CN117560104A - Construction method of interpretable machine learning-assisted channel model in mixed traffic - Google Patents
Construction method of interpretable machine learning-assisted channel model in mixed traffic Download PDFInfo
- Publication number
- CN117560104A CN117560104A CN202311505555.7A CN202311505555A CN117560104A CN 117560104 A CN117560104 A CN 117560104A CN 202311505555 A CN202311505555 A CN 202311505555A CN 117560104 A CN117560104 A CN 117560104A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- channel
- channel model
- data set
- infrastructure
- 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
- 238000010801 machine learning Methods 0.000 title claims abstract description 35
- 238000010276 construction Methods 0.000 title claims abstract description 9
- 238000004891 communication Methods 0.000 claims abstract description 91
- 238000000034 method Methods 0.000 claims abstract description 76
- 238000012549 training Methods 0.000 claims abstract description 28
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 18
- 238000012360 testing method Methods 0.000 claims abstract description 13
- 241000725585 Chicken anemia virus Species 0.000 claims description 42
- 239000013598 vector Substances 0.000 claims description 33
- 238000004458 analytical method Methods 0.000 claims description 32
- 238000003066 decision tree Methods 0.000 claims description 18
- 208000037262 Hepatitis delta Diseases 0.000 claims description 16
- 241000724709 Hepatitis delta virus Species 0.000 claims description 16
- 208000029570 hepatitis D virus infection Diseases 0.000 claims description 16
- 230000035699 permeability Effects 0.000 claims description 14
- 230000005540 biological transmission Effects 0.000 claims description 13
- 230000006870 function Effects 0.000 claims description 12
- 239000000654 additive Substances 0.000 claims description 11
- 230000000996 additive effect Effects 0.000 claims description 11
- 238000005516 engineering process Methods 0.000 claims description 7
- 238000013507 mapping Methods 0.000 claims description 7
- 239000011159 matrix material Substances 0.000 claims description 6
- 230000002159 abnormal effect Effects 0.000 claims description 4
- 238000012217 deletion Methods 0.000 claims description 3
- 230000037430 deletion Effects 0.000 claims description 3
- 230000003137 locomotive effect Effects 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 2
- 238000004088 simulation Methods 0.000 description 12
- 238000010586 diagram Methods 0.000 description 8
- 230000008901 benefit Effects 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 230000000052 comparative effect Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 206010039203 Road traffic accident Diseases 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 239000000470 constituent Substances 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000006855 networking Effects 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B17/00—Monitoring; Testing
- H04B17/30—Monitoring; Testing of propagation channels
- H04B17/391—Modelling the propagation channel
- H04B17/3912—Simulation models, e.g. distribution of spectral power density or received signal strength indicator [RSSI] for a given geographic region
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
- G08G1/0125—Traffic data processing
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W24/00—Supervisory, monitoring or testing arrangements
- H04W24/06—Testing, supervising or monitoring using simulated traffic
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
- H04W4/30—Services specially adapted for particular environments, situations or purposes
- H04W4/40—Services specially adapted for particular environments, situations or purposes for vehicles, e.g. vehicle-to-pedestrians [V2P]
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Electromagnetism (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- General Physics & Mathematics (AREA)
- Traffic Control Systems (AREA)
Abstract
A construction method of an interpretable machine learning assisted channel model in mixed traffic constructs a mixed traffic scene, and sets input characteristics and output characteristics of the channel model; constructing a mixed traffic channel data set T, and dividing the mixed traffic channel data set T into two sub data sets T= { T V ,T I Dividing training set and test set; light gbm using machine learning ML algorithm respectively for vehicle-to-vehicle V2V communication dataset T V And vehicle-to-infrastructure V2I communication data set T I Training the training set of the system, and constructing a vehicle-to-vehicle V2V channel model and a vehicle-to-infrastructure V2I channel model in a mixed traffic environment; SHAP method is additively interpreted by saprolidine, and T is utilized V And T I Respectively analyzing the V2V channel model and the V2I channel model, acquiring key features which have obvious influence on communication quality according to SHAP values, and generating an interpretable model g (x) V′ ) And g (x) I′ ) Constructing a lightweight mixed traffic channel model; the invention realizes the safety and high efficiency of traffic transportation and provides powerful support for mixed traffic communication.
Description
Technical Field
The invention belongs to the technical field of mixed traffic communication, and particularly relates to a method for constructing an interpretable machine learning-assisted channel model in mixed traffic.
Background
In recent years, with the gradual advancement of the urban process, urban traffic is under tremendous pressure, and traffic accidents are frequently happened. In particular, the continual advances in automation and intelligence technology have made automatic driving vehicle CAVs increasingly part of the traffic system. These autonomous vehicles have shown tremendous potential in improving transportation systems using advanced technologies such as sensors, high-precision maps, computer vision, and the like. According to predictions, a hybrid traffic pattern in which CAVs and human-driven vehicle HDVs coexist with each other will continue to prevail for about 30 years until the fully automatic driving era is entered.
For hybrid traffic, reliable vehicle communication is critical. Traffic safety services, intelligent driving assistance, vehicle infotainment and other applications are all independent of communication. Notably, traffic safety services have received widespread attention from researchers due to their primary concern for ensuring the personal safety of users. In vehicle communication, vehicle-to-vehicle V2V and vehicle-to-infrastructure V2I are used as main constituent forms, and information sharing between vehicles and between road infrastructure is realized by collecting sensor data, aiming at improving driving safety and traffic efficiency. Thus, the role of accurate channel models in improving communication reliability of CAVs in complex propagation environments in mixed traffic is not trivial.
