CN111559378A - Four-wheel independent-drive electric vehicle dynamics control method considering driver characteristics - Google Patents

Four-wheel independent-drive electric vehicle dynamics control method considering driver characteristics Download PDF

Info

Publication number
CN111559378A
CN111559378A CN202010321277.XA CN202010321277A CN111559378A CN 111559378 A CN111559378 A CN 111559378A CN 202010321277 A CN202010321277 A CN 202010321277A CN 111559378 A CN111559378 A CN 111559378A
Authority
CN
China
Prior art keywords
driver
wheel
neural network
characteristic
data
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202010321277.XA
Other languages
Chinese (zh)
Inventor
李刚
姬晓
李宁
陈双
申彩英
曹景胜
赵德阳
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Liaoning University of Technology
Original Assignee
Liaoning University of Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Liaoning University of Technology filed Critical Liaoning University of Technology
Priority to CN202010321277.XA priority Critical patent/CN111559378A/en
Publication of CN111559378A publication Critical patent/CN111559378A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W40/00—Estimation or calculation of non-directly measurable driving parameters for road vehicle drive control systems not related to the control of a particular sub unit, e.g. by using mathematical models
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W40/00—Estimation or calculation of non-directly measurable driving parameters for road vehicle drive control systems not related to the control of a particular sub unit, e.g. by using mathematical models
    • B60W40/08—Estimation or calculation of non-directly measurable driving parameters for road vehicle drive control systems not related to the control of a particular sub unit, e.g. by using mathematical models related to drivers or passengers
    • B60W40/09—Driving style or behaviour
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Mathematical Physics (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Automation & Control Theory (AREA)
  • Transportation (AREA)
  • Mechanical Engineering (AREA)
  • Probability & Statistics with Applications (AREA)
  • Control Of Driving Devices And Active Controlling Of Vehicle (AREA)

Abstract

The invention discloses a four-wheel independent-drive electric vehicle dynamics control method considering the characteristics of a driver, which comprises the following steps of firstly, acquiring experimental data, and resampling and filtering the acquired experimental data; extracting characteristic values, carrying out cluster analysis, and classifying the characteristics of the driver; establishing a driver characteristic identification model based on the BP neural network under different experimental conditions, and analyzing the identification precision of the identification model; and step four, determining the characteristic type of the driver, inputting the target force and moment into the force and motion distributor, performing dynamic control based on the characteristic of the driver, and distributing power. The classification is carried out based on the characteristics of the driver through the cluster analysis, and the verification and the precision analysis are carried out based on the BP neural network under different test conditions, so that the driver characteristic identification is easy to identify, and the dynamic control is realized according to the characteristics of the driver.

