CN113378479A - Intelligent standard method and system based on automatic driving test intelligent platform vehicle - Google Patents
Intelligent standard method and system based on automatic driving test intelligent platform vehicle Download PDFInfo
- Publication number
- CN113378479A CN113378479A CN202110744950.5A CN202110744950A CN113378479A CN 113378479 A CN113378479 A CN 113378479A CN 202110744950 A CN202110744950 A CN 202110744950A CN 113378479 A CN113378479 A CN 113378479A
- Authority
- CN
- China
- Prior art keywords
- intelligent
- module
- automatic driving
- index
- vehicle
- 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
Links
- 238000012360 testing method Methods 0.000 title claims abstract description 55
- 238000010561 standard procedure Methods 0.000 title claims abstract description 13
- 238000011156 evaluation Methods 0.000 claims abstract description 66
- 238000012549 training Methods 0.000 claims abstract description 30
- 238000013528 artificial neural network Methods 0.000 claims abstract description 23
- 210000002569 neuron Anatomy 0.000 claims abstract description 23
- 238000012545 processing Methods 0.000 claims abstract description 19
- 238000010606 normalization Methods 0.000 claims abstract description 8
- 230000006870 function Effects 0.000 claims description 26
- 238000000034 method Methods 0.000 claims description 26
- 230000004913 activation Effects 0.000 claims description 8
- 238000013507 mapping Methods 0.000 claims description 6
- 238000012937 correction Methods 0.000 claims description 3
- 238000005516 engineering process Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 238000011161 development Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000012854 evaluation process Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000011158 quantitative evaluation Methods 0.000 description 2
- 238000012827 research and development Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 241000408659 Darpa Species 0.000 description 1
- 238000012098 association analyses Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 210000002364 input neuron Anatomy 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 210000004205 output neuron Anatomy 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000004445 quantitative analysis Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 238000011425 standardization method Methods 0.000 description 1
- 238000010998 test method Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- 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
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Entrepreneurship & Innovation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Strategic Management (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- Educational Administration (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Development Economics (AREA)
- Geometry (AREA)
- Computer Hardware Design (AREA)
- Medical Informatics (AREA)
- Game Theory and Decision Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention provides an intelligent standard method based on an automatic driving test intelligent platform truck, which comprises the following steps: step 1: establishing an automobile intelligent index system of the automatic driving test intelligent platform vehicle, and analyzing related evaluation indexes of the automatic driving test intelligent platform vehicle; step 2: determining a network structure of a BP neural network according to an automobile intelligent standard of an automatic driving test intelligent platform vehicle; and step 3: determining the number of hidden neurons; and 4, step 4: carrying out normalization dimensionless processing on the data; and 5: determining the intelligent level of the automatic driving intelligent platform vehicle; step 6: and carrying out BP network training and determining the weight range of the sample index. The intelligent evaluation index is fitted based on the BP neural network, and the true value of the intelligent index value can be more approximated.
Description
Technical Field
The invention relates to the technical field of intelligent driving tests, in particular to an intelligent standard method and system based on an automatic driving test intelligent platform vehicle.
Background
The automatic driving vehicle is an intelligent vehicle, also called as a wheel type automatic robot, and mainly senses the road environment by a vehicle-mounted sensing system, automatically plans a driving route and controls the vehicle to reach a preset target.
To facilitate the development of the unmanned vehicle technology, a test evaluation technology for the unmanned vehicle is also beginning to emerge. At present, most evaluation tests are carried out by each research and development unit according to the condition of each research and development unit to carry out single index evaluation tests on single or partial functional requirements. With the maturity of single technologies and integrated systems, evaluation test single evaluation develops towards complex comprehensive capability evaluation, and evaluation of a third party gradually appears. DARPA in the united states organizes three unmanned vehicle games, the Grand Challenge in 2004, 2005 and the Urban Challenge in 2007 to complete all the prescribed items and the length of elapsed time as an evaluation test index.
In the prior art, the evaluation test method of the unmanned vehicle has great artificial subjectivity and tendency, and the evaluation result is inaccurate. The intelligent level grading of the unmanned vehicle is not carried out, the evaluation result lacks scientificity and accuracy, and the influence of factors such as the unmanned vehicle, the driving environment and human intervention is not considered in the evaluation process. Although the merits of the unmanned vehicle performance can be evaluated, the shortcomings of specific indexes and the direction of improvement in future are not pointed out, which is not favorable for the development of the unmanned vehicle technology.
