Torque distribution method of distributed driving electric vehicle based on INPDOA

Abstract

The invention is suitable for the technical field of distributed driving electric vehicle control, and provides an INPDOA-based distributed driving electric vehicle torque distribution method, which comprises the following steps of: establishing a seven-degree-of-freedom dynamic model and a linear two-degree-of-freedom dynamic model, and determining a state input quantity, a control quantity, an output quantity and an expected output quantity; establishing an improved neural population dynamic optimization algorithm architecture, and initializing parameters; a weight coefficient matrix of the NMPC controller is optimized through INPDOA, and the optimized NMPC controller is used for carrying out yaw moment control; and optimizing a torque distribution system by improving a neural population dynamic optimization algorithm to obtain an optimal torque distribution strategy. According to the method, the control precision and robustness are effectively improved, the application range is wide, and the response speed is high.

Description

Technical Field

[0001] The present invention belongs to the technical field of distributed drive electric vehicle control, and in particular relates to a torque distribution method for a distributed drive electric vehicle based on INPDOA. Background Art

[0002] As the world pays more and more attention to environmental protection and energy security, electric vehicles have developed rapidly as a clean and efficient means of transportation. Among them, distributed drive electric vehicles have higher flexibility and control accuracy due to their four-wheel independent drive and independent steering characteristics.

[0003] In the current research on stability control of distributed drive electric vehicles, the MPC method is mainly used for direct yaw moment control. This method has good robustness. The nonlinear MPC (NMPC) method is derived for the highly complex nonlinear system of the vehicle. The performance of the MPC method mainly depends on the selection of its weight coefficient matrix P, Q, and R. Most of the existing methods only set a set of initial constants, and no in-depth research has been conducted on the solution of the weight coefficient matrix.

[0004] In addition, in the study of torque distribution control, most studies simplify the solution problem to a linear quadratic solution problem (LQR), which has poor adaptability to nonlinear systems, or use methods such as particle swarm optimization (PSO) for optimization. However, these methods have limited applicability and cannot fully meet application requirements. Therefore, an improved neural population dynamic optimization algorithm (INPDOA) is proposed to solve the above problems. Summary of the invention

[0005] The purpose of the embodiment of the present invention is to provide a torque distribution method for a distributed drive electric vehicle based on INPDOA, aiming to solve the problems raised in the above background technology.

[0006] The embodiment of the present invention is implemented as follows: a torque distribution method for a distributed drive electric vehicle based on INPDOA, comprising the following steps: Step 1: Establish a seven-degree-of-freedom dynamic model and a linear two-degree-of-freedom dynamic model to determine the state input , control amount , output , expected output ; Step 2: Optimize and improve the basic neural population dynamic optimization algorithm, establish the improved neural population dynamic optimization algorithm architecture, and initialize the parameters; Step 3: Optimize the weight coefficient matrix of the NMPC controller by improving the neural population dynamic optimization algorithm (At this time, the algorithm ), use the optimized NMPC controller to perform yaw moment control; Step 4: Optimize the torque distribution coefficient by improving the neural population dynamic optimization algorithm (At this time, the algorithm ), and obtain the optimal torque distribution strategy.

[0007] An embodiment of the present invention provides a torque distribution method for a distributed drive electric vehicle based on INPDOA. The method addresses the problem of difficulty in selecting the weight coefficient matrix of the MPC control algorithm. The improved INPDOA is used to optimize the weight matrices P, Q, and R of the MPC method to achieve the goal of improving control accuracy and robustness. For torque distribution control, the INPDOA algorithm is used to optimize its torque distribution coefficient to achieve the goal of enhancing the scope of application and improving the response speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 A flowchart of a torque distribution method for a distributed drive electric vehicle based on INPDOA provided in an embodiment of the present invention; Figure 2 is the vehicle dynamics model; Figure 3 It is a schematic diagram of the linear two-degree-of-freedom reference model; Figure 4 Flowchart for the basic neural population dynamics optimization algorithm; Figure 5 is the comparison diagram of yaw angular velocity; Figure 6 This is a comparison chart of the center of mass sideslip angle. DETAILED DESCRIPTION

[0009] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0010] The specific implementation of the present invention is described in detail below in conjunction with specific embodiments.

