A method for constructing a functional architecture for autonomous transportation systems
By using a multi-attribute text clustering algorithm and an optimized density peak clustering model, an autonomous transportation system functional architecture is constructed, which solves the problem of insufficient adaptability of the transportation system framework and achieves efficient functional architecture optimization and clustering accuracy.
Patent Information
- Application Number
- CN202210237148.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-03-11
- Publication Date
- 2024-12-17
- Estimated Expiration
- 2042-03-11
AI Technical Summary
The existing transportation system framework is difficult to adapt to the development from "passive intelligence" to "active intelligence" under emerging technologies, and there is an urgent need to build a functional architecture that adapts to the next generation of transportation systems.
A multi-attribute text clustering algorithm is adopted. By optimizing the density peak clustering model, combining the Gaussian kernel method and decision value optimization, the multi-attribute text is transformed into spatial dimension coordinates using the TF-IDF algorithm and text vector space model. The clustering results are evaluated by the silhouette coefficient method, and the functional architecture of the autonomous transportation system is constructed.
It optimizes clustering performance, provides a new approach to the functional architecture of autonomous transportation systems, offers a reference for the construction of next-generation transportation systems, and improves the accuracy and effectiveness of clustering.
Smart Images
Figure CN114579748B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of road traffic, and in particular relates to a method for constructing a functional architecture of an autonomous traffic system. Background Art
[0002] At present, we need to build a modern integrated transportation system that is "safe, convenient, efficient, green and economical". The existing transportation system framework is difficult to adapt to the development of transportation from "passive intelligence" to "active intelligence" under emerging technologies, and it is urgent to build a transportation system that adapts to the development of the new generation of transportation. In this context, the country has clearly proposed to build an autonomous transportation system (ATS) to improve the applicability of the current intelligent transportation system to emerging technologies and guide the development of my country's transportation in a more scientific and reasonable way. Summary of the invention
[0003] In view of this, the present invention proposes a method for constructing a functional architecture of an autonomous transportation system, which promotes the construction of a new generation of transportation system functional architecture and promotes the development of a theoretical system of an autonomous transportation system.
[0004] The method for constructing the functional architecture of an autonomous transportation system disclosed in the present invention comprises the following steps:
[0005] Step 1: Aiming at the problem of multi-attribute text clustering, an optimized density peak clustering model for multi-attribute text is constructed;
[0006] Step 2: Calculate the word attribute frequency and the inverse document frequency of the attribute according to the characteristics of the multi-attribute text, and convert the multi-attribute text into a numerical value according to the word attribute frequency and the inverse document frequency of the attribute;
[0007] Step 3: According to the numerical weights of different attribute texts, the text vector space model is used to convert the numerical values into space dimension coordinates;
[0008] Step 4: Optimize the calculation method of density peak: Use Gaussian kernel method to optimize the calculation method of local density of data points;
[0009] The decision value is used to optimize the selection of cluster centers, and the samples with larger decision values are selected as the density peak cluster centers to obtain the clustering results;
[0010] Step 5: Based on the clustering results of step 4, use the silhouette coefficient method to evaluate the clustering results, and repeat step 4 for the data points with lower silhouette coefficients;
[0011] Step 6: Verify the feasibility of the algorithm. In the road autonomous driving scenario of the autonomous transportation system, the functions under the three service domains are divided into four layers of "autonomous perception, autonomous learning, autonomous decision-making, and autonomous response" based on the final clustering results. Draw a functional architecture diagram based on the relationship between the layers.
[0012] Furthermore, the specific steps of step 2 are as follows:
[0013] (1) Calculate the frequency of word attributes: Among them, s represents the attribute of a text, is the number of times the attribute s appears in the document, W n (n=1,2,...,N, N is the total number of documents) represents the attribute library consisting of text attributes;
[0014] (2) According to the needs of text attribute conversion, assuming that the word must appear in the attribute library, the inverse document frequency of the attribute is calculated using the smoothed inverse document frequency:
[0015]
[0016] (3) The calculation formula of the TF-IDF algorithm for multi-attribute text is TF·IDF(s,W n ,W)=TF(s,W n )·IDF(s,W), when the TF-IDF value is larger, it means that the word can better represent the document, that is, when the TF-IDF value is larger, it means that the attribute can better represent the text.