On-board channel models can be generally classified into empirical models, stochastic models, and deterministic models. While empirical and stochastic models are insensitive to physical information changes in the actual propagation environment, the ray tracing method in deterministic models can acquire accurate channel characteristics through accurate environmental information and is widely applied to channel modeling in specific environments. Unfortunately, however, the computational complexity of ray tracing increases dramatically due to the high degree of randomness exhibited by wireless channels in mixed traffic, resulting in reduced efficiency, thereby limiting its effectiveness. With the rapid development of ML, researchers began to apply it to channel modeling to overcome the limitations faced by traditional channel models in terms of performance such as accuracy and efficiency. The application of multi-layer perceptron MLP, convolutional neural network CNN and random forest RF in channel modeling in urban vehicle environment was studied in article "Machine LearningApproaches for Radio Propagation Modeling in Urban Vehicular Channels" by Ahmad et al (K.Ahmad and S.Hussain, "Machine Learning Approaches for Radio Propagation Modeling in Urban Vehicular Channels," IEEE Access, vol.10, pp.113690-113698,2022.) and the effectiveness of ML technique in vehicle-mounted channel modeling was verified. Ramya et al (P.M.Ramya, M.Boban, C.Zhou, and S.Sta ń czak, "Using Learning Methods forV2V Path Loss Prediction," in 2019IEEE Wireless Communications and Networking Conference (WCNC), pp.1-6, IEEE, 2019.) in article "Using Learning Methods for V2V Path Loss Prediction," propose an RF-based path loss prediction model with V2V communication datasets under different traffic environments that exhibits superior results in performance compared to conventional log-distance path loss models. In addition, rumelhart and McClelland et al introduced a scene recognition model based on a back propagation neural network BPNN (M.Yang, B.Ai, R.He, C.Shen, M.Wen, C.Huang, J.Li, Z.Ma, L.Chen, X.Li, et al, "Machine-Learning-based Scenario Identification using Channel Characteristics in Intelligent Vehicular Communications," IEEE Transactions on Intelligent Transportation Systems, vol.22, no.7, pp.3961-3974,2020.) that utilized key channel properties for accurate scene recognition in a vehicle communication environment.
The traditional experience channel model can cause interference of various factors on a communication channel in a complex traffic environment, and the advantages of the traditional experience channel model are difficult to fully develop.
Current research has conducted channel modeling in a vehicle environment, and has made remarkable progress in support of machine learning ML technology, playing a positive role in the development of vehicle-mounted communication. However, the uniqueness of the hybrid traffic scene, marked by the integration of CAVs and HDVs, blurs the advantages of existing vehicle channel models to some extent. Moreover, due to experimental data limitations, these models are only applicable to certain proposed scenarios, but may face some challenges when attempting to extend them to other environments.
Under the complex environment of mixed traffic, the radio propagation environment has obvious randomness under the comprehensive influence of a plurality of environmental factors such as surrounding scatterers, dynamic traffic conditions, antenna configuration and the like. In addition, unpredictability and randomness of the on-board channel are further amplified due to the high speed movement of the vehicle. Notably, the vehicle antenna is typically located lower than in conventional cellular networks, while there may be shadowing of the HDVs between the CAVs. In this case, the presence of a large number of scatterers around the transmit and receive antennas causes a more serious non-line-of-sight NLoS problem (M.Yang, B.Ai, R.He, L.Chen, X.Li, J.Li, B.Zhang, C.Huang, and Z.Zhong, "A Cluster-based Three-dimensional Channel Model for Vehicle-to-vehicle Communications," IEEE transactions on vehiculartechnology, vol.68, no.6, pp.5208-5220,2019.).
Disclosure of Invention
In order to overcome the defects of the prior art, the invention aims to provide a construction method of an interpretable machine learning assisted channel model in mixed traffic, which is characterized in that a plurality of mixed traffic scenes are designed, a large amount of vehicle-to-vehicle V2V and vehicle-to-infrastructure V2I channel data are collected from the designed mixed traffic scenes, and a rich traffic channel data set is formed; training the data set by using a lightweight gradient hoist algorithm LightGBM, constructing a mixed traffic channel model, and realizing the universality modeling of a mixed traffic channel; deep analysis is carried out on the model by applying a saprolitic additive interpretation SHAP method, and a lightweight channel model based on the most critical characteristics is constructed so as to realize efficient channel characteristic prediction; the universal mixed traffic channel model provided by the invention can accurately and efficiently predict the channel characteristics in various traffic environments, can be used for improving the communication quality in different mixed traffic environments, realizes the safety and the high efficiency of traffic transportation, and provides powerful support for mixed traffic communication.
In order to achieve the above purpose, the technical scheme adopted by the invention is as follows:
a construction method of an interpretable machine learning aided channel model in mixed traffic comprises the following steps:
step 1, constructing a mixed traffic scene, and setting input features and output features of a channel model;
step 2, constructing a mixed traffic channel data set T based on the mixed traffic scene constructed in the step 1, and dividing the mixed traffic channel data set T into two sub-data sets: vehicle-to-vehicle V2V communication data set T V And vehicle-to-infrastructure V2I communication data set T I I.e. t= { T V ,T I -a }; dividing a training set and a testing set respectively;
step 3, using a machine learning ML algorithm LightGBM to respectively obtain the vehicle-to-vehicle V2V communication data set T obtained in step 2 V And vehicle-to-infrastructure V2I communication data set T I Training the training set of the system, capturing complex association and nonlinear characteristics in channel data, and constructing a vehicle-to-vehicle V2V channel model and a vehicle-to-infrastructure V2I channel model in a mixed traffic environment;
step 4, adopting a saprolidine additive interpretation SHAP method, and utilizing the vehicle-to-vehicle V2V communication data set T obtained in the step 2 V And vehicle-to-infrastructure V2I communication data set T I The test set of (2) respectively analyzes the vehicle-to-vehicle V2V channel model and the vehicle-to-infrastructure V2I channel model in the mixed traffic environment constructed in the step 3, obtains key features which have obvious influence on the communication quality according to SHAP values, and generates an interpretable model g (x V′ ) And g (x) I′ ) And constructing a lightweight mixed traffic channel model.
The specific method of the step 1 is as follows:
step 1.1, constructing a diversity traffic scene;
simulating network topology of urban, suburban and highway traffic scenes, wherein the scene construction of the urban and suburban areas is jointly determined by three statistical parameters of a ratio alpha of building floor area to total area, building density beta and building height obeying Rayleigh distribution with standard deviation gamma, and the statistical parameters of the highway buildings are all 0;
step 1.2, configuring antennas in the diversity traffic scene constructed in the step 1.1;
the transceivers on the road side equipment RSUs and the automatic driving vehicles CAVs are configured into a multi-antenna system, and key parameters are introduced at the same time: transmitting power P, transmitting antenna number N t Number of receiving antennas N r Height H of transmitting antenna t Height H of receiving antenna r An antenna array pitch d;
step 1.3, analyzing traffic conditions based on the diversity traffic scene constructed in the step 1.1, and respectively calculating traffic flow densityAnd permeability->
Step 1.4, analyzing the output channel characteristics based on the input channel characteristics of step 1.1, step 1.2 and step 1.3.