[0011] like Figure 1 As shown, a torque distribution method for a distributed drive electric vehicle based on INPDOA is provided in one embodiment of the present invention, comprising the following steps: Step 1: Establish a seven-degree-of-freedom dynamic model and a linear two-degree-of-freedom dynamic model to determine the state input , control amount , output , expected output ; Step 2: Optimize and improve the basic neural population dynamic optimization algorithm, establish the improved neural population dynamic optimization algorithm architecture, and initialize the parameters; Step 3: Optimize the weight coefficient matrix of the NMPC controller by improving the neural population dynamic optimization algorithm (At this time, the algorithm ), use the optimized NMPC controller to perform yaw moment control; Step 4: Optimize the torque distribution coefficient by improving the neural population dynamic optimization algorithm (At this time, the algorithm ), and obtain the optimal torque distribution strategy.

[0012] like Figure 2 As shown, as a preferred embodiment of the present invention, in step 1, the method for constructing the seven-degree-of-freedom dynamics model of the distributed drive electric vehicle is as follows: Using the body coordinate system of the distributed wire-controlled vehicle as a reference, fix the origin at the center of mass of the vehicle Based on the “right-hand rule”, the center axis of the vehicle body along the direction of travel is Positive axis, perpendicular to the vehicle's direction of travel The axis points to the left. Axis positive direction; is the velocity at the center of mass of the vehicle ( ), is the vehicle speed in the vehicle body coordinate system The component in the axial direction ( ), is the vehicle speed in the vehicle body coordinate system The component in the axial direction ( ), is the center of mass sideslip angle ( ), is the vehicle yaw angle ( ), is the yaw angular velocity ( ), is the yaw angular acceleration ( ), is the distance from the vehicle's center of mass to the front axle ( ), is the distance from the vehicle's center of mass to the rear axle ( ), is the vehicle width ( ), They represent the left front wheel, right front wheel, left rear wheel and right rear wheel respectively. is the longitudinal force of the tire ( ), is the lateral force of the tire ( ), is the turning angle of the front wheels ( ), is the tire slip angle ( ), is the direction of the tire speed ( ).

[0013] According to the geometric relationship and force analysis, the following seven-degree-of-freedom dynamic equation can be obtained: Longitudinal Movement: (1); Lateral Movement: (2); Yaw motion: (3); Each wheel rotates independently: (4); in, is the moment of inertia about the z-axis of the vehicle body coordinate system ( ), is the vehicle curb mass ( ), is the moment of inertia of the wheel about its axis ( ), are the angular velocities of the left front wheel, right front wheel, left rear wheel, and right rear wheel respectively ( ), The driving torques of the left front wheel, right front wheel, left rear wheel, and right rear wheel are respectively ( ), is the turning radius of each wheel ( ).

[0014] The angular velocity of each wheel can be expressed as: (5); The vertical load on each wheel is: (6); in, is the longitudinal acceleration, is the lateral acceleration; The formula for tire slip angle is as follows: (7); The formula for each wheel slip rate is as follows: (8); in, are the vertical loads on the left front wheel, right front wheel, left rear wheel, and right rear wheel, respectively. is the height of the vehicle's center of mass, They are the longitudinal slip rates of the left front wheel, right front wheel, left rear wheel, and right rear wheel respectively.

[0015] For any tire of a distributed wire-controlled vehicle, the tire expression based on the "magic" formula is as follows: (9); In the formula, is the lateral force, is the slip angle, is the stiffness factor, is the shape factor, is the peak factor, is the curvature factor, is the horizontal deformation, is the ground normal deformation, is the auxiliary slip angle, To assist the lateral force; the variables of formula 9 can be represented by formula 10; (10); In the formula, is the tire vertical force, is a constant determined by the properties of the vehicle's tires.

[0016] like Figure 3 As shown, as a preferred embodiment of the present invention, in step 1, the method for constructing a linear two-degree-of-freedom reference model is as follows: Lateral Movement: (11); Yaw motion: (12); in, and are the front and rear axle cornering stiffnesses respectively; combining formulas 11 and 12, we can get the following formula: (13); in, is the sideslip angle of the center of mass, is the distance from the vehicle's center of mass to the front axle, is the distance from the vehicle's center of mass to the rear axle, is the vehicle speed in the vehicle body coordinate system The component in the axial direction, m is the vehicle mass, is the vehicle yaw angle, is the turning angle of the front wheels, represents the sideslip angular velocity of the center of mass, is the yaw angular velocity, is the yaw angular acceleration.