[0017] Furthermore, the text space dimension coordinates are obtained based on the text vector space model. The text vector space model converts a given text into a vector of a certain dimension, which takes feature items as the basic unit of text representation. Each dimension of the vector corresponds to a feature item in the text, and each dimension itself represents the weight of the corresponding feature item in the text. The weight represents the importance of the feature item to the text, that is, to what extent the feature item can reflect the category of the document in which it is located.
[0018] Furthermore, by optimizing the density peak clustering algorithm, we can find the clustering center that meets the conditions, cluster the remaining points to be clustered, and improve the clustering performance. The specific steps are as follows:
[0019] The local density truncation kernel calculation method is optimized to the Gaussian kernel calculation method to significantly improve the precision of density calculation; the Gaussian kernel calculation method calculates the Euclidean distance and the Gaussian function value, and the sum of all point values is used as the density. The Gaussian function value defines the sample local density calculation formula as follows:
[0020]
[0021] Among them, dij For sample x i and x j The distance between c is the cutoff distance, n is the number of samples;
[0022] Calculate the local density distance: Calculate the minimum distance ξ from each point to a point with a higher local density value than itself i ,ξ i The calculation formula is as follows: For the sample with the highest density, ξ i Defined as ξ i =max j (d ij );where p i For sample x i The local density, p j For sample x j The local density of .
[0023] The selection of cluster center points is optimized by using the decision value method: select the decision value γ i The larger sample is used as the density peak cluster center, that is, the point where the product of local density and local density distance is larger, γ i The calculation formula of γ is as follows: i =ρ i ·ξ i .
[0024] Furthermore, the clustering performance is tested by the silhouette coefficient, and the data points with lower silhouette coefficients are clustered repeatedly. The specific algorithm flow is as follows:
[0025] Calculate the average distance a from sample i to other samples in the same cluster i , a i is called the intra-cluster dissimilarity of sample i; a i The smaller it is, the more sample i should be clustered into this cluster. The a of all samples in cluster C i The mean is called the cluster dissimilarity of cluster C;
[0026] Calculate the number of samples i to another cluster C j The average distance b of all samples i,j , called sample i and cluster C j The dissimilarity between clusters of sample i is defined as: b i = {b i1 ,b i2 ,...,b ik}, b i The larger it is, the less sample i belongs to other clusters;
[0027] According to the intra-cluster dissimilarity a of sample i i and the inter-cluster dissimilarity bi , define the silhouette coefficient of sample i:
[0028]
[0029]
[0030] Judgment: i If it is close to 1, it means that the clustering of sample i is reasonable and effective; s i If s is close to -1, it means that sample i should be classified into another cluster. i If it is approximately 0, it means that sample i is on the boundary of two clusters.
[0031] Furthermore, for the construction of the functional architecture of the autonomous transportation system, the functions in the service domain under the scenario are clustered and analyzed, and the functional architecture is built based on the clustering results, and the functional architecture diagram is drawn.
[0032] Furthermore, the functions of the autonomous transportation system have three attributes: provider, process object, and service object, and an attribute library is established.
[0033] The beneficial effects of the present invention are as follows:
[0034] 1) Determine the construction method of the functional architecture of the autonomous transportation system and provide new ideas for the construction of the autonomous transportation system;
[0035] 2) Optimize the density peak clustering algorithm, calculate the data point density through the Gaussian function value, and optimize the cluster center selection through the decision value;
[0036] 3) Provide methods for converting multi-attribute texts in various fields into usable data;
[0037] 4) The present invention provides a method for constructing a functional architecture of an autonomous transportation system, which is an innovation of the functional architecture of the transportation system and can provide a reference for the construction of the next generation of autonomous transportation systems. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] Figure 1 Flow chart of the method of the present invention;
[0039] Figure 2 Schematic diagram of the density peak clustering algorithm;
[0040] Figure 3 Road transport operation service domain;
[0041] Figure 4 Transportation infrastructure management service domain;
[0042] Figure 5 Traffic safety management service domain;
[0043] Figure 6Find the cluster center scatter plot;
[0044] Figure 7 Preliminary clustering result diagram;
[0045] Figure 8 Preliminary clustering silhouette coefficient plot;
[0046] Fig. 9 Functional architecture diagram of collision management sub-service. DETAILED DESCRIPTION
[0047] The present invention is further described below in conjunction with the accompanying drawings, but the present invention is not limited in any way. Any changes or substitutions made based on the teachings of the present invention belong to the protection scope of the present invention.
