A distributed supervised full vector motion control method for electric truck skateboard chassis

Abstract

The present invention discloses a distributed supervised full vector motion control method for an electric truck skateboard chassis, including: establishing a six-degree-of-freedom model of a vehicle composed of subsystems; establishing a distributed supervisory controller for each subsystem; the distributed supervisory controller supervises the state of the subsystem, and when a subsystem generates a large disturbance or a performance degradation event, the constraint condition is updated online according to the working state of the neural network controller and transmitted to the domain controller; the weight value is modified by transmitting a tag signal combination to the distributed supervisory controller; the distributed supervisory controller configures the working parameters of the actuator of each subsystem according to the input tag signal combination, and performs full vector motion control on the electric truck skateboard chassis. The present invention implements the intervention and exit of the neural network control through the distributed supervisory controller when the vehicle state and the external environment change, and at the same time applies the solved constraints to the domain controller, and the domain controller performs a global solution to achieve coordination between subsystems and optimal global performance.

Description

Technical Field

[0001] The invention belongs to the technical field of vehicle motion control, and in particular relates to a distributed supervised full-vector motion control method for an electric truck skateboard chassis. Background Art

[0002] At present, the control subsystems for the electric truck skateboard chassis system are usually developed separately. When multiple control subsystems coexist, there are functional overlaps / conflicts and structural interferences between the systems. Therefore, it is necessary to design a domain controller to consider the global control objectives for full vector motion control. Due to the complex driving conditions and changeable environmental conditions of the electric truck skateboard chassis, most of the existing motion control methods do not consider the interference caused by parameter perturbations and external disturbances, and cannot achieve the optimal global performance specifications of each subsystem function. Therefore, it is necessary to design a supervisory controller with online adaptive adjustment function to perform necessary interventions on each subsystem of the chassis system. Summary of the invention

[0003] In view of the above-mentioned deficiencies in the prior art, the purpose of the present invention is to provide a distributed supervisory full vector motion control method for an electric truck skateboard chassis. The present invention uses a distributed supervisory controller to implement the intervention and exit of neural network control when the vehicle state and external environment change, and at the same time imposes the constraints of the solution on the domain controller, which performs global solution to achieve coordination between subsystems and global performance optimization.

[0004] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0005] The distributed supervised full vector motion control method of an electric truck skateboard chassis of the present invention comprises the following steps:

[0006] 1) Modeling is performed based on the kinematic equations of the electric truck skateboard chassis, and a six-degree-of-freedom model of the vehicle consisting of subsystems is established;

[0007] 2) According to the kinematic model of each subsystem, configure the weight function and establish the distributed supervisory controller of each subsystem;

[0008] 3) The distributed supervisory controller monitors the subsystem status. When a subsystem experiences a large disturbance or a performance degradation event, the constraint conditions are updated online according to the working status of the neural network controller and transmitted to the domain controller;

[0009] 4) Solve the generalized control force of each distributed supervisory controller, perform global optimization to achieve optimal performance, and transmit the marker signal combination to the distributed supervisory controller to modify the weight;

[0010] 5) The distributed supervisory controller configures the operating parameters of the actuators of each subsystem according to the input marker signal combination and performs full vector motion control on the electric truck skateboard chassis.

[0011] Furthermore, the vehicle six-degree-of-freedom model established in step 1) is:

[0012] The vehicle six-degree-of-freedom model formula is as follows:

[0013]

[0014]

[0015]

[0016]

[0017]

[0018]

[0019]

[0020] In the formula, m s is the mass of the electric truck, v x 、v y are the speeds of the chassis in the x and y directions, δ f is the front wheel turning angle, is the pitch angular acceleration, is the roll angular acceleration, is the yaw angular acceleration, I x ,I y ,I z are the moments of inertia of the electric truck around the x, y, and z axes, respectively, l f is the horizontal distance from the center of mass to the front axle, l r is the horizontal distance from the center of mass to the rear axle, h is the height distance from the center of mass to the chassis, and F txf is the force component of the front tire in the x-axis direction, F txr is the force component in the x-axis direction of the rear tire, F tyf is the y-axis force of the front tire, F tyr is the y-axis force of the rear tire, F zrl 、F zrr 、F zfl 、F zfr are the vertical forces of the left wheel of the rear axle tire, the right wheel of the rear axle tire, the left wheel of the front axle tire, and the right wheel of the front axle tire, respectively, t f ,t r are front and rear axle track, F zf 、F zr 、F dz are the vertical reaction forces of the front axle, rear axle and suspension respectively; m usij is the mass of a single wheel, is the vertical acceleration of a single wheel, where i represents the front axle (f) or the rear axle (r), j represents the left wheel (l) or the right wheel (r), and M dz is the steering moment acting on the center of mass.