Therefore, it is necessary to implement an intelligent standard of an autonomous vehicle, which can scientifically and accurately perform quantitative analysis on the unmanned vehicle, so as to implement quantitative evaluation test of the unmanned vehicle. The evaluation process not only considers the influence of various factors, retains all information of each level of evaluation, but also can better reflect the actual situation by the quantitative result and can be conveniently converted into a visual comparison or sequencing result. Finally, the defects of various indexes of the automatic driving vehicle can be found out, the direction needing to be improved later is indicated, and the development of the automatic driving vehicle technology is better guided.
Disclosure of Invention
Aiming at the defects in the prior art, the invention aims to provide an intelligent standard method and system based on an automatic driving test intelligent platform truck.
The invention provides an intelligent standard method based on an automatic driving test intelligent platform truck, which comprises the following steps:
step 1: establishing an automobile intelligent index system of the automatic driving test intelligent platform vehicle, and analyzing related evaluation indexes of the automatic driving test intelligent platform vehicle;
step 2: determining a network structure of a BP neural network according to an automobile intelligent standard of an automatic driving test intelligent platform vehicle;
and step 3: determining the number of hidden neurons;
and 4, step 4: carrying out normalization dimensionless processing on the data;
and 5: determining the intelligent level of the automatic driving intelligent platform vehicle;
step 6: and carrying out BP network training and determining the weight range of the sample index.
Preferably, the step 1 comprises:
step 1.1: selecting indexes related to intelligent evaluation of the automobile as comparison series, taking the intelligent indexes of the intelligent platform car as reference series, carrying out non-dimensionalization processing on variables, and then calculating a correlation coefficient;
note deltai(k)＝|y(k)-xi(k) I, then
ρ ∈ (0, ∞), which is called the resolution coefficient, and the smaller ρ, the larger the resolution, and the value of ρ is (0, 1).
Step 1.2: after the correlation coefficient is obtained, calculating the average value of the correlation coefficient as the number representation of the correlation degree between the comparison number series and the reference number series; the degree of association ri is formulated as follows:
step 1.3: and sorting according to the relevance and the size, and selecting the index with high relevance as a final evaluation index.
Preferably, the step 2 includes:
step 2.1: determining the number of nodes of an input layer and an output layer; and determining the number of neurons of the input layer according to the number of the selected detailed test function indexes, wherein the number of the neurons of the output layer is one, and the intelligent index value of the intelligent vehicle is output.
Step 2.2: determining the number of layers of the BP neural network, fitting a function containing continuous mapping from one finite space to another finite space by a single hidden layer, and expressing any decision boundary of any progress by two layers of activation functions.
Preferably, the step 3 comprises:
the number of the implicit neurons is determined by mainly adopting an empirical formula:
in the formula: q is the number of hidden layer nodes; a is the number of nodes of an input layer; b is the number of output layer nodes; c is any integer of [ a,10 ].
When the overfitting condition occurs, the correction is carried out by adopting early-stopping or adding a regularization term.
Preferably, the step 5 comprises:
the intelligent index value of the intelligent vehicle is reasonably divided into regions, and data with special conditions in the test data of the intelligent vehicle is subjected to other processing.
Preferably, the step 6 comprises:
step 6.1: creating a network object, and setting an activation function, a training function, a learning function and a performance function of each layer;
step 6.2: adjusting the weight and threshold of the network;
step 6.3: setting training parameters of the network; forming a training sample by the normalized data and the target value, and repeatedly training the established BP network;
step 6.4: obtaining an initial weight bi of the normalized evaluation index value through an entropy weight method, scoring the set intelligent index through k experts and calculating the weight;
step 6.5: calculating according to the weights calculated by an expert evaluation method and an entropy weight method to obtain a comprehensive weight;
step 6.6: after calculating the intelligent index value of the intelligent vehicle through the comprehensive weight, comparing the ideal intelligent index value output by the BP network;
step 6.7: and adjusting the comprehensive weight in the weight range obtained by an expert evaluation method and an entropy weight method until the intelligent index value error of the automobile reaches an ideal value.
The invention also provides an intelligent standard system based on the automatic driving test intelligent platform vehicle, which comprises the following modules:
module M1: establishing an automobile intelligent index system of the automatic driving test intelligent platform vehicle, and analyzing related evaluation indexes of the automatic driving test intelligent platform vehicle;
module M2: determining a network structure of a BP neural network according to an automobile intelligent standard of an automatic driving test intelligent platform vehicle;
module M3: determining the number of hidden neurons;
module M4: carrying out normalization dimensionless processing on the data;
module M5: determining the intelligent level of the automatic driving intelligent platform vehicle;
module M6: and carrying out BP network training and determining the weight range of the sample index.
