A multi-agent based active suspension output consensus control method

Abstract

The application discloses a kind of based on multi-agent active suspension output consistency control method, comprising: establishing vehicle dynamics model, tire dynamics model;Establish the vehicle attitude control model and vehicle vibration coupling model;Design vehicle attitude controller, determine the vehicle vibration decoupling performance index;Get the desired output of active suspension system under different driving conditions;Establish the dynamics model of active suspension system based on multi-agent;The desired output of active suspension is obtained, and the inconsistent error between suspension subsystems is eliminated through force synchronization controller, to ensure the consistency of active suspension system output.The application weakens the mutual influence of suspension system vibration in each direction on the basis of guaranteeing the stability and safety of vehicle, which is conducive to damping performance, and improves the comfort of vehicle.

Description

Technical Field

[0001] This invention belongs to the field of active suspension system control technology, specifically relating to an active suspension output consistency control method based on multiple agents. Background Technology

[0002] As a crucial component of a vehicle's driving system, the performance of the suspension system directly impacts the vehicle's stability and safety. Vehicles equipped with active suspension systems can control the vehicle's posture in real time based on terrain information, thereby improving ride comfort, lateral stability, and ultimately enhancing driving safety.

[0003] In existing vehicle active suspension system control technologies, Chinese invention patent application number CN202011120014.9, entitled "A Top-Level Vehicle Body Attitude Control Method Based on Pitch and Roll Force Compensation," discloses a method that uses a suspension controller to determine whether the ideal damping force is within the effective damping range provided by the shock absorber. For ideal damping forces exceeding the effective damping range, corrections are made. A logic threshold control method is used to determine the acceleration state, and based on this, pitch force weighting coefficients and roll force weighting coefficients are calculated to ensure that the ideal damping force remains within the effective range provided by the shock absorber. This solves the problem of deteriorating vehicle body attitude during driving, improving ride comfort and safety. Chinese invention patent application number CN201911197599.1, entitled "An Active Suspension Control Method and System," discloses a method that estimates the desired vehicle posture by observing the driving road and the vehicle's posture to achieve ride comfort. Combined with the current vehicle posture state, a suspension posture command is calculated. The chassis controller independently adjusts the actuation mechanisms of each suspension, enabling the vehicle to adapt to changes in road surface undulations, maintain good overall vehicle ride comfort, and improve vehicle passenger comfort.

[0004] However, none of the aforementioned patented technologies consider the multi-directional vibration coupling problem during vehicle operation or the inconsistency in output between suspension subsystems. When a vehicle is driving on the road, the vibrations generated in multiple directions exhibit coupling behavior, with vibrations in each direction influencing each other, which is detrimental to improving driving comfort and vehicle safety. On the other hand, during the control of the vehicle's active suspension system, the asynchronous responses between the various suspension actuators prevent the desired suspension output from being achieved simultaneously. This results in inconsistent force output in the active suspension system, affecting the overall vehicle ride comfort and thus impacting vehicle stability and safety. Therefore, further research and development require analyzing the coupling behavior of vehicle body vibrations, considering the asynchronous responses between suspension actuators, and taking corresponding measures to address the inconsistency in force output, in order to achieve superior overall vehicle ride comfort and further improve vehicle stability and safety. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the present invention aims to provide a multi-agent-based active suspension output consistency control method to solve the problems of multi-directional vibration coupling and inconsistent outputs between suspension subsystems. This invention proposes a layered vehicle attitude control strategy. The upper layer performs decoupling analysis on vertical vibration, pitch vibration, and roll vibration of the vehicle and designs a vehicle attitude controller. The lower layer, based on multi-agent theory, treats the vehicle's active suspension system as a multi-agent system. By designing a suitable multi-agent system consistency control protocol, the active suspension system can ensure consistent final state while achieving optimal system performance, thereby improving vehicle stability and safety during driving.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] The present invention provides a multi-agent-based active suspension output consistency control method, comprising the following steps:

[0008] (1) Establish the vehicle dynamics model and the tire dynamics model;

[0009] (2) Using the vehicle dynamics model and tire dynamics model established in step (1), establish the vehicle attitude control model and the vehicle vibration coupling model.

[0010] (3) Design a vehicle attitude controller using the vehicle attitude control model and vehicle vibration coupling model established in step (2); perform decoupling analysis on the vehicle vertical vibration, pitch vibration and roll vibration according to the vehicle vibration coupling model, and determine the vehicle vibration decoupling performance index; the vehicle vibration decoupling performance index and the vehicle desired attitude performance index form the controller optimization target, and obtain the desired output of the active suspension system under different driving conditions by solving the optimization target in real time.

[0011] (4) Establish a dynamic model of the active suspension system based on multi-agent systems;

[0012] (5) Using the dynamic model of the active suspension system established in step (4) and the vehicle attitude controller designed in step (3), design an output consistency controller; the output consistency controller includes a force tracking controller and a force synchronization controller. The force tracking controller tracks the expected output of the active suspension obtained by solving the vehicle attitude controller designed in step (3), and the force synchronization controller eliminates the inconsistency error between the suspension subsystems, thereby ensuring the consistency of the output of the active suspension system.

[0013] Furthermore, the specific steps for establishing the vehicle dynamics model and tire dynamics model in step (1) are as follows:

[0014] (11) Vehicle dynamics model:

[0015] Ignoring crosswinds and longitudinal motion of the vehicle during driving, and considering only front wheel steering, the equations of motion for each degree of freedom in the vehicle dynamics model can be obtained by applying d'Alembert's principle as follows:

[0016] The equation of motion for vehicle body lateral slip is:

[0017] (1);

[0018] In the formula, For the overall vehicle weight; For the sprung mass; For vehicle speed; It is the centroid sideslip angle; This refers to the yaw rate; This is the height of the roll center; This refers to the vehicle body roll angle; These are the lateral forces of the left front tire, right front tire, left rear tire, and right rear tire, respectively.

