CN113012424B - Dynamic evolution prediction method for open type unmanned vehicle group in expressway scene - Google Patents

Dynamic evolution prediction method for open type unmanned vehicle group in expressway scene Download PDF

Info

Publication number
CN113012424B
CN113012424B CN202110195975.4A CN202110195975A CN113012424B CN 113012424 B CN113012424 B CN 113012424B CN 202110195975 A CN202110195975 A CN 202110195975A CN 113012424 B CN113012424 B CN 113012424B
Authority
CN
China
Prior art keywords
vehicle group
unmanned vehicle
vehicle
group
unmanned
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202110195975.4A
Other languages
Chinese (zh)
Other versions
CN113012424A (en
Inventor
程久军
魏超
原桂远
毛其超
刘登程
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tongji University
Original Assignee
Tongji University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Tongji University filed Critical Tongji University
Priority to CN202110195975.4A priority Critical patent/CN113012424B/en
Publication of CN113012424A publication Critical patent/CN113012424A/en
Application granted granted Critical
Publication of CN113012424B publication Critical patent/CN113012424B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/20—Monitoring the location of vehicles belonging to a group, e.g. fleet of vehicles, countable or determined number of vehicles
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/243—Classification techniques relating to the number of classes
    • G06F18/2431—Multiple classes
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/243—Classification techniques relating to the number of classes
    • G06F18/24323—Tree-organised classifiers
    • 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]
    • H04W4/44—Services specially adapted for particular environments, situations or purposes for vehicles, e.g. vehicle-to-pedestrians [V2P] for communication between vehicles and infrastructures, e.g. vehicle-to-cloud [V2C] or vehicle-to-home [V2H]
    • 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]
    • H04W4/46—Services specially adapted for particular environments, situations or purposes for vehicles, e.g. vehicle-to-pedestrians [V2P] for vehicle-to-vehicle communication [V2V]
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W84/00—Network topologies
    • H04W84/18—Self-organising networks, e.g. ad-hoc networks or sensor networks

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Traffic Control Systems (AREA)

Abstract

The purpose of the invention is as follows: how to predict the following vehicle group movement behaviors according to the dynamic evolution of the unmanned vehicle group is an urgent problem to be solved to ensure that the unmanned vehicle group movement behaviors are always stable and orderly. The invention discloses a dynamic evolution prediction method of an open unmanned vehicle group in a highway scene, which comprises the following steps: and combining a highway scene, firstly, extracting the characteristics of the structure of the unmanned vehicle group, then generating a sample according to the generated evolution event, and finally predicting the to-be-generated evolution event of the unmanned vehicle group by using a classification prediction method. The problem is solved, the subsequent evolution behavior of the unmanned vehicle cluster can be accurately predicted, the establishment of a new vehicle cluster or a vehicle cluster can be reasonably guided, the movement behavior of the unmanned vehicle cluster is ensured to be stable and orderly, and the unmanned vehicle can be widely applied to the expressway.