[0048] The present invention establishes an optimized density peak clustering algorithm (Density Peaks Clustering, DPC) model for multi-attribute text to construct the functional architecture under autonomous transportation. The algorithm combines the basic characteristics of the functional architecture of the transportation system, and transforms the multi-attribute text into spatial dimension coordinates through an improved term frequency-inverse document frequency (Term Frequency-Inverse Document Frequency, TF-IDF) algorithm and a text vector space model. Then, the Gaussian function and decision value are used to optimize the DPC algorithm for clustering, and the clustering results are evaluated in combination with the silhouette coefficient. The algorithm construction ideas are as follows Figure 1 shown.
[0049] TF stands for word frequency. If a word t appears in document D n (n=1,2,...,N, N is the total number of documents) The word frequency Considering the different lengths of each document, the frequency of the word is adjusted according to the document length: Considering that this model is for multi-attribute text clustering, each word has more attributes. Now we perform cluster analysis based on word attributes. To optimize the cluster analysis results, we improve the word frequency TF calculation formula to obtain the word attribute frequency:
[0050]
[0051] Among them, s represents the attribute of a text, W n (n=1,2,...,N, N is the total number of documents) represents the attribute library consisting of text attributes;
[0052] IDF stands for inverse document frequency, which can be used to measure the general importance of a word. IDF(t,D) can be expressed as the total number of documents divided by the logarithm of the number of documents in which the word appears in the word library. The calculation formula is: Where t represents a word, D represents the entire corpus, N represents the total number of documents in the corpus, and n t It indicates the frequency of texts with t in the entire corpus. The value of l is determined by experiments and is generally 0.01. According to the needs of text attribute conversion, assuming that the word must appear in the attribute library, the smoothed inverse document frequency is used to calculate the inverse document frequency of the text attribute:
[0053]
[0054] The calculation formula of the TF-IDF algorithm for multi-attribute text is TF·IDF(s,W n ,W)=TF(s,W n )·IDF(s,W), when the TF-IDF value is larger, it means that the word is more representative of the document, that is, in the present invention, the larger the TF-IDF value is, the more representative the attribute is of the text. Thus, the data value converted from the text is calculated according to the attribute characteristics of each text, thereby providing a scientific basis for converting multi-attribute text into data;
[0055] According to the TF-IDF algorithm, the numerical weights of texts with different attributes are obtained. In order to use the density peak clustering algorithm for spatial clustering analysis, it is necessary to convert the attribute weights of the text into spatial dimension coordinates. Therefore, the text vector space model is used to convert the attribute weights into spatial dimension coordinates.
[0056] The text vector space model (VSM) can convert a given text into a vector of a certain dimension. It uses feature items as the basic unit of text representation. Each dimension of the vector corresponds to a feature item in the text, and each dimension itself represents the weight of the corresponding feature item in the text. The weight represents the importance of the feature item to the text, that is, to what extent the feature item can reflect the category of the document it is in. The model is defined as: In a given document, D = D(s1, w1; s2, w2; ...; s n ,w n ), D complies with the following two agreements: each characteristic item s k (1≤k≤n) are mutually different; each characteristic item s k There is no order relationship, and a space vector is constructed. That is, the attribute weight TF·IDF(s,W) of the text is obtained by the TF-IDF algorithm. n ,W), multiple attributes of the text are converted into spatial dimension coordinates of the text according to VSM.