Preferably, said module M1 comprises:
module M1.1: selecting indexes related to intelligent evaluation of the automobile as comparison series, taking the intelligent indexes of the intelligent platform car as reference series, carrying out non-dimensionalization processing on variables, and then calculating a correlation coefficient;
note deltai(k)＝|y(k)-xi(k) I, then
ρ ∈ (0, ∞), which is called the resolution coefficient, and the smaller ρ, the larger the resolution, and the value of ρ is (0, 1).
Module M1.2: after the correlation coefficient is obtained, calculating the average value of the correlation coefficient as the number representation of the correlation degree between the comparison number series and the reference number series; the degree of association ri is formulated as follows:
module M1.3: sorting according to the relevance and the size, and selecting an index with high relevance as a final evaluation index;
the module M2 includes:
module M2.1: determining the number of nodes of an input layer and an output layer; and determining the number of neurons of the input layer according to the number of the selected detailed test function indexes, wherein the number of the neurons of the output layer is one, and the intelligent index value of the intelligent vehicle is output.
Module M2.2: determining the number of layers of the BP neural network, fitting a function containing continuous mapping from one finite space to another finite space by a single hidden layer, and expressing any decision boundary of any progress by two layers of activation functions.
Preferably, said module M3 comprises:
the number of the implicit neurons is determined by mainly adopting an empirical formula:
in the formula: q is the number of hidden layer nodes; a is the number of nodes of an input layer; b is the number of output layer nodes; c is any integer of [ a,10 ].
When the overfitting condition occurs, the correction is carried out by adopting early-stopping or adding a regularization term.
The module M5 includes:
the intelligent index value of the intelligent vehicle is reasonably divided into regions, and data with special conditions in the test data of the intelligent vehicle is subjected to other processing.
Preferably, said module M6 comprises:
module M6.1: creating a network object, and setting an activation function, a training function, a learning function and a performance function of each layer;
module M6.2: adjusting the weight and threshold of the network;
module M6.3: setting training parameters of the network; forming a training sample by the normalized data and the target value, and repeatedly training the established BP network;
module M6.4: obtaining an initial weight bi of the normalized evaluation index value through an entropy weight method, scoring the set intelligent index through k experts and calculating the weight;
module M6.5: calculating according to the weights calculated by an expert evaluation method and an entropy weight method to obtain a comprehensive weight;
module M6.6: after calculating the intelligent index value of the intelligent vehicle through the comprehensive weight, comparing the ideal intelligent index value output by the BP network;
module M6.7: and adjusting the comprehensive weight in the weight range obtained by an expert evaluation method and an entropy weight method until the intelligent index value error of the automobile reaches an ideal value.
Compared with the prior art, the invention has the following beneficial effects:
1. the intelligent evaluation index is fitted based on the BP neural network, and the true value of the intelligent index value can be more approximated.
2. The invention combines the entropy weight method and the expert evaluation method, which can reflect the real situation of the evaluation index and has excellent adaptability.
3. The existing bp neural network is used as a basis, big data are used for training, and the result can be widely applied to different platform intelligent vehicles and has strong expandability.
4. The evaluation indexes of the invention can be continuously expanded by a plurality of indexes, and the invention has transverse expandability. And the data is normalized, the numerical value of the index has no practical meaning, the network can continue to train and change through new data, and the longitudinal direction can continue to deepen.
Drawings
Other features, objects and advantages of the invention will become more apparent upon reading of the detailed description of non-limiting embodiments with reference to the following drawings:
FIG. 1 is a schematic diagram of a BP neural network training process;
FIG. 2 is a schematic diagram of a BP neural network model;
FIG. 3 illustrates an intelligent indicator of the intelligent platform vehicle;
fig. 4 is an evaluation test flowchart of the intelligent platform vehicle.
Detailed Description
The present invention will be described in detail with reference to specific examples. The following examples will assist those skilled in the art in further understanding the invention, but are not intended to limit the invention in any way. It should be noted that it would be obvious to those skilled in the art that various changes and modifications can be made without departing from the spirit of the invention. All falling within the scope of the present invention.
Referring to fig. 1 and 2, the present invention provides an intelligent standard method and system based on an automatic driving test intelligent platform vehicle, comprising the following steps:
step 1: establishing an automobile intelligent index system of the automatic driving test intelligent platform vehicle; determining a required evaluation criterion and a corresponding detailed evaluation index according to the characteristics of the intelligent automobile or the requirements of a test environment; and selecting an index which has a large influence on the intelligent platform truck as a final evaluation index according to the index association degree obtained by the grey association analysis model.