[0017] Determine the reference center of mass sideslip angle of distributed drive electric vehicles based on the linear two-degree-of-freedom model With reference yaw rate As shown below: (14); in, is the understeer coefficient; Based on the stability constraints, the desired center of mass slip angle is obtained and the desired yaw rate : (15); in, is the road adhesion coefficient, is the acceleration due to gravity .

[0018] As a preferred embodiment of the present invention, in step 3, the method for establishing the NMPC controller based on the seven-degree-of-freedom dynamics model is as follows: Its state quantity , control amount , output , expected output ,in, is the additional yaw moment, is the desired longitudinal speed, from the driver.

[0019] remember The derivative of , the dynamic analysis of the seven-degree-of-freedom model can obtain the following differential equation: (16); (17); (18); Based on the Euler method discretization, is the step length, is the sampling time, the continuous differential equations described by (16) to (18) are discretized to obtain the discrete differential equation shown in (19): (19); remember To control the time domain, To predict the time domain, according to the current time The state quantity is based on the prediction of formula (19) The state of a step can be expressed as: (20); Assume that when the sampling time exceeds the control time domain, the control quantity remains unchanged, that is: (twenty one); in, Representation variables In the The value of the step.

[0020] Then the control input sequence , control input change sequence , control output sequence , expected output sequence As shown below: (twenty two); Considering that the system is a multi-input and multi-output nonlinear system with inequality constraints, for the tracking target to be achieved and the control amount to be as stable as possible, the cost function is constructed: (twenty three); in, is the weight coefficient matrix to be optimized.

[0021] Defining tire utilization of distributed drive vehicles based on the "adhesion ellipse" theory , the following relationship should be satisfied: (twenty four); At this point, the longitudinal force and additional yaw moment The problem of solving is transformed into solving the following constrained nonlinear programming problem: (25); Subsequent use of INPDOA to optimize NMPC parameters .

[0022] As a preferred embodiment of the present invention, in step 4, the specific method of torque distribution is as follows: The control objective of the torque distribution layer is to provide the longitudinal force through the aforementioned yaw moment control layer. and additional yaw moment , which is expressed as follows: (26); in, As shown in Formula 4, the torque is about the wheel angular velocity function.

[0023] Total demand torque The PID controller is based on the longitudinal speed error The calculation results in: (27); in, They are proportional gain constant, integral gain constant and differential gain constant respectively; define torque distribution coefficient as follows: (28); In order to ensure the stability of distributed drive electric vehicles, the minimum square sum of tire utilization is set as the objective function , defined as follows: (29); To ensure the economy of distributed drive electric vehicles, the power of each drive motor must be determined. formula: (30); In order to unify the magnitude of stability and economy indicators, the drive motor power is normalized: (31); in, is the power of each driving motor, is the power proportion after normalization, is the rated power of the motor.

[0024] To set economic performance indicators: (32); The final performance index of the torque distribution layer is for: (33).

[0025] like Figure 4 As shown, as a preferred embodiment of the present invention, in step 2, the basic neural population dynamic optimization algorithm is specifically as follows: initialization: (34); in, is the variable to be optimized, Representation variables No. Quantity, for The lower limit of for The upper limit of Indicates the generation of a random number between 0 and 1; for the NMPC control layer, is the weight coefficient matrix ; For the torque distribution control layer, is the torque distribution coefficient for each wheel ; Attractor Trend Strategy: (35); in, is the trend of the attractor's scaling factor, Represents a random number between 0 and 1. yes The attractor, is the proportion of attractors among all neural states of the neural population, is the number of neural populations, is the number of attractors, indicating Round down, The covariance is A zero-mean Gaussian noise vector, is the peak value of the neural state, is the minimum boundary of the neural state, It has a mean of 0 and a variance of The normal distribution of .