[0057] Density Peaks Clustering (DPC) is a density-based spatial clustering algorithm. The principle of the algorithm is that the local density of the cluster center is large, and the cluster center is far away from other points with higher local density. The steps of the algorithm are as follows:
[0058] (1) Set the cutoff distance dc, based on statistical experience, dc=0.02 or dc=0.04.
[0059] (2) For a given data set X N·M =[x1,x2,...,x N ] T , where x i =[x i1 ,x i2 ,...,x iN ], N is the number of samples, M is the sample dimension. Sample x i The local density is denoted as ρ i , the calculation formula is as follows: i =∑ i≠j χ(d ij -d c ), Among them, d ij For sample x i and x j The distance between them is calculated using the Euclidean distance, which is i The number of points whose distance is less than dc is directly used as the density of the i-th point. Euclidean distance calculation formula (i,j=1,2,...,n,i≠j,α=1,2,...,M), where, x iα Represents the αth dimension coordinate of the i-th point.
[0060] (3) Calculate the local density distance: Calculate the minimum distance ξ from each point to a point with a higher local density value than itself i ,ξ i The calculation formula is as follows: For the sample with the highest density, ξ i Defined as ξ i =max j (d ij ), where p i For sample x i The local density, p j For sample x j The local density of .
[0061] (4) Estimate the center point based on density and distance: For each point to be clustered, draw a scatter plot of local density and the distance to points with higher local density. The density peak is usually the sample with higher local density and larger relative distance.
[0062] (5) Divide the remaining points to be clustered: classify each data point into the category to which the closest cluster center with a higher density belongs. The principle of the density peak clustering algorithm is as follows: Figure 2 shown.
[0063] By optimizing the DPC algorithm, it is possible to more accurately find cluster centers that meet the conditions, cluster the remaining points to be clustered, and improve clustering performance. Optimizing the local density truncation kernel calculation method to the Gaussian kernel calculation method can significantly improve the precision of density calculation, thereby better optimizing the clustering effect. The Gaussian kernel calculation method calculates the Euclidean distance and the Gaussian function value, and takes the sum of all point values as the density. The Gaussian function value defines the sample local density calculation formula as: Since the density peak clustering algorithm initially manually selected the cluster center based on the scatter plot of the local density and the distance between the high local density points, the manual selection of the cluster center is prone to large deviations, so the selection of the cluster center point is optimized. The decision value method is used for optimization: the decision value γ is selected i The larger sample is used as the density peak cluster center, that is, the point where the product of local density and local density distance is larger, γ i The calculation formula of γ is as follows: i =ρ i ·ξ i
[0064] The silhouette coefficient is used to evaluate the effect of clustering, and its value range is [-1,1]. The larger the silhouette coefficient, the better the clustering effect. The specific algorithm flow is as follows:
[0065] (1) Calculate the average distance a between sample i and other samples in the same cluster i , a i It is called the intra-cluster dissimilarity of sample i. i The smaller it is, the more likely sample i should be clustered into this cluster. i The mean is called the cluster dissimilarity of cluster C. The average distance a i That is, calculate the sum of the distances from sample i to other samples in the same cluster and divide it by the number of all samples.
[0066] (2) Calculate the number of samples i to another cluster C j The average distance b of all samples i,j , called sample i and cluster C j The dissimilarity between clusters of sample i is defined as: b i = {b i1 ,b i2 ,...,b ik}, b i The larger the value is, the less sample i belongs to other clusters. Average distance bi,j That is, calculate the sample i to another cluster C j The sum of the distances of all samples is divided by the number of samples.
[0067] (3) According to the intra-cluster dissimilarity a of sample i i and the inter-cluster dissimilarity b i , define the silhouette coefficient of sample i:
[0068]
[0069]
[0070] (4) Judgment: s i If it is close to 1, it means that the clustering of sample i is reasonable and effective; s i If s is close to -1, it means that sample i should be classified into another cluster. i If it is approximately 0, it means that sample i is on the boundary of two clusters.