Step 2: determining a network structure of the BP neural network; according to the automobile intelligent standard of the automatic driving test intelligent platform vehicle, n layers of BP networks can be determined to be used, and the BP networks comprise 1 layer of input layer, n-2 layers of hidden layers and 1 layer of output layer, wherein the number of input neurons is the number of the automobile intelligent standard, and the number of output neurons is 1.
And step 3: determining the number of hidden neurons; the number of the hidden layer neurons has direct relation to the learning time, the complexity and the output progress of the bp neural network.
And 4, step 4: carrying out data dimensionless processing; carrying out normalization processing on the sample; after the evaluation index is determined, preprocessing the index data by a normalization method, and processing the data acquired in the experiment to be used as sample data.
And 5: referring to fig. 3, determining an intelligentization level of the intelligent platform vehicle; after the specific numerical value of the intelligent index of the intelligent vehicle is obtained, the evaluation value interval of the intelligent vehicle can be divided, and the intelligent grade and the corresponding score interval of the intelligent vehicle are set.
Step 6: BP network training, inputting corresponding learning samples to train the network after setting training parameters; and when the output error meets the requirement, the BP neural network training result of intelligent vehicle index evaluation is reliable.
And 7: determining the weight of each index of the sample; the method combines an entropy weight method and an expert evaluation method to assign the weight of the automobile intelligent index of the automatic driving test intelligent platform vehicle, and adjusts each parameter by taking the intelligent index value of the intelligent vehicle as ideal output.
Selecting evaluation indexes of the automatic driving intelligent platform vehicle, including signal lamp identification, GPS navigation performance, intelligent system delay and the like, and screening by calculating the correlation degree between each index and the intelligent level; through strong nonlinear mapping capability, self-learning capability, self-adaption capability and good generalization capability of the BP neural network, the reasonable solving rule can be continuously approached through learning of the standard sample, so that the evaluation indexes can be automatically calculated through the BP neural network to obtain reasonable intelligent values, and quantitative evaluation of the automatic driving intelligent vehicle is realized; the weight range of each index in the sample is determined by combining each algorithm, so that an acceptable weight value of each index is obtained.
Fig. 4 is a flowchart of the test evaluation of the intelligent flatcar according to the present invention, which is described in detail below with reference to fig. 4. The automobile intelligent index based on the automatic driving test intelligent platform vehicle mainly comprises the following steps:
various intelligent evaluation indexes capable of reflecting the intelligent platform truck are selected, and the indexes are reflected through a plurality of testable tasks. Because too many complex unquantized and predictable indexes exist in the real traffic environment of the intelligent platform vehicle, and simultaneously, a plurality of indexes detected in multiple aspects of the intelligent vehicle exist, such as the adaptability of the road environment (expressway, rural road, field mountain road and the like), the avoidance of obstacles, the identification of traffic signs, lane keeping, environmental perception and other factors, the embodiment comprehensively and reasonably selects a plurality of evaluation indexes to perform experiments, obtains the intelligent value of each group of data through expert evaluation, and determines the correlation degree of each index through a grey correlation analysis model.
Setting the intelligentized value of the sample as a reference number sequence:
Y＝Y(k)|k＝1，2，...n；
taking the evaluation index selected from the sample as a comparison sequence
Xi＝Xi(k)|k＝1，2...n，i＝1，2...m。
Because the data of each index may not be accurately compared due to different dimensions when the evaluation index is selected, the data can be initialized:
wherein k corresponds to a time period and i corresponds to a row in the comparison sequence;
the index data may be initialized and then the correlation coefficient may be calculated:
note deltai(k)＝|y(k)-xi(k) I, then
ρ ∈ (0, ∞), which is called the resolution coefficient, and the smaller ρ, the larger the resolution, and the value of ρ is (0, 1).
Then we can get the correlation degree through the correlation coefficient, and sort the index according to the correlation degree.
Table 1 shows an exemplary automobile intelligent index association table of an intelligent platform vehicle
TABLE 1 automobile intelligent index association table
By using this table, several evaluation indexes with the highest degree of correlation can be selected as the final evaluation index.
And starting to construct the BP neural network after the evaluation index is selected.
In this embodiment, three neural networks are mainly used, which correspond to an input layer, a hidden layer, and an output layer, respectively. The number of the input layer neurons is 9, and the number of the output layer neurons is 1.
It should be noted that when determining the number of hidden layer neurons, we can determine that the number of hidden layer nodes is an integer in [4,13] by selecting according to the following formula by mainly using an empirical formula,
in the formula: q is the number of hidden layer nodes; a is the number of nodes of an input layer; b is the number of output layer nodes; c is any integer of [ a,10 ].
Thus, two BP neural networks of 9-4-1 and 9-7-1 are established respectively.