[0019] The equation of motion for the vehicle's yaw is:

[0020] (2);

[0021] In the formula, This represents the yaw moment of inertia of the entire vehicle. These are the distances from the vehicle's center of gravity to the front and rear axles, respectively.

[0022] The equation of motion for vehicle body roll is:

[0023] (3);

[0024] In the formula, The moment of inertia due to vehicle roll; The lateral slope angle of the road surface; These are the vertical forces acting on the vehicle body from the left front suspension, right front suspension, left rear suspension, and right rear suspension, respectively. The wheelbase of the vehicle;

[0025] The equation of motion for vehicle body pitch is:

[0026] (4);

[0027] In the formula, The moment of inertia of the vehicle body during pitch; The vehicle body pitch angle;

[0028] The equation of motion for the vertical motion at the center of mass of the vehicle body is:

[0029] (5);

[0030] In the formula, This represents the vertical displacement of the center of mass of the sprung mass.

[0031] (12) Tire dynamics model:

[0032] Ignoring changes in tire characteristics due to load variations and the effect of tire self-aligning torque, under small steering angle conditions, the tire characteristics are linear, and the tire's lateral force is expressed as:

[0033] (6);

[0034] (7);

[0035] (8);

[0036] In the formula, This refers to the front wheel steering angle; and These are the lateral stiffness of the front and rear wheels, respectively. and These are the side slip angles of the front and rear wheels, respectively. and These are the roll steering coefficients for the front and rear wheels, respectively.

[0037] Tire vertical load is expressed as:

[0038] (9);

[0039] In the formula, , , , The vertical loads are for the left front tire, right front tire, left rear tire, and right rear tire, respectively. , , , These are the stiffnesses of the left front tire, right front tire, left rear tire, and right rear tire, respectively. , , , The road surface input excitations are for the left front tire, right front tire, left rear tire, and right rear tire, respectively. , , , The unsprung mass vertical displacements at the left front tire, right front tire, left rear tire, and right rear tire are respectively.

[0040] Furthermore, the specific steps for establishing the vehicle attitude control model and the vehicle vibration coupling model in step (2) are as follows:

[0041] (21) Establish a vehicle attitude control model;

[0042] (211) The specific steps for establishing the whole vehicle system prediction model are as follows;

[0043] Using the dynamic model established in steps (11) and (12), the prediction model of the whole vehicle system is established as follows:

[0044] (10);

[0045] The whole vehicle system prediction model is represented in state space, and the system's state variables are taken. for:

[0046] ;

[0047] Control input for:

[0048] ;

[0049] Control output for:

[0050] ;

[0051] The state space representation of the whole vehicle system prediction model is as follows:

[0052] (11);

[0053] In the formula, It is an 8×8 dimensional system matrix; It is an 8×4 dimensional control input matrix; The output state matrix is ​​5×8 dimensional; each matrix can be obtained from the vehicle system prediction model.

[0054] (212) Establish a vehicle attitude control model;

[0055] The state space of the vehicle system prediction model is discretized using the first-order difference quotient method. The discretized state space equation is expressed as:

[0056] (12);

[0057] in,

[0058] ;

[0059] In the formula, Indicates and An identity matrix of the same matrix dimensions , , These are the corresponding discretized coefficient matrices; Discretization time;

[0060] (22) The vehicle vibration coupling model is established as follows:

[0061] The vehicle vibration coupling model is established as follows:

[0062] (13);

[0063] The vertical displacement of the sprung mass at the suspension is expressed as follows:

[0064] (14);

[0065] In the formula, , , , These are the vertical displacements of the left front spring load, right front spring load, left rear spring load, and right rear spring load, respectively.

[0066] The output force of the suspension subsystem is expressed as follows:

[0067] (15);

[0068] Combining equations (14) and (15), the vehicle vibration coupling model can be represented in state space as follows:

[0069] (16);

[0070] In the formula, , , , , , , , , , , , , , , , , , , , ,in These are the equivalent elastic coefficients of the four suspension subsystems. These are the equivalent damping coefficients of the four suspension subsystems; The damping matrix of the vehicle vibration coupling model. The stiffness matrix of the vehicle vibration coupling model is given. The equivalent elastic coefficients and damping coefficients mentioned above can be obtained in real time through the system identification algorithm. Specific methods include, but are not limited to, least squares method and Kalman filtering.

[0071] Furthermore, the design steps for the vehicle attitude controller in step (3) are as follows:

[0072] (31) Vehicle vibration decoupling analysis;

[0073] Based on the vehicle vibration coupling model established in step (22), the vertical vibration, pitch vibration and roll vibration of the vehicle are decoupled and analyzed to obtain the damping matrix of equation (16). and stiffness matrix When the off-diagonal element is not zero, the vertical vibration, pitch vibration, and roll vibration of the vehicle exhibit coupling behavior. When the off-diagonal element is zero, the coupling term does not exist, and the vibrations in all directions of the vehicle are no longer coupled, as shown below:

[0074] (17);

[0075] Substituting the specific expression in equation (16) into equation (17), we can simplify it as follows:

[0076] (18);

[0077] (32) Determine the vibration decoupling performance indicators of the whole vehicle;

[0078] By optimizing the vehicle vibration decoupling performance indicators and reducing vehicle vibration coupling behavior, the vehicle vibration decoupling performance indicators are improved. It is expressed as follows:

[0079] (19);

[0080] (33) Determine the controller inputs and outputs;

[0081] The vehicle body state parameters detected in real time during vehicle operation are used as the input to the controller, and the expected output force of each subsystem in the active suspension system is used as the output of the controller.