[0026] Coupling interference strategy: The neural state of a particular neural population is also affected by the neural states of other neural populations. It is usually assumed that the interaction between neural populations occurs through additive or diffusive coupling. In this approach, both types of coupling are considered to exist in neural population dynamics. In additive coupling, the input to a neural population is a function of the sum of the neural states of other neural populations. To simplify the calculations, the neural states of other neural populations are expressed as the average of the sum of all neural states. The formula used is as follows: (36); in, is the neural state addition function, is a random number between 0 and 0.5. is the number of neural populations.

[0027] In diffuse coupling, the input to a neural population is a function of the sum of its neural state and the neural states of other neural populations, using the following formula: (37); in, is the neural state diffusion function, is a random number between 0 and 0.5.

[0028] A combination of two couplings is used to simulate the effects of other neural populations. The formula is as follows: (38); in, is the proportionality factor of the coupled interference.

[0029] Equation (38) shows that one neural population is affected by another neural population through two types of coupling. The interference of other neural populations causes the neural population to deviate from the influence of the attractor. This search strategy based on the coupling between neural populations can help enhance the exploration ability of the algorithm.

[0030] Information projection strategy: As mentioned above, information exchange and interaction between neural populations are carried out through the communication subspace. In this method, the matrix representing adjacency and communication strength is used to represent the communication subspace of information projection between neural populations.

[0031] In the attractor trend, the information transfer between the attractor and another neural state of the neural population is represented by an adjacency matrix with binary random values and the communication strength matrix with randomly generated values ​​between 0 and 1 The formula used is as follows: (39); Perturbations in the neural state should have a greater effect early in the process, before the neural state stabilizes; therefore, the effect of coupled perturbations gradually decreases, using the following formula: (40); in, is an adjacency matrix with binary random values, is a communication strength matrix with randomly generated values ​​between 0 and 1, is a random number between 0 and 1. is the function evaluation degree, The maximum value of the function evaluation. As the iteration process proceeds, The influence of coupling interference gradually increases, while the influence of coupling interference gradually decreases. Eventually, the neural state of each neural group will be affected by the attractors and neural states of other neural groups. The neural state update is as follows: (41); Use the above strategies to simulate information projection between neural populations and achieve the transition from exploration to exploitation.

[0032] As a preferred embodiment of the present invention, in step 2, the specific method of improving the neural population dynamic optimization algorithm is as follows: In order to make up for the shortcomings of the basic neural population dynamic optimization algorithm, the second-order central moment is used in the coupling interference strategy to replace the arithmetic mean in the original formulas (36) and (37). The changes (36*) and (37*) are as follows: (36*); (37*); The simple linear combination shown in formula (38) does not take into account and The correlation between them, this method changes the covariance formula (38*) as follows: (38*); The modified formulas 36*, 37* and 38* are used to replace the original formulas to obtain an improved neural population dynamic optimization algorithm.

[0033] As a preferred embodiment of the present invention, in step 3, for the NMPC control layer, the optimization parameter is the weight coefficient matrix The output of the control is The expected output is , the control quantity is The control target is Under the control of As close as possible to the desired output , the output error is , which is characterized as follows: (42); Normalize each error amount: (43); In the control system performance evaluation, the integral of the absolute error and the time product (ITAE) indicator comprehensively considers the size and duration of the system error, and relatively comprehensively evaluates the dynamic performance of the control strategy. Define the fitness function of the algorithm to optimize the NMPC control The expression is as follows: (44); Where t is time. The weight coefficient matrix of the NMPC control layer is optimized using the improved neural population dynamic optimization algorithm. , and get the optimal solution . Using this optimal solution to perform NMPC control, the optimal control quantity is obtained , providing longitudinal force for subsequent torque distribution control With additional yaw moment .

[0034] As a preferred embodiment of the present invention, in step 4, for the torque distribution control layer, the control goal is to optimize the torque distribution coefficient , so that the tire utilization rate As small as possible to ensure that each tire has a large load margin to avoid the vehicle from slipping and becoming unstable under extreme conditions; minimize the total power of each motor to achieve the goal of energy saving. Fitness function of torque distribution control The same as its performance index (33), the expression is as follows: (45); Optimize the optimal torque distribution coefficient through INPDOA algorithm .