[0071] By selecting functional data sets in the road autonomous driving scenario in the next generation autonomous transportation system, an example analysis is conducted to verify the performance of the algorithm. The autonomous transportation system is divided into nine service domains: traveler information service domain, road vehicle operation service domain, freight transportation service domain, transportation infrastructure management service domain, public transportation service domain, transportation operation management service domain, comprehensive transportation service domain, traffic safety management service domain, and data management and security service domain. According to the needs of building a road autonomous driving scenario, three service domains are studied in the scenario: road vehicle operation service domain, transportation infrastructure management service domain, and traffic safety management service domain. Each service domain has corresponding services, sub-services, and functions. There are 3 services and 10 sub-services in the road vehicle operation service domain. The distribution of this service domain is as follows: Figure 3 The traffic infrastructure management service domain has a total of 1 service and 2 sub-services. The distribution of this service domain is as follows: Figure 4 The traffic safety management service domain has 2 services and 3 sub-services. The distribution of this service domain is as follows: Figure 5 shown.
[0072] Each sub-service in the three service domains contains multiple functions. To facilitate clustering, the attributes of the functions are first defined as follows and an attribute library is established.
[0073] (1) Provider: Determines the component entities, associated technology combinations, and associated requirements combinations. Use a to create a library for attribute providers. i (i=1,2,......,10) as shown in Table 1.
[0074] Table 1 Property provider library
[0075]
[0076]
[0077] Process object: An object used or processed by a function. i (i=1,2,......,25) as shown in Table 2.
[0078] Table 2 Attribute Process Object Library
[0079] symbol name symbol name <![CDATA[b1]]> Basic vehicle information <![CDATA[b 14 ]]> Basic road surface information <![CDATA[b2]]> Vehicle data <![CDATA[b 15 ]]> Road network information <![CDATA[b3]]> Vehicle Information <![CDATA[b 16 ]]> Task dispatch information <![CDATA[b4]]> Vehicle operation information <![CDATA[b 17 ]]> Equipment operation information <![CDATA[a5]]> Vehicle status data <![CDATA[b 18 ]]> Accident Information <![CDATA[b6]]> Scheduling plan <![CDATA[b 19 ]]> Collaborative facility information <![CDATA[b7]]> Driver information <![CDATA[b 20 ]]> Maintenance plan <![CDATA[b8]]> Traffic regulations information <![CDATA[b 21 ]]> Early warning plan <![CDATA[b9]]> Traffic status information <![CDATA[b 22 ]]> Warning information <![CDATA[b 10 ]]> Emergency Vehicle Information <![CDATA[b 23 ]]> Support facility information <![CDATA[b 11 ]]> Emergency Information <![CDATA[b 24 ]]> Surrounding vehicle information <![CDATA[b 12 ]]> Emergency warning information <![CDATA[b 25 ]]> Surrounding environment information <![CDATA[b 13 ]]> Route navigation information
[0080] Service object: An object that can directly use a function or utilize the output of a function. i (i=1,2,......,10) as shown in Table 3.
[0081] Table 3 Attribute service object library
[0082]
[0083]
[0084] Combined with the attribute library, the attributes of each function are studied, and the various functions of the sub-services in the road autonomous driving scenario are represented in the form of attribute combinations. The required functions of the sub-services are scientifically determined according to their needs, as follows:
[0085] Table 4 Service function table under the scenario