Description

Four-wheel independent-drive electric vehicle dynamics control method considering driver characteristics
Technical Field
The invention relates to a four-wheel independent-drive electric vehicle dynamics control method considering driver characteristics, and belongs to the field of automobile control.
Background
The lack of petroleum resources and the urgency of human environment protection, as well as the increasing living standard and the improvement of automobile technology, the more careful and comprehensive requirements of people on automobiles, all of which push the automobile industry to develop towards the direction of electromotion and intellectualization. With the arrival of the era of automobile electromotion and intelligence, electronic control systems on automobiles are more and more abundant, but the traffic accident rate is rapidly increased along with the emergence of the era, wherein 45% to 75% of road collision accidents are attributed to the mistake of drivers, and the judgment mistake of the drivers becomes a large reason of most traffic accidents. The driving habits of different drivers and the expected response to a car are certainly different, for example, racing drivers prefer a vehicle with quick acceleration and little oversteer for pursuit of excitement, and novice drivers prefer a vehicle with linear response for easy self-control, while old people prefer a vehicle with slow response. At the present stage, the design of the general electric control system does not consider individual differences of drivers, the defects of the general electric control system are more and more obvious, and the driving experience of the automobile is reduced and frequent traffic accidents are possibly caused by continuous use.
Disclosure of Invention
The invention designs and develops a four-wheel independent drive electric vehicle dynamics control method considering the characteristics of a driver, classifies based on the characteristics of the driver through cluster analysis, and performs verification and precision analysis under different test conditions based on a BP neural network, so that the driver characteristic identification is easy to identify, and dynamics control is realized according to the characteristics of the driver.
The technical scheme provided by the invention is as follows:
a four-wheel individual drive electric vehicle dynamics control method considering driver characteristics, comprising:
step one, acquiring experimental data, and resampling and filtering the acquired experimental data;
extracting characteristic values, carrying out cluster analysis, and classifying the characteristics of the driver;
establishing a driver characteristic identification model based on the BP neural network under different experimental conditions, and analyzing the identification precision of the identification model;
and step four, determining the characteristic type of the driver, inputting the target force and moment into the force and motion distributor, performing dynamic control based on the characteristic of the driver, and distributing power.
Preferably, in the first step, the collected experimental data includes: accelerator pedal travel, brake pedal travel, accelerator pedal travel rate of change, brake pedal travel rate of change, longitudinal velocity, longitudinal acceleration, steering wheel angle, and yaw rate;
the resampling time is set to be 0.1s, the sampling interval is 0.001s, and median filtering is adopted for filtering.
Preferably, in the second step, a K-means clustering algorithm is selected to classify the drivers, and:
the clustering number k is 3;
randomly creating central points of k initial centroid positions;
the clustering data dimension is 5;
after clustering, the characteristics of the drivers are classified into A, B, C types, wherein A is an aggressive type, B is a general type, and C is a cautious type.
Preferably, the third step includes:
respectively carrying out a steering experiment, an acceleration experiment and a braking experiment, and establishing a corresponding driver characteristic identification model based on a BP neural network, wherein the method comprises the following steps:
step 1, data normalization;
step 2, carrying out data classification, and selecting training data, variable data and test data;
step 3, establishing a neural network, and setting the number of layers of the neural network, the number of nodes, the transmission function of a hidden layer and the like;
step 4, appointing training parameters for training;
step 5, finishing training, inputting test data, and carrying out accuracy test on the trained network;
step 6, carrying out inverse normalization on the data;
and 7, error analysis and result classification.
It is preferable that the first and second liquid crystal layers are formed of,
the BP neural network adopts a double hidden layer BP neural network;
the threshold value of the hidden layer is [5, 1], the number of nodes of the hidden layer is 5, and the transfer function is S-shaped;
the transfer function of the output layer is a purelin function;
setting the maximum iteration number to be 1000, the learning rate to be 0.01 and the error performance to be 0.001;
and selecting a CGP algorithm in the gradient-varying algorithm as a training function to train the BP neural network.
Preferably, in the fourth step, when performing dynamic control, a dynamic control allocation and a two-stage optimization method are adopted.
Preferably, the dual-stage optimization method includes:
taking the torque and the rotation angle of an actuator of the whole vehicle as independent controllable variables, introducing the maximum torque, the highest rotation speed and the maximum rotation angle of the actuator as constraint conditions, taking dynamic control based on the characteristics of a driver as a main target, and performing first optimization distribution of the force and motion of the actuator subjected to multi-target constraint;
and taking the maximum tire adhesion margin as an optimization target, taking the first optimization distribution result as a constraint condition, performing secondary optimization distribution of the force and the motion of the actuator, and outputting the optimization result to corresponding actuators of each wheel.
Preferably, the empirical formula of the tire adhesion margin is:
Figure BDA0002461512780000031
where μ is the coefficient of adhesion between the tire and the ground, FxiIs the longitudinal force of the i-th wheel, FyiIs the lateral force of the ith wheel.