The specific method of the step 2 is as follows:
step 2.1, comprehensively acquiring channel data of vehicle-to-vehicle V2V and vehicle-to-infrastructure V2I communication to construct a mixed traffic channel data set T:
wherein,representing input feature vectors, ++>Representing the output value corresponding to the input feature vector, namely the channel capacity C;
step 2.2, willThe mixed traffic channel data set T constructed in step 2.1 is divided into two sub data sets t= { T V ,T I Vehicle-to-vehicle V2V communication data set T V And vehicle-to-infrastructure V2I communication data set T I Vehicle-to-vehicle V2V communication data set T V And vehicle-to-infrastructure V2I communication data set T I Input feature vector x in (a) V And x I The method comprises the following steps:
vehicle-to-vehicle V2V communication data set T V Input feature vector x in (a) V Feature set of (a)The method comprises the following steps:
wherein, (x) t ,y t ) And (x) r ,y r ) Coordinates of a transmitting antenna and a receiving antenna on CAVs for transmitting and receiving signals, respectively, h t And h r Representing the heights of the transmitter and the receiver, D being the transmission distance;
vehicle-to-infrastructure V2I communication data set T I Input feature vector x in (a) I Feature set of (a)The method comprises the following steps:
wherein, (x) v ,y v ) Representing the antenna coordinates on CAVs, h v And h s Representing the heights of CAVs and RSUs, respectively, N v And N s The number of antennas on CAVs and RSUs, respectively;
step 2.3, for the vehicle-to-vehicle V2V communication data set T obtained in step 2.2 V And vehicle-to-infrastructure V2I communication data set T I After abnormal value deletion and standardization processing, the training set and the test set are respectively divided intoA collection.
The specific method of the step 3 is as follows:
the specific method for constructing the vehicle-to-vehicle V2V channel model in the mixed traffic environment comprises the following steps:
step 3.1.1, slave vehicle-to-vehicle V2V communication data set T V Extracting partial data as training setPredictive value f (x) of GBDT for gradient boosting decision tree algorithm V ) From a set of decision tree models h (x V ) The representation is:
wherein W represents the number of decision trees;
step 3.1.2 the GBDT model is constructed with the aim of finding an approximation functionThereby achieving a loss function L minimization:
step 3.1.3, dividing the data according to the absolute value of the gradient, and taking a part with larger absolute value of the gradient, namely the top a multiplied by 100%, as a subset A; the rest samples are randomly sampled in a proportion of bx100%, and form a subset B, and the variance gain V of the characteristic j is calculated j (d) Dividing nodes corresponding to data points d in the set A U B:
wherein, g i a negative gradient representing the loss function output at each iteration;
step 3.1.4, after training by the lightweight gradient hoist algorithm LightGBM, obtaining an original mixed traffic channel model of the vehicle-to-vehicle V2V, namely an original vehicle-to-vehicle V2V channel model
The specific method for constructing the vehicle-to-infrastructure V2I channel model in the mixed traffic environment comprises the following steps:
step 3.2.1 communicating the data set T from the vehicle to the infrastructure V2I I Extracting partial data as training setPredictive value f (x) of GBDT for gradient boosting decision tree algorithm I ) From a set of decision tree models h (x I ) The representation is:
wherein W represents the number of decision trees;
step 3.2.2 the GBDT model is built with the aim of finding an approximation functionThereby achieving a loss function L minimization:
step 3.2.3, dividing the data according to the absolute value of the gradient, and taking a part with larger absolute value of the gradient, namely the top a multiplied by 100%, as a subset A; the rest samples are randomly sampled in a proportion of bx100%, and form a subset B, and the variance gain V of the characteristic j is calculated j (d) Corresponding to the data point d in the set A U BIs divided into nodes:
wherein, g i a negative gradient representing the loss function output at each iteration;
step 3.2.4, after training by the lightweight gradient hoist algorithm LightGBM, obtaining an original mixed traffic channel model from the vehicle to the infrastructure V2I, namely an original vehicle to the infrastructure V2I channel model
The specific method of the step 4 is as follows:
the specific method for acquiring key features and constructing the lightweight vehicle-to-vehicle V2V channel model comprises the following steps:
step 4.1.1, generating an interpretable model by an additive feature attribution technology, and generating an original vehicle-to-vehicle V2V channel modelInterpretation model g (x V′ ) The expression is as follows:
wherein phi is 0 Is a constant value, M represents the input feature vector x V′ Is phi j The SHAP method attributes the contribution of feature j to the prediction result to the saprolitic value representing feature jx V′ To simplify the later transfusionAn input feature vector, which is identical to the original input feature vector x V A mapping relation exists between the two;
step 4.1.2, using the SHAP method, utilizing the vehicle-to-vehicle V2V communication data set T obtained in step 2 V The input features of the constructed vehicle-to-vehicle V2V channel model in the step 3 are subjected to feature influence analysis, feature importance analysis and feature dependency analysis, and according to analysis results, a group of key features including a transmission distance D and a transmitter height h are selected when the vehicle-to-vehicle V2V channel is modeled t Transmitting power P, transmitting antenna number N t Number of receiving antennas N r Permeability and permeability of
The specific method for acquiring key features and constructing a lightweight vehicle-to-infrastructure V2I channel model comprises the following steps:
step 4.2.1 generating an interpretable model by additive feature-attribution technique, the raw vehicle-to-infrastructure V2I channel modelInterpretation model g (x I′ ) The expression is as follows:
wherein phi is 0 Is a constant value, M represents the input feature vector x I′ Is phi j The SHAP method attributes the contribution of feature j to the prediction result to the saprolitic value representing feature jx I′ To simplify the input feature vector, it is compared with the original input feature vector x I A mapping relation exists between the two;
step 4.2.2, using the SHAP method, communicating the data set T using the vehicle-to-infrastructure V2I obtained in step 2 I Is set of test pairs of (a)Performing feature influence analysis, feature importance analysis and feature dependency analysis on the input features of the constructed vehicle-to-infrastructure V2I channel model in the step 3, and selecting a group of key features comprising a transmission distance D, a ratio alpha of building floor area to total area and a height h of RSUs in the vehicle-to-infrastructure V2I channel modeling according to analysis results s Number of antennas N on RSUs s Transmit power P, and antenna array spacing d.
The specific method of the step 1.3 is as follows:
1.3.1 calculating the traffic density in a Mixed traffic Environment where vehicles interfere with each otherThe formula is as follows:
wherein,is the average locomotive spacing;
1.3.2 permeability in Mixed trafficThe ratio of the CAVs to the HDVs in the vehicle, that is, the ratio of the number of the automatic driving vehicle CAVs equipped with the in-vehicle communication device to the number of the entire vehicle is reflected by the following formula:
wherein N is CAV For the number of autonomous vehicles CAVs, N HDV The number of HDVs for a human driven vehicle.