[0082] (34) Determine the cost function;

[0083] The controller cost function is determined based on the vehicle body attitude performance indicators. Factors affecting the vehicle body attitude performance indicators include: vehicle vertical acceleration, roll angle, pitch angle, and tire dynamic displacement. Combined with the vehicle vibration decoupling performance indicators determined in step (32), the designed controller cost function is as follows:

[0084] (20);

[0085] In the formula, This represents the controller cost function value. , , , , The weight matrix represents the state variables. The weight matrix represents the control quantity. Indicates the prediction time domain, Represents the control time domain, relaxation factor , These are the weighting coefficients of the relaxation factor;

[0086] (35) Solve for the system cost function;

[0087] Using the cost function in step (34) of the vehicle attitude control model in step (21), the required control output is obtained by solving the following optimal problem:

[0088] ;

[0089] ;

[0090] The above optimization problem is transformed into a quadratic programming problem, which is solved using the effective set method or the interior point method, thereby obtaining a series of output variables of the controller in the control time domain:

[0091] ;

[0092] The first element of the control sequence is used as the actual output variable of the controller. The active suspension system executes control until the next discretization time. The controller predicts the state of the next discretization time based on the vehicle attitude control model and solves again to generate a new sequence of control output variables. This process is repeated to obtain the real-time expected output of the active suspension system under different driving conditions.

[0093] Furthermore, the specific steps for establishing the dynamic model of the active suspension system based on multi-agent systems in step (4) are as follows:

[0094] (41) Dynamic model of active suspension system based on multi-agent: The four suspension subsystems of the active suspension system are regarded as four agents, and the agents communicate with each other. The dynamic equation of the i-th agent is:

[0095] (twenty one);

[0096] In the formula, Output force for the i-th agent. ; Let be the suspension stiffness coefficient of the i-th agent; Let be the suspension damping coefficient of the i-th agent; The actuator force of the i-th intelligent agent; Let be the displacement at the tire of the i-th agent;

[0097] Vertical motion of the i-th agent:

[0098] (twenty two);

[0099] In the formula, Let the mass of the i-th agent be denoted as 'i'. Let be the vertical displacement of the i-th agent; This represents the total vertical load of the i-th agent; This represents the suspension mass of the i-th agent; Indicates the longitudinal slope angle of the road surface; Let be the vertical load on the tire of the i-th intelligent agent;

[0100] (42) State-space equation of the agent: Based on the dynamic model established in step (41), the state-space equation of the i-th agent is obtained as follows:

[0101] (twenty three);

[0102] In the formula, , , These are the state variables, control input variables, and system output variables of the i-th agent, respectively. , , , , These represent the corresponding coefficient matrices, which are obtained by solving the dynamic model established in step (41); This represents the road surface stimulus received by the i-th agent.

[0103] Furthermore, the specific steps for designing the output consistency controller in step (5) are as follows:

[0104] The communication topology between the agents is represented using graph theory, and the active suspension system is an undirected graph;

[0105] Undirected graph A set of 4 vertices and edge set Composition; undirected graph The weighted adjacency matrix is ​​denoted as When the side hour, Then it is called a node. and nodes The nodes that obtain each other's state information are called neighbor nodes; the edge set ;

[0106] (51) The consistency control strategy is designed as follows:

[0107] Considering the output consistency requirements among suspension subsystems, an output consistency control strategy for the active suspension system is designed based on multi-agent theory. The final output of agent i consists of two parts, as follows:

[0108] (twenty four);

[0109] In the formula, The total output of the output consistency controller for agent i; The output of the force tracking controller for agent i is obtained by solving the force tracking controller. The output of the force synchronization controller for agent i is obtained by solving the force synchronization controller.

[0110] Calculate the average value of the output force of all agents, then take the deviation value by subtracting the average value from the output force of each agent, and use the deviation value of each agent as the input of the force synchronization controller; normalize the output force of the agents, and use the normalized output force to perform the above calculation.

[0111] The formula for calculating the average value is as follows:

[0112] (25);

[0113] In the formula, This represents the average output force of the active suspension system. This represents the actual output force of the i-th agent; This represents the expected output force of the i-th agent;

[0114] The formula for calculating the deviation value of each agent is as follows:

[0115] (26);

[0116] In the formula, This represents the deviation value of the i-th agent;

[0117] Define the output consistency error of agent i to reflect the output consistency error among agents, as follows:

[0118] (27);

[0119] In the formula, , This is the weight coefficient matrix;

[0120] The following is a consensus control strategy for a multi-agent system:

[0121] (28);

[0122] In the formula, , representing the force tracking error of the i-th agent; Represents the 2-norm of a matrix or vector;

[0123] By solving for the expected output of the agent that satisfies the above strategy, the consistency of output of the multi-agent system can be ensured.

[0124] (52) Design of a force tracking controller:

[0125] The desired output force of the active suspension system is obtained by solving the vehicle attitude controller designed in step (3), which are respectively The expected output force of the suspension subsystem is taken as the expected output force of the corresponding intelligent agent:

[0126] (29);

[0127] The control cost function is designed as follows:

[0128] ;

[0129] In the formula, This represents the control cost function value of the force tracking controller; This indicates the time domain prediction of the force tracking controller; This indicates the time domain of the force tracking controller. Let each represent the state weight matrix of agent i; This represents the weighting matrix of the force tracking controller output.

[0130] Combining the control cost function described above and the agent state space equation established in step (42), the required control output is obtained by solving the following optimal problem:

[0131] ;

[0132] The above optimization problem is transformed into a quadratic programming problem, which can be solved using the effective set method or the interior point method to obtain the output of the i-th intelligent physical tracking controller. ;

[0133] (53) Design of a force synchronization controller:

[0134] Based on the consistency control strategy designed in step (51), the control cost function of the force synchronization controller is designed as follows:

[0135] ;

[0136] In the formula, This represents the control cost function value of the force synchronization controller; This indicates the predictive time domain of the force synchronization controller; This indicates the time domain of the force synchronization controller. These represent the weight matrices of the relevant state variables; This represents the weighting matrix of the force synchronization controller output.