The invention has the following beneficial effects: the method provides a layered integrated control, force and motion two-stage optimization distribution algorithm aiming at the dynamics control based on the characteristics of a driver, and takes human factors, namely the characteristics of the driver, into consideration when an electric control system is designed, so that the automobile has the vehicle dynamics performance according with the preference of the driver through control, the vehicle dynamics response characteristics expected by the driver are realized, and the conversion from a human-adaptive automobile to a vehicle-adaptive human is realized. Meanwhile, the algorithm performs unified control and distribution on the four-wheel torque and the four-wheel turning angle, so that the coupling influence among all subsystems can be well avoided, and the development potential of the subsystems is improved. In addition, the hierarchical structure algorithm designed by the method designs a function with a weight distribution target in a distribution layer, and takes the tire adhesion margin into consideration, so that each tire force of a vehicle is far away from the limit boundary of the tire as far as possible, the abrasion degree of the tire is reduced, and the driving safety is improved to a certain extent. The control stability and the driving comfort of the automobile are integrally improved, and a theoretical basis is laid for improving the control performance and the intelligent level of a new-generation electric automobile and accelerating the industrialization.
Drawings
Fig. 1 is a flow chart of a driver characteristic classification method according to the present invention.
Fig. 2(a) is a schematic diagram of the acquired data before filtering according to the present invention.
Fig. 2(b) is a schematic diagram of the collected data after filtering according to the present invention.
Fig. 3 is a flow chart of the steering characteristic identification model establishment according to the present invention.
FIG. 4 is a flow chart of a brake characteristic identification model according to the present invention.
FIG. 5 is a flowchart illustrating the establishment of an acceleration characteristic identification model according to the present invention.
Fig. 6(a) is a driving characteristic network training error performance curve according to the present invention.
Fig. 6(b) is a driving characteristic error probability distribution diagram according to the present invention.
FIG. 7 is a graph comparing the characteristic class identification result of the present invention with known results.
Fig. 8 shows the driver characteristic-based dynamics control principle according to the present invention.
FIG. 9 is a diagram of a vehicle dynamics model according to the present invention.
FIG. 10 is a schematic diagram of a vehicle control according to the present invention.
FIG. 11 illustrates the actuator force and motion bi-level optimal distribution principles of the present invention.
Fig. 12 is a schematic view of a friction circle according to the present invention.
Detailed Description
The present invention is further described in detail below with reference to the attached drawings so that those skilled in the art can implement the invention by referring to the description text.
As shown in fig. 1 to 12, the present invention provides a four-wheel independent drive electric vehicle dynamics control method considering driver characteristics, which selects a plurality of drivers to perform test data acquisition, processes the acquired test data, extracts characteristic values, classifies the drivers through cluster analysis, and performs verification and precision analysis under different test conditions through a BP neural network, so that the driver characteristics are easy to identify, and dynamics control is performed according to the driver characteristics, and the method specifically includes:
selecting a plurality of drivers to collect test data, and resampling and filtering the collected test data;
select a plurality of personnel that have certain driving experience, after being familiar with the use of driving the simulator, transfer the experimental operating mode that designs in advance, set for the same highway section, carry out the collection of experimental data through the sensor, the experimental data of gathering include: accelerator pedal travel, brake pedal travel, accelerator pedal travel rate of change, brake pedal travel rate of change, longitudinal velocity, longitudinal acceleration, steering wheel angle, and yaw rate;
after the data acquisition is finished, resampling is carried out, the sampling interval is set to be 0.001s, the resampling time is 0.1s, due to the fact that the sampling interval of experimental data is increased, a large amount of data can be derived everywhere even if the sampling time is short, redundant data are more, unnecessary workload can be increased, and meanwhile the extraction of a follow-up noise elimination characteristic value can be influenced, therefore, filtering is carried out, algae removal is carried out, filtering is carried out by adopting median filtering, and the influence of white noise can be effectively reduced, and the method is shown in fig. 2(a) and 2 (b).
In the data acquisition process of the sensor, abnormal values can occur due to the aging of the sensor or the interference of electromagnetic waves and the like, the number of the abnormal values is not huge, and manual elimination is carried out.
Extracting characteristic values, carrying out cluster analysis, and classifying the characteristics of the driver;
extracting characteristic values of experimental data, and extracting representative data; since the driver characteristics are greatly affected by the maximum value of the data, the method for extracting the characteristic value is specifically as follows:
the acceleration performance of the vehicle is determined by the opening degree and the change rate of the accelerator pedal:
maximum values of speed, acceleration, drive torque, and steering wheel angle are extracted near a time point corresponding to a maximum value of accelerator pedal opening, and a brake pedal opening, brake deceleration, and vehicle speed corresponding to a time point when a rate of change of brake pedal opening is a maximum value are extracted.
The acceleration performance of the vehicle is determined by the magnitude of the steering wheel angle and the magnitude of the change rate: and extracting extreme values of the steering wheel angle and the yaw velocity during each turning and the vehicle speed corresponding to the moment when the steering wheel angle reaches the extreme values.
In the present invention, as a preferable example, the clustering analysis is performed by using a K-means clustering algorithm.
The K-means clustering algorithm is a typical clustering algorithm which adopts distance as an evaluation index of similarity, and is characterized in that data is divided into K types of clusters with the same characteristics according to point-to-point distance. The K-means algorithm is a widely applied heuristic partitioning method in cluster analysis and has the advantages of simplicity and quickness. And after the characteristic values are extracted, classifying the characteristics of the driver by adopting a K-means clustering algorithm.