[0086] Serve Function Automatic parking <![CDATA[{A1;A2;A3;A4;A5;A6;A7;A8;A9}]]> Adaptive cruise control <![CDATA[{A1;A 10 ;A 11 ;A 12 ;A 13 ;A5;A 14 ;A 15 ;A7;A 16 ;A 17 ;A 18 }]]> Automatic Lane Change <![CDATA[{A2;A4;A 10 ;A 19 ;A 14 ;A7;A 16 ;A 20 ;A 21 }]]> Collision Management <![CDATA[{A 10 ;A1;A 19 ;A 16 ;A 16 ;A 22 ;A 23 ;A 24 }]]> Vehicle platooning services <![CDATA[{A1;A2;A 10 ;A 25 ;A7;A 16 ;A 26 ;A 27 ;A 28 ;A 29 }]]> Vehicle and signal light collaborative driving service <![CDATA[{A 30 ;A 13 ;A 31 ;A 15 ;A 32 ;A 33 ;A 34 }]]> Traffic service at intersections without traffic lights <![CDATA[{A 10 ;A 35 ;A 36 ;A 16 ;A 36 ;A 37 ;A 38 ;A 39 ;A 40 ;A 41 }]]> Vehicle annual inspection <![CDATA[{A 42 ;A 43 ;A 44 ;A 45 ;A 46 ;A 47 }]]> Unmanned vehicle performance test <![CDATA[{A 48 ;A 42 ;A 44 ;A 45 ;A 49 ;A 50 }]]> Unmanned vehicle scenario testing <![CDATA[{A 48 ;A 42 ;A 51 ;A 44 ;A 45 ;A 52 ;A 49 ;A 50 }]]> Communications technology collaborative facility management <![CDATA[{A 53 ;A 54 ;A 55 ;A 56 ;A 57 }]]> Artificial Intelligence Information Support Facility Management <![CDATA[{A 58 ;A 59 ;A 60 ;A 61 ;A 62 }]]> Vehicle monitoring and warning <![CDATA[{A 63 ;A 48 ;A 45 ;A 44 ;A 64 ;A 65 ;A 66 }]]> Emergency Detection and Automatic Identification <![CDATA[{A 67 ;A 68 ;A 69 ;A 70 ;A 71 ;A 72 ;A 73 ;A 74 }]]> Emergency Vehicle Dispatch <![CDATA[{A 75 ;A 60 ;A 72 ;A 76 ;A 77 ;A 78 }]]>
[0087] The TF-IDF algorithm is used to convert the attributes of each service function into data, and then converted into spatial dimension coordinates based on the VSM algorithm, as shown in Table 5.
[0088] Table 5 Functional space dimension coordinate table under the scenario
[0089]
[0090]
[0091] Since some functions with different names have the same attributes, the above 78 data are deleted after duplicate values, resulting in a total of 37 spatial dimension coordinate data. Then, MATLAB software is used for programming, and the optimized density peak clustering algorithm is adopted, where dc is taken as 0.02. The density and local density distance of each function point are calculated, and the scatter plot for finding the cluster center is obtained, as shown in the figure below: Figure 6 shown.
[0092] From the above figure, we can see that the data density and local density distance are calculated for each function point. The decision value method is used to calculate the product of the data density and local density distance of each function point. The data are arranged in descending order. Starting from the sixth place, the decision value of the data is lower than 0.1, so the first 5 places with larger decision values are recorded as shown in Table 6. The data in the table show that it is more appropriate to use these 5 points as the clustering centers for the clustering points.
[0093] Table 6 Decision values
[0094]
[0095]
[0096] Then, according to the principle of density peak clustering algorithm, continue to cluster the data and get the following through MATLAB programming: Figure 7 Then, the silhouette coefficient is used to evaluate the clustering results of each function point obtained by the density peak clustering algorithm, and the following is obtained: Figure 8 The silhouette coefficient diagram of each functional point is shown in Table 7. The overall mean of the silhouette coefficient after clustering is 0.84, and the clustering effect is good, but the silhouette coefficients of three data points are lower than 0.6. Therefore, cluster analysis is performed again on clusters 1 and 4 where these three data points are located to improve the performance of clustering results. After cluster analysis of clusters 1 and 4 again, the clustering results of 4 clusters are obtained as shown in Table 7.
[0097] Table 7 Final clustering results
[0098]
[0099] According to the clustering results, cluster 1 is named the autonomous perception layer, which is mainly responsible for acquiring information from the physical world; cluster 2 is named the autonomous learning layer, which is mainly responsible for processing perception data and acquiring knowledge; cluster 3 is named the autonomous decision-making layer, which is mainly responsible for forming the optimal control plan through full evaluation; cluster 4 is named the autonomous response layer, which is mainly responsible for the final execution stage.
[0100] Draw a functional architecture diagram such as Fig. 9 As shown in the figure, the collision management sub-service under the automatic vehicle driving service in the road transport operation service domain is taken as an example. In this sub-service, there are four layers: autonomous perception layer, autonomous learning layer, autonomous decision layer and autonomous response layer. Finally, the collision management sub-service is realized through a variety of multi-attribute functions, thereby supporting the automatic vehicle driving service.