[0021] Furthermore, the subsystems include: a steering subsystem, a driving and braking subsystem, and a suspension subsystem, which respectively correspond to the six degrees of freedom of movement: lateral, longitudinal, vertical, yaw, roll, and pitch.

[0022] Furthermore, the specific method of establishing a distributed supervisory controller in step 2) is:

[0023] 21) Design a state feedback controller for a distributed supervisory controller;

[0024] 22) Design a neural network controller for distributed supervisory controller.

[0025] Furthermore, the step 21) specifically includes:

[0026] For the subsystem dynamics model, the robust H2 / H ∞ Mixed sensitivity control, solve the state feedback controller, the corresponding mathematical description is as follows:

[0027] Min:||W3KS||2

[0028]

[0029] Where K is the state feedback controller; S is the sensitivity function, T is the complementary sensitivity function, W1 is the sensitivity weighting function, W2 is the complementary sensitivity weighting function, and W3 is the performance index weight function.

[0030] Furthermore, the neural network in step 22) comprises: an input layer, a hidden layer and an output layer, wherein the input layer has 2 neurons, the hidden layer has 5 neurons, and the output layer has 1 neuron;

[0031] The activation function of the neurons in the hidden layer is a Gaussian function, expressed as:

[0032]

[0033] In the formula, represents the input vector of the neural network; c j =[c j1 c j2 ] T represents the center vector value of the jth node; b j =[b j1 b j2 ] TThe vector representing the basis width values ​​of the Gaussian basis function of the jth node.

[0034] Furthermore, in step 22), the RBF neural network is corrected using the training sample set, and the weights (w1, w2, . . . , w5) and the center vector c of the neural network are corrected using the gradient descent method. j =[c j1 c j2 ] T and base width vector b j =[b j1 b j2 ] T The value of is corrected, and the mathematical expression is as follows;

[0035] The error performance index is defined as:

[0036]

[0037] Among them, u n is the output of the RBF neural network controller, and u is the output of the distributed supervisory controller;

[0038] The adjustment of the neural network is done by adjusting the weights, center vector, and base width vector until E reaches the minimum. The correction process of the gradient descent method is:

[0039]

[0040]

[0041]

[0042] In the formula, χ is the learning rate, is the partial derivative operator, ΔW, Δc j , Δb j are the adjustment amounts of weight, center vector, and base width value vector respectively;

[0043] The output of the RBF neural network controller is as follows:

[0044] u n =h1w1+h2w2+h3w3+h4w4+h5w5

[0045] Therefore, the total control output of the distributed supervisory controller is:

[0046] u=u n +u p

[0047] In the formula, u p is the control output of the state feedback controller.

[0048] Furthermore, the performance indicator weight function W3 in step 21) specifically includes:

[0049] The lateral displacement tracking control weight function set for the steering subsystem is:

[0050] W steer =k e (T d1 s+1) / (T d2 s+1)

[0051] Where W steer is the lateral displacement tracking control weight, T di is the time constant related to the performance requirements, k e is the steady-state error indicator, and the steady-state value of the tracking error is lower than 1 / k e , s is the complex frequency domain of the Laplacian transform;

[0052] For the driving and braking subsystem, the longitudinal displacement tracking control weight function is set as:

[0053] W brake =φ a (T b1 s+1) / (T b2 s+1)

[0054] Where W brake is the longitudinal displacement tracking control weight, T bi is the time constant related to the performance requirements, φ a is the lateral acceleration gain;

[0055] For the suspension subsystem, the vertical acceleration, suspension stiffness, and tire stiffness controller weight functions are set as follows:

[0056] W az =φ az (T s1 s+1) / (T s2 s+1)

[0057] W sd =φ sd (T s3 s+1) / (T s4 s+1)

[0058] W td =k td (T s6 s+1) / (T s6 s+1)

[0059] Where W az , W sd , W tdare the controller weights of vertical acceleration, suspension stiffness, and tire stiffness, respectively, T si , k td is the time constant related to the performance requirements, φ az ,φ sd are the vertical acceleration and suspension stiffness gain.