[0035] The following is a simulation example to verify the effectiveness of this method: In order to simplify the simulation process, the upper yaw moment control layer is simulated and verified to obtain the comparison between the actual yaw rate and the expected yaw rate, and the comparison between the actual center of mass sideslip angle and the expected center of mass sideslip angle. The vehicle parameters are as follows: vehicle mass , the moment of inertia around the z-axis of the vehicle body coordinate system , centroid height , Distance from center of mass to front axle , Distance from center of mass to rear axle , wheelbase , Tire rolling radius , Wheel moment of inertia , road adhesion coefficient , Reference longitudinal speed , the initial weight matrices of the NMPC controller are: (46); The double lane change condition was used for simulation verification, and the comparison between the yaw angular velocity and the sideslip angle of the center of mass was obtained. Figure 5 and 6 As shown, it can be seen that the yaw angular velocity and center of mass sideslip angle curves of the NMPC controller optimized by INPDOA are closer to the expected values, indicating that the method proposed in the present invention has certain advantages over the traditional NMPC controller, proving the feasibility of the present invention.

[0036] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A torque distribution method for a distributed drive electric vehicle based on INPDOA, characterized in that: The following steps are involved: Step 1: Establish a seven-degree-of-freedom dynamic model and a linear two-degree-of-freedom dynamic model to determine the state input , control amount , output , expected output ; Step 2: Optimize and improve the basic neural population dynamic optimization algorithm, establish the improved neural population dynamic optimization algorithm architecture, and initialize the parameters; Step 3: Optimize the weight coefficient matrix of the NMPC controller by improving the neural population dynamic optimization algorithm , at this time in the algorithm , the yaw moment control is performed using the optimized NMPC controller; Step 4: Optimize the torque distribution coefficient by improving the neural population dynamic optimization algorithm , at this time in the algorithm , and obtain the optimal torque distribution strategy.

2. The torque distribution method of the distributed drive electric vehicle based on INPDOA according to claim 1 is characterized in that: In step 1, the method for constructing the seven-degree-of-freedom dynamics model of the distributed drive electric vehicle is as follows: Using the body coordinate system of the distributed wire-controlled vehicle as a reference, fix the origin at the center of mass of the vehicle Based on the "right-hand rule", the center axis of the vehicle body along the direction of travel is Positive axis, perpendicular to the vehicle's direction of travel The axis points to the left. Axis positive direction; is the velocity at the center of mass of the vehicle, is the vehicle speed in the vehicle body coordinate system The component in the axial direction, is the vehicle speed in the vehicle body coordinate system The component in the axial direction, is the sideslip angle of the center of mass, is the vehicle yaw angle, is the yaw angular velocity, is the yaw angular acceleration, is the distance from the vehicle's center of mass to the front axle, is the distance from the vehicle's center of mass to the rear axle, is the vehicle width, They represent the left front wheel, right front wheel, left rear wheel and right rear wheel respectively. is the longitudinal force of the tire, is the lateral force of the tire, is the turning angle of the front wheels, is the tire's slip angle, is the speed direction of the tire; According to the geometric relationship and force analysis, the following seven-degree-of-freedom dynamic equation can be obtained: Longitudinal Movement: （1）； Lateral Movement: （2）； Yaw motion: （3）； Each wheel rotates independently: （4）； in, The coordinate system around the vehicle body z The moment of inertia of the shaft, is the vehicle curb weight, is the moment of inertia of the wheel about the axis ( ), are the angular velocities of the left front wheel, right front wheel, left rear wheel, and right rear wheel, respectively. are the driving torques of the left front wheel, right front wheel, left rear wheel, and right rear wheel respectively, is the rotation radius of each wheel; The angular velocity of each wheel is expressed as: （5）； The vertical load on each wheel is: （6）； in, is the longitudinal acceleration, is the lateral acceleration; The formula for tire slip angle is as follows: （7）； The formula for each wheel slip rate is as follows: （8）； in, are the vertical loads on the left front wheel, right front wheel, left rear wheel, and right rear wheel, respectively. is the height of the vehicle's center of mass, are the longitudinal slip rates of the left front wheel, right front wheel, left rear wheel, and right rear wheel respectively; For any tire of a distributed wire-controlled vehicle, the tire expression based on the "magic" formula is as follows: （9）； In the formula, is the lateral force, is the slip angle, is the stiffness factor, is the shape factor, is the peak factor, is the curvature factor, is the horizontal deformation, is the ground normal deformation, is the auxiliary slip angle, To assist the lateral force; the variables of formula 9 can be represented by formula 10; （10）； In the formula, is the tire vertical force, is a constant determined by the properties of the vehicle's tires.