The data is subjected to non-dimensionalization processing, so that the influence of numerical value difference on the data analysis result caused by different dimensions of all indexes can be eliminated, the data is subjected to normalization processing to be limited within a certain range, and meanwhile, the network can be rapidly converged and neuron saturation can be avoided.
Specifically, the calculation process of the non-dimensionalization processing of the data is as follows:
obtaining data in a sample, linearly changing original data by adopting a maximum-minimum standardization method, setting minA and maxA as the maximum value and the minimum value of the attribute A respectively, and mapping an original value x of A to a value x' of an interval [0,1] through maximum-minimum standardization, wherein the formula is as follows:
after the sample data is normalized, the score interval corresponding to the first level of the automobile intelligent level can be roughly judged according to the sample data.
| Grade | Grade 5 | 4 stage | Grade 3 | Stage 2 | Level 1 |
| Intelligent level | Height of | Is higher than | Medium and high grade | Is lower than | Is low in |
| Corresponding score | [0.8,1] | [0.6,0.8] | [0.4,0.6] | [0.2,0.4] | [0,0.2] |
TABLE 2 Intelligent grade and score interval for automobile
After the intelligent level is determined, a certain training parameter is set, and the intelligent index data of the automobile is trained, wherein the training process is shown in fig. 1. The sample data are divided into two groups, 80 are taken for network training, the other 20 are used for verification, and the dispersity of the verification samples is particularly and fully noticed when the samples are grouped. And when the operation result meets the condition, the network training is finished, and at the moment, the network model can be shaped and used for judging and identifying the evaluation object.
Since there may not be a definite linear correlation between the input and the output of the BP neural network, we can approximate the final intelligent value as much as possible by determining the weight interval when determining the weight of each index.
And (4) obtaining the initial weight bi by the normalized evaluation index value through an entropy weight method.
Specifically, the calculation process of the intelligent index by the entropy weight method is as follows:
firstly, determining the entropy value of the j index:
wherein k is 1/ln (n) > 0, satisfies ej≥0；
Wherein pij is the proportion of the ith sample value in the j index:
then, the information entropy redundancy is calculated:
dj＝1-ej，j＝1，…，m
and finally, calculating the weight of each index, wherein the weight is only distributed for the weight under an ideal state:
and scoring the set intelligent indexes by k experts and calculating the weight, wherein the parameters of the intelligent indexes are set as the scoring interval of the ith expert in [ u1k, u2k ], and ui belongs to [0,1].
To eliminate the evaluation difference between each expert individual, we can calculate the weight average by the expert score of each index as:
in the formula
Objectivity representing a safety risk index; k represents the total of experts
Number of
Represents the lowest score given by the h expert;
represents the h-th expert
Highest score given
Further, the comprehensive weight is calculated according to the weights calculated by an expert evaluation method and an entropy weight method.
And comparing the intelligent value of the intelligent vehicle calculated by the comprehensive weight with the ideal intelligent value output by the BP network. And adjusting the comprehensive weight in the weight range obtained by an expert evaluation method and an entropy weight method until the error of the intelligent value of the automobile reaches an ideal value. The weights are shown in table 3.
| Serial number | Automobile intelligent index | Weighted value |
| 1 | Traffic signal light identification | 0.22 |
| 2 | Automatic pedestrian avoidance | 0.19 |
| 3 | Emergency brake | 0.17 |
| 4 | GPS navigation performance | 0.06 |
| 5 | Vehicle distance keeping | 0.11 |
| 6 | Intelligent system delay | 0.08 |
| 7 | Straight lane keeping | 0.12 |
| 8 | Parking vehicle | 0.03 |
| 9 | Speed of normal running | 0.02 |
TABLE 3 Intelligent indicator weight table for intelligent vehicle
By the weight, the influence degree of the automobile intelligent index on the automobile intelligence can be known, and meanwhile, the algorithm process can be simplified, so that the approximate value of the automobile intelligent value can be obtained directly through a linear equation.
The intelligent degree of the intelligent platform vehicle can be analyzed through the automobile intelligent value output by the BP neural network.
Those skilled in the art will appreciate that, in addition to implementing the system and its various devices, modules, units provided by the present invention as pure computer readable program code, the system and its various devices, modules, units provided by the present invention can be fully implemented by logically programming method steps in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers and the like. Therefore, the system and various devices, modules and units thereof provided by the invention can be regarded as a hardware component, and the devices, modules and units included in the system for realizing various functions can also be regarded as structures in the hardware component; means, modules, units for performing the various functions may also be regarded as structures within both software modules and hardware components for performing the method.