[0060] Furthermore, in the step 3), when the subsystem generates a large disturbance or a performance degradation event, the neural network controller no longer plays a leading role, the error performance index E of the RBF neural network controller fluctuates greatly, the working state of the neural network controller changes, and the distributed supervisory controller calculates and updates the constraints based on the neural network controller error performance index and sensor information and applies them to the domain controller; the constraints refer to constraints that take into account vehicle stability and performance index requirements; the sensor information includes vehicle state information and environmental road information.

[0061] Furthermore, in step 4), the generalized control force of each distributed supervisory controller is solved online, specifically comprising the following steps:

[0062] 41) Aiming at the problem of optimizing the longitudinal and lateral forces of tires, formulate the objective functions of minimizing the maximum load rate, minimizing the load rate variance, and averaging the load rates of each wheel;

[0063] 42) Use multi-objective optimization algorithm to solve the generalized control force;

[0064] 43) The domain controller selects a tag signal combination based on the generalized control power, and the tag signal combination value is transmitted to the distributed supervisory controller to modify the performance indicator weight.

[0065] Furthermore, the step 5) specifically includes the following steps:

[0066] 51) The distributed supervisory controller modifies supervisory controller parameters based on the combination of marker signals provided by the domain controller;

[0067] 52) The distributed supervisory controller calculates the operating parameters of the actuator according to the states of each actuator in the subsystem;

[0068] 53) Each actuator executes according to the control instructions output by the distributed supervisory controller.

[0069] Among them, the domain controller is used to coordinate components, perform multi-objective optimization and update performance indicator weights according to the external constraints transmitted by the distributed supervisory controller; the domain controller specifies the configured specific value combination of the marking signal to the distributed supervisory controller, and the marking signal is used to supervise the scheduling variables at the controller level. Different combinations of marking signals represent the options of the domain controller when dealing with multi-objective or conflict scenarios.

[0070] Among them, the distributed supervisory controller is used to supervise performance degradation events and execute domain controller marking signal instructions to control the actuator, and impose external constraints on the domain controller after the event occurs; performance degradation events include emergency vehicle operations, vehicle failures, controller performance degradation, performance degradation caused by actuator saturation and other events.

[0071] The components include: a distributed supervisory controller, an actuator and a sensor, wherein the sensor is used to detect vehicle performance degradation events.

[0072] The distributed supervisory controller design of each subsystem is based on the motion vector control target, using control methods such as H2 / H ∞ Control is performed, and based on the performance indicators of the control process, linear constraints are provided to the domain controller. The domain controller makes global planning decisions based on sensor information and linear constraints, allocates generalized control forces, specifies marker signal combinations, and improves the accuracy of full vector motion control.

[0073] The distributed supervisory controller of each subsystem is constructed based on the RBF neural network supervisory control algorithm. Feedback control is used in the initial stage, and then gradually transitions to neural network control. During the control process, if a large error occurs, the designed feedback controller plays a leading role and the neural network controller plays a regulating role.

[0074] Beneficial effects of the present invention:

[0075] 1. The present invention designs a distributed supervisory controller for each subsystem. On the one hand, it ensures the steady-state performance and dynamic performance of the control during the operation of the subsystem. On the other hand, it observes the changes in the subsystem performance indicators through the intervention and exit of the neural network controller, providing a basis for the weight update of the domain controller.

[0076] 2. The method of the present invention aims at the characteristic that the design of the domain controller of the skateboard chassis needs to consider the global state for solution, and realizes the control and supervision of each subsystem with the help of a distributed supervisory controller. The distributed supervisory controller solves the constraint changes caused by the performance degradation of the subsystem. The domain controller only needs to perform online optimization and solution for the constraints and adjust the weights to realize the coordinated global optimal control of each subsystem.

[0077] 3. The method of the present invention introduces supervisory control into the full vector motion control of the electric truck skateboard chassis. The supervisory controller is independently designed to ensure the stability and robustness of the control task and the global performance indicators. At the same time, the distributed solution reduces the model structure and calculation complexity, saving system resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 It is a principle framework diagram of the method of the present invention;

[0079] Figure 2 A six-degree-of-freedom model diagram of the electric truck skateboard chassis of the present invention;

[0080] Figure 3 It is the principle diagram of the distributed supervisory controller in the present invention;

[0081] Figure 4 Flow chart of the control method of the present invention. DETAILED DESCRIPTION

[0082] In order to facilitate the understanding of those skilled in the art, the present invention is further described below in conjunction with embodiments and drawings. The contents mentioned in the implementation modes are not intended to limit the present invention.