[0137] Combining the control cost function described above and the agent state space equation established in step (42), the required control output is obtained by solving the following optimal problem:

[0138] ;

[0139] The above optimization problem is transformed into a quadratic programming problem, which can be solved using the effective set method or the interior point method to obtain the output of the i-th intelligent physical strength synchronization controller. .

[0140] The beneficial effects of this invention are:

[0141] 1. This invention proposes a layered vehicle body posture control strategy, which improves vehicle comfort, driving stability, and safety;

[0142] 2. This invention addresses the multi-directional vibration coupling problem in existing technologies by designing an upper-level vehicle attitude controller, establishing a vehicle vibration coupling model, performing decoupling analysis on the vehicle's vertical vibration, pitch vibration, and roll vibration, determining the vehicle's overall vibration decoupling performance index, and combining the vehicle's overall vibration decoupling performance index with the vehicle's desired attitude performance index. This reduces the mutual influence of vibrations in various directions of the suspension system while ensuring vehicle driving stability and safety, which is beneficial for vibration reduction performance and improves vehicle comfort.

[0143] 3. This invention addresses the problem of inconsistent outputs between suspension subsystems in existing technologies by designing a lower-level output consistency controller. Based on multi-agent theory, the active suspension system is treated as a multi-agent system, and an output consistency control strategy for the active suspension system is designed. Based on the desired output of the active suspension system obtained from the vehicle attitude controller, the outputs of the force tracking controller and the force synchronization controller are obtained, which together constitute the total output of the suspension subsystem. This ensures that the active suspension system can guarantee output consistency while achieving global optimization of system performance indicators, greatly improving the stability and safety of the vehicle during driving. Attached Figure Description

[0144] Figure 1 This is a schematic diagram of the method of the present invention.

[0145] Figure 2 This is a structural diagram of the consistency controller in this invention. Detailed Implementation

[0146] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to embodiments and accompanying drawings. The content mentioned in the embodiments is not intended to limit the present invention.

[0147] Reference Figure 1 , Figure 2 As shown, the present invention provides a multi-agent-based active suspension output consistency control method, comprising the following steps:

[0148] (1) Establish the vehicle dynamics model and tire dynamics model; the specific steps are as follows:

[0149] (11) Vehicle dynamics model:

[0150] Ignoring crosswinds and longitudinal motion of the vehicle during driving, and considering only front wheel steering, the equations of motion for each degree of freedom in the vehicle dynamics model can be obtained by applying d'Alembert's principle as follows:

[0151] The equation of motion for vehicle body lateral slip is:

[0152] (1);

[0153] In the formula, For the overall vehicle weight; For the sprung mass; For vehicle speed; It is the centroid sideslip angle; This refers to the yaw rate; This is the height of the roll center; This refers to the vehicle body roll angle; These are the lateral forces of the left front tire, right front tire, left rear tire, and right rear tire, respectively.

[0154] The equation of motion for the vehicle's yaw is:

[0155] (2);

[0156] In the formula, This represents the yaw moment of inertia of the entire vehicle. These are the distances from the vehicle's center of gravity to the front and rear axles, respectively.

[0157] The equation of motion for vehicle body roll is:

[0158] (3);

[0159] In the formula, The moment of inertia due to vehicle roll; The lateral slope angle of the road surface; These are the vertical forces acting on the vehicle body from the left front suspension, right front suspension, left rear suspension, and right rear suspension, respectively. The wheelbase of the vehicle;

[0160] The equation of motion for vehicle body pitch is:

[0161] (4);

[0162] In the formula, The moment of inertia of the vehicle body during pitch; The vehicle body pitch angle;

[0163] The equation of motion for the vertical motion at the center of mass of the vehicle body is:

[0164] (5);

[0165] In the formula, This represents the vertical displacement of the center of mass of the sprung mass.

[0166] (12) Tire dynamics model:

[0167] Ignoring changes in tire characteristics due to load variations and the effect of tire self-aligning torque, under small steering angle conditions, the tire characteristics are linear, and the tire's lateral force is expressed as:

[0168] (6);

[0169] (7);

[0170] (8);

[0171] In the formula, This refers to the front wheel steering angle; and These are the lateral stiffness of the front and rear wheels, respectively. and These are the side slip angles of the front and rear wheels, respectively. and These are the roll steering coefficients for the front and rear wheels, respectively.

[0172] Tire vertical load is expressed as:

[0173] (9);

[0174] In the formula, , , , The vertical loads are for the left front tire, right front tire, left rear tire, and right rear tire, respectively. , , , These are the stiffnesses of the left front tire, right front tire, left rear tire, and right rear tire, respectively. , , , The road surface input excitations are for the left front tire, right front tire, left rear tire, and right rear tire, respectively. , , , The unsprung mass vertical displacements at the left front tire, right front tire, left rear tire, and right rear tire are respectively.

[0175] (2) Using the vehicle dynamics model and tire dynamics model established in step (1), establish the vehicle attitude control model and the vehicle vibration coupling model; the specific steps are as follows:

[0176] (21) Establish a vehicle attitude control model;

[0177] (211) The specific steps for establishing the whole vehicle system prediction model are as follows;

[0178] Using the dynamic model established in steps (11) and (12), the prediction model of the whole vehicle system is established as follows:

[0179] (10);

[0180] The whole vehicle system prediction model is represented in state space, and the system's state variables are taken. for:

[0181] ;

[0182] Control input for:

[0183] ;

[0184] Control output for:

[0185] ;

[0186] The state space representation of the whole vehicle system prediction model is as follows:

[0187] (11);

[0188] In the formula, It is an 8×8 dimensional system matrix; It is an 8×4 dimensional control input matrix; The output state matrix is ​​5×8 dimensional; each matrix can be obtained from the vehicle system prediction model.

