CN115204417B - Vehicle weight prediction method and system based on ensemble learning and storage medium - Google Patents

Vehicle weight prediction method and system based on ensemble learning and storage medium Download PDF

Info

Publication number
CN115204417B
CN115204417B CN202211109476.XA CN202211109476A CN115204417B CN 115204417 B CN115204417 B CN 115204417B CN 202211109476 A CN202211109476 A CN 202211109476A CN 115204417 B CN115204417 B CN 115204417B
Authority
CN
China
Prior art keywords
vehicle
vehicle weight
learner
trained
predicting
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.)
Active
Application number
CN202211109476.XA
Other languages
Chinese (zh)
Other versions
CN115204417A (en
Inventor
张楠
吴碧磊
张中磊
张文
陈金城
张志武
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.)
Yukuai Chuangling Intelligent Technology Nanjing Co ltd
Original Assignee
Yukuai Chuangling Intelligent Technology Nanjing Co ltd
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 Yukuai Chuangling Intelligent Technology Nanjing Co ltd filed Critical Yukuai Chuangling Intelligent Technology Nanjing Co ltd
Priority to CN202211109476.XA priority Critical patent/CN115204417B/en
Publication of CN115204417A publication Critical patent/CN115204417A/en
Application granted granted Critical
Publication of CN115204417B publication Critical patent/CN115204417B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G06N20/20—Ensemble learning
    • 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

Abstract

The application provides a vehicle weight prediction method, a vehicle weight prediction system and a storage medium based on ensemble learning. The vehicle weight prediction method includes: predicting to obtain a first vehicle weight based on vehicle running state parameters and environment parameters through a longitudinal vehicle force balance equation; predicting, by at least one trained base learner, a second vehicle weight based at least on non-constant ones of the vehicle travel state parameters and the environmental parameters; predicting, by a trained meta learner, a vehicle weight based on the first vehicle weight and the second vehicle weight.