[0101] The beneficial effects of the present invention are as follows:
[0102] 1) Determine the construction method of the functional architecture of the autonomous transportation system and provide new ideas for the construction of the autonomous transportation system;
[0103] 2) Optimize the density peak clustering algorithm, calculate the data point density through the Gaussian function value, and optimize the cluster center selection through the decision value;
[0104] 3) Provide methods for converting multi-attribute texts in various fields into usable data;
[0105] 4) The present invention provides a method for constructing a functional architecture of an autonomous transportation system, which is an innovation of the functional architecture of the transportation system and can provide a reference for the construction of the next generation of autonomous transportation systems.
[0106] As used herein, the word "preferred" is intended to be used as an example, instance, or illustration. Any aspect or design described herein as "preferred" is not necessarily to be construed as being more advantageous than other aspects or designs. On the contrary, the use of the word "preferred" is intended to present concepts in a specific way. The term "or" as used in this application is intended to mean an inclusive "or" rather than an exclusive "or". That is, unless otherwise specified or clear from the context, "X uses A or B" means any one of the naturally included permutations. That is, if X uses A; X uses B; or X uses both A and B, then "X uses A or B" is satisfied in any of the foregoing examples.
[0107] Moreover, although the present disclosure has been shown and described with respect to one or implementations, those skilled in the art will think of equivalent variations and modifications based on the reading and understanding of this specification and the accompanying drawings. The present disclosure includes all such modifications and variations, and is limited only by the scope of the appended claims. In particular, with respect to the various functions performed by the above-mentioned components (such as elements, etc.), the terms used to describe such components are intended to correspond to any component (unless otherwise indicated) that performs the specified function of the component (such as it is functionally equivalent), even if the structure is not equivalent to the disclosed structure of the function in the exemplary implementation of the present disclosure shown herein. In addition, although the specific features of the present disclosure have been disclosed with respect to only one of several implementations, such features can be combined with one or other features of other implementations that may be desired and advantageous for a given or specific application. Moreover, insofar as the terms "including", "having", "containing" or their variations are used in specific embodiments or claims, such terms are intended to be included in a manner similar to the term "comprising".
[0108] The functional units in the embodiments of the present invention may be integrated into a processing module, or each unit may exist physically separately, or multiple or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium. The above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc. The above-mentioned devices or systems may execute the storage method in the corresponding method embodiment.
[0109] To sum up, the above embodiment is an implementation mode of the present invention, but the implementation mode of the present invention is not limited by the embodiment, and any other changes, modifications, substitutions, combinations, and simplifications that deviate from the spirit and principles of the present invention should be equivalent replacement methods and are included in the protection scope of the present invention.
Claims
1. A method for constructing a functional architecture of an autonomous transportation system, characterized in that: The following steps are involved: Step 1: Aiming at the problem of multi-attribute text clustering, an optimized density peak clustering model for multi-attribute text is constructed; Step 2: Calculate the word attribute frequency and the inverse document frequency of the attribute according to the characteristics of the multi-attribute text, and convert the multi-attribute text into a numerical value according to the word attribute frequency and the inverse document frequency of the attribute; Step 3: According to the numerical weights of different attribute texts, the text vector space model is used to convert the numerical values into space dimension coordinates; Step 4: Optimize the calculation method of density peak: Use Gaussian kernel method to optimize the calculation method of local density of data points; The decision value is used to optimize the selection of cluster centers, and the samples with larger decision values are selected as the density peak cluster centers to obtain the clustering results; Step 5: Based on the clustering results of step 4, use the silhouette coefficient method to evaluate the clustering results, and repeat step 4 for the data points with lower silhouette coefficients; Step 6: Verify the feasibility of the algorithm. In the road autonomous driving scenario of the autonomous transportation system, the functions under the three service domains are divided into four layers of "autonomous perception, autonomous learning, autonomous decision-making, and autonomous response" based on the final clustering results. Draw a functional architecture diagram based on the relationship between the layers.