The specific method of the step 1.4 is as follows:
regarding the channel capacity as the output characteristic of the channel model, the existence of N is set r Root receive antenna and N t Root emissionIn the case of antennas, the capacity C of the multiple-input multiple-output MIMO channel is calculated:
wherein,represents N r The real identity matrix of the order, ρ represents the average signal-to-noise ratio SNR of each receiving branch, H is N r ×N t Channel matrix of (H) * Is the conjugate transpose of H.
Compared with the prior art, the invention has the beneficial effects that:
first, the present invention encompasses radio propagation environments with different urban features, traffic environments and antenna configurations in mixed traffic, and the combined consideration of these factors helps to improve the applicability and universality of the proposed mixed traffic channel model.
Secondly, the mixed traffic environment of the invention widely considers various characteristics affecting communication quality including building height, antenna number, vehicle spacing and permeability, and collects a large amount of V2V and V2I communication data, thereby forming a rich channel data set and providing powerful support for constructing a mixed traffic channel model.
Thirdly, the mixed traffic channel model enabled by the LightGBM constructed by the invention can excavate the mapping relation between the channel capacity and the mixed traffic environment characteristics, and can realize high-accuracy channel capacity prediction in various different environments.
Fourth, the invention determines key characteristics which have significant influence on the channel capacity through the SHAP method interpretability analysis, creates a lightweight mixed traffic channel model based on the key characteristics, and provides a feasible method for accurately and efficiently predicting the characteristics of the mixed traffic channel.
In summary, the invention provides a method for constructing an interpretable machine learning assisted channel model in mixed traffic, which comprises the steps of constructing a mixed traffic environment, collecting data, training by Machine Learning (ML), and performing feature analysis and performance verification by using saprolidine for additively interpreting SHAP.
Drawings
FIG. 1 is a flow chart of an implementation of the present invention.
Fig. 2 is a schematic diagram of a hybrid traffic scenario of the present invention.
FIG. 3 is a schematic diagram of a network topology simulation of three common traffic scenarios of the present invention; fig. 3 (a) is a schematic diagram of network topology simulation of urban traffic scene, fig. 3 (b) is a schematic diagram of network topology simulation of suburban traffic scene, and fig. 3 (c) is a schematic diagram of network topology simulation of highway traffic scene.
Fig. 4 is a comparative simulation diagram of the feature impact analysis of the V2V and V2I channel models using SHAP according to the present invention, where fig. 4 (a) is a SHAP value distribution diagram of all input features in V2V in a mixed traffic channel model test set, and fig. 4 (b) is a SHAP value distribution diagram of all input features in V2I in a mixed traffic channel model test set.
Fig. 5 is a comparative simulation of feature importance analysis using SHAP according to the present invention, wherein fig. 5 (a) is the average SHAP value in V2V and fig. 5 (b) is the average SHAP value in V2I.
FIG. 6 is a comparative simulation graph of the invention analyzing the performance of different channel models.
Detailed Description
The technical scheme adopted by the invention is described in detail below.
The invention aims to solve the problem that the traditional empirical channel model is difficult to competence in the mixed traffic environment. For this purpose, an interpretable machine learning ML-assisted channel model in mixed traffic is built, which takes into account various influencing factors such as urban environment, traffic conditions and antenna configuration. The implementation scheme is as follows: multiple mixed traffic scenes are designed to cover different city characteristics, antenna configurations and traffic conditions; collecting a large amount of V2V and V2I channel data from a designed mixed traffic scene to form a rich traffic channel data set; training the data set by using a lightweight gradient hoist algorithm LightGBM, constructing a mixed traffic channel model, and realizing the universality modeling of a mixed traffic channel; deep analysis is carried out on the model by applying a saprolitic additive interpretation SHAP method, and a lightweight channel model based on the most critical features is constructed so as to realize efficient channel characteristic prediction.
As shown in FIG. 2, the scene used by the invention fuses CAVs and HDVs, and simultaneously comprehensively considers V2V communication and V2I communication; under the condition of CAVs, the transmitter and the receiver are arranged at the top position of the vehicle, and the height difference of different vehicle types is considered, so that the height of the antenna is correspondingly adjusted, the accuracy and universality of a channel model are ensured, and the key point is to construct a channel data set of a credible mixed traffic scene; for this purpose, the accuracy of the channel characteristics in the dataset and the comprehensiveness of the covered propagation environment must be ensured, which is a process that takes into account various key factors such as scene diversity, antenna configuration and traffic conditions.
The embodiment builds the universal mixed traffic channel model based on the mixed traffic scene so as to accurately and efficiently predict the characteristics of the mixed traffic channel and improve the communication quality.
Specifically, as shown in fig. 1, a method for constructing an interpretable machine learning-assisted channel model in mixed traffic includes the following steps:
step 1, constructing a mixed traffic scene, and setting input features and output features of a channel model; ensuring the accuracy of the channel characteristics in the dataset and the comprehensiveness of the covered propagation environment, this process takes into full account various key factors such as scene diversity, antenna configuration and traffic conditions.
Step 1.1, constructing a diversity traffic scene;
using ray tracing software Wireless institute to simulate the network topology of three common traffic scenes, covering urban areas, suburban areas and highways, and displaying simulation results as shown in figure 3; the scene construction of urban and suburban areas is consistent with a standardized urban model ITU-RRec.P.1410, and is jointly determined by three statistical parameters, including: the ratio α of the building floor area to the total area, the building density β, and the building height following the rayleigh distribution with standard deviation γ are all 0 for the statistical parameters of the highway in view of its general distance from the building group;
step 1.2, configuring antennas in the diversity traffic scene constructed in the step 1.1;
in view of the significant advantages of the MIMO technology in terms of improving data throughput and communication reliability, both the transceivers on the road side devices RSUs and the autonomous vehicles CAVs are configured as multi-antenna systems, while introducing key parameters: transmitting power P, transmitting antenna number N t Number of receiving antennas N r Height H of transmitting antenna t Height H of receiving antenna r And antenna array spacing d, in vehicle-to-infrastructure V2I communications, the influence of the antenna height at the vehicle roof on the communication quality is limited in view of the significant height advantage of road side devices RSUs over vehicles. In this context, the invention only has to be focused on taking into account the antenna height H located on the road side devices RSUs RSU ；
Step 1.3, analyzing traffic conditions based on the diversity traffic scene constructed in the step 1.1, and respectively calculating traffic flow densityAnd permeability->
1.3.1 calculating the traffic density in a Mixed traffic Environment where vehicles interfere with each otherThe formula is as follows:
wherein the method comprises the steps of，Is the average locomotive spacing;
1.3.2 permeability in Mixed trafficThe ratio of the CAVs to the HDVs in the vehicle, that is, the ratio of the number of the automatic driving vehicle CAVs equipped with the in-vehicle communication device to the number of the entire vehicle is reflected by the following formula:
wherein N is CAV For the number of autonomous vehicles CAVs, N HDV Number of HDVs for human driven vehicles;
1.3.3 in the vehicle-to-vehicle V2V communication, when two autonomous vehicles CAVs communicate, the human-driven vehicle HDVs may act as a main intermediate obstacle or surrounding scatterer, which has a significant influence on the communication quality, and the permeability needs to be considered as a key feature; in contrast, in the vehicle-to-infrastructure V2I communication scenario involving communication of the autonomous vehicle CAVs with the roadside apparatuses RSUs, there is no difference in communication influence between the human-driven vehicle HDVs and other autonomous vehicles CAVs, without taking into consideration the feature of the vehicle permeability.