In the classification of the steering characteristics, when steering to the left and the right, the signs of all indexes are different, absolute value processing is firstly carried out, and according to the actual situation, the clustering number k is set to be 3 to represent three characteristics; the initial centroid position applies random function to create k central points, and data dimension 5 is clustered; the clustering program written using MATLAB is as follows:
[Idx,C,sumD,D]＝kmeans(Data,3,'dist','sqEuclidean','rep',4)
after clustering, dividing the characteristics of the drivers into three types ABC, wherein A is marked as an aggressive type, B is marked as a general type, and C is marked as a cautious type. Aggressive drivers tend to accelerate and decelerate quickly in the driving style, while prudent drivers tend to accelerate and decelerate slowly, with the average being in between.
Establishing a driver characteristic identification model based on the BP neural network under different experimental conditions, and analyzing the identification precision of the identification model;
as shown in fig. 3-5, a steering experiment, a braking experiment and an acceleration experiment are respectively performed, and a driver characteristic online identification model is established and analyzed based on the establishment of the BP neural network.
The BP neural network is one of artificial neural networks, and is a multilayer feedforward network trained by adopting an error inverse propagation algorithm, so that the problem that a single-layer perception neural network cannot process nonlinearity can be well solved, and the BP neural network is widely applied to all neural network models. The BP network can learn the mode mapping relation of input and output in a data set without a definite mathematical relation between the input and the output, has arbitrary complex mode classification capability and excellent multidimensional function mapping capability, and solves the problems of XOR and other problems which cannot be solved by a simple sensor.
And (3) establishing an online identification model of the driver characteristics by adopting a BP neural network toolbox in MATLAB software to realize online identification of the driver characteristics. The basic steps of training a BP neural network by using MATLAB software are as follows: (in the windows 7 system, the software version is MATLAB R2016b, using MATLAB native m language, with default system parameters.)
(1) Normalizing the data;
(2) carrying out data classification, and selecting training data, variable data and test data;
(3) establishing a neural network, and setting the number of layers of the neural network, the number of nodes, a transmission function of a hidden layer and the like;
(4) appointing a training parameter for training;
(5) completing training, inputting test data, and testing the accuracy of the trained network;
(6) carrying out inverse normalization on the data;
(7) error analysis, result prediction or classification, and mapping.
And (3) training and testing the BP Neural Network identification module by using Neural Network Tools (Neural Network toolbox) in MATLAB software and utilizing the classified driver characteristic data. And the classified data set is divided into a training set and a testing set, the training set is used for training the BP neural network identification module, and the testing set is used for verifying the accuracy of the training module.
As shown in fig. 3, a steering experiment is performed, a driver steering characteristic identification model is established, the number of nodes of an input layer and an output layer of the established steering characteristic identification model is 3 input and 1 output respectively, and the input layer is as follows: the characteristic values of the steering wheel angle of the driver during turning, the vehicle speed of the driver during turning, and the yaw rate of the vehicle during turning are as follows: a corresponding driver steering characteristic type, i.e. one of discreet type, general type, aggressive type.
As shown in fig. 4, a braking experiment is performed, a driver braking characteristic identification model is established, the number of nodes of an input layer and an output layer of the established braking characteristic identification model is 5 input and 1 output respectively, and the input layer is as follows: the brake pedal opening degree, the brake pedal opening degree change rate, the speed during braking, the deceleration during braking and the relative distance from the front vehicle during braking of the driver, the output layer is as follows: a corresponding driver braking characteristic type, i.e. one of a cautious type, a generic type, an aggressive type.
As shown in fig. 5, an acceleration experiment is performed to establish a driver acceleration characteristic identification model, the number of nodes of an input layer and an output layer of the established acceleration characteristic identification model is 5 inputs and 1 output, the input layer is the longitudinal acceleration, the longitudinal speed, the pedal opening change rate and the total driving torque of the driver, and the output layer is one of the driver's acceleration characteristic types, namely, a cautious type, a normal type and an aggressive type.
In order to improve the model training efficiency and the model precision, in the invention, as an optimal selection, a double hidden layer BP neural network is selected, the hidden layer threshold value of the BP neural network is set to [51], and the number of hidden layer nodes is determined to be 5. The transfer function of the hidden layer selects an S-type function tan-sigmoid commonly used by a BP neural network, the transfer function of the output layer is a purelin function, the maximum iteration number is set to be 1000, the learning rate is 0.01, and the error performance is 0.001. In order to obtain the performance that the convergence rate is higher than the standard BP neural network speed and the pattern recognition effect is almost the same as the standard BP neural network, in the invention, the CGP algorithm in the gradient-variable algorithm is preferably selected as the training function. The training results of the BP neural network are shown in fig. 6(a) and 6 (b).
After the identification model is established, the accuracy of the identification model is verified by using a reserved test set. Inputting the five-dimensional characteristic value data of the test set into the established identification model, wherein the output result is a numerical value between 0 and 3.5, and the numerical value between 0.5 and 1.5 is determined as an output numerical value 1, namely the identification result is regarded as a cautious driver; determining the value between 1.5 and 2.5 as an output value 2, namely, regarding the identification result as a general driver; determining the value between 2.5 and 3.5 as an output value 3, namely, regarding the identification result as an aggressive driver; this result is compared with the original classification results of the test set, and the comparison graph is shown in fig. 7. As can be seen from the comparison graph, the output result of the identification model has extremely high matching degree with the known characteristic class, so that the identification model has good performance in the off-line simulation stage.