The foregoing description of specific embodiments of the present invention has been presented. It is to be understood that the present invention is not limited to the specific embodiments described above, and that various changes or modifications may be made by one skilled in the art within the scope of the appended claims without departing from the spirit of the invention. The embodiments and features of the embodiments of the present application may be combined with each other arbitrarily without conflict.
Claims (10)
1. An intelligent standard method based on an automatic driving test intelligent platform vehicle is characterized by comprising the following steps:
step 1: establishing an automobile intelligent index system of the automatic driving test intelligent platform vehicle, and analyzing related evaluation indexes of the automatic driving test intelligent platform vehicle;
step 2: determining a network structure of a BP neural network according to an automobile intelligent standard of an automatic driving test intelligent platform vehicle;
and step 3: determining the number of hidden neurons;
and 4, step 4: carrying out normalization dimensionless processing on the data;
and 5: determining the intelligent level of the automatic driving intelligent platform vehicle;
step 6: and carrying out BP network training and determining the weight range of the sample index.
2. The intelligent standard method based on the automatic driving test intelligent platform truck as claimed in claim 1, wherein the step 1 comprises:
step 1.1: selecting indexes related to intelligent evaluation of the automobile as comparison series, taking the intelligent indexes of the intelligent platform car as reference series, carrying out non-dimensionalization processing on variables, and then calculating a correlation coefficient;
note deltai(k)＝|y(k)-xi(k) I, then
Rho belongs to (0, infinity), namely a resolution coefficient, and the smaller rho is, the larger the resolution is, and the value of rho is (0, 1);
step 1.2: after the correlation coefficient is obtained, calculating the average value of the correlation coefficient as the number representation of the correlation degree between the comparison number series and the reference number series; the degree of association ri is formulated as follows:
step 1.3: and sorting according to the relevance and the size, and selecting the index with high relevance as a final evaluation index.
3. The intelligent standard method based on the automatic driving test intelligent platform truck as claimed in claim 1, wherein the step 2 comprises:
step 2.1: determining the number of nodes of an input layer and an output layer; determining the number of neurons of an input layer according to the number of the selected detailed test function indexes, wherein the number of the neurons of an output layer is one, and the intelligent index value of the intelligent vehicle is output;
step 2.2: determining the number of layers of the BP neural network, fitting a function containing continuous mapping from one finite space to another finite space by a single hidden layer, and expressing any decision boundary of any progress by two layers of activation functions.
4. The intelligent standard method based on the automatic driving test intelligent platform truck as claimed in claim 1, wherein the step 3 comprises:
the number of the implicit neurons is determined by mainly adopting an empirical formula:
in the formula: q is the number of hidden layer nodes; a is the number of nodes of an input layer; b is the number of output layer nodes; c, taking any integer of [ a,10 ];
when the overfitting condition occurs, the correction is carried out by adopting early-stopping or adding a regularization term.
5. The intelligent standard method based on the automatic driving test intelligent platform truck as claimed in claim 1, wherein the step 5 comprises:
the intelligent index value of the intelligent vehicle is reasonably divided into regions, and data with special conditions in the test data of the intelligent vehicle is subjected to other processing.
6. The intelligent standard method based on the automatic driving test intelligent platform truck as claimed in claim 1, wherein the step 6 comprises:
step 6.1: creating a network object, and setting an activation function, a training function, a learning function and a performance function of each layer;
step 6.2: adjusting the weight and threshold of the network;
step 6.3: setting training parameters of the network; forming a training sample by the normalized data and the target value, and repeatedly training the established BP network;
step 6.4: obtaining an initial weight bi of the normalized evaluation index value through an entropy weight method, scoring the set intelligent index through k experts and calculating the weight;
step 6.5: calculating according to the weights calculated by an expert evaluation method and an entropy weight method to obtain a comprehensive weight;
step 6.6: after calculating the intelligent index value of the intelligent vehicle through the comprehensive weight, comparing the ideal intelligent index value output by the BP network;
step 6.7: and adjusting the comprehensive weight in the weight range obtained by an expert evaluation method and an entropy weight method until the intelligent index value error of the automobile reaches an ideal value.
7. The utility model provides an intelligent standard system based on automatic driving tests intelligent platform car which characterized in that, the system includes following module:
module M1: establishing an automobile intelligent index system of the automatic driving test intelligent platform vehicle, and analyzing related evaluation indexes of the automatic driving test intelligent platform vehicle;
module M2: determining a network structure of a BP neural network according to an automobile intelligent standard of an automatic driving test intelligent platform vehicle;
module M3: determining the number of hidden neurons;
module M4: carrying out normalization dimensionless processing on the data;
module M5: determining the intelligent level of the automatic driving intelligent platform vehicle;
module M6: and carrying out BP network training and determining the weight range of the sample index.
