows: A phase plane with the sideslip angle and sideslip angular velocity of the center of mass as coordinate axes is constructed based on the two-degree-of-freedom dynamics model of the vehicle and the magic tire model. The stability boundary of the phase plane is determined by the biline method, and the phase plane is divided into stable and unstable regions. Based on the vehicle's current sideslip angle and sideslip velocity on the phase plane, calculate its distance to the center of stability and the nearest distance to the stability boundary equation, and calculate the weighting coefficients based on the distances.

3. The fuzzy neural network H∞ yaw moment controller based on phase plane analysis according to claim 1, characterized in that, The fuzzy neural network weight self-adjustment module includes a five-layer network structure: The first layer is the input layer, which is used to fuzzify the vehicle speed and road surface adhesion coefficient. The Gaussian membership function is used to calculate the membership degree corresponding to the input quantity. The second layer is the rule layer, which is used to calculate the trigger strength of each rule based on preset fuzzy rules; The third layer is the normalization layer, which is used to normalize the trigger strength calculation; The fourth layer is the consequent layer, which is used to calculate the output of each rule based on the normalized trigger strength and adjustable parameters; The fifth layer is the output layer, which is used to calculate the final output state variable weight parameters and control variable weight parameters using a weighted average method.

4. The fuzzy neural network H∞ yaw moment controller based on phase plane analysis according to claim 1, characterized in that, The H∞ yaw moment controller is further configured as follows: The system state variables are defined as the difference between the actual center-of-mass sideslip angle and the ideal center-of-mass sideslip angle, and the difference between the actual yaw rate and the ideal yaw rate. Using the additional yaw moment as the control variable and the front and rear wheel steering angles as disturbance terms, the system state-space equations are constructed. The state feedback gain matrix is ​​obtained by solving a convex optimization problem that satisfies the linear matrix inequality constraint. The additional yaw moment is calculated based on the state feedback gain matrix and the system state variables.

5. The fuzzy neural network H∞ yaw moment controller based on phase plane analysis according to claim 4, characterized in that, The ideal yaw rate and ideal center-of-gravity sideslip angle are determined based on the dynamic relationship of the vehicle during steady-state steering and the road surface adhesion conditions.

6. The fuzzy neural network H∞ yaw moment controller based on phase plane analysis according to claim 1, characterized in that, The H∞ yaw moment controller is also configured to use the weighting coefficients output by the phase plane analysis module as weighting factors to weight the state variables in the system state space equation.

7. The fuzzy neural network H∞ yaw moment controller based on phase plane analysis according to claim 1, characterized in that, The vehicle speed ranges from 30 km / h to 120 km / h, and the road surface adhesion coefficient ranges from 0.3 to 0.

8.

8. The fuzzy neural network H∞ yaw moment controller based on phase plane analysis according to claim 1, characterized in that, The vehicle's two-degree-of-freedom dynamics model is a linearized model based on Newton's second law, used to describe the vehicle's lateral and yaw motions.

9. The fuzzy neural network H∞ yaw moment controller based on phase plane analysis according to claim 1, characterized in that, The magic tire model is used to calculate tire lateral forces, and its inputs include tire vertical load and tire slip angle.

10. A vehicle yaw stability control system, characterized in that, Includes the fuzzy neural network H∞ yaw moment controller based on phase plane analysis as described in any one of claims 1 to 9.

Images