Step 1.4, analyzing the output channel characteristics based on the input channel characteristics of step 1.1, step 1.2 and step 1.3;
regarding the channel capacity as the output characteristic of the channel model, the existence of N is set r Root receive antenna and N t In the case of a root transmit antenna, the capacity C of the MIMO channel is calculated:
wherein,represents N r The real identity matrix of the order, ρ represents the average signal-to-noise ratio SNR of each receiving branch, H is N r ×N t Channel matrix of (H) * Is the conjugate transpose of H.
Step 2, constructing a mixed traffic channel data set T based on the mixed traffic scene constructed in the step 1, and dividing the mixed traffic channel data set T into two sub-data sets: vehicle-to-vehicle V2V communication data set T V And vehicle-to-infrastructure V2I communication data set T I I.e. t= { T V ,T I -a }; dividing a training set and a testing set respectively;
step 2.1, comprehensively acquiring channel data of vehicle-to-vehicle V2V and vehicle-to-infrastructure V2I communication to construct a mixed traffic channel data set T:
wherein,representing input feature vectors, ++>Representing the output value corresponding to the input feature vector, namely the channel capacity C;
step 2.2, dividing the mixed traffic channel data set T constructed in step 2.1 into two sub data sets t= { T V ,T I Vehicle-to-vehicle V2V communication data set T V And vehicle-to-infrastructure V2I communication data set T I Vehicle-to-vehicle V2V communication data set T V And vehicle-to-infrastructure V2I communication data set T I Input feature vector x in (a) V And x I The method comprises the following steps:
vehicle-to-vehicle V2V communication data set T V Input feature vector x in (a) V Feature set of (a)The method comprises the following steps:
wherein, (x) t ,y t ) And (x) r ,y r ) Coordinates of a transmitting antenna and a receiving antenna on CAVs for transmitting and receiving signals, respectively, h t And h r Representing the heights of the transmitter and the receiver, D being the transmission distance;
vehicle-to-infrastructure V2I communication data set T I Input feature vector x in (a) I Feature set F XI The method comprises the following steps:
wherein, (x) v ,y v ) Representing the antenna coordinates on CAVs, h v And h s Representing the heights of CAVs and RSUs, respectively, N v And N s The number of antennas on CAVs and RSUs, respectively;
step 2.3, for the vehicle-to-vehicle V2V communication data set T obtained in step 2.2 V And vehicle-to-infrastructure V2I communication data set T I After abnormal value deletion and standardization treatment, the abnormal value is divided into a training set and a testing set according to the proportion of 9:1.
Step 3, using a machine learning ML algorithm LightGBM to respectively obtain the vehicle-to-vehicle V2V communication data set T obtained in step 2 V And vehicle-to-infrastructure V2I communication data set T I Capturing complex correlations and non-linear features in the channel data, constructing highly reliable vehicle-to-vehicle V2V and vehicle-to-infrastructure V2I channel models in a mixed traffic environment,
the specific method for constructing the vehicle-to-vehicle V2V channel model in the mixed traffic environment comprises the following steps:
step 3.1.1, slave vehicle-to-vehicle V2V communication data set T V 90% of the data are extracted as training setPredictive value f (x) of GBDT for gradient boosting decision tree algorithm V ) From a set of decision tree models h (x V ) The representation is:
wherein W represents the number of decision trees;
step 3.1.2 the GBDT model is constructed with the aim of finding an approximation functionThereby achieving a loss function L minimization:
step 3.1.3, dividing the data according to the absolute value of the gradient, and taking a part with larger absolute value of the gradient, namely the top a multiplied by 100%, as a subset A; the rest samples are randomly sampled in a proportion of bx100%, and form a subset B, and the variance gain V of the characteristic j is calculated j (d) Dividing nodes corresponding to data points d in the set A U B:
wherein, g i a negative gradient representing the loss function output at each iteration;
step 3.1.4, after training by the lightweight gradient hoist algorithm LightGBM, obtaining an original mixed traffic channel model of the vehicle-to-vehicle V2V, namely an original vehicle-to-vehicle V2V channel model
The specific method for constructing the vehicle-to-infrastructure V2I channel model in the mixed traffic environment comprises the following steps:
step 3.2.1 communicating the data set T from the vehicle to the infrastructure V2I I Extracting partial data as training setPredictive value f (x) of GBDT for gradient boosting decision tree algorithm I ) From a set of decision tree models h (x I ) The representation is:
wherein W represents the number of decision trees;
step 3.2.2 the GBDT model is built with the aim of finding an approximation functionThereby achieving a loss function L minimization:
step 3.2.3, dividing the data according to the absolute value of the gradient, and taking a part with larger absolute value of the gradient, namely the top a multiplied by 100%, as a subset A; the rest samples are randomly sampled in a proportion of bx100%, and form a subset B, and the variance gain V of the characteristic j is calculated j (d) Dividing nodes corresponding to data points d in the set A U B:
wherein, g i a negative gradient representing the loss function output at each iteration;
step 3.2.4, after training by the lightweight gradient hoist algorithm LightGBM, obtaining an original mixed traffic channel model from the vehicle to the infrastructure V2I, namely an original vehicle to the infrastructure V2I channel model
Step 4, adopting a saprolidine additive interpretation SHAP method, and utilizing the vehicle-to-vehicle V2V communication data set T obtained in the step 2 V And vehicle-to-infrastructure V2I communication data set T I The test set of (2) respectively analyzes the vehicle-to-vehicle V2V channel model and the vehicle-to-infrastructure V2I channel model in the mixed traffic environment constructed in the step 3, obtains key features which have obvious influence on the communication quality according to SHAP values, and generates an interpretable model g (x V′ ) And g (x) I′ ) A more lightweight hybrid traffic channel model is constructed.