And step four, determining the characteristic type of the driver, inputting the target force and moment into the force and motion distributor, performing dynamic control based on the characteristic of the driver, and distributing power.
The four-wheel independent driving and steering electric automobile has the advantages that the torque and the steering angle of the four wheels are independently controllable, the driving characteristics can be changed by controlling the torque of the four wheels, the steering characteristics can be changed by controlling the steering angle of the four wheels, and the steering angle can be controlled by controlling the force of the four wheels. When the automobile runs, according to the characteristics of a driver, the individual requirements of the driver are met through hierarchical cooperative control based on the characteristics of the driver, and the driving comfort is improved.
As shown in fig. 10, in the actual vehicle control, the driving condition needs to be identified. The running working condition is divided into a limit working condition and a safe working condition, the working condition identification is mainly carried out aiming at the limit working condition, namely, the controller identifies whether the vehicle enters the limit working condition or not and the reason for entering the limit working condition according to information such as driver operation, vehicle running state, environmental condition, driver characteristics and the like. When the automobile is identified not to be in the limit working condition, the controller considers that the automobile runs under the safe working condition.
As shown in FIG. 8, the invention mainly researches the automobile under the safe working condition, the four-wheel independent drive and steering electric automobile control system inputs the reference model of the driver operation and the current vehicle motion state, outputs the longitudinal vehicle speed, the lateral vehicle speed and the yaw angular velocity which are combined with the characteristics of the driver to the integrated controller in equal quantity, the integrated controller outputs the total longitudinal force, the total lateral force, the yaw moment, the pitch moment and the roll moment which meet the control target, and the force and motion optimal distributor distributes the longitudinal force and the lateral force of each wheel, wherein the distributed lateral force of each wheel is finally converted into a wheel corner by adopting an inverse tire model, the wheel corner is controlled by a driving motor to realize, and the selection of the inverse tire model is the key for accurately solving the wheel corner. The tire has nonlinear characteristics, and the force distribution also takes full consideration of the coupling relation between the tire longitudinal force and the tire lateral force attachment ellipse, as shown in fig. 12, which is also an important constraint condition in the force distribution, and specifically includes:
(1) establishing a vehicle dynamic model;
as shown in fig. 9; the model primarily considers lateral, longitudinal, and yaw motions, neglecting roll, vertical, pitch motions, and disregarding the effects of suspension mechanisms. The model comprises seven degrees of freedom of rotation, transverse direction, longitudinal direction and transverse direction of four wheels. An equilibrium equation is established according to newton dynamics:
Figure BDA0002461512780000091
Figure BDA0002461512780000092
Figure BDA0002461512780000093
in the formula: fxiIs the longitudinal force of the ith wheel; fyiIs the lateral force of the ith wheel;ii is 1,2,3,4 f of the steering angle of the ith wheell，fr，rl，rr(ii) a lf is the distance of the center of mass to the front axis; lr is the distance from the center of mass to the rear axis; ls — distance of centroid to wheel; m is the vehicle mass; v. ofxIs the vehicle longitudinal speed; v. ofyIs the vehicle lateral speed, r is the vehicle yaw rate, Mz is the vehicle yaw moment, and β is the vehicle centroid side slip angle.
(2) An integrated control structure;
as shown in FIG. 10, the whole vehicle control idea is to realize a dynamic response target considering the characteristics of a driver through multi-target hierarchical cooperative control. The sensor layer obtains driving signal data according to pedal operation, steering wheel operation and the like of a driver; the identification and estimation layer identifies the characteristics of the driver, obtains the characteristics of the driver, generates corresponding characteristic factors and transmits the characteristic factors to the control layer; the integrated control layer implements control considering the characteristics of a driver, and firstly, decides the total lateral force, the longitudinal force and the yaw moment of the whole vehicle for realizing the control of the driver demand; the force and motion distribution layer reasonably distributes the four-wheel turning angle and the four-wheel torque of the vehicle according to the required target quantity by using a two-stage optimization distribution method; the work of the relevant components of the vehicle in the layer is performed to produce the desired response corresponding to the steering input by the driver. Because the whole vehicle is a multi-input multi-output nonlinear strong coupling uncertain system, the integrated controller must adopt a multi-input multi-output nonlinear adaptive controller.
(3) A control algorithm and a control strategy;
sliding Mode Control (SMC) is a control process in which the control quantity is not continuously changed, and is essentially a special nonlinear control method. Usually, the state of the system (e.g. the deviation and its derivative) is constantly changing, and the structure of the control system can be regularly adjusted according to the current system state, so as to restrict the state trajectory thereof to a predetermined region (referred to as a sliding mode). The sliding mode can be defined by self and is independent of the parameters of system objects and external disturbance, so that the control mode has better robustness.
For a general system:
Figure BDA0002461512780000101
x∈Rn
the existence of a switching surface s (x) is 0 to divide the state space of the system into two parts of s less than 0 and s more than 0, and one type of points moving on the switching surface can approach the switching surface when the switching surface s is close to 0, and the moving points meet the following requirements:
Figure BDA0002461512780000102