[0083] Reference Figure 1 , Figure 4 As shown, a distributed supervised full vector motion control method for an electric truck skateboard chassis of the present invention comprises the following steps:

[0084] 1) Modeling is performed based on the kinematic equations of the electric truck skateboard chassis, and a six-degree-of-freedom model of the vehicle consisting of subsystems is established;

[0085] Among them, the established vehicle six-degree-of-freedom model is:

[0086] Reference Figure 2 As shown, the vehicle six-degree-of-freedom model formula is as follows:

[0087]

[0088]

[0089]

[0090]

[0091]

[0092]

[0093]

[0094] In the formula, m s is the mass of the electric truck, v x 、v y are the speeds of the chassis in the x and y directions, δ f is the front wheel turning angle, is the pitch angular acceleration, is the roll angular acceleration, is the yaw angular acceleration, I x ,I y ,I zare the moments of inertia of the electric truck around the x, y, and z axes, respectively, l f is the horizontal distance from the center of mass to the front axle, l r is the horizontal distance from the center of mass to the rear axle, h is the height distance from the center of mass to the chassis, and F txf is the force component of the front tire in the x-axis direction, F txr is the force component in the x-axis direction of the rear tire, F tyf is the y-axis force of the front tire, F tyr is the y-axis force of the rear tire, F zrl 、F zrr 、F zfl 、F zfr are the vertical forces of the left wheel of the rear axle tire, the right wheel of the rear axle tire, the left wheel of the front axle tire, and the right wheel of the front axle tire, respectively, t f ,t r are front and rear axle track, F zf 、F zr 、F dz are the vertical reaction forces of the front axle, rear axle and suspension respectively; m usij is the mass of a single wheel, is the vertical acceleration of a single wheel, where i represents the front axle (f) or the rear axle (r), j represents the left wheel (l) or the right wheel (r), and M dz is the steering moment acting on the center of mass.

[0095] Specifically, the subsystems include: a steering subsystem, a driving and braking subsystem, and a suspension subsystem, which respectively correspond to the six degrees of freedom of movement: lateral, longitudinal, vertical, yaw, roll, and pitch.

[0096] 2) According to the kinematic model of each subsystem, configure the weight function and establish the distributed supervisory controller of each subsystem;

[0097] Among them, refer to Figure 3 As shown, the specific method of establishing a distributed supervisory controller is:

[0098] 21) Design a state feedback controller for a distributed supervisory controller;

[0099] 22) Design a neural network controller for distributed supervisory controller.

[0100] Specifically, the step 21) specifically includes:

[0101] For the subsystem dynamics model, the robust H2 / H ∞ Mixed sensitivity control, solve the state feedback controller, the corresponding mathematical description is as follows:

[0102] Min:||W3KS||2

[0103]

[0104] Where K is the state feedback controller; S is the sensitivity function, T is the complementary sensitivity function, W1 is the sensitivity weighting function, W2 is the complementary sensitivity weighting function, and W3 is the performance index weight function.

[0105] Specifically, the neural network in step 22) includes: an input layer, a hidden layer and an output layer, wherein the input layer has 2 neurons, the hidden layer has 5 neurons, and the output layer has 1 neuron;

[0106] The activation function of the neurons in the hidden layer is a Gaussian function, expressed as:

[0107]

[0108] In the formula, represents the input vector of the neural network; c j =[c j1 c j2 ] T represents the center vector value of the jth node; b j =[b j1 b j2 ] T The vector representing the basis width values ​​of the Gaussian basis function of the jth node.

[0109] In addition, in step 22), the RBF neural network is corrected using the training sample set, and the weights (w1, w2, ···, w5) and the center vector c of the neural network are adjusted using the gradient descent method. j =[c j1 c j2 ] T and base width vector b j =[b j1 b j2 ] T The value of is corrected, and the mathematical expression is as follows;

[0110] The error performance index is defined as:

[0111]

[0112] Among them, u n is the output of the RBF neural network controller, and u is the output of the distributed supervisory controller;

[0113] The adjustment of the neural network is done by adjusting the weights, center vector, and base width vector until E reaches the minimum. The correction process of the gradient descent method is:

[0114]

[0115]

[0116]

[0117] In the formula, χ is the learning rate, is the partial derivative operator, ΔW, Δc j , Δb j are the adjustment amounts of weight, center vector, and base width value vector respectively;

[0118] The output of the RBF neural network controller is as follows:

[0119] u n =h1w1+h2w2+h3w3+h4w4+h5w5

[0120] Therefore, the total control output of the distributed supervisory controller is:

[0121] u=u n +u p

[0122] In the formula, u p is the control output of the state feedback controller.