The specific method for acquiring key features and constructing the lightweight vehicle-to-vehicle V2V channel model comprises the following steps:
step 4.1.1, generating an interpretable model by an additive feature attribution technology, and generating an original vehicle-to-vehicle V2V channel modelInterpretation model g (x V′ ) The expression is as follows:
/>
wherein phi is 0 Is a constant value, M represents the input feature vector x V′ Is phi j The SHAP method attributes the contribution of feature j to the prediction result to the saprolitic value representing feature jx V′ To simplify the input feature vector, it is compared with the original input feature vector x V A mapping relation exists between the two;
step 4.1.2, in the simulation software, using the SHAP method, utilizing the vehicle-to-vehicle V2V communication data set T obtained in the step 2 V The input features of the constructed vehicle-to-vehicle V2V channel model in the step 3 are subjected to feature influence analysis, feature importance analysis and feature dependency analysis, and according to analysis results, a group of key features including a transmission distance D and a transmitter height h are selected when the vehicle-to-vehicle V2V channel is modeled t Transmitting power P, transmitting antenna number N t Number of receiving antennas N r Permeability and permeability of
The specific method for acquiring key features and constructing a lightweight vehicle-to-infrastructure V2I channel model comprises the following steps:
step 4.2.1 generating an interpretable model by additive feature-attribution technique, the raw vehicle-to-infrastructure V2I channel modelInterpretation model g (x I′ ) The expression is as follows:
wherein phi is 0 Is a constant value, M represents the input feature vector x I′ Is phi j The SHAP method attributes the contribution of feature j to the prediction result to the saprolitic value representing feature jx I′ To simplify the input feature vector, it is compared with the original input feature vector x I A mapping relation exists between the two;
step 4.2.2 usingSHAP method, vehicle-to-infrastructure V2I communication data set T obtained by step 2 I The input characteristics of the constructed vehicle-to-infrastructure V2I channel model in the step 3 are subjected to characteristic influence analysis, characteristic importance analysis and characteristic dependency analysis, and a group of key characteristics are selected in the process of carrying out vehicle-to-infrastructure V2I channel modeling according to analysis results, wherein the key characteristics comprise a transmission distance D, a ratio alpha of the occupied area of a building to the total area and the height h of RSUs s Number of antennas N on RSUs s Transmit power P, and antenna array spacing d.
In addition, to better conform to the actual situation, the invention uses the Wireless Insite software to simulate:
three mixed traffic scenarios, urban, suburban and highway, were constructed in a Wireless institute, as shown in fig. 3. Wherein, the occupied areas of the urban area and the suburban area are the same and are 200m multiplied by 200m; whereas the footprint of the highway is 50m x 400m. In each simulation scenario, a MIMO antenna array RSU is randomly placed on the road side. In addition, a large number of CAVs and HDVs are placed in an equally spaced manner in each scene, with transceiver antennas being configured on the CAVs to enable them to communicate with other CAVs or RSUs. Based on the 802.11p standard, the bandwidth is 30MHz, the carrier frequency is 5.9GHz, and all the characteristic value ranges covered by the mixed traffic channel model are shown in table 1:
TABLE 1 simulation key parameters
With the above simulation scenario and simulation conditions, the present invention simulates the V2V and V2I channel models in the scenario of fig. 3 using SHAP to perform feature impact analysis, respectively. The result is shown in FIG. 4, in the V2V channel model, the four most influential features are the transmitter abscissa x t TransmitterHeight h t Distance D and receiver ordinate y r The method comprises the steps of carrying out a first treatment on the surface of the In the V2I channel model, the five features that have the most significant impact on channel capacity are respectively the abscissa x of the transmission distance D, CAVs v Ordinate y of CAVs v A ratio of building floor area to total area, a, and a height h of RSUs s 。
In addition, the present invention was simulated using SHAP for feature importance analysis in the scenario of fig. 3, and the results are shown in fig. 5. As can be seen from fig. 5, in the V2V communication field, the transmitter height h is in addition to the coordinates of the vehicle t Transmission distance D, transmission power P, number of transmitting antennas N t Permeability and permeability ofThe key influencing characteristics are formed by factors such as the like; in V2I communication, in addition to taking into account the coordinates of the vehicle and the RSU, the transmission distance D, the ratio α of the building footprint to the total area, the height h of the RSU s Number of antennas N on RSU s And the transmit power P are considered critical features.
Finally, the invention simulates the performance of different channel models in the scene of fig. 3, and examines the root mean square error RMSE and the prediction duration result, and the result is shown in fig. 6.
As can be seen from fig. 6, the conventional empirical channel model is poor in prediction accuracy and calculation efficiency because it is insensitive to physical information changes in the actual propagation environment. The two mixed traffic channel models provided by the invention are similar to the urban vehicle-mounted channel model based on ML in terms of calculation efficiency, but are obviously superior to the urban vehicle-mounted channel model based on ML in terms of prediction accuracy. The ML-based urban on-board channel model performs poorly in terms of prediction accuracy, mainly because it is limited to a limited data set and is difficult to expand in different environments. In summary, the hybrid traffic channel model of the present invention has excellent accuracy and high efficiency, and can reliably predict channel characteristics in various hybrid traffic scenes.
The invention considers various influencing factors such as urban environment, traffic condition, antenna configuration and the like, designs various mixed traffic environments, builds a rich traffic channel data set, creatively designs an ML model with resolvable property, can be suitable for different traffic environments, and realizes the universal modeling of the mixed traffic channels. On the basis, the invention constructs a lightweight channel model which is composed of the most critical features so as to realize efficient channel characteristic prediction.
Claims (7)
1. The construction method of the interpretable machine learning assisted channel model in the mixed traffic is characterized by comprising the following steps of: the method comprises the following steps:
step 1, constructing a mixed traffic scene, and setting input features and output features of a channel model;
step 2, constructing a mixed traffic channel data set T based on the mixed traffic scene constructed in the step 1, and dividing the mixed traffic channel data set T into two sub-data sets: vehicle-to-vehicle V2V communication data set T V And vehicle-to-infrastructure V2I communication data set T I I.e. t= { T V ,T I -a }; dividing a training set and a testing set respectively;
step 3, using a machine learning ML algorithm LightGBM to respectively obtain the vehicle-to-vehicle V2V communication data set T obtained in step 2 V And vehicle-to-infrastructure V2I communication data set T I Training the training set of the system, capturing complex association and nonlinear characteristics in channel data, and constructing a vehicle-to-vehicle V2V channel model and a vehicle-to-infrastructure V2I channel model in a mixed traffic environment;
step 4, adopting a saprolidine additive interpretation SHAP method, and utilizing the vehicle-to-vehicle V2V communication data set T obtained in the step 2 V And vehicle-to-infrastructure V2I communication data set T I The test set of (2) respectively analyzes the vehicle-to-vehicle V2V channel model and the vehicle-to-infrastructure V2I channel model in the mixed traffic environment constructed in the step 3, obtains key features which have obvious influence on the communication quality according to SHAP values, and generates an interpretable model g (x V′ ) And g (x) I′ ) And constructing a lightweight mixed traffic channel model.