therefore, the sliding mode variable structure control is to find the switching surface and set a corresponding control function.
First, assume a control system where X is f (X, U, T), where X is a state vector of the system, U is a system control vector, U is U (X, T), and X ∈ Rn，U∈RnT is a time variable, T ∈ Rn；。
Then determining the switching surface function S (x, t) ═ 0S ∈ Rn
Finally, solving the control function:
Ui(X,t)＝Ui -(X,t),Si(X,t)>0i＝1,2,…,m
Ui -(X,t),Si(X,t)<0
in this method, a sliding mode function is defined as:
eβ＝β-βd，er＝r-rd，eu＝u-ud(4)
Figure BDA0002461512780000103
Figure BDA0002461512780000104
in the formula: cr,Cβ,CuAre all sliding mode constants, satisfying the Helwitz condition, i.e. Cr>0，Cβ>0， Cu>0，udA target vehicle speed; e.g. of the typeβThe deviation of the actual mass center slip angle of the vehicle and the ideal slip angle is obtained; e.g. of the typerThe deviation of the actual yaw rate of the vehicle from the ideal angular rate; e.g. of the typeuIs the deviation of the actual vehicle speed from the driver's target vehicle speed.
In order to ensure that the actual tracking error track is limited on a tracking error sliding mode surface and the system is asymptotically stable, the Lyapunov function is defined as V-S2(ii)/2, using exponential approach:
Figure BDA0002461512780000105
at this time
Figure BDA0002461512780000106
The method is always true, and the accessibility of the sliding mode surface is ensured. According to the Lyapunov theorem, the system is asymptotically stable.
Because the vehicle does not consider the roll and the pitch of the vehicle body, the following formula (1-3) shows that:
Figure BDA0002461512780000111
and (3) obtaining the total lateral force, the longitudinal force and the yaw moment required to be applied to the vehicle according to the formulas (4-8):
Figure BDA0002461512780000112
Figure BDA0002461512780000113
Figure BDA0002461512780000114
(4) lower layer control allocation algorithm
Since four-wheel independent drive and steering electric vehicles are typical overdrive control systems and the control distribution method is an important method for dealing with the problem of overdrive redundancy control, the force and motion optimal distribution method adopts the control distribution method.
In the invention, as a preferable mode, the dynamic control distribution and two-stage optimization theory is adopted to solve the problem that the control solution of the overdrive system is not unique.
As shown in fig. 11, the torque and the rotation angle of the actuator of the whole vehicle are used as independent controllable variables, the maximum torque, the maximum rotation speed and the maximum rotation angle of the actuator are introduced as constraint conditions, and the problem of 'first optimization distribution' of the force and the motion of the actuator, which is based on the dynamics control of the characteristics of the driver as a main target, is abstracted into a multi-target constrained optimization control problem to be solved; and (3) taking the maximum stability margin as an optimization target, taking the result of the first optimization distribution as a constraint condition, performing the second optimization distribution of the force and the motion of the actuator, and outputting the optimization result to a corresponding actuator, so that the automobile has good stability while the characteristics of the driver are met.
The maximum stability margin index mainly considers the tire load rate, so that the tire load rate of each wheel is minimum, and the tire force of each wheel for finishing the target force and the movement is minimum.
The calculation formula of the tire adhesion margin is as follows:
Figure BDA0002461512780000121
where μ is the coefficient of adhesion between the tire and the ground, FxiIs the longitudinal force of the i-th wheel, FyiIs the lateral force of the ith wheel.
As shown in fig. 12, the tire adhesion margin expresses how close the actual tire force is to the boundary of the friction circle, the coupling relationship of the vehicle longitudinal force and the measured force is expressed by the friction circle,
setting the adhesion coefficient between the tire and the ground to be mu, wherein the total force generated by the tire can not exceed mu g, and when the vehicle is accelerated or braked emergently, the longitudinal force is increased and the lateral force is reduced; when the vehicle turns at a high speed, the lateral force is increased to cause the longitudinal force to be reduced, no matter what the working condition of the vehicle is, the vehicle can slip even generate the phenomena of tail flicking and sideslip as long as the resultant force exceeds the limit allowed by a friction circle, so that the life safety of a driver and passengers is influenced, and the ideal response of the characteristics of the driver is realized and the stability of the vehicle is ensured by using a two-stage optimization allocation algorithm.
Designing an objective function: j. the design is a square3＝J2+J1Wherein J1＝[BF-U]TQ[BF-U]，J2＝FTWF。
In the formula: j. the design is a square1For the dynamic behavior of vehicles, J2For the stability of the vehicle, U is the target desired value obtained by the control layer, U ═ FxFyMZ]TB is a constant coefficient matrix relating to the steering angle of the tire and vehicle parameters, and Q is J1W is J2The weight matrix of (2). F is the longitudinal force and the lateral force of each wheel, and F is ═ Fx1，Fx2，Fx3，Fx4Fx1Fy1Fy2Fy3Fy4]T。
The second derivative of the objective function is:
Figure BDA0002461512780000122
the minimum of the objective function can therefore be found when the first derivative is zero:
Figure BDA0002461512780000123
the following can be obtained:
F*＝(BTQB+W)-1BTQU (10)
Figure BDA0002461512780000124
in the formula: f*Is the longitudinal and lateral forces of each wheel at the minimum of the objective function. The F*Constrained by the matrix W, ensures that the individual tire forces are within the friction circle stability region. Actual lateral force Fyi-realCan be obtained by sensors, the desired lateral force, F, for reasonable controlyi *The steering angle required by the wheel with certain lateral force can be obtained by the following formula, and the finally obtained tire force can realize driving control through 4 hub motors.iThe desired tire force is obtained by substituting the matrix B obtained by equation (11) and finally obtained by equation (10).
While embodiments of the invention have been described above, it is not limited to the applications set forth in the description and the embodiments, which are fully applicable in various fields of endeavor to which the invention pertains, and further modifications may readily be made by those skilled in the art, it being understood that the invention is not limited to the details shown and described herein without departing from the general concept defined by the appended claims and their equivalents.