[0123] The performance indicator weight function W3 in step 21) specifically includes:

[0124] The lateral displacement tracking control weight function set for the steering subsystem is:

[0125] W steer =k e (T d1 s+1) / (T d2 s+1)

[0126] Where W steer is the lateral displacement tracking control weight, T di is the time constant related to the performance requirements, k e is the steady-state error indicator, and the steady-state value of the tracking error is lower than 1 / k e , s is the complex frequency domain of the Laplacian transform;

[0127] For the driving and braking subsystem, the longitudinal displacement tracking control weight function is set as:

[0128] W brake =φ a (T b1 s+1) / (T b2 s+1)

[0129] Where W brake is the longitudinal displacement tracking control weight, T bi is the time constant related to the performance requirements, φa is the lateral acceleration gain;

[0130] For the suspension subsystem, the vertical acceleration, suspension stiffness, and tire stiffness controller weight functions are set as follows:

[0131] W az =φ az (T s1 s+1) / (T s2 s+1)

[0132] W sd =φ sd (T s3 s+1) / (T s4 s+1)

[0133] W td =k td (T s6 s+1) / (T s6 s+1)

[0134] Where W az , W sd , W td are the controller weights of vertical acceleration, suspension stiffness, and tire stiffness, respectively, T si , k td is the time constant related to the performance requirements, φ az and φ sd are the vertical acceleration and suspension stiffness gain.

[0135] 3) The distributed supervisory controller monitors the subsystem status. When a subsystem experiences a large disturbance or a performance degradation event, the constraint conditions are updated online according to the working status of the neural network controller and transmitted to the domain controller;

[0136] Among them, in the step 3), when the subsystem generates a large disturbance or a performance degradation event, the neural network controller no longer plays a leading role, the error performance index E of the RBF neural network controller produces a large fluctuation, the working state of the neural network controller changes, and the distributed supervisory controller calculates and updates the constraints based on the neural network controller error performance index and sensor information and applies them to the domain controller; the constraints refer to constraints that take into account vehicle stability and performance index requirements; the sensor information includes vehicle status information and environmental road information.

[0137] 4) Solve the generalized control force of each distributed supervisory controller online, perform global optimization to achieve optimal performance, and transmit the marker signal combination to the distributed supervisory controller to modify the weight;

[0138] Wherein, in step 4), solving the generalized control force of each distributed supervisory controller online specifically includes the following steps:

[0139] 41) Aiming at the problem of optimizing the distribution of longitudinal and lateral forces of tires, the objective functions of minimizing the maximum load rate, minimizing the load rate variance, and averaging the load rates of each wheel are formulated:

[0140]

[0141] Where ΣF is the matrix of total longitudinal force, total lateral force and total yaw moment to be distributed; W f , W n , W k are weight coefficients, representing the minimum load variance, the average load of tire force distribution, and the minimum maximum load; F d is the expected force matrix.

[0142] At the same time, the following constraints are met:

[0143] ΣF=M f f

[0144] f∈U

[0145] Where ΣF is the matrix of the total longitudinal force, total lateral force and total yaw moment to be distributed, U represents the feasible domain of the tire force f, which is determined by the constraints, and M f is the control efficiency matrix;

[0146] 42) Use multi-objective optimization algorithm to solve the generalized control force;

[0147] 43) The domain controller selects a tag signal combination based on the generalized control power, and the tag signal combination value is transmitted to the distributed supervisory controller to modify the performance indicator weight.

[0148] 5) The distributed supervisory controller configures the operating parameters of the actuators of each subsystem according to the input tag signal combination, and performs full vector motion control on the electric truck skateboard chassis;

[0149] Wherein, the step 5) specifically includes the following steps:

[0150] 51) The distributed supervisory controller modifies supervisory controller parameters based on the combination of marker signals provided by the domain controller;

[0151] 52) The distributed supervisory controller calculates the operating parameters of the actuator according to the states of each actuator in the subsystem;

[0152] 53) Each actuator executes according to the control instructions output by the distributed supervisory controller.