[0189] (212) Establish a vehicle attitude control model;

[0190] The state space of the vehicle system prediction model is discretized using the first-order difference quotient method. The discretized state space equation is expressed as:

[0191] (12);

[0192] in,

[0193] ;

[0194] In the formula, Indicates and An identity matrix of the same matrix dimensions , , These are the corresponding discretized coefficient matrices; Discretization time;

[0195] (22) The vehicle vibration coupling model is established as follows:

[0196] The vehicle vibration coupling model is established as follows:

[0197] (13);

[0198] The vertical displacement of the sprung mass at the suspension is expressed as follows:

[0199] (14);

[0200] In the formula, , , , These are the vertical displacements of the left front spring load, right front spring load, left rear spring load, and right rear spring load, respectively.

[0201] The output force of the suspension subsystem is expressed as follows:

[0202] (15);

[0203] Combining equations (14) and (15), the vehicle vibration coupling model can be represented in state space as follows:

[0204] (16);

[0205] In the formula, , , , , , , , , , , , , , , , , , , , ,in These are the equivalent elastic coefficients of the four suspension subsystems. These are the equivalent damping coefficients of the four suspension subsystems; The damping matrix of the vehicle vibration coupling model. The stiffness matrix of the vehicle vibration coupling model is given. The equivalent elastic coefficients and damping coefficients mentioned above can be obtained in real time through the system identification algorithm. Specific methods include, but are not limited to, least squares method and Kalman filtering.

[0206] (3) Design a vehicle attitude controller using the vehicle attitude control model and vehicle vibration coupling model established in step (2); perform decoupling analysis on the vehicle vertical vibration, pitch vibration and roll vibration according to the vehicle vibration coupling model, and determine the vehicle vibration decoupling performance index; the vehicle vibration decoupling performance index and the vehicle desired attitude performance index form the controller optimization target, and obtain the desired output of the active suspension system under different driving conditions by solving the optimization target in real time.

[0207] The design steps for the vehicle attitude controller are as follows:

[0208] (31) Vehicle vibration decoupling analysis;

[0209] Based on the vehicle vibration coupling model established in step (22), the vertical vibration, pitch vibration and roll vibration of the vehicle are decoupled and analyzed to obtain the damping matrix of equation (16). and stiffness matrix When the off-diagonal element is not zero, the vertical vibration, pitch vibration, and roll vibration of the vehicle exhibit coupling behavior. When the off-diagonal element is zero, the coupling term does not exist, and the vibrations in all directions of the vehicle are no longer coupled, as shown below:

[0210] (17);

[0211] Substituting the specific expression in equation (16) into equation (17), we can simplify it as follows:

[0212] (18);

[0213] (32) Determine the vibration decoupling performance indicators of the whole vehicle;

[0214] By optimizing the vehicle vibration decoupling performance indicators and reducing vehicle vibration coupling behavior, the vehicle vibration decoupling performance indicators are improved. It is expressed as follows:

[0215] (19);

[0216] (33) Determine the controller inputs and outputs;

[0217] The vehicle body state parameters detected in real time during vehicle operation are used as the input to the controller, and the expected output force of each subsystem in the active suspension system is used as the output of the controller.

[0218] (34) Determine the cost function;

[0219] The controller cost function is determined based on the vehicle body attitude performance indicators. Factors affecting the vehicle body attitude performance indicators include: vehicle vertical acceleration, roll angle, pitch angle, and tire dynamic displacement. Combined with the vehicle vibration decoupling performance indicators determined in step (32), the designed controller cost function is as follows:

[0220] (20);

[0221] In the formula, This represents the controller cost function value. , , , , The weight matrix represents the state variables. The weight matrix represents the control quantity. Indicates the prediction time domain, This represents the control time domain. Since the system model changes in real time, it cannot be guaranteed that the optimization objective will have a feasible solution at every moment. Therefore, a relaxation factor is added to the cost function. , These are the weighting coefficients of the relaxation factor;

[0222] (35) Solve for the system cost function;

[0223] Using the cost function in step (34) of the vehicle attitude control model in step (21), the required control output is obtained by solving the following optimal problem:

[0224] ;

[0225] ;

[0226] The above optimization problem is transformed into a quadratic programming problem, which is solved using the effective set method or the interior point method, thereby obtaining a series of output variables of the controller in the control time domain:

[0227] ;

[0228] The first element of the control sequence is used as the actual output variable of the controller. The active suspension system executes control until the next discretization time. The controller predicts the state of the next discretization time based on the vehicle attitude control model and solves again to generate a new sequence of control output variables. This process is repeated to obtain the real-time expected output of the active suspension system under different driving conditions.

[0229] (4) Establish a dynamic model of the active suspension system based on multi-agent systems; the specific steps are as follows:

[0230] (41) Dynamic model of active suspension system based on multi-agent: The four suspension subsystems of the active suspension system are regarded as four agents, and the agents communicate with each other. The dynamic equation of the i-th agent is:

[0231] (twenty one);

[0232] In the formula, Output force for the i-th agent. ; Let be the suspension stiffness coefficient of the i-th agent; Let be the suspension damping coefficient of the i-th agent; The actuator force of the i-th intelligent agent; Let be the displacement at the tire of the i-th agent;

[0233] Vertical motion of the i-th agent:

[0234] (twenty two);

[0235] In the formula, Let the mass of the i-th agent be denoted as 'i'. Let be the vertical displacement of the i-th agent; This represents the total vertical load of the i-th agent; This represents the suspension mass of the i-th agent; Indicates the longitudinal slope angle of the road surface; Let be the vertical load on the tire of the i-th intelligent agent;

[0236] (42) State-space equation of the agent: Based on the dynamic model established in step (41), the state-space equation of the i-th agent is obtained as follows:

[0237] (twenty three);

[0238] In the formula, , , These are the state variables, control input variables, and system output variables of the i-th agent, respectively. , , , , These represent the corresponding coefficient matrices, which are obtained by solving the dynamic model established in step (41); This represents the road surface stimulus received by the i-th agent.