8. The intelligent standard system based on the automated driving test intelligent platform vehicle of claim 7, wherein the module M1 comprises:
module M1.1: selecting indexes related to intelligent evaluation of the automobile as comparison series, taking the intelligent indexes of the intelligent platform car as reference series, carrying out non-dimensionalization processing on variables, and then calculating a correlation coefficient;
note deltai(k)＝|y(k)-xi(k) I, then
Rho belongs to (0, infinity), namely a resolution coefficient, and the smaller rho is, the larger the resolution is, and the value of rho is (0, 1);
module M1.2: after the correlation coefficient is obtained, calculating the average value of the correlation coefficient as the number representation of the correlation degree between the comparison number series and the reference number series; the degree of association ri is formulated as follows:
module M1.3: sorting according to the relevance and the size, and selecting an index with high relevance as a final evaluation index;
the module M2 includes:
module M2.1: determining the number of nodes of an input layer and an output layer; determining the number of neurons of an input layer according to the number of the selected detailed test function indexes, wherein the number of the neurons of an output layer is one, and the intelligent index value of the intelligent vehicle is output;
module M2.2: determining the number of layers of the BP neural network, fitting a function containing continuous mapping from one finite space to another finite space by a single hidden layer, and expressing any decision boundary of any progress by two layers of activation functions.
9. The intelligent standard system based on the automated driving test intelligent platform vehicle of claim 7, wherein the module M3 comprises:
the number of the implicit neurons is determined by mainly adopting an empirical formula:
in the formula: q is the number of hidden layer nodes; a is the number of nodes of an input layer; b is the number of output layer nodes; c, taking any integer of [ a,10 ];
when the overfitting condition occurs, correcting in an early-stopping or regularization item adding mode;
the module M5 includes:
the intelligent index value of the intelligent vehicle is reasonably divided into regions, and data with special conditions in the test data of the intelligent vehicle is subjected to other processing.
10. The intelligent standard system based on the automated driving test intelligent platform vehicle of claim 7, wherein the module M6 comprises:
module M6.1: creating a network object, and setting an activation function, a training function, a learning function and a performance function of each layer;
module M6.2: adjusting the weight and threshold of the network;
module M6.3: setting training parameters of the network; forming a training sample by the normalized data and the target value, and repeatedly training the established BP network;
module M6.4: obtaining an initial weight bi of the normalized evaluation index value through an entropy weight method, scoring the set intelligent index through k experts and calculating the weight;
module M6.5: calculating according to the weights calculated by an expert evaluation method and an entropy weight method to obtain a comprehensive weight;
module M6.6: after calculating the intelligent index value of the intelligent vehicle through the comprehensive weight, comparing the ideal intelligent index value output by the BP network;
module M6.7: and adjusting the comprehensive weight in the weight range obtained by an expert evaluation method and an entropy weight method until the intelligent index value error of the automobile reaches an ideal value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110744950.5A CN113378479A (en) | 2021-06-30 | 2021-06-30 | Intelligent standard method and system based on automatic driving test intelligent platform vehicle |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110744950.5A CN113378479A (en) | 2021-06-30 | 2021-06-30 | Intelligent standard method and system based on automatic driving test intelligent platform vehicle |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113378479A true CN113378479A (en) | 2021-09-10 |
Family
ID=77580663
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110744950.5A Pending CN113378479A (en) | 2021-06-30 | 2021-06-30 | Intelligent standard method and system based on automatic driving test intelligent platform vehicle |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113378479A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114971225A (en) * | 2022-05-10 | 2022-08-30 | 东南大学 | Unmanned bus ride comfort evaluation method based on distributed intelligent comprehensive decision |
| CN114996859A (en) * | 2022-07-19 | 2022-09-02 | 山东贞元汽车车轮有限公司 | Intelligent equipment control system for intelligent manufacturing of wheel steel ring |
| CN117689266A (en) * | 2023-12-27 | 2024-03-12 | 中国汽车工程研究院股份有限公司 | New energy automobile index evaluation system and method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103234763A (en) * | 2013-04-09 | 2013-08-07 | 北京理工大学 | System and method for quantitatively evaluating unmanned vehicles |