[0153] Among them, the domain controller is used to coordinate components, perform multi-objective optimization and update performance indicator weights according to the external constraints transmitted by the distributed supervisory controller; the domain controller specifies the configured specific value combination of the marking signal to the distributed supervisory controller, and the marking signal is used to supervise the scheduling variables at the controller level. Different combinations of marking signals represent the options of the domain controller when dealing with multi-objective or conflict scenarios.

[0154] Among them, the distributed supervisory controller is used to supervise performance degradation events and execute domain controller marking signal instructions to control the actuator, and impose external constraints on the domain controller after the event occurs; performance degradation events include emergency vehicle operations, vehicle failures, controller performance degradation, performance degradation caused by actuator saturation and other events.

[0155] The components include: a distributed supervisory controller, an actuator and a sensor, wherein the sensor is used to detect vehicle performance degradation events.

[0156] The distributed supervisory controller design of each subsystem is based on the motion vector control target, using control methods such as H2 / H ∞ Control is performed, and based on the performance indicators of the control process, linear constraints are provided to the domain controller. The domain controller makes global planning decisions based on sensor information and linear constraints, allocates generalized control forces, specifies marker signal combinations, and improves the accuracy of full vector motion control.

[0157] The distributed supervisory controller of each subsystem is constructed based on the RBF neural network supervisory control algorithm. Feedback control is used in the initial stage, and then gradually transitions to neural network control. During the control process, if a large error occurs, the designed feedback controller plays a leading role and the neural network controller plays a regulating role.

[0158] The present invention has many specific application paths. The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements can be made without departing from the principle of the present invention. These improvements should also be regarded as the protection scope of the present invention.