[0239] (5) Using the dynamic model of the active suspension system established in step (4) and the vehicle attitude controller designed in step (3), design an output consistency controller; the output consistency controller includes a force tracking controller and a force synchronization controller. The force tracking controller tracks the expected output of the active suspension obtained by solving the vehicle attitude controller designed in step (3), and the force synchronization controller eliminates the inconsistency error between the suspension subsystems, thereby ensuring the consistency of the output of the active suspension system.

[0240] The specific steps for designing the output consistency controller in step (5) are as follows:

[0241] The communication topology between the agents is represented using graph theory, and the active suspension system is an undirected graph;

[0242] Undirected graph A set of 4 vertices and edge set Composition; undirected graph The weighted adjacency matrix is ​​denoted as When the side hour, Then it is called a node. and nodes The nodes that obtain each other's state information are called neighbor nodes; the edge set ;

[0243] (51) The consistency control strategy is designed as follows:

[0244] Considering the output consistency requirements among suspension subsystems, an output consistency control strategy for the active suspension system is designed based on multi-agent theory. The final output of agent i consists of two parts, as follows:

[0245] (twenty four);

[0246] In the formula, The total output of the output consistency controller for agent i; The output of the force tracking controller for agent i is obtained by solving the force tracking controller. The output of the force synchronization controller for agent i is obtained by solving the force synchronization controller.

[0247] The average output force of all agents is calculated, and then the output force of each agent is subtracted from the average to obtain the deviation value. The deviation value of each agent is used as the input of the force synchronization controller. Since the deviation value is obtained by the difference between the average output force of all agents and the output force of each agent, the deviation value connects all agents together, minimizing the influence of each agent on other agents in the system, thereby ensuring optimal system output consistency. Since the expected output force of each agent is different, the agent output force is normalized, and the above calculation is performed using the normalized output force.

[0248] The formula for calculating the average value is as follows:

[0249] (25);

[0250] In the formula, This represents the average output force of the active suspension system. This represents the actual output force of the i-th agent; This represents the expected output force of the i-th agent;

[0251] The formula for calculating the deviation value of each agent is as follows:

[0252] (26);

[0253] In the formula, This represents the deviation value of the i-th agent;

[0254] Define the output consistency error of agent i to reflect the output consistency error among agents, as follows:

[0255] (27);

[0256] In the formula, , This is the weight coefficient matrix;

[0257] In the defined output consistency error, in addition to the tracking error of agent i itself, there is also the output error between agent i and agent j; this can ensure that even if the expected output of each suspension agent is different, the expected output can be achieved at the same time, thus ensuring that the vehicle reaches the desired attitude.

[0258] The following is a consensus control strategy for a multi-agent system:

[0259] (28);

[0260] In the formula, , representing the force tracking error of the i-th agent; Represents the 2-norm of a matrix or vector;

[0261] By solving for the expected output of the agent that satisfies the above strategy, the consistency of output of the multi-agent system can be ensured.

[0262] (52) Design of a force tracking controller:

[0263] The desired output force of the active suspension system is obtained by solving the vehicle attitude controller designed in step (3), which are respectively The expected output force of the suspension subsystem is taken as the expected output force of the corresponding intelligent agent:

[0264] (29);

[0265] The main objective of the force tracking controller is to control the agent's actual output force to track the desired output force in real time, ensuring the system's tracking accuracy. Simultaneously, to guarantee driving comfort, the vertical velocity and acceleration of the suspension system need to be minimized. Therefore, the control cost function is designed as follows:

[0266] ;

[0267] In the formula, This represents the control cost function value of the force tracking controller; This indicates the time domain prediction of the force tracking controller; This indicates the time domain of the force tracking controller. Let each represent the state weight matrix of agent i; This represents the weighting matrix of the force tracking controller output.

[0268] Combining the control cost function described above and the agent state space equation established in step (42), the required control output is obtained by solving the following optimal problem:

[0269] ;

[0270] The above optimization problem is transformed into a quadratic programming problem, which can be solved using the effective set method or the interior point method to obtain the output of the i-th intelligent physical tracking controller. ;

[0271] (53) Design of a force synchronization controller:

[0272] Based on the consistency control strategy designed in step (51), the control cost function of the force synchronization controller is designed as follows:

[0273] ;

[0274] In the formula, This represents the control cost function value of the force synchronization controller; This indicates the predictive time domain of the force synchronization controller; This indicates the time domain of the force synchronization controller. These represent the weight matrices of the relevant state variables; This represents the weighting matrix of the force synchronization controller output.

[0275] The cost function designed above can achieve consistency among the force output, displacement output, speed output, and acceleration output of the suspension subsystems, thereby solving the problem of decreased vehicle stability and safety caused by inconsistent output due to asynchronous response between actuators in the existing active suspension system.

[0276] Combining the control cost function described above and the agent state space equation established in step (42), the required control output is obtained by solving the following optimal problem:

[0277] ;

[0278] The above optimization problem is transformed into a quadratic programming problem, which can be solved using the effective set method or the interior point method to obtain the output of the i-th intelligent physical strength synchronization controller. .

[0279] This invention has many specific applications. The above description is only a preferred embodiment of this invention. It should be noted that for those skilled in the art, several improvements can be made without departing from the principle of this invention, and these improvements should also be considered within the scope of protection of this invention.