Description

Vehicle weight prediction method and system based on ensemble learning and storage medium
Technical Field
The present application relates to the field of computing and calculating, and in particular, to a vehicle weight prediction method, system and storage medium based on ensemble learning.
Background
In recent years, the transportation industry in China is developing vigorously. During the transportation of people and goods by vehicles, it is often desirable for a passenger or freight manager to be able to know the weight information of the vehicle in real time or near real time.
For example, in the background of current internet of vehicles and rapid development of big data technology, if an enterprise can monitor the weight of a vehicle in real time in the management process of logistics vehicles, the enterprise can effectively perform task scheduling and monitoring management on the vehicle, so that unnecessary resource loss or risks can be avoided. In addition, for the government traffic control department, if the real-time dynamic monitoring of the vehicle weight can be realized, the problem of vehicle overload can be conveniently and effectively treated, so that the service life of a road is prolonged, and the driving safety is effectively guaranteed.
In this large context, conventional vehicle weight measurement using static wagon balance has not met the needs of current social development. The market has a high demand for a new generation of vehicle weight detection analysis solutions.
Disclosure of Invention
The application provides a vehicle weight prediction method based on ensemble learning, which comprises the following steps: predicting to obtain a first vehicle weight based on the vehicle running state parameters and the environmental parameters through a longitudinal vehicle force balance equation; predicting, by at least one trained base learner, a second vehicle weight based at least on non-constant ones of the vehicle travel state parameters and the environmental parameters; predicting a vehicle weight based on the first vehicle weight and the second vehicle weight by a trained meta-learner, wherein the meta-learner is trained using a gradient descent algorithm based on an output of the longitudinal vehicle force balance equation and an output of the at least one trained base learner as inputs, the meta-learner using a linear regression model of
Figure DEST_PATH_IMAGE002A
，
Wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE004A
、
Figure DEST_PATH_IMAGE006A
、
Figure DEST_PATH_IMAGE008A
is a weight value of the weight value,
Figure DEST_PATH_IMAGE010A
in order to be an error term, the error term,
Figure DEST_PATH_IMAGE012A
is the output of the longitudinal vehicle force balance equation,
Figure DEST_PATH_IMAGE014A
to
Figure DEST_PATH_IMAGE016A
Is the output of the at least one trained base learner,
the objective function is:
Figure DEST_PATH_IMAGE018A
。
according to an embodiment of the present application, the vehicle running state parameter includes: frontal projection area of vehicle in windward direction
Figure DEST_PATH_IMAGE020A
Speed of vehicle
Figure DEST_PATH_IMAGE022A
Rotational speed of engine
Figure DEST_PATH_IMAGE024A
Torque of engine
Figure DEST_PATH_IMAGE026A
Mechanical efficiency of vehicle drive train
Figure DEST_PATH_IMAGE028A
Acceleration of vehicle
Figure DEST_PATH_IMAGE030A
Pressure in the tyre
Figure DEST_PATH_IMAGE032A
Angle between vehicle and horizontal plane
Figure DEST_PATH_IMAGE034A
(ii) a The environmental parameters include: air pressure
Figure DEST_PATH_IMAGE036A
Relative humidity of air
Figure DEST_PATH_IMAGE038A
Air temperature in Kelvin
Figure DEST_PATH_IMAGE040A
Temperature in centigrade
Figure DEST_PATH_IMAGE042A
Coefficient of air resistance
Figure DEST_PATH_IMAGE044A
Velocity of wind
Figure DEST_PATH_IMAGE046A
Acceleration of gravity
Figure DEST_PATH_IMAGE048A
。
According to an embodiment of the application, the longitudinal vehicle force balance equation is:
Figure DEST_PATH_IMAGE050A
wherein m is the first vehicle weight,
coefficient of rolling resistance
Figure DEST_PATH_IMAGE052A
The expression of (a) is:
Figure DEST_PATH_IMAGE054A
air density of the location of the vehicle
Figure DEST_PATH_IMAGE056A
The expression of (a) is:
Figure DEST_PATH_IMAGE058A
。
according to an embodiment of the present application, the at least one trained base learner comprises a plurality of base learners trained based on different initialization parameters.
According to the embodiment of the application, each base learner in the plurality of base learners comprises at least one hidden layer, and the nonlinear activation function adopted by the hidden layer is a Sigmoid function.
According to the embodiment of the application, each base learner in the plurality of base learners is trained by adopting a K-fold cross validation method, wherein K is a natural number not less than 2.
According to the embodiment of the application, the at least one trained base learner comprises three trained base learners based on different initialization parameters, and each base learner is trained by adopting a four-fold cross validation method.
The present application further provides a vehicle weight prediction system based on ensemble learning, comprising: a memory storing executable instructions; and one or more processors in communication with the memory to execute the executable instructions to: predicting to obtain a first vehicle weight based on vehicle running state parameters and environment parameters through a longitudinal vehicle force balance equation; predicting, by at least one trained base learner, a second vehicle weight based at least on non-constant ones of the vehicle travel state parameters and the environmental parameters; predicting, by a trained meta learner, a vehicle weight based on the first vehicle weight and the second vehicle weight.
The present application further proposes a computer-readable storage medium storing executable instructions that are executable by one or more processors to: predicting to obtain a first vehicle weight based on the vehicle running state parameters and the environmental parameters through a longitudinal vehicle force balance equation; predicting, by at least one trained base learner, a second vehicle weight based at least on non-constant ones of the vehicle travel state parameters and the environmental parameters; predicting, by a trained meta learner, a vehicle weight based on the first vehicle weight and the second vehicle weight.
The present application also proposes a computer program product for vehicle weight prediction, comprising a computer program characterized in that said computer program when processed performs the following operations: predicting to obtain a first vehicle weight based on vehicle running state parameters and environment parameters through a longitudinal vehicle force balance equation; predicting, by at least one trained base learner, a second vehicle weight based at least on non-constant ones of the vehicle travel state parameters and the environmental parameters; predicting, by a trained meta learner, a vehicle weight based on the first vehicle weight and the second vehicle weight.
The present application further provides a vehicle weight prediction device, including: an equation prediction module that predicts a first vehicle weight based on a vehicle driving state parameter and an environmental parameter through a longitudinal vehicle force balance equation; a neural network prediction module that predicts a second vehicle weight based on at least a non-constant of the vehicle driving state parameter and the environmental parameter through at least one trained base learner; a quadratic fitting module that predicts a vehicle weight based on the first vehicle weight and the second vehicle weight by a trained meta learner.