2. The method for constructing an interpretable machine learning-assisted channel model in mixed traffic according to claim 1, wherein: the specific method of the step 1 is as follows:
step 1.1, constructing a diversity traffic scene;
simulating network topology of urban, suburban and highway traffic scenes, wherein the scene construction of the urban and suburban areas is jointly determined by three statistical parameters of a ratio alpha of building floor area to total area, building density beta and building height obeying Rayleigh distribution with standard deviation gamma, and the statistical parameters of the highway buildings are all 0;
step 1.2, configuring antennas in the diversity traffic scene constructed in the step 1.1;
the transceivers on the road side equipment RSUs and the automatic driving vehicles CAVs are configured into a multi-antenna system, and key parameters are introduced at the same time: transmitting power P, transmitting antenna number N t Number of receiving antennas N r Height H of transmitting antenna t Height H of receiving antenna r An antenna array pitch d;
step 1.3, analyzing traffic conditions based on the diversity traffic scene constructed in the step 1.1, and respectively calculating traffic flow densityAnd permeability->
Step 1.4, analyzing the output channel characteristics based on the input channel characteristics of step 1.1, step 1.2 and step 1.3.
3. The method for constructing an interpretable machine learning-assisted channel model in mixed traffic according to claim 1, wherein: the specific method of the step 2 is as follows:
step 2.1, comprehensively acquiring channel data of vehicle-to-vehicle V2V and vehicle-to-infrastructure V2I communication to construct a mixed traffic channel data set T:
wherein,representing input feature vectors, ++>Representing the output value corresponding to the input feature vector, namely the channel capacity C;
step 2.2, dividing the mixed traffic channel data set T constructed in step 2.1 into two sub data sets t= { T V ,T I Vehicle-to-vehicle V2V communication data set T V And vehicle-to-infrastructure V2I communication data set T I Vehicle-to-vehicle V2V communication data set T V And vehicle-to-infrastructure V2I communication data set T I Input feature vector x in (a) V And x I The method comprises the following steps:
vehicle-to-vehicle V2V communication data set T V Input feature vector x in (a) V Feature set of (a)The method comprises the following steps:
wherein, (x) t ,y t ) And (x) r ,y r ) Coordinates of a transmitting antenna and a receiving antenna on CAVs for transmitting and receiving signals, respectively, h t And h r Representing the heights of the transmitter and the receiver, D being the transmission distance;
vehicle-to-infrastructure V2I communication data set T I Input feature vector x in (a) I Feature set of (a)The method comprises the following steps:
wherein, (x) v ,y v ) Representing the antenna coordinates on CAVs, h v And h s Representing the heights of CAVs and RSUs, respectively, N v And N s The number of antennas on CAVs and RSUs, respectively;
step 2.3, for the vehicle-to-vehicle V2V communication data set T obtained in step 2.2 V And vehicle-to-infrastructure V2I communication data set T I After abnormal value deletion and standardization processing, the training set and the testing set are respectively divided.
4. The method for constructing an interpretable machine learning-assisted channel model in mixed traffic according to claim 1, wherein: the specific method of the step 3 is as follows:
the specific method for constructing the vehicle-to-vehicle V2V channel model in the mixed traffic environment comprises the following steps:
step 3.1.1, slave vehicle-to-vehicle V2V communication data set T V Extracting partial data as training setPredictive value f (x) of GBDT for gradient boosting decision tree algorithm V ) From a set of decision tree models h (x V ) The representation is:
wherein W represents the number of decision trees;
step 3.1.2 the GBDT model is constructed with the aim of finding an approximation functionThereby realizing the loss function LmostAnd (3) miniaturization:
step 3.1.3, dividing the data according to the absolute value of the gradient, and taking a part with larger absolute value of the gradient, namely the top a multiplied by 100%, as a subset A; the rest samples are randomly sampled in a proportion of bx100%, and form a subset B, and the variance gain V of the characteristic j is calculated j (d) Dividing nodes corresponding to data points d in the set A U B:
wherein, g i a negative gradient representing the loss function output at each iteration;
step 3.1.4, after training by the lightweight gradient hoist algorithm LightGBM, obtaining an original mixed traffic channel model of the vehicle-to-vehicle V2V, namely an original vehicle-to-vehicle V2V channel model
The specific method for constructing the vehicle-to-infrastructure V2I channel model in the mixed traffic environment comprises the following steps:
step 3.2.1 communicating the data set T from the vehicle to the infrastructure V2I I Extracting partial data as training setPredictive value f (x) of GBDT for gradient boosting decision tree algorithm I ) From a set of decision tree models h (x I ) The representation is:
wherein W represents the number of decision trees;
step 3.2.2 the GBDT model is built with the aim of finding an approximation functionThereby achieving a loss function L minimization:
step 3.2.3, dividing the data according to the absolute value of the gradient, and taking a part with larger absolute value of the gradient, namely the top a multiplied by 100%, as a subset A; the rest samples are randomly sampled in a proportion of bx100%, and form a subset B, and the variance gain V of the characteristic j is calculated j (d) Dividing nodes corresponding to data points d in the set A U B:
wherein, g i a negative gradient representing the loss function output at each iteration;
step 3.2.4, after training by the lightweight gradient hoist algorithm LightGBM, obtaining an original mixed traffic channel model from the vehicle to the infrastructure V2I, namely an original vehicle to the infrastructure V2I channel model
5. The method for constructing an interpretable machine learning-assisted channel model in mixed traffic according to claim 1, wherein: the specific method of the step 4 is as follows:
the specific method for acquiring key features and constructing the lightweight vehicle-to-vehicle V2V channel model comprises the following steps:
step 4.1.1, generating an interpretable model by an additive feature attribution technology, and generating an original vehicle-to-vehicle V2V channel modelInterpretation model g (x V′ ) The expression is as follows:
wherein phi is 0 Is a constant value, M represents the input feature vector x V′ Is phi j The SHAP method attributes the contribution of feature j to the prediction result to the saprolitic value representing feature jx V′ To simplify the input feature vector, it is compared with the original input feature vector x V A mapping relation exists between the two;
step 4.1.2, using the SHAP method, utilizing the vehicle-to-vehicle V2V communication data set T obtained in step 2 V The input features of the constructed vehicle-to-vehicle V2V channel model in the step 3 are subjected to feature influence analysis, feature importance analysis and feature dependency analysis, and according to analysis results, a group of key features including a transmission distance D and a transmitter height h are selected when the vehicle-to-vehicle V2V channel is modeled t Transmitting power P, transmitting antenna number N t Number of receiving antennas N r Permeability and permeability of
The specific method for acquiring key features and constructing a lightweight vehicle-to-infrastructure V2I channel model comprises the following steps:
step 4.2.1 generating an interpretable model by additive feature-attribution technique, the raw vehicle-to-infrastructure V2I channel modelInterpretation model g (x I′ ) The expression is as follows:
wherein phi is 0 Is a constant value, M represents the input feature vector x I′ Is phi j The SHAP method attributes the contribution of feature j to the prediction result to the saprolitic value representing feature jx I′ To simplify the input feature vector, it is compared with the original input feature vector x I A mapping relation exists between the two;
step 4.2.2, using the SHAP method, communicating the data set T using the vehicle-to-infrastructure V2I obtained in step 2 I The input features of the constructed vehicle-to-infrastructure V2I channel model in the step 3 are subjected to feature influence analysis, feature importance analysis and feature dependency analysis, and a group of key features are selected in the vehicle-to-infrastructure V2I channel modeling according to analysis results, wherein the key features comprise a transmission distance D, a ratio alpha of the building floor area to the total area and a height h of RSUs s Number of antennas N on RSUs s Transmit power P, and antenna array spacing d.