3. The torque distribution method of the distributed drive electric vehicle based on INPDOA according to claim 2 is characterized in that: In step 1, the construction method of the linear two-degree-of-freedom reference model is as follows: Lateral Movement: （11）； Yaw motion: （12）； in, and are the front and rear axle cornering stiffnesses respectively; combining formulas 11 and 12, we can get the following formula: （13）； in, is the sideslip angle of the center of mass, is the distance from the vehicle's center of mass to the front axle, is the distance from the vehicle's center of mass to the rear axle, is the vehicle speed in the vehicle body coordinate system The component in the axial direction, m is the vehicle mass, is the vehicle yaw angle, is the turning angle of the front wheels, represents the sideslip angular velocity of the center of mass, is the yaw angular velocity, is the yaw angular acceleration; Determine the reference center of mass sideslip angle of distributed drive electric vehicles based on the linear two-degree-of-freedom model With reference yaw rate As shown below: （14）； in, is the understeer coefficient; Based on the stability constraints, the desired center of mass slip angle is obtained and the desired yaw rate : （15）； in, is the road adhesion coefficient, is the acceleration due to gravity.

4. The torque distribution method of the distributed drive electric vehicle based on INPDOA according to claim 3 is characterized in that: In step 3, the method for establishing the NMPC controller based on the seven-degree-of-freedom dynamics model is as follows: Its state quantity , control amount , output , expected output ,in, is the additional yaw moment, is the desired longitudinal speed, from the driver; remember The derivative of , the dynamic analysis of the seven-degree-of-freedom model yields the following differential equation: （16）； （17）； （18）； Based on the Euler method discretization, is the step length, is the sampling time, and the continuous differential equations described by formulas 16 to 18 are discretized to obtain the discrete differential equations shown below: （19）； remember To control the time domain, To predict the time domain, according to the current time The state quantity is based on the prediction of formula (19) The state of a step can be expressed as: （20）； Assume that when the sampling time exceeds the control time domain, the control quantity remains unchanged, that is: （21）； in, Representation variables In the The value of the step time; Then the control input sequence , control input change sequence , control output sequence , expected output sequence As shown below: （22）； Considering that the system is a multi-input and multi-output nonlinear system with inequality constraints, for the tracking target to be achieved and the control amount to be as stable as possible, the cost function is constructed: （23）； in, is the weight coefficient matrix to be optimized; Defining tire utilization of distributed drive vehicles based on the "adhesion ellipse" theory , the following relationship should be satisfied: （24）； At this point, the longitudinal force and additional yaw moment The problem of solving is transformed into solving the following constrained nonlinear programming problem: （25）； Subsequent use of INPDOA to optimize NMPC parameters .

5. The torque distribution method of the distributed drive electric vehicle based on INPDOA according to claim 4 is characterized in that: In step 4, the specific method of torque distribution is as follows: The control objective of the torque distribution layer is to provide the longitudinal force through the aforementioned yaw moment control layer. and additional yaw moment , which is expressed as follows: （26）； in, As shown in Formula 4, the torque is about the wheel angular velocity Function of Total demand torque The PID controller is based on the longitudinal speed error The calculations show that: （27）； in, They are proportional gain constant, integral gain constant and differential gain constant respectively; define torque distribution coefficient as follows: （28）； In order to ensure the stability of distributed drive electric vehicles, the minimum square sum of tire utilization is set as the objective function , defined as follows: （29）； To ensure the economy of distributed drive electric vehicles, the power of each drive motor must be determined. formula: （30）； In order to unify the magnitude of stability and economy indicators, the drive motor power is normalized: （31）； in, is the power of each driving motor, is the power proportion after normalization, is the rated power of the motor; To set economic performance indicators: （32）； The final performance index of the torque distribution layer is for: （33）。 6. The torque distribution method of the distributed drive electric vehicle based on INPDOA according to claim 5 is characterized in that: In step 2, the basic neural population dynamic optimization algorithm is specifically as follows: initialization: （34）； in, is the variable to be optimized, Representation variables No. Quantity, for The lower limit of for The upper limit of Indicates the generation of a random number between 0 and 1; for the NMPC control layer, is the weight coefficient matrix ; For the torque distribution control layer, is the torque distribution coefficient for each wheel ; Attractor Trend Strategy: （35）； in, is the trend of the attractor's scaling factor, Represents a random number between 0 and 1. yes The attractor, is the proportion of attractors among all neural states of the neural population, is the number of neural populations, is the number of attractors, indicating Round down, The covariance is A zero-mean Gaussian noise vector, is the peak value of the neural state, is the minimum boundary of the neural state, It has a mean of 0 and a variance of Normal distribution of Coupling interference strategy: Assume that the interaction between neural populations occurs through additive or diffusive coupling. In additive coupling, the input of a neural population is a function of the sum of the neural states of other neural populations. To simplify the calculation, the neural states of other neural populations are expressed as the average of the sum of all neural states. The formula used is as follows: （36）； in, is the neural state addition function, is a random number between 0 and 0.