The technical scheme for predicting the vehicle weight effectively combines a method for predicting the vehicle weight based on a longitudinal vehicle force balance equation and a method for predicting the vehicle weight based on a neural network (base learner). On one hand, the combination can utilize a force balance equation to predict the vehicle weight, and the prediction generally does not have the defect of poor overfitting or generalization capability possibly occurring in a neural network model; on the other hand, the capability of universal approximation of the neural network can be effectively utilized to obtain more accurate prediction results. Considering that the generalization capability of the ensemble learning model is generally better than that of a single strong learner, and that strong learners are generally difficult to acquire, the present application utilizes ensemble learning to improve the prediction accuracy of an easily-acquired but relatively weak learner to a prediction accuracy comparable to that of a strong learner. According to the technical scheme of the vehicle weight prediction, the integrated learning method based on the stacking method is introduced, and the prediction result has the advantages of high robustness and low prediction error.
Drawings
Other features, objects and advantages of the present application will become more apparent upon reading of the following detailed description of non-limiting embodiments thereof, made with reference to the accompanying drawings in which:
FIG. 1 is a flow chart of a vehicle weight prediction method according to an embodiment of the present application;
FIG. 2 is a schematic diagram of an integrated training model training process according to an embodiment of the present application;
fig. 3 is a schematic block diagram of a vehicle weight prediction system according to an embodiment of the present application.
Detailed Description
For a better understanding of the present application, the technical solutions of the present application will be described in more detail with reference to the accompanying drawings. It should be understood that the detailed description is merely illustrative of exemplary embodiments of the present application and is not intended to limit the scope of the present application in any way. Like reference numerals refer to like elements throughout the specification. The expression "and/or" includes any and all combinations of one or more of the associated listed items.
It should be noted that in the present specification, expressions "first", "second", "third", etc. are used only for distinguishing one feature from another, and do not represent any limitation on the features. Thus, the first vehicle weight discussed below may also be referred to as a second vehicle weight without departing from the teachings of the present application. And vice versa.
In the drawings, the size, proportion and shape of the illustrations have been adjusted slightly for the convenience of illustration. The figures are purely diagrammatic and not drawn to scale. As used herein, the terms "approximately," "about," and the like are used as terms of table approximation and not as terms of table degree, and are intended to account for inherent deviations in measured or calculated values that will be recognized by those of ordinary skill in the art.
It will be further understood that expressions such as "comprising," "including," "having," "including," and/or "containing" are open-ended and not closed-ended expressions in this specification that indicate the presence of stated features, elements, and/or components, but do not preclude the presence or addition of one or more other features, elements, components, and/or groups thereof. Moreover, when a statement such as "at least one of" appears after a list of listed features, it modifies the entire list of features, rather than just a single feature in the list. Furthermore, when describing embodiments of the present application, the use of "may" mean "one or more embodiments of the present application. Also, the word "exemplary" is intended to mean exemplary or illustrative.
Unless otherwise defined, all terms (including engineering and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
In addition, the features of the embodiments and examples in the present application may be combined with each other without conflict. In addition, unless explicitly defined or contradicted by context, the specific steps included in the methods described herein are not necessarily limited to the order described, but can be performed in any order or in parallel. The present application will be described in detail below with reference to the accompanying drawings in conjunction with embodiments.
Fig. 1 is a flowchart of a vehicle weight prediction method according to an embodiment of the present application.
Referring to fig. 1, the present application proposes a vehicle weight prediction method 1000 based on ensemble learning. According to an embodiment of the present application, the vehicle weight prediction method 1000 includes the following steps.
In step S1010, a first vehicle weight is predicted based on the vehicle running state parameter and the environmental parameter by the longitudinal vehicle force balance equation. The vehicle running state parameter may be a physical quantity related to the running state of the vehicle, and the environment parameter may be a physical quantity related to the environment in which the vehicle is currently located. The longitudinal vehicle force balance equation may be an equation derived based on a force balance of the vehicle in a vertical direction (i.e., longitudinal direction). The vehicle driving state parameter and the environmental parameter may be acquired by an on-board sensor, but the present application is not limited thereto. For example, the environmental parameters may also be obtained by a driving computer from an external server via a network.
In step S1020, a second vehicle weight is predicted by at least one trained base learner based on at least the vehicle driving state parameter and the non-constant term of the environmental parameter. The base learner can be a neural network model built based on Keras or TensorFlow, the input quantity of the neural network model can be vehicle driving state parameters and environment parameters, and the output quantity of the neural network model can be the weight of the vehicle. Since the base learner can well adapt to the constant deviation, it is generally unnecessary to input the constant terms in the vehicle running state parameter and the environmental parameter to the base learner.
In some application scenarios, some vehicle driving state parameters and environmental parameters may have large noise. These noises may be due to electrical noise of the sensor or other factors. For example, during continuous operation of the vehicle, the speed information jumps at one instant and returns to normal at the next instant. The instantaneous abnormal velocity information is noise information. According to the embodiment of the application, the parameters can be subjected to data cleaning and then input into a longitudinal vehicle force balance equation or a base learner. For example, when a parameter is collected to have a larger jump compared with other parameters of the same class, the parameter of the jump can be replaced by the parameter of the same class collected at the last time. Further, low pass filtering may also be employed to filter the parameters to screen out noisy data.