Claims

1. A distributed supervised full vector motion control method for an electric truck skateboard chassis, characterized in that: Here are the steps: 1) Modeling is performed based on the kinematic equations of the electric truck skateboard chassis, and a six-degree-of-freedom model of the vehicle consisting of subsystems is established; 2) According to the kinematic model of each subsystem, configure the weight function and establish the distributed supervisory controller of each subsystem; 3) The distributed supervisory controller monitors the subsystem status. When a subsystem experiences a large disturbance or a performance degradation event, the constraint conditions are updated online according to the working status of the neural network controller and transmitted to the domain controller; 4) Solve the generalized control force of each distributed supervisory controller, perform global optimization to achieve optimal performance, and transmit the marker signal combination to the distributed supervisory controller to modify the weight; 5) The distributed supervisory controller configures the operating parameters of the actuators of each subsystem according to the input tag signal combination, and performs full vector motion control on the electric truck skateboard chassis; The vehicle six-degree-of-freedom model established in step 1) is: The vehicle six-degree-of-freedom model formula is as follows: In the formula, m s is the mass of the electric truck, v x 、v y are the speeds of the chassis in the x and y directions, δ f is the front wheel turning angle, is the pitch angular acceleration, is the roll angular acceleration, is the yaw angular acceleration, I x ,I y ,I z are the moments of inertia of the electric truck around the x, y, and z axes, respectively, l f is the horizontal distance from the center of mass to the front axle, l r is the horizontal distance from the center of mass to the rear axle, h is the height distance from the center of mass to the chassis, and F txf is the force component of the front tire in the x-axis direction, F txr is the force component in the x-axis direction of the rear tire, F tyf is the y-axis force of the front tire, F tyr is the y-axis force of the rear tire, F zrl 、F zrr 、F zfl 、F zfr are the vertical forces of the left wheel of the rear axle tire, the right wheel of the rear axle tire, the left wheel of the front axle tire, and the right wheel of the front axle tire, respectively, t f ,t r are front and rear axle track, F zf 、F zr 、F dz are the vertical reaction forces of the front axle, rear axle and suspension respectively; m usij is the mass of a single wheel, is the vertical acceleration of a single wheel, where i represents the front axle or rear axle, j represents the left wheel or right wheel, and M dz is the turning moment acting on the center of mass; The specific method of establishing a distributed supervisory controller in step 2) is: 21) Design a state feedback controller for a distributed supervisory controller; 22) Design a neural network controller for distributed supervisory controllers; The step 21) specifically includes: For the subsystem dynamics model, the robust H2 / H ∞ Mixed sensitivity control, solve the state feedback controller, the corresponding mathematical description is as follows: Min:||W3KS||2 Where K is the state feedback controller; S is the sensitivity function, T is the complementary sensitivity function, W1 is the sensitivity weighting function, W2 is the complementary sensitivity weighting function, and W3 is the performance index weight function; The performance indicator weight function W3 in step 21) specifically includes: The lateral displacement tracking control weight function set for the steering subsystem is: W steer =k e (T d1 s+1) / (T d2 s+1) Where W steer is the lateral displacement tracking control weight, T di is the time constant related to the performance requirements, k e is the steady-state error indicator, and the steady-state value of the tracking error is lower than 1 / k e , s is the complex frequency domain of the Laplacian transform; For the driving and braking subsystem, the longitudinal displacement tracking control weight function is set as: W brake =φ a (T b1 s+1) / (T b2 s+1) Where W brake is the longitudinal displacement tracking control weight, T bi is the time constant related to the performance requirements, φ a is the lateral acceleration gain; For the suspension subsystem, the vertical acceleration, suspension stiffness, and tire stiffness controller weight functions are set as follows: W az =φ az (T s1 s+1) / (T s2 s+1) W sd =φ sd (T s3 s+1) / (T s4 s+1) W td =k td (T s6 s+1) / (T s6 s+1) Where W az , W sd , W td are the controller weights of vertical acceleration, suspension stiffness, and tire stiffness, respectively, T si , k td is the time constant related to the performance requirements, φ az and φ sd is the vertical acceleration and suspension stiffness gain; The neural network in step 22) comprises: an input layer, a hidden layer and an output layer, wherein the input layer has 2 neurons, the hidden layer has 5 neurons, and the output layer has 1 neuron; The activation function of the neurons in the hidden layer is a Gaussian function, expressed as: In the formula, represents the input vector of the neural network; c j =[c j1 c j2 ] T represents the center vector value of the jth node; b j =[b j1 b j2 ] T represents the basis width value vector of the Gaussian basis function of the jth node; In step 22), the RBF neural network is corrected using the training sample set, and the weights (w1, w2, ..., w5) and the center vector c of the neural network are adjusted using the gradient descent method. j =[c j1 c j2 ] T and base width vector b j =[b j1 b j2 ] T The value of is corrected, and the mathematical expression is as follows; The error performance index is defined as: Among them, u n is the output of the RBF neural network controller, and u is the output of the distributed supervisory controller; The adjustment of the neural network is done by adjusting the weights, center vector, and base width vector until E reaches the minimum. The correction process of the gradient descent method is: In the formula, χ is the learning rate, is the partial derivative operator, ΔW, Δc j , Δb j are the adjustment amounts of weight, center vector, and base width value vector respectively; The output of the RBF neural network controller is as follows: u n =h1w1+h2w2+h3w3+h4w4+h5w5 Therefore, the total control output of the distributed supervisory controller is: in=in n +in p In the formula, u p is the control output of the state feedback controller; In the step 3), when a large disturbance or a performance degradation event occurs in the subsystem, the neural network controller no longer plays a leading role, the error performance index E of the RBF neural network controller fluctuates greatly, the working state of the neural network controller changes, and the distributed supervisory controller calculates and updates the constraint conditions based on the error performance index of the neural network controller and the sensor information and applies them to the domain controller; the constraint conditions refer to the constraints considering the vehicle stability and performance index requirements; the sensor information includes vehicle state information and environmental road information; In step 4), the generalized control force of each distributed supervisory controller is solved online, which specifically includes the following steps: 41) Aiming at the problem of optimizing the longitudinal and lateral forces of tires, formulate the objective functions of minimizing the maximum load rate, minimizing the load rate variance, and averaging the load rates of each wheel; 42) Use multi-objective optimization algorithm to solve the generalized control force; 43) The domain controller selects a tag signal combination based on the generalized control power, and the tag signal combination value is transmitted to the distributed supervisory controller to modify the performance indicator weight.

2. The distributed supervised full vector motion control method for the electric truck skateboard chassis according to claim 1 is characterized in that: The step 5) specifically comprises the following steps: 51) The distributed supervisory controller modifies supervisory controller parameters based on the combination of marker signals provided by the domain controller; 52) The distributed supervisory controller calculates the operating parameters of the actuator according to the states of each actuator in the subsystem; 53) Each actuator executes according to the control instructions output by the distributed supervisory controller.

Images