| CN112465395A (en) * | 2020-12-15 | 2021-03-09 | 同济大学 | Multi-dimensional comprehensive evaluation method and device for automatically-driven automobile |
| DE102019134309A1 (en) * | 2019-12-13 | 2021-06-17 | Valeo Schalter Und Sensoren Gmbh | METHOD OF OPERATING A VEHICLE, PARKING ASSISTANCE SYSTEM AND VEHICLE |
-
2021
- 2021-06-30 CN CN202110744950.5A patent/CN113378479A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103234763A (en) * | 2013-04-09 | 2013-08-07 | 北京理工大学 | System and method for quantitatively evaluating unmanned vehicles |
| DE102019134309A1 (en) * | 2019-12-13 | 2021-06-17 | Valeo Schalter Und Sensoren Gmbh | METHOD OF OPERATING A VEHICLE, PARKING ASSISTANCE SYSTEM AND VEHICLE |
| CN112465395A (en) * | 2020-12-15 | 2021-03-09 | 同济大学 | Multi-dimensional comprehensive evaluation method and device for automatically-driven automobile |
Non-Patent Citations (7)
| Title |
|---|
| FANGJING DU等: "Evaluation of the Autonomous Parking System Based on BP Neural Network", 《IEEE XPLORE》 * |
| 修海林: "有条件自动驾驶汽车测试与综合评价研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 * |
| 北京市科学技术协会主编: "《人工智能》", 30 September 2018 * |
| 卢霄娟: "雨天城市快速路短时速度预测与交通安全评价研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 * |
| 汪懋华等: "《现代精细农业理论与实践》", 31 October 2012 * |
| 熊光明等: "《无人驾驶车辆智能行为及其测试与评价》", 30 November 2015 * |
| 闫文周: "《绿色建筑部品评价***》", 30 November 2009 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114971225A (en) * | 2022-05-10 | 2022-08-30 | 东南大学 | Unmanned bus ride comfort evaluation method based on distributed intelligent comprehensive decision |
| CN114971225B (en) * | 2022-05-10 | 2024-07-12 | 东南大学 | Unmanned public transportation smoothness evaluation method based on distributed intelligent comprehensive decision |
| CN114996859A (en) * | 2022-07-19 | 2022-09-02 | 山东贞元汽车车轮有限公司 | Intelligent equipment control system for intelligent manufacturing of wheel steel ring |
| CN117689266A (en) * | 2023-12-27 | 2024-03-12 | 中国汽车工程研究院股份有限公司 | New energy automobile index evaluation system and method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113378479A (en) | Intelligent standard method and system based on automatic driving test intelligent platform vehicle | |
| CN108492555B (en) | Urban road network traffic state evaluation method and device | |
| CN106874597B (en) | highway overtaking behavior decision method applied to automatic driving vehicle | |
| CN110188797B (en) | Intelligent automobile rapid test method based on Bayesian optimization | |
| CN105946861A (en) | NAR neural network vehicle speed prediction method based on driving intention recognition | |
| CN111738337B (en) | Driver distraction state detection and identification method in mixed traffic environment | |
| CN112668172B (en) | Following behavior modeling method considering heterogeneity of vehicle type and driving style and model thereof | |
| US20210166085A1 (en) | Object Classification Method, Object Classification Circuit, Motor Vehicle | |
| CN113158560A (en) | Intelligent driving vehicle autonomous capacity testing method based on scene impedance | |
| CN108875812B (en) | Driver behavior classification method based on branch convolutional neural network | |
| CN111563555A (en) | Driver driving behavior analysis method and system | |
| CN106570560A (en) | Driving style quantitative evaluation method based on standardized driving behaviors and phase space reconstruction | |
| CN114971009A (en) | Vehicle insurance risk prediction method and computer equipment | |
| CN113406955A (en) | Complex network-based automatic driving automobile complex environment model, cognitive system and cognitive method | |
| CN113706860B (en) | Intelligent timing traffic light control method based on raspberry group | |
| CN113642114A (en) | Modeling method for humanoid random car following driving behavior capable of making mistakes | |
| CN116946183A (en) | Commercial vehicle driving behavior prediction method considering driving capability and vehicle equipment | |
| CN114066288B (en) | Intelligent data center-based emergency detection method and system for operation road | |
| CN113954844B (en) | Intelligent automobile man-machine driving mode switching system | |
| Mumcuoglu et al. | Driver evaluation in heavy duty vehicles based on acceleration and braking behaviors | |
| Tang et al. | Study on comprehensive evaluation of L3 automated vehicles | |
| CN116884225B (en) | Method and device for evaluating influence of proportion of automatic driving vehicles on traffic efficiency | |
| Li et al. | Research on Personalized AEB Strategies Based on Self-Supervised Contrastive Learning | |
| CN112215520B (en) | Multi-target fusion passing method and device, computer equipment and storage medium | |
| CN112560354B (en) | Car following behavior modeling method based on Gaussian process regression |
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 |
Application publication date: 20210910 |
|
| RJ01 | Rejection of invention patent application after publication |