Claims

1. A method for active suspension output consistency control based on multi-agent systems, characterized in that, The steps are as follows: (1) Establish the vehicle dynamics model and the tire dynamics model; (2) Establish a vehicle attitude control model and a vehicle vibration coupling model; (3) Design a vehicle attitude controller, and perform decoupling analysis on the vertical vibration, pitch vibration and roll vibration of the vehicle according to the vehicle vibration coupling model to determine the whole vehicle vibration decoupling performance index; the whole vehicle vibration decoupling performance index and the vehicle desired attitude performance index form the controller optimization target, and the desired output of the active suspension system under different driving conditions is obtained by solving the optimization target in real time. (4) Establish a dynamic model of the active suspension system based on multi-agent systems; (5) Using the dynamic model of the active suspension system established in step (4) and the vehicle attitude controller designed in step (3), design an output consistency controller; the output consistency controller includes a force tracking controller and a force synchronization controller. The force tracking controller tracks the expected output of the active suspension obtained by solving the vehicle attitude controller designed in step (3), and the force synchronization controller eliminates the inconsistency error between the suspension subsystems, thereby ensuring the consistency of the output of the active suspension system. The design steps for the vehicle attitude controller in step (3) are as follows: (31) Vehicle vibration decoupling analysis; Based on the vehicle vibration coupling model established in step (2), the vertical vibration, pitch vibration and roll vibration of the vehicle are decoupled and analyzed. It is found that the off-diagonal elements of the damping matrix and stiffness matrix are not zero. The vertical vibration, pitch vibration and roll vibration of the vehicle have coupling behavior. When the off-diagonal element is zero, the coupling term does not exist, and the vibration in each direction of the vehicle is no longer coupled. (32) Determine the vibration decoupling performance index of the whole vehicle; By optimizing the vehicle vibration decoupling performance indicators, the vehicle vibration coupling behavior is reduced; (33) Determine the controller inputs and outputs; The vehicle body state parameters detected in real time during vehicle operation are used as the input to the controller, and the expected output force of each subsystem in the active suspension system is used as the output of the controller. (34) Determine the cost function; The controller cost function is determined based on the vehicle body attitude performance indicators. Factors affecting the vehicle body attitude performance indicators include: vehicle vertical acceleration, roll angle, pitch angle, and tire dynamic displacement. Combined with the vehicle vibration decoupling performance index determined in step (32), the controller cost function is designed; (35) Solve for the system cost function; Using the vehicle attitude control model in step (2) and the controller cost function in step (34), the required control output is obtained by solving the optimization problem: The above optimization problem is transformed into a quadratic programming problem, which is solved by the effective set method or the interior point method, thereby obtaining a series of output variables of the controller in the control time domain; The first element of the control sequence is used as the actual output variable of the controller. The active suspension system executes control until the next discretization time. The controller predicts the state of the next discretization time based on the vehicle attitude control model and solves again to generate a new sequence of control output variables. This process is repeated to obtain the real-time expected output of the active suspension system under different driving conditions.

2. The active suspension output consistency control method based on multi-agent systems according to claim 1, characterized in that, The specific steps for establishing the vehicle dynamics model and tire dynamics model in step (1) are as follows: (11) Vehicle dynamics model: Ignoring crosswinds and longitudinal motion of the vehicle during driving, and considering only front wheel steering, the equations of motion for each degree of freedom in the vehicle dynamics model can be obtained by applying d'Alembert's principle as follows: The equation of motion for vehicle body lateral slip is: （1）； In the formula, For the overall vehicle weight; For the sprung mass; For vehicle speed; It is the centroid sideslip angle; This refers to the yaw rate; This is the height of the roll center; This refers to the vehicle body roll angle; These are the lateral forces of the left front tire, right front tire, left rear tire, and right rear tire, respectively. The equation of motion for the vehicle's yaw is: （2）； In the formula, This represents the yaw moment of inertia of the entire vehicle. These are the distances from the vehicle's center of gravity to the front and rear axles, respectively. The equation of motion for vehicle body roll is: （3）； In the formula, The moment of inertia due to vehicle roll; The lateral slope angle of the road surface; These are the vertical forces acting on the vehicle body from the left front suspension, right front suspension, left rear suspension, and right rear suspension, respectively. The wheelbase of the vehicle; The equation of motion for vehicle body pitch is: （4）； In the formula, The moment of inertia of the vehicle body during pitch; The vehicle body pitch angle; The equation of motion for the vertical motion at the center of mass of the vehicle body is: （5）； In the formula, This represents the vertical displacement of the center of mass of the sprung mass. (12) Tire dynamics model: Ignoring changes in tire characteristics due to load variations and the effect of tire self-aligning torque, under small steering angle conditions, the tire characteristics are linear, and the tire's lateral force is expressed as: （6）； （7）； （8）； In the formula, This refers to the front wheel steering angle; and These are the lateral stiffness of the front and rear wheels, respectively. and These are the side slip angles of the front and rear wheels, respectively. and These are the roll steering coefficients for the front and rear wheels, respectively. Tire vertical load is expressed as: （9）； In the formula, , , , The vertical loads are for the left front tire, right front tire, left rear tire, and right rear tire, respectively. , , , These are the stiffnesses of the left front tire, right front tire, left rear tire, and right rear tire, respectively. , , , The road surface input excitations are for the left front tire, right front tire, left rear tire, and right rear tire, respectively. , , , The unsprung mass vertical displacements at the left front tire, right front tire, left rear tire, and right rear tire are respectively.