Claims (8)

1. A four-wheel individual drive electric vehicle dynamics control method taking into account driver characteristics, characterized by comprising:
step one, acquiring experimental data, and resampling and filtering the acquired experimental data;
extracting characteristic values, carrying out cluster analysis, and classifying the characteristics of the driver;
establishing a driver characteristic identification model based on the BP neural network under different experimental conditions, and analyzing the identification precision of the identification model;
and step four, determining the characteristic type of the driver, inputting the target force and moment into the force and motion distributor, performing dynamic control based on the characteristic of the driver, and distributing power.
2. A four-wheel individual drive electric vehicle dynamics control method in consideration of driver characteristics according to claim 1, wherein the experimental data collected in the first step comprises: accelerator pedal travel, brake pedal travel, accelerator pedal travel rate of change, brake pedal travel rate of change, longitudinal velocity, longitudinal acceleration, steering wheel angle, and yaw rate;
the resampling time is set to be 0.1s, the sampling interval is 0.001s, and median filtering is adopted for filtering.
3. A four-wheel individual drive electric vehicle dynamics control method in consideration of driver characteristics as claimed in claim 2, wherein in the second step, a K-means clustering algorithm is selected to classify the driver, and set:
the clustering number k is 3;
randomly creating central points of k initial centroid positions;
the clustering data dimension is 5;
after clustering, the characteristics of the drivers are classified into A, B, C types, wherein A is an aggressive type, B is a general type, and C is a cautious type.
4. A four-wheel individual drive electric vehicle dynamics control method in consideration of driver characteristics according to claim 2, characterized in that the third step comprises:
respectively carrying out a steering experiment, an acceleration experiment and a braking experiment, and establishing a corresponding driver characteristic identification model based on a BP neural network, wherein the method comprises the following steps:
step 1, data normalization;
step 2, carrying out data classification, and selecting training data, variable data and test data;
step 3, establishing a neural network, and setting the number of layers of the neural network, the number of nodes, the transmission function of a hidden layer and the like;
step 4, appointing training parameters for training;
step 5, finishing training, inputting test data, and carrying out accuracy test on the trained network;
step 6, carrying out inverse normalization on the data;
and 7, error analysis and result classification.
5. The four-wheel individual drive electric vehicle dynamics control method taking into account driver characteristics according to claim 4, characterized in that,
the BP neural network adopts a double hidden layer BP neural network;
the threshold value of the hidden layer is [5, 1], the number of nodes of the hidden layer is 5, and the transfer function is S-shaped;
the transfer function of the output layer is a purelin function;
setting the maximum iteration number to be 1000, the learning rate to be 0.01 and the error performance to be 0.001;
and selecting a CGP algorithm in the gradient-varying algorithm as a training function to train the BP neural network.
6. A four-wheel individual drive electric vehicle dynamics control method taking into account driver characteristics according to claim 5, wherein in the fourth step, dynamic control distribution and two-stage optimization methods are adopted when performing dynamics control.
7. A four-wheel individual drive electric vehicle dynamics control method in consideration of driver characteristics according to claim 6, characterized in that the two-stage optimization method comprises:
taking the torque and the rotation angle of an actuator of the whole vehicle as independent controllable variables, introducing the maximum torque, the highest rotation speed and the maximum rotation angle of the actuator as constraint conditions, taking dynamic control based on the characteristics of a driver as a main target, and performing first optimization distribution of the force and motion of the actuator subjected to multi-target constraint;
and taking the maximum tire adhesion margin as an optimization target, taking the first optimization distribution result as a constraint condition, performing secondary optimization distribution of the force and the motion of the actuator, and outputting the optimization result to corresponding actuators of each wheel.
8. A four-wheel individual drive electric vehicle dynamics control method in consideration of driver characteristics according to claim 7, wherein the empirical formula of the tire adhesion margin is:
Figure FDA0002461512770000021
where μ is the coefficient of adhesion between the tire and the ground, FxiIs the longitudinal force of the i-th wheel, FyiIs the lateral force of the ith wheel.
CN202010321277.XA 2020-04-22 2020-04-22 Four-wheel independent-drive electric vehicle dynamics control method considering driver characteristics Pending CN111559378A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010321277.XA CN111559378A (en) 2020-04-22 2020-04-22 Four-wheel independent-drive electric vehicle dynamics control method considering driver characteristics