In step S1030, a vehicle weight is predicted based on the first vehicle weight and the second vehicle weight by the trained meta learner. The method is based on a Stacking method (Stacking) integrated learning model, a model for predicting vehicle weight based on a longitudinal vehicle force balance equation and a model for predicting the vehicle weight based on a neural network are used as base learners located on a first layer, and a meta-learner located on a second layer is used for carrying out quadratic fitting on results of the base learners located on the first layer, so that the vehicle weight with high robustness and low prediction error is obtained.
In the present application, step S1010 and step S1020 are not limited in order. Step S1010 may be performed before step S1020, after step S1020, or in parallel with step S1020.
The vehicle weight prediction method 1000 described with reference to fig. 1 effectively combines a method of predicting the weight of a vehicle based on a longitudinal vehicle force balance equation with a method of predicting the weight of a vehicle based on a neural network (base learner). On one hand, the combination can utilize a force balance equation to predict the vehicle weight, and the prediction generally does not have the defect of poor overfitting or generalization capability possibly occurring in a neural network model; on the other hand, the capability of universal approximation of the neural network can be effectively utilized to obtain more accurate prediction results. Considering that the generalization capability of the ensemble learning model is generally better than that of a single strong learner, and that strong learners are generally difficult to obtain, the present application utilizes ensemble learning to improve the prediction accuracy of an easily obtained, but relatively weak learner to a prediction accuracy comparable to that of a strong learner. According to the vehicle weight prediction method 1000 of the embodiment of the application, an ensemble learning method based on a stacking method is introduced, and the prediction result has the advantages of high robustness and low prediction error.
According to an embodiment of the present application, the vehicle running state parameter includes: frontal projection area of vehicle in windward direction
Figure DEST_PATH_IMAGE020AA
Speed of vehicle
Figure DEST_PATH_IMAGE022AA
Rotational speed of engine
Figure DEST_PATH_IMAGE024AA
Torque of engine
Figure DEST_PATH_IMAGE026AA
Mechanical efficiency of vehicle drive train
Figure DEST_PATH_IMAGE028AA
Acceleration of vehicle
Figure DEST_PATH_IMAGE030AA
Pressure in the tyre
Figure DEST_PATH_IMAGE032AA
Angle between vehicle and horizontal plane
Figure DEST_PATH_IMAGE034AA
(ii) a The environmental parameters include: air pressure
Figure DEST_PATH_IMAGE036AA
Relative humidity of air
Figure DEST_PATH_IMAGE038AA
Air temperature in Kelvin
Figure DEST_PATH_IMAGE040AA
Temperature in centigrade
Figure DEST_PATH_IMAGE042AA
Coefficient of air resistance
Figure DEST_PATH_IMAGE044AA
Velocity of wind
Figure DEST_PATH_IMAGE046AA
Acceleration of gravity
Figure DEST_PATH_IMAGE048AA
。
According to an embodiment of the application, the longitudinal vehicle force balance equation may be:
Figure DEST_PATH_IMAGE060A
，
wherein m is the first vehicle weight,
coefficient of rolling resistance
Figure DEST_PATH_IMAGE062A
The expression of (c) is:
Figure DEST_PATH_IMAGE064
air density of the location of the vehicle
Figure DEST_PATH_IMAGE066A
The expression of (c) is:
Figure DEST_PATH_IMAGE068A
。
it should be noted that although the prior art proposes a plurality of longitudinal vehicle force balance equations, the parameters and approximation assumptions considered in different solutions are different, and thus the longitudinal vehicle force balance equations are different.
The present application considers the force balance between vehicle traction, rolling resistance, air resistance, and vehicle gravity when developing a longitudinal vehicle force balance equation. The following specifically describes the establishment of the longitudinal vehicle force balance equation.
Vehicle traction
Figure DEST_PATH_IMAGE070A
The formula of (1) is as follows:
Figure DEST_PATH_IMAGE072A
，
Figure DEST_PATH_IMAGE074A
is the engine torque, in units:
Figure DEST_PATH_IMAGE076A
；
Figure DEST_PATH_IMAGE078A
is the wheel radius, in units:
Figure DEST_PATH_IMAGE080A
；
Figure DEST_PATH_IMAGE082
is the main reducer transmission ratio, unit: none;
Figure DEST_PATH_IMAGE084
is the transmission ratio of a certain gear of the speed changer, and the unit is as follows: none;
Figure DEST_PATH_IMAGE086
is the mechanical efficiency of the drive train, in units:
Figure DEST_PATH_IMAGE088
。
the relationship between the engine speed n (in rpm, rpm) and the vehicle speed v (in km/h) is:
Figure DEST_PATH_IMAGE090
，
by comprehensively considering the expression, the vehicle traction force can be obtained
Figure DEST_PATH_IMAGE091
The formula of (1) is:
Figure DEST_PATH_IMAGE093
，
when the wheels of the vehicle rotate, deformation occurs between the wheels and the ground. Such deformation causes some irreversible energy loss, which is reflected in the force applied as rolling resistance during the movement of the vehicle. Rolling resistance as used herein
Figure DEST_PATH_IMAGE095
The formula of (1) is:
Figure DEST_PATH_IMAGE097
，
Figure DEST_PATH_IMAGE099
represents the angle between the vehicle and the ground level, i.e. the angle between the road on which the vehicle is travelling and the horizontal plane, in units:
Figure DEST_PATH_IMAGE101
；
Figure DEST_PATH_IMAGE103
represents the weight of the vehicle at this time, in units:
Figure DEST_PATH_IMAGE105
；
Figure DEST_PATH_IMAGE107
is the rolling resistance coefficient, in units: none.
Since the rolling resistance coefficient has a large relationship with the vehicle speed and the pressure inside the tire, the present application selects the following formula as an approximate formula of the rolling resistance coefficient:
Figure DEST_PATH_IMAGE109
in the above-mentioned formula, the compound has the following structure,
Figure DEST_PATH_IMAGE111
in the unit of (a) is psi,
Figure DEST_PATH_IMAGE113
in mph.
Therefore, rolling resistance
Figure DEST_PATH_IMAGE095A
The formula of (a) can evolve as:
Figure DEST_PATH_IMAGE115
，
the air resistance formula according to the embodiment of the present application is selected as:
Figure DEST_PATH_IMAGE117
，
wherein, the first and the second end of the pipe are connected with each other,
Figure DEST_PATH_IMAGE119
air density indicating the position of the vehicle, unit:
Figure DEST_PATH_IMAGE121
;
Figure DEST_PATH_IMAGE020AAA
represents the orthographic projection area of the vehicle in the windward direction, unit:
Figure DEST_PATH_IMAGE123
;
Figure DEST_PATH_IMAGE046AAA
representing wind speed, the positive direction of the vector is the same as the vehicle speed, and the unit is:
Figure DEST_PATH_IMAGE125
。C d represents the air resistance coefficient, in units: none. Since the air density varies greatly in real life, the following approximation can be used:
Figure DEST_PATH_IMAGE127
，
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE129
represents the relative humidity of the air, in units: percent;
Figure DEST_PATH_IMAGE131
air temperature in kelvin (K);
Figure DEST_PATH_IMAGE132
is the temperature in degrees Celsius;
Figure DEST_PATH_IMAGE133