3. The active suspension output consistency control method based on multi-agents according to claim 2, characterized in that, The specific steps for establishing the vehicle attitude control model and the vehicle vibration coupling model in step (2) are as follows: (21) Establish a vehicle attitude control model; (211) The specific steps for establishing the whole vehicle system prediction model are as follows; Using the dynamic model established in steps (11) and (12), the prediction model of the whole vehicle system is established as follows: （10）； The whole vehicle system prediction model is represented in state space, and the system's state variables are taken. for: ； Control input for: ； Control output for: ； The state space representation of the whole vehicle system prediction model is as follows: （11）； In the formula, It is an 8×8 dimensional system matrix; It is an 8×4 dimensional control input matrix; The output state matrix is ​​5×8 dimensional; each matrix can be obtained from the vehicle system prediction model. (212) Establish a vehicle attitude control model; The state space of the vehicle system prediction model is discretized using the first-order difference quotient method. The discretized state space equation is expressed as: （12）； in, ； In the formula, Indicates and An identity matrix of the same matrix dimensions , , These are the corresponding discretized coefficient matrices; Discretization time; (22) The vehicle vibration coupling model is established as follows: The vehicle vibration coupling model is established as follows: （13）； The vertical displacement of the sprung mass at the suspension is expressed as follows: （14）； In the formula, , , , These are the vertical displacements of the left front spring load, right front spring load, left rear spring load, and right rear spring load, respectively. The output force of the suspension subsystem is expressed as follows: （15）； Combining equations (14) and (15), the vehicle vibration coupling model can be represented in state space as follows: （16）； In the formula, , , , , , , , , , , , , , , , , , , , ,in These are the equivalent elastic coefficients of the four suspension subsystems. These are the equivalent damping coefficients of the four suspension subsystems; The damping matrix of the vehicle vibration coupling model. The stiffness matrix of the vehicle vibration coupling model is given. The equivalent elastic coefficients and damping coefficients mentioned above can be obtained in real time through system identification algorithms, including the least squares method and Kalman filtering.

4. The active suspension output consistency control method based on multi-agent as described in claim 1, characterized in that, The specific steps for establishing the dynamic model of the active suspension system based on multi-agent systems in step (4) are as follows: (41) Dynamic model of active suspension system based on multi-agent: The four suspension subsystems of the active suspension system are regarded as four agents, and the agents communicate with each other. The dynamic equation of the i-th agent is: （21）； In the formula, Output force for the i-th agent. ; Let be the suspension stiffness coefficient of the i-th agent; Let be the suspension damping coefficient of the i-th agent; The actuator force of the i-th intelligent agent; Let be the displacement at the tire of the i-th agent; Vertical motion of the i-th agent: （22）； In the formula, Let the mass of the i-th agent be denoted as . Let be the vertical displacement of the i-th agent; This represents the total vertical load of the i-th agent; This represents the suspension mass of the i-th agent; Indicates the longitudinal slope angle of the road surface; Let be the vertical load on the tire of the i-th intelligent agent; (42) State-space equation of the agent: Based on the dynamic model established in step (41), the state-space equation of the i-th agent is obtained as follows: （23）； In the formula, , , These are the state variables, control input variables, and system output variables of the i-th agent, respectively. , , , , These represent the corresponding coefficient matrices, which are obtained by solving the dynamic model established in step (41); This represents the road surface stimulus received by the i-th agent.

5. The active suspension output consistency control method based on multi-agents according to claim 4, characterized in that, The specific steps for designing the output consistency controller in step (5) are as follows: The communication topology between the agents is represented using graph theory, and the active suspension system is an undirected graph; Undirected graph A set of 4 vertices and edge set Composition; undirected graph The weighted adjacency matrix is ​​denoted as When the side hour, Then it is called a node. and nodes The nodes that obtain each other's state information are called neighbor nodes; the edge set ; (51) The consistency control strategy is designed as follows: Considering the output consistency requirements between suspension subsystems, an output consistency control strategy for the active suspension system is designed based on multi-agent theory. The final output of agent i consists of two parts, as follows: （24）； In the formula, The total output of the output consistency controller for agent i; The output of the force tracking controller for agent i is obtained by solving the force tracking controller. The output of the force synchronization controller for agent i is obtained by solving the force synchronization controller. Calculate the average value of the output force of all agents, then take the deviation value by subtracting the average value from the output force of each agent, and use the deviation value of each agent as the input of the force synchronization controller; normalize the output force of the agents, and use the normalized output force to perform the above calculation. The formula for calculating the average value is as follows: （25）； In the formula, This represents the average output force of the active suspension system. This represents the actual output force of the i-th agent; This represents the expected output force of the i-th agent; The formula for calculating the deviation value of each agent is as follows: （26）； In the formula, This represents the deviation value of the i-th agent; Define the output consistency error of agent i to reflect the output consistency error among agents, as follows: （27）； In the formula, , This is the weight coefficient matrix; The following is a consensus control strategy for a multi-agent system: （28）； In the formula, , representing the force tracking error of the i-th agent; Represents the 2-norm of a matrix or vector; By solving for the expected output of the agent that satisfies the above strategy, the consistency of output of the multi-agent system can be ensured. (52) Design of a force tracking controller: The desired output force of the active suspension system is obtained by solving the vehicle attitude controller designed in step (3), which are respectively The expected output force of the suspension subsystem is taken as the expected output force of the corresponding intelligent agent: （29）； The control cost function is designed as follows: ； In the formula, This represents the control cost function value of the force tracking controller; This indicates the time domain prediction of the force tracking controller; This indicates the time domain of the force tracking controller. Let each represent the state weight matrix of agent i; This represents the weighting matrix of the force tracking controller output. Combining the control cost function described above and the agent state space equation established in step (42), the required control output is obtained by solving the following optimal problem: ； The above optimal solution is transformed into a quadratic programming problem, which can be solved using the effective set method or the interior point method to obtain the output of the i-th intelligent physical tracking controller. ; (53) Design of a force synchronization controller: Based on the consistency control strategy designed in step (51), the control cost function of the force synchronization controller is designed as follows: ； In the formula, This represents the control cost function value of the force synchronization controller; This indicates the predictive time domain of the force synchronization controller; This indicates the time domain of the force synchronization controller. These represent the weight matrices of the relevant state variables; This represents the weighting matrix of the force synchronization controller output. Combining the control cost function described above and the agent state space equation established in step (42), the required control output is obtained by solving the following optimal problem: ； The above optimal solution is transformed into a quadratic programming problem, which can be solved using the effective set method or the interior point method to obtain the output of the i-th intelligent physical synchronization controller. .

Images