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010321277.XA CN111559378A (en) 2020-04-22 2020-04-22 Four-wheel independent-drive electric vehicle dynamics control method considering driver characteristics

Publications (1)

Publication Number Publication Date
CN111559378A true CN111559378A (en) 2020-08-21

Family

ID=72067812

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010321277.XA Pending CN111559378A (en) 2020-04-22 2020-04-22 Four-wheel independent-drive electric vehicle dynamics control method considering driver characteristics

Country Status (1)

Country Link
CN (1) CN111559378A (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112632707A (en) * 2020-12-29 2021-04-09 浙江天行健智能科技有限公司 Working condition fusion road feel simulation method based on ANN algorithm
CN112632706A (en) * 2020-12-29 2021-04-09 浙江天行健智能科技有限公司 Road feel simulation method based on GMM and CART regression tree
CN112660109A (en) * 2020-12-25 2021-04-16 浙江吉利控股集团有限公司 Four-wheel drive torque limiting method and device
CN112836722A (en) * 2020-12-26 2021-05-25 浙江天行健智能科技有限公司 Road feel simulation method based on data driving
CN112896185A (en) * 2021-01-25 2021-06-04 北京理工大学 Intelligent driving behavior decision planning method and system for vehicle-road cooperation

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101633359A (en) * 2008-07-24 2010-01-27 通用汽车环球科技运作公司 Adaptive vehicle control system with driving style recognition
CN108280484A (en) * 2018-01-30 2018-07-13 辽宁工业大学 A kind of driver's accelerating performance online classification and discrimination method
CN108407814A (en) * 2018-03-08 2018-08-17 辽宁工业大学 A kind of driver characteristics discrimination method
CN108482481A (en) * 2018-03-21 2018-09-04 辽宁工业大学 Four motorized wheels and the four-wheel steering control method for turning to electric vehicle
EP3547215A1 (en) * 2018-03-26 2019-10-02 Cohda Wireless Pty Ltd. Systems and methods for automatically training neural networks

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101633359A (en) * 2008-07-24 2010-01-27 通用汽车环球科技运作公司 Adaptive vehicle control system with driving style recognition
CN108280484A (en) * 2018-01-30 2018-07-13 辽宁工业大学 A kind of driver's accelerating performance online classification and discrimination method
CN108407814A (en) * 2018-03-08 2018-08-17 辽宁工业大学 A kind of driver characteristics discrimination method
CN108482481A (en) * 2018-03-21 2018-09-04 辽宁工业大学 Four motorized wheels and the four-wheel steering control method for turning to electric vehicle
EP3547215A1 (en) * 2018-03-26 2019-10-02 Cohda Wireless Pty Ltd. Systems and methods for automatically training neural networks

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
刘经文: "四轮独立电动车驱动/转向/制动稳定性集成控制算法研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 *
孟腾飞等: "4WD_WS电动车辆稳定性分层控制与轮胎力优化分配算法研究", 《机电工程》 *

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112660109A (en) * 2020-12-25 2021-04-16 浙江吉利控股集团有限公司 Four-wheel drive torque limiting method and device
CN112660109B (en) * 2020-12-25 2022-08-30 浙江吉利控股集团有限公司 Four-wheel drive torque limiting method and device
CN112836722A (en) * 2020-12-26 2021-05-25 浙江天行健智能科技有限公司 Road feel simulation method based on data driving
CN112632707A (en) * 2020-12-29 2021-04-09 浙江天行健智能科技有限公司 Working condition fusion road feel simulation method based on ANN algorithm
CN112632706A (en) * 2020-12-29 2021-04-09 浙江天行健智能科技有限公司 Road feel simulation method based on GMM and CART regression tree
CN112632707B (en) * 2020-12-29 2023-08-01 浙江天行健智能科技有限公司 Working condition fusion road feel simulation method based on ANN algorithm
CN112632706B (en) * 2020-12-29 2024-03-05 浙江天行健智能科技有限公司 Road feel simulation method based on GMM and CART regression tree
CN112896185A (en) * 2021-01-25 2021-06-04 北京理工大学 Intelligent driving behavior decision planning method and system for vehicle-road cooperation

Similar Documents

Publication Publication Date Title
CN111559378A (en) Four-wheel independent-drive electric vehicle dynamics control method considering driver characteristics
CN105741637B (en) Four-wheel hub motor electric car automated steering control method
Zhang et al. Chassis coordinated control for full X-by-wire vehicles-A review
Zhu et al. Typical-driving-style-oriented personalized adaptive cruise control design based on human driving data
CN111775949B (en) Personalized driver steering behavior auxiliary method of man-machine co-driving control system
CN111332362B (en) Intelligent steer-by-wire control method integrating individual character of driver
CN108482481B (en) Four-wheel steering control method for four-wheel independent drive and steering electric automobile
Zhao et al. Identification of driver’s braking intention based on a hybrid model of GHMM and GGAP-RBFNN
CN109733474A (en) A kind of intelligent vehicle steering control system and method based on piecewise affine hierarchical control
CN111717217B (en) Driver intention identification method based on probability correction
CN112046489B (en) Driving style identification algorithm based on factor analysis and machine learning
Guo et al. Improved car-following strategy based on merging behavior prediction of adjacent vehicle from naturalistic driving data
Deng et al. Shared control for intelligent vehicle based on handling inverse dynamics and driving intention
CN115257819A (en) Decision-making method for safe driving of large-scale commercial vehicle in urban low-speed environment
CN115303289A (en) Vehicle dynamics model based on depth Gaussian, training method, intelligent vehicle trajectory tracking control method and terminal equipment
CN114987537A (en) Neural network dynamics-based road adaptive drift control system and method for automatic driving vehicle
Sousa et al. Nonlinear tire model approximation using machine learning for efficient model predictive control
CN113341994B (en) Intelligent automobile path tracking control method based on optimal control of piecewise affine system
CN113184040B (en) Unmanned vehicle line-controlled steering control method and system based on steering intention of driver
CN117227834B (en) Man-machine cooperative steering control method for special vehicle
Rath et al. A lane keeping assist design: Adaptation to driving style based on aggressiveness
CN112810602B (en) Intelligent line control chassis personalized stability control method and system based on clustering
Shojaeefard et al. Weather-aware fuzzy adaptive cruise control: Dynamic reference signal design
CN114148349A (en) Vehicle personalized following control method based on generation countermeasure simulation learning
Nacpil et al. Surface electromyography-controlled pedestrian collision avoidance: A driving simulator study

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20200821