is total air pressure, unit:
Figure DEST_PATH_IMAGE135
. The expression of the final air resistance is therefore:
Figure DEST_PATH_IMAGE137
，
by comprehensively considering the force balance among vehicle traction, rolling resistance, air resistance and vehicle gravity, the following longitudinal vehicle force balance equation can be obtained:
Figure DEST_PATH_IMAGE139
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE030AAA
represents the acceleration of the vehicle, in units:
Figure DEST_PATH_IMAGE141
；
Figure DEST_PATH_IMAGE048AAA
is the gravitational acceleration, unit:
Figure DEST_PATH_IMAGE143
. The vehicle weight can thus be found to be:
Figure DEST_PATH_IMAGE145
。
according to an embodiment of the present application, the at least one trained base learner includes a plurality of base learners trained based on different initialization parameters. Under the arrangement, the base learners have certain diversity, and the output errors of different base learners also have certain independence. This is helpful to enhance the robustness of the system.
In an exemplary embodiment of the present application, the at least one trained base learner includes three trained base learners based on different initialization parameters. To achieve full utilization of the training samples, a K-fold cross validation method (e.g., a four-fold cross validation method) may be employed to train the respective base learners.
Table 1 shows a training scheme combination according to an embodiment of the present application, wherein NN 1 、NN 2 、NN 3 Three basis learners built based on Keras are shown separately.
TABLE 1
Figure DEST_PATH_IMAGE147
Figure DEST_PATH_IMAGE149
Figure DEST_PATH_IMAGE151
Each iteration batch _ size: 50 100 200
epoch: 3000 3000 3000
beta_1,beta_2： 0.9,0.999 0.9,0.999 0.9,0.999
learning rate, attenuation rate: 0.001，0.0 0.001，0.0 0.001，0.0
each base learner may include a deep neural network structure including a plurality of convolutional layers, pooling layers, hidden layers, fully-connected layers connected in sequence, wherein: the convolution layer is used for carrying out feature extraction on the input data of the previous layer; the pooling layer is used to down-sample the data, preventing some over-fitting and providing some non-linearity; the hidden layer contains a non-linear activation function, such as a Sigmoid function, to provide non-linearity for the entire base learner; the full connection layer is used for integrating all local features, so that the input information can be comprehensively perceived.
According to the embodiment of the application, the results of the base learner and the longitudinal vehicle force balance equation are fitted secondarily by using the meta learner. The meta learner is based on as inputs the output of the longitudinal vehicle force balance equation and the output of the at least one trained base learner. The meta learner uses a linear regression model of
Figure DEST_PATH_IMAGE002AA
Wherein, in the step (A),
Figure DEST_PATH_IMAGE004AA
、
Figure DEST_PATH_IMAGE152
、
Figure DEST_PATH_IMAGE008AA
is a weight value of the weight value,
Figure DEST_PATH_IMAGE010AA
in order to be an error term, the error term,
Figure DEST_PATH_IMAGE012AA
is the output of the longitudinal vehicle force balance equation,
Figure DEST_PATH_IMAGE014AA
to
Figure DEST_PATH_IMAGE016AA
Is the output of the at least one trained base learner. The risk of overfitting can be effectively reduced by using a simpler linear regression model as a basic model of the meta-learner.
The meta learner according to embodiments of the present application is trained using a gradient descent algorithm, wherein the objective function is
Figure DEST_PATH_IMAGE154
。
The following exemplarily illustrates a process of training an ensemble learning model implementing a vehicle weight prediction method according to an embodiment of the present application with reference to fig. 2. The ensemble learning model described hereinafter includes a base learner and a longitudinal vehicle force balance equation at a first level and a meta learner at a second level. It should be noted that although three base learners are shown in fig. 2 and a four-fold cross-validation training scheme is shown, the present application is not limited thereto. The ensemble learning model according to the embodiment of the present application may include other numbers of base learners, and may employ a training scheme of K-fold cross validation (K is a natural number not less than 2).
First, a sample set with a sufficient number of samples 2100 needs to be collected and collated. Each sample in the sample set includes a vehicle driving state parameter, an environmental parameter, and a vehicle true weight.
Then, dividing the sample set into two parts S1 and S2, wherein the S1 is used for training the whole ensemble learning model and is called a training set S1 hereinafter; and S2 is used for testing the prediction accuracy of the trained ensemble learning model, and is called a test set S2 hereinafter. The data size of S1 is typically larger than the data size of S2.
According to the embodiment of the application, the training of the ensemble learning model is carried out in two steps, firstly, the base learner is trained, and the network parameters of the base learner are solidified; the meta-learner is then trained to consolidate its parameters, e.g.
Figure DEST_PATH_IMAGE004AAA
、
Figure DEST_PATH_IMAGE006AA
、
Figure DEST_PATH_IMAGE156
。
Taking the four-fold cross validation method as an example, when training the base learner, firstly, the training set S1 needs to be divided equally into four equal parts, that is, a first sub-training set D1, a second sub-training set D2, a third sub-training set D3, and a fourth sub-training set D4. Thereafter, the first, second, and third basis learners 2210, 2220, 2230 are trained using the training set S1. According to the embodiment of the present application, although the longitudinal vehicle force balance equation 2240 does not need to be trained, it can be regarded as a base learner after training.
In training the first base learner 2210, the second sub-training set D2, the third sub-training set D3, and the fourth sub-training set D4 are trained as training sets, for example, using an adaptive moment estimation optimization algorithm (ADAM). According to the embodiment of the present application, the training process may be automatically terminated and the network parameters of the first base learner 2210 may be fixed after 3000 cycles of iterative training. Then, using the first sub-training set D1 as a test set, an output D11 is obtained.
In training the first base learner 2210, the first sub-training set D1, the third sub-training set D3, and the fourth sub-training set D4 may also be used as training sets for training, and then the second sub-training set D2 is used as a test set to obtain an output D21.
In training the first base learner 2210, the first sub-training set D1, the second sub-training set D2, and the fourth sub-training set D4 may also be used as training sets for training, and then the third sub-training set D3 may be used as a test set to obtain an output D31.
When training the first base learner 2210, the first sub-training set D1, the second sub-training set D2, and the third sub-training set D3 may also be used as training sets for training, and then the fourth sub-training set D4 may be used as a test set to obtain an output D41.
Thus, after training first base learner 2210, a first set of intermediate outputs 2310 may be obtained, which includes (D) 11 ，D 21 ，D 31 ，D 41 ）。