2. The method for constructing the functional architecture of an autonomous transportation system according to claim 1, characterized in that: The specific steps of step 2 are as follows: (1) Calculate the frequency of word attributes: Among them, s represents the attribute of a text, is the number of times the attribute s appears in the document, W n (n=1,2,...,N, N is the total number of documents) represents the attribute library consisting of text attributes; (2) According to the needs of text attribute conversion, assuming that the word must appear in the attribute library, the inverse document frequency of the attribute is calculated using the smoothed inverse document frequency: (3) The calculation formula of the TF-IDF algorithm for multi-attribute text is TF·IDF(s,W n ,W)=TF(s,W n )·IDF(s,W), when the TF-IDF value is larger, it means that the word can better represent the document, that is, when the TF-IDF value is larger, it means that the attribute can better represent the text.
3. The method for constructing the functional architecture of an autonomous transportation system according to claim 1, characterized in that: The text space dimension coordinates are obtained based on the text vector space model. The text vector space model converts a given text into a vector of a certain dimension. The feature item is used as the basic unit of text representation. Each dimension of the vector corresponds to a feature item in the text, and each dimension itself represents the weight of the corresponding feature item in the text. The weight represents the importance of the feature item to the text, that is, to what extent the feature item can reflect the category of the document in which it is located.
4. The method for constructing a functional architecture of an autonomous transportation system according to claim 1, characterized in that: By optimizing the density peak clustering algorithm, we can find the clustering center that meets the conditions, cluster the remaining points to be clustered, and improve the clustering performance. The specific steps are as follows: The local density truncation kernel calculation method is optimized to the Gaussian kernel calculation method to significantly improve the precision of density calculation; the Gaussian kernel calculation method calculates the Euclidean distance and the Gaussian function value, and the sum of all point values is used as the density. The Gaussian function value defines the sample local density calculation formula as follows: Among them, d ij For sample x i and x j The distance between c is the cutoff distance, n is the number of samples; Calculate the local density distance: Calculate the minimum distance ξ from each point to a point with a higher local density value than itself i ,ξ i The calculation formula is as follows: For the sample with the highest density, ξ i Defined as ξ i =max j (d ij ), where p i For sample x i The local density, p j For sample x j The local density of The selection of cluster center points is optimized by using the decision value method: select the decision value γ i The larger sample is used as the density peak cluster center, that is, the point where the product of local density and local density distance is larger, γ i The calculation formula of γ is as follows: i =ρ i ·ξ i .
5. The method for constructing the functional architecture of an autonomous transportation system according to claim 1, characterized in that: The clustering performance is tested by the silhouette coefficient, and the data points with lower silhouette coefficients are clustered repeatedly. The specific algorithm flow is as follows: Calculate the average distance a between sample i and other samples in the same cluster i , a i is called the intra-cluster dissimilarity of sample i; a i The smaller it is, the more sample i should be clustered into this cluster. The a of all samples in cluster C i The mean is called the cluster dissimilarity of cluster C; Calculate the number of samples i to another cluster C j The average distance b of all samples i,j , called sample i and cluster C j The dissimilarity between clusters of sample i is defined as: b i = {b i1 ,b i2 ,...,b ik }, b i The larger it is, the less sample i belongs to other clusters; According to the intra-cluster dissimilarity a of sample i i and the inter-cluster dissimilarity b i , define the silhouette coefficient of sample i: Judgment: i If it is close to 1, it means that the clustering of sample i is reasonable and effective; s i If s is close to -1, it means that sample i should be classified into another cluster. i If it is approximately 0, it means that sample i is on the boundary of two clusters.
6. The method for constructing the functional architecture of an autonomous transportation system according to claim 1, characterized in that: In order to construct the functional architecture of the autonomous transportation system, the functions in the service domain under the scenario are clustered and analyzed, and the functional architecture is built based on the clustering results, and the functional architecture diagram is drawn.
7. The method for constructing the functional architecture of an autonomous transportation system according to claim 6, characterized in that: The functions of the autonomous transportation system have three attributes: provider, process object, and service object, and an attribute library is established.