6. The method for constructing an interpretable machine learning-assisted channel model in hybrid traffic of claim 2, wherein: the specific method of the step 1.3 is as follows:
1.3.1 calculating the traffic density in a Mixed traffic Environment where vehicles interfere with each otherThe formula is as follows:
wherein,is the average locomotive spacing;
1.3.2 permeability in Mixed trafficThe ratio of the CAVs to the HDVs in the vehicle, that is, the ratio of the number of the automatic driving vehicle CAVs equipped with the in-vehicle communication device to the number of the entire vehicle is reflected by the following formula:
wherein N is CAV For the number of autonomous vehicles CAVs, N HDV The number of HDVs for a human driven vehicle.
7. The method for constructing an interpretable machine learning-assisted channel model in hybrid traffic of claim 2, wherein: the specific method of the step 1.4 is as follows:
regarding the channel capacity as the output characteristic of the channel model, the existence of N is set r Root receive antenna and N t In the case of a root transmit antenna, the capacity C of the MIMO channel is calculated:
wherein,represents N r The real identity matrix of the order, ρ represents the average signal-to-noise ratio SNR of each receiving branch, H is N r ×N t Channel matrix of (H) * Is the conjugate transpose of H.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311505555.7A CN117560104A (en) | 2023-11-13 | 2023-11-13 | Construction method of interpretable machine learning-assisted channel model in mixed traffic |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311505555.7A CN117560104A (en) | 2023-11-13 | 2023-11-13 | Construction method of interpretable machine learning-assisted channel model in mixed traffic |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117560104A true CN117560104A (en) | 2024-02-13 |
Family
ID=89822655
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311505555.7A Pending CN117560104A (en) | 2023-11-13 | 2023-11-13 | Construction method of interpretable machine learning-assisted channel model in mixed traffic |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117560104A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118114574A (en) * | 2024-04-26 | 2024-05-31 | 安徽大学 | Vegetation growth attribution analysis method and system using interpretable machine learning |
-
2023
- 2023-11-13 CN CN202311505555.7A patent/CN117560104A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118114574A (en) * | 2024-04-26 | 2024-05-31 | 安徽大学 | Vegetation growth attribution analysis method and system using interpretable machine learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Sun et al. | Fingerprint-based localization for massive MIMO-OFDM system with deep convolutional neural networks | |
| CN105790813B (en) | Code book selection method based on deep learning under a kind of extensive MIMO | |
| Zhou et al. | Deep-learning-based spatial–temporal channel prediction for smart high-speed railway communication networks | |
| CN110166154B (en) | Software radio frequency spectrum monitoring and identifying method based on neural network | |
| Yang et al. | Machine-learning-based scenario identification using channel characteristics in intelligent vehicular communications | |
| CN110446160B (en) | Deep learning method for vehicle position estimation based on multipath channel state information | |
| CN117560104A (en) | Construction method of interpretable machine learning-assisted channel model in mixed traffic | |
| Lee et al. | Path loss exponent prediction for outdoor millimeter wave channels through deep learning | |
| WO2022021602A1 (en) | 5g reference signal received power prediction method based on geographic information | |
| CN113435247A (en) | Intelligent identification method, system and terminal for communication interference | |
| CN113283169B (en) | Three-dimensional group exploration method based on multi-head attention asynchronous reinforcement learning | |
| Ding et al. | Digital-twin-enabled city-model-aware deep learning for dynamic channel estimation in urban vehicular environments | |
| Wang et al. | Multi-classification of UWB signal propagation channels based on one-dimensional wavelet packet analysis and CNN | |
| CN114925720A (en) | Small sample modulation signal identification method based on space-time mixed feature extraction network | |
| CN113037668B (en) | Millimeter wave point-to-point communication channel equalization method | |
| CN113114399B (en) | Three-dimensional spectrum situation complementing method and device based on generation countermeasure network | |
| CN110290466A (en) | Floor method of discrimination, device, equipment and computer storage medium | |
| CN112867021A (en) | Improved TrAdaBoost-based indoor positioning method for transfer learning | |
| CN114724245B (en) | Incremental learning human body action recognition method based on CSI | |
| CN112566117B (en) | Vehicle node identity recognition method and device based on metric learning | |
| Wu et al. | Enhanced path loss model by image-based environmental characterization | |
| CN113596757A (en) | Rapid high-precision indoor fingerprint positioning method based on integrated width learning | |
| CN115002703A (en) | Passive indoor people number detection method based on Wi-Fi channel state information | |
| Nagao et al. | A study on path loss modeling using ResNet and pre-training with free space path loss | |
| Yang et al. | SNR estimation of UAV control signal based on convolutional neural network |
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 |