The training process for the second and third base learners 2220, 2230 is similar to the training process for the first base learner 2210. Second intermediate output set 2320 and third intermediate output set 2330 may be derived after second basis learner 2220 and third basis learner 2230 are trained. Second intermediate output set 2320 contains (D) 12 ，D 22 ，D 32 ，D 42 ) The third intermediate output set comprises (D) 13 ，D 23 ，D 33 ，D 43 ）。
Although the longitudinal vehicle force balance equation 2240 need not be trained, it can be input as samples D1, D2, D3, D4, respectively, and a fourth intermediate output set 2340 is derived, which contains (D) 14 ，D 24 ，D 34 ，D 44 ）。
In training the meta learner, the inputs are first intermediate output set 2310, second intermediate output set 2320, third intermediate output set 2330, fourth intermediate output set 2340 and the true weight of the respective vehicle. Then, the local minimum of the objective function can be solved by a gradient descent method, wherein the objective function is:
Figure DEST_PATH_IMAGE158
。
according to the embodiment of the present application, the maximum number of iterations may be set to 1000. After training is complete, the accuracy of the prediction of the vehicle weight by the ensemble learning model may be tested using test set S2.
The application also provides a vehicle weight prediction system which can be realized in the forms of a mobile terminal, a Personal Computer (PC), a tablet computer, a server and the like. Referring now to FIG. 3, a schematic diagram of a vehicle weight prediction system suitable for use in implementing embodiments of the present application is shown.
As shown in fig. 3, the computer system includes one or more processors, communication section, and the like, for example: one or more Central Processing Units (CPUs) 301, and/or one or more image processors (GPUs) 313, etc., which may perform various appropriate actions and processes according to executable instructions stored in a Read Only Memory (ROM) 302 or loaded from a storage section 308 into a Random Access Memory (RAM) 303. The communication section 312 may include, but is not limited to, a network card, which may include, but is not limited to, an IB (Infiniband) network card.
The processor may communicate with the read-only memory 302 and/or the random access memory 303 to execute the executable instructions, connect with the communication part 312 through the bus 304, and communicate with other target devices through the communication part 312, so as to complete the operations corresponding to any one of the methods proposed in the embodiments of the present application, for example: predicting to obtain a first vehicle weight based on the vehicle running state parameters and the environmental parameters through a longitudinal vehicle force balance equation; predicting, by at least one trained base learner, a second vehicle weight based at least on non-constant ones of the vehicle travel state parameters and the environmental parameters; a meta learner, completed by training, predicts a vehicle weight based on the first vehicle weight and the second vehicle weight.
In addition, in the RAM 303, various programs and data necessary for the operation of the apparatus can also be stored. The CPU 301, ROM 302, and RAM 303 are connected to each other via a bus 304. The ROM 302 is an optional module in the presence of the RAM 303. The RAM 303 stores or writes executable instructions into the ROM 302 at runtime, and the executable instructions cause the processor 301 to perform operations corresponding to the above-described communication method. An input/output interface (I/O interface) 305 is also connected to the bus 304. The communication unit 312 may be integrated, or may be provided with a plurality of sub-modules (e.g., a plurality of IB network cards) and connected to the bus link.
The following components are connected to the I/O interface 305: an input portion 306 including a keyboard, a mouse, and the like; an output section 307 including a display such as a Cathode Ray Tube (CRT), a Liquid Crystal Display (LCD), and the like, and a speaker; a storage unit 308 including a hard disk and the like; and a communication section 309 including a network interface card such as a LAN card, a modem, or the like. The communication section 309 performs communication processing via a network such as the internet. A drive 310 is also connected to the I/O interface 305 as needed. A removable medium 311 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is mounted on the drive 310 as necessary.
It should be noted that the architecture shown in fig. 3 is only an optional implementation manner, and in a specific practical process, the number and types of the components in fig. 3 may be selected, deleted, added, or replaced according to actual needs; in different functional component settings, separate settings or integrated settings may also be used, for example, the GPU and the CPU may be separately set or the GPU may be integrated on the CPU, the communication unit 312 may be separately set or integrated on the CPU or the GPU, and so on. These alternative embodiments are all within the scope of the present disclosure.
In particular, the process described with reference to the flowchart of fig. 1 may be implemented as a computer program product according to the present application. For example, the present application proposes a computer program product comprising computer readable instructions which, when executed by a processor, implement the following: predicting to obtain a first vehicle weight based on vehicle running state parameters and environment parameters through a longitudinal vehicle force balance equation; predicting, by at least one trained base learner, a second vehicle weight based at least on non-constant ones of the vehicle travel state parameters and the environmental parameters; predicting, by a trained meta learner, a vehicle weight based on the first vehicle weight and the second vehicle weight.
In such embodiments, the computer program product may be downloaded and installed from a network via the communication section 309, and/or read and installed from the removable media 311. The above-described functions as defined in the method of the present application are performed when the computer program product is executed by a Central Processing Unit (CPU) 301.
The solution of the present application may be implemented in many ways. For example, the technical solutions of the present application may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The order of the steps used to describe the method is provided for clarity of description of the embodiments only. Unless specifically limited, the method steps of the present application are not limited to the order specifically described above. Furthermore, in some embodiments, the present application may also be implemented as a storage medium storing a computer program product.
The above description is only an embodiment of the present application and an illustration of the principles of the technology employed. It will be appreciated by a person skilled in the art that the scope of protection covered by the present application is not limited to the embodiments with a specific combination of the features described above, but also covers other embodiments with any combination of the features described above or their equivalents without departing from the technical idea described above. For example, the above features may be replaced with (but not limited to) features having similar functions disclosed in the present application.