5. is the number of neural populations; In diffuse coupling, the input to a neural population is a function of the sum of its neural state and the neural states of other neural populations, using the following formula: （37）； in, is the neural state diffusion function, is a random number between 0 and 0.5; A combination of two couplings is used to simulate the effects of other neural populations, as shown in the following formula: （38）； in, is the proportionality factor of the coupled interference; Information projection strategy: As mentioned above, information exchange and interaction between neural populations are carried out through the communication subspace, and the communication subspace of information projection between neural populations is represented by a matrix representing adjacency and communication strength; In the attractor trend, the information transfer between the attractor and another neural state of the neural population is represented by an adjacency matrix with binary random values and the communication strength matrix with randomly generated values ​​between 0 and 1 The formula used is as follows: （39）； Perturbations in the neural state should have a greater effect early in the process, before the neural state stabilizes; therefore, the effect of coupled perturbations gradually decreases, using the following formula: （40）； in, is an adjacency matrix with binary random values, is a communication strength matrix with randomly generated values ​​between 0 and 1, is a random number between 0 and 1. is the function evaluation degree, The maximum value of the function evaluation degree; as the iteration process proceeds, gradually increases, while the influence of coupling interference gradually decreases; eventually, the neural state of each neural group will be affected by the attractors and neural states of other neural groups, and the current neural population The neural state update is as follows: （41）； Use the above strategies to simulate information projection between neural populations and achieve the transition from exploration to exploitation.

7. The torque distribution method of a distributed drive electric vehicle based on INPDOA according to claim 6 is characterized in that: In step 2, the specific method of improving the neural population dynamic optimization algorithm is as follows: In the coupling interference strategy, the second-order central moment is used to replace the arithmetic mean in the original formulas 36 and 37, and is changed to 36* and 37* as shown below: （36*）； （37*）； The analogy covariance formula changes formula 38 to 38* as follows: （38*）； The modified formulas 36*, 37* and 38* are used to replace the original formulas to obtain an improved neural population dynamic optimization algorithm.

8. The torque distribution method of the distributed drive electric vehicle based on INPDOA according to claim 7 is characterized in that: In step 3, for the NMPC control layer, the optimization parameter is the weight coefficient matrix The output of the control is The expected output is , the control quantity is The control target is Under the control of As close as possible to the desired output , the output error is , which is characterized as follows: （42）； Normalize each error amount: （43）； In the control system performance evaluation, the integral index of the product of absolute error and time is used to evaluate the dynamic performance of the control strategy, and the fitness function of the algorithm to optimize the NMPC control is defined. The expression is as follows: （44）； in t For time; The weight coefficient matrix of the NMPC control layer is optimized using an improved neural population dynamic optimization algorithm , and get the optimal solution ; Use this optimal solution to perform NMPC control and obtain the optimal control quantity , providing longitudinal force for subsequent torque distribution control With additional yaw moment .

9. The torque distribution method of the distributed drive electric vehicle based on INPDOA according to claim 7, characterized in that: In step 4, for the torque distribution control layer, the control goal is to optimize the torque distribution coefficient , the fitness function of torque distribution control With its performance indicators The same expression is as follows: （45）； Optimize the optimal torque distribution coefficient through INPDOA algorithm .

Images