Claims (9)

1. A vehicle weight prediction method based on ensemble learning, characterized by comprising:
predicting to obtain a first vehicle weight based on vehicle running state parameters and environment parameters through a longitudinal vehicle force balance equation;
predicting, by at least one trained base learner, a second vehicle weight based at least on non-constant ones of the vehicle travel state parameters and the environmental parameters;
predicting a vehicle weight based on the first vehicle weight and the second vehicle weight by a trained meta learner,
wherein the meta learner performs training using a gradient descent algorithm based on the output of the longitudinal vehicle force balance equation and the output of the at least one trained basis learner as inputs, and the meta learner employs a linear regression model of
Figure 330702DEST_PATH_IMAGE002
，
Wherein the content of the first and second substances,
Figure 8546DEST_PATH_IMAGE004
、
Figure 752511DEST_PATH_IMAGE006
、
Figure 333665DEST_PATH_IMAGE008
is a weight value of the weight value,
Figure 606515DEST_PATH_IMAGE010
in order to be an error term, the error term,
Figure 69857DEST_PATH_IMAGE012
is the output of the longitudinal vehicle force balance equation,
Figure 319310DEST_PATH_IMAGE014
to
Figure 551709DEST_PATH_IMAGE016
Is the output of the at least one trained base learner,
the objective function is:
Figure 995459DEST_PATH_IMAGE018
。
2. the vehicle weight prediction method according to claim 1,
the vehicle running state parameters include: frontal projection area of vehicle in windward direction
Figure 883781DEST_PATH_IMAGE020
Speed of vehicle
Figure 500707DEST_PATH_IMAGE022
Rotational speed of engine
Figure 283550DEST_PATH_IMAGE024
Torque of engine
Figure 898202DEST_PATH_IMAGE026
Mechanical efficiency of vehicle drive train
Figure 70557DEST_PATH_IMAGE028
Acceleration of vehicle
Figure 163278DEST_PATH_IMAGE030
Pressure in the tyre
Figure 370269DEST_PATH_IMAGE032
Angle between vehicle and horizontal plane
Figure 919937DEST_PATH_IMAGE034
；
The environmental parameters include: air pressure
Figure 517271DEST_PATH_IMAGE036
Relative humidity of air
Figure 476000DEST_PATH_IMAGE038
Air temperature in Kelvin
Figure 6338DEST_PATH_IMAGE040
Temperature in centigrade
Figure 228372DEST_PATH_IMAGE042
Coefficient of air resistance
Figure 77117DEST_PATH_IMAGE044
Velocity of wind
Figure 573958DEST_PATH_IMAGE046
Acceleration of gravity
Figure 693223DEST_PATH_IMAGE048
。
3. The vehicle weight prediction method according to claim 2,
the longitudinal vehicle force balance equation is:
Figure 555000DEST_PATH_IMAGE050
，
wherein m is the first vehicle weight,
coefficient of rolling resistance
Figure 891041DEST_PATH_IMAGE052
The expression of (a) is:
Figure 394835DEST_PATH_IMAGE054
air density of the location of the vehicle
Figure 430924DEST_PATH_IMAGE056
The expression of (a) is:
Figure 932444DEST_PATH_IMAGE058
。
4. the vehicle weight prediction method of claim 1, wherein the at least one trained base learner comprises a plurality of base learners trained based on different initialization parameters.
5. The vehicle weight prediction method according to claim 4, wherein each of the plurality of base learners includes at least one hidden layer, and the non-linear activation function employed by the hidden layer is a Sigmoid function.
6. The vehicle weight prediction method according to claim 5, wherein each of the plurality of base learners is trained using a K-fold cross validation method, where K is a natural number not less than 2.
7. The vehicle weight prediction method according to claim 6, characterized in that the at least one trained base learner comprises three trained base learners based on different initialization parameters, and each base learner is trained using a four-fold cross-validation method.
8. A vehicle weight prediction system based on ensemble learning, the vehicle weight prediction system comprising:
a memory storing executable instructions; and
one or more processors in communication with the memory to execute the executable instructions to:
predicting to obtain a first vehicle weight based on the vehicle running state parameters and the environmental parameters through a longitudinal vehicle force balance equation;
predicting, by at least one trained base learner, a second vehicle weight based at least on non-constant ones of the vehicle travel state parameters and the environmental parameters;
predicting a vehicle weight based on the first vehicle weight and the second vehicle weight by a trained meta learner,
wherein the meta learner performs training using a gradient descent algorithm based on the output of the longitudinal vehicle force balance equation and the output of the at least one trained basis learner as inputs, and the meta learner employs a linear regression model of
Figure 761640DEST_PATH_IMAGE059
，
Wherein the content of the first and second substances,
Figure 69125DEST_PATH_IMAGE004
、
Figure 694141DEST_PATH_IMAGE006
、
Figure 163300DEST_PATH_IMAGE008
is a weight value of the weight value,
Figure 709819DEST_PATH_IMAGE010
in order to be an error term, the error term,
Figure 116267DEST_PATH_IMAGE012
is the output of the longitudinal vehicle force balance equation,
Figure 64632DEST_PATH_IMAGE014
to is that
Figure 704692DEST_PATH_IMAGE016
Is the output of the at least one trained base learner,
the objective function is:
Figure 535244DEST_PATH_IMAGE018
。
9. a computer-readable storage medium for vehicle weight prediction, the computer-readable storage medium storing executable instructions that are executable by one or more processors to:
predicting to obtain a first vehicle weight based on vehicle running state parameters and environment parameters through a longitudinal vehicle force balance equation;
predicting, by at least one trained base learner, a second vehicle weight based at least on non-constant ones of the vehicle travel state parameters and the environmental parameters;
predicting a vehicle weight based on the first vehicle weight and the second vehicle weight by a trained meta learner,
wherein the meta learner performs training using a gradient descent algorithm based on the output of the longitudinal vehicle force balance equation and the output of the at least one trained basis learner as inputs, and the meta learner employs a linear regression model of
Figure 184531DEST_PATH_IMAGE059
，
Wherein the content of the first and second substances,
Figure 220358DEST_PATH_IMAGE004
、
Figure 93636DEST_PATH_IMAGE006
、
Figure 614748DEST_PATH_IMAGE008
is a weight value of the weight value,
Figure 536567DEST_PATH_IMAGE010
in order to be an error term, the error term,
Figure 958059DEST_PATH_IMAGE012
is the output of the longitudinal vehicle force balance equation,
Figure 2238DEST_PATH_IMAGE014
to
Figure 745066DEST_PATH_IMAGE016
Is the output of the at least one trained base learner,
the objective function is:
Figure 736156DEST_PATH_IMAGE060
。
CN202211109476.XA 2022-09-13 2022-09-13 Vehicle weight prediction method and system based on ensemble learning and storage medium Active CN115204417B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211109476.XA CN115204417B (en) 2022-09-13 2022-09-13 Vehicle weight prediction method and system based on ensemble learning and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211109476.XA CN115204417B (en) 2022-09-13 2022-09-13 Vehicle weight prediction method and system based on ensemble learning and storage medium

Publications (2)

Publication Number Publication Date
CN115204417A CN115204417A (en) 2022-10-18
CN115204417B true CN115204417B (en) 2022-12-27

Family

ID=83572329

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211109476.XA Active CN115204417B (en) 2022-09-13 2022-09-13 Vehicle weight prediction method and system based on ensemble learning and storage medium

Country Status (1)

Country Link
CN (1) CN115204417B (en)

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109118787A (en) * 2018-08-20 2019-01-01 浙江工业大学 A kind of car speed prediction technique based on deep neural network
WO2019199561A1 (en) * 2018-04-09 2019-10-17 Cambridge Mobile Telematics Inc. Vehicle classification based on telematics data
CN111047085A (en) * 2019-12-06 2020-04-21 北京理工大学 Hybrid vehicle working condition prediction method based on meta-learning
CN111121939A (en) * 2020-01-02 2020-05-08 深圳市汉德网络科技有限公司 High-precision vehicle-mounted area weighing method
CN111707343A (en) * 2020-06-23 2020-09-25 北京经纬恒润科技有限公司 Method and device for determining weight of vehicle
CN112258251A (en) * 2020-11-18 2021-01-22 北京理工大学 Grey correlation-based integrated learning prediction method and system for electric vehicle battery replacement demand
CN115009288A (en) * 2022-07-19 2022-09-06 潍柴动力股份有限公司 Method and device for determining vehicle weight

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7783450B2 (en) * 2007-11-14 2010-08-24 Ut-Battelle, Llc Method and system for reducing errors in vehicle weighing systems
CN113386781A (en) * 2021-05-24 2021-09-14 江苏大学 Intelligent vehicle trajectory tracking control method based on data-driven vehicle dynamics model

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2019199561A1 (en) * 2018-04-09 2019-10-17 Cambridge Mobile Telematics Inc. Vehicle classification based on telematics data
CN109118787A (en) * 2018-08-20 2019-01-01 浙江工业大学 A kind of car speed prediction technique based on deep neural network
CN111047085A (en) * 2019-12-06 2020-04-21 北京理工大学 Hybrid vehicle working condition prediction method based on meta-learning
CN111121939A (en) * 2020-01-02 2020-05-08 深圳市汉德网络科技有限公司 High-precision vehicle-mounted area weighing method
CN111707343A (en) * 2020-06-23 2020-09-25 北京经纬恒润科技有限公司 Method and device for determining weight of vehicle
CN112258251A (en) * 2020-11-18 2021-01-22 北京理工大学 Grey correlation-based integrated learning prediction method and system for electric vehicle battery replacement demand
CN115009288A (en) * 2022-07-19 2022-09-06 潍柴动力股份有限公司 Method and device for determining vehicle weight

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
Hierarchical reinforcement learning based energy management strategy for hybrid electric vehicle;ChunyangQi等;《Energy》;20220101;第238卷;1-10 *
高海拔环境下轻型车滑行阻力建模及实验研究;陈伟;《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》;20200315(第03期);C035-499 *
高精度高速公路车辆动态称重系统的研究;王丽丽;《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》;20111215(第S2期);C030-23 *

Also Published As

Publication number Publication date
CN115204417A (en) 2022-10-18

Similar Documents

Publication Publication Date Title
Li et al. Intelligent fault diagnosis of rolling bearings under imbalanced data conditions using attention-based deep learning method
CN106095963B (en) Vehicle driving behavior analysis big data public service platform under internet + era
CN110657984B (en) Planetary gearbox fault diagnosis method based on reinforced capsule network
CN111275660B (en) Flat panel display defect detection method and device
CN110378346A (en) Establish the method, apparatus, equipment and computer storage medium of Text region model
CN110929839A (en) Method and apparatus for training neural network, electronic device, and computer storage medium
Jiang et al. A novel fusion algorithm for estimation of the side-slip angle and the roll angle of a vehicle with optimized key parameters
CN111047078A (en) Traffic characteristic prediction method, system and storage medium
CN113139594A (en) Airborne image unmanned aerial vehicle target self-adaptive detection method
CN110988263B (en) Vehicle exhaust concentration estimation method based on improved Stacking model
CN115204417B (en) Vehicle weight prediction method and system based on ensemble learning and storage medium
Dong et al. A fast sparsity-free compressive sensing approach for vibration data reconstruction using deep convolutional GAN
CN114722520A (en) Planetary bearing intelligent fault diagnosis method based on enhanced dictionary learning
JP2021526696A (en) Systems and methods for processing data from online on-demand service platforms
Wang et al. Uncertainty-controlled remaining useful life prediction of bearings with a new data-augmentation strategy
Du et al. Pavement friction evaluation based on vehicle dynamics and vision data using a multi-feature fusion network
Yang et al. Simultaneous estimation of vehicle mass and unknown road roughness based on adaptive extended Kalman filtering of suspension systems
Sun et al. Fault diagnosis of train traction motor bearing based on improved deep residual network
Fraikin et al. A fast and accurate hybrid simulation model for the large-scale testing of automated driving functions
Li et al. Efficient method based on recurrent neural networks for pavement evenness detection
US11525728B1 (en) Systems and methods for determining an estimated weight of a vehicle
CN116242631A (en) Method and device for measuring running resistance of whole vehicle, computer equipment, medium and product
Moghadam et al. Application of smartphones in pavement deterioration identification using artificial neural network
CN114997214A (en) Fault diagnosis method and device for residual error intensive network
TW201913528A (en) Method and processing device for driving risk assessment based on multiple kernel learning

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
GR01 Patent grant
GR01 Patent grant