Description

Dynamic evolution prediction method for open type unmanned vehicle group in expressway scene
Technical Field
The present invention relates to the field of unmanned driving.
Background
The prior art of the existing internet of vehicles interconnection:
currently, the internet of vehicles refers to vehicles driven by people. Related researchers have conducted related research in the aspects of Vehicular Ad-hoc networks (VANET), Infrastructure-based Vehicular networking networks (VINET), and VANET and VINET hybrid networks.
The internet of vehicles (manned) is from the perspective of the information field, and does not consider external factors of the surrounding environment, such as interference of obstacles, manned vehicle bodies (here, moving obstacle nodes), traffic lights, and the like, with respect to direct information interaction between vehicles, vehicle-road infrastructure, and background servers.
Disclosure of Invention
In view of the above problems, the inventor of chengdu et al applies for the dynamic evolution method of the open unmanned vehicle group in the expressway scene at 12/16/2020 (applicant: university of Tongji, patent application No. 202011484672.6): researching a meta vehicle group sequence based on vehicle group similarity, microscopically analyzing node change events such as appearance, addition, disappearance and departure of nodes, and macroscopically analyzing unmanned vehicle group evolution events including vehicle group formation, death, survival, splitting and merging events, and providing an unmanned vehicle group dynamic evolution method; this is disclosed in example 1.
The purpose of the invention is as follows: how to predict the following vehicle group movement behaviors according to the dynamic evolution of the unmanned vehicle group is an urgent problem to be solved to ensure that the unmanned vehicle group movement behaviors are always stable and orderly.
Based on the embodiment 1, the invention further discloses a method for predicting the motion behavior of the unmanned vehicle group after dynamic evolution, which comprises the following steps: and combining a highway scene, firstly, extracting the characteristics of the structure of the unmanned vehicle group, then generating a sample according to the generated evolution event, and finally predicting the to-be-generated evolution event of the unmanned vehicle group by using a classification prediction method. The problem is solved, the subsequent evolution behavior of the unmanned vehicle cluster can be accurately predicted, the establishment of a new vehicle cluster or a vehicle cluster can be reasonably guided, the movement behavior of the unmanned vehicle cluster is ensured to be stable and orderly, and the unmanned vehicle can be widely applied to the expressway.
Therefore, the technical scheme of the embodiment 2 of the invention is as follows:
a dynamic evolution prediction method for an open unmanned vehicle group in an expressway scene specifically comprises the following steps:
step 1. correlation definition
Step 2, predicting the dynamic evolution of the unmanned vehicle group
Step 2.1 vehicle group feature extraction
Step 2.2 fleet sample Generation
Step 2.3 vehicle group dynamic evolution prediction
The invention aims to disclose a method which is oriented to the driving environment of an open unmanned vehicle group and can accurately predict the future evolution trend of the vehicle group after dynamic evolution in real time under the condition of considering an expressway scene, effectively meet the requirement that the future unmanned vehicle group always keeps motion behavior intellectualization, and further enable unmanned vehicles to be applied to the expressway scene. The technical scheme of the invention is especially suitable for highway scenes, and is not suitable for closed scenes such as ports, logistics and the like, and is also not suitable for urban scenes.
Description of the attached tables
TABLE 1 legends
TABLE 2 driverless vehicle consist architecture features
TABLE 3WEKA support classifier
TABLE 4 simulation test parameters
Drawings
FIG. 1 example 2 fleet formation and extinction samples
FIG. 2 example 2 train survival and fragmentation samples
FIG. 3 example 2 vehicle consist merge and split samples
FIG. 4 embodiment 2 dynamic evolution sample example of train group
FIG. 5 embodiment 2 is a flow chart of a prediction algorithm for dynamic evolution of unmanned vehicle group (Algorithm 1)
FIG. 6 embodiment 2 prediction accuracy of dynamic evolution of unmanned vehicle group
FIG. 7 embodiment 2 prediction recall rate of dynamic evolution of unmanned vehicle cluster
FIG. 8 embodiment 2 unmanned vehicle group dynamic evolution prediction F1 value comparison diagram
FIG. 9 is a general flowchart of the method of example 2 of the present invention
FIG. 10 is a network topology structure diagram of the skip car group of embodiment 1
FIG. 11 is a frame diagram of the state transition of the unmanned vehicle according to embodiment 1
Detailed Description
Example 1
Example 1 discloses an open unmanned vehicle group model in a highway scene and a vehicle group formation process research method
The research method is characterized by comprising the following steps:
one, predefining steps (including unmanned vehicle direct connectivity, neighbor nodes, neighbor node set)
In order to study the unmanned vehicle group formation algorithm based on the highway scene, the invention provides the following definitions:
defining 1 driverless Vehicle direct connectivity dvc (driverless Vehicle connectivity) to represent the stability of direct connection of two driverless Vehicle nodes, and the mathematical expression is (1):
Figure GDA0003398469070000031
wherein the content of the first and second substances,
dcr (driverless Communication range) represents the maximum Communication range of the unmanned vehicle Communication;
distt(vi,vj) Node v representing unmanned vehicle at time tiNode v with another unmanned vehiclejThe distance between them; when the distance between the nodes is larger than the maximum communication range, the DVC is 0, which indicates that the two unmanned vehicle nodes are not connected, namely, the topological graph shows that no edge exists between the two nodes; when the distance between the unmanned vehicle nodes is less than or equal to the maximum transmission range, the DVC is inversely related to the distance between the vehicles. The closer the distance between the nodes is, the larger the DVC is, the higher the reliability of direct connection between two unmanned vehicle nodes is, the tighter the connection is, and the larger the weight reflected to the upper side of the topological graph is.
Defining 2 neighbor nodes NeiNode: if unmanned vehicle node viAnd another does notHuman-driven vehicle node vjSatisfy DVC (v)i,vj)>0, then viAnd vjThe nodes which are adjacent to each other are reflected in the topological graph, namely viAnd vjAn edge is arranged between the two edges; the neighbor node NeiNode is characterized by a mathematical expression as (2):
NeiNode(vi,vj)＝1 if DVC(vi,vj)>0 (2)
defining 3 neighbor node set Ni,t: representing an unmanned node viSet V of neighbor nodes at time tjThe mathematical expression is (3):
Figure GDA0003398469070000041
step two, constructing network topological structure of unmanned vehicle group
The unmanned vehicle group is the neighbor vehicle group NeiVG, if two unmanned vehicle groups VGiAnd VGjIf a connecting edge exists between the nodes in the unmanned vehicle cluster, the two vehicle clusters are mutually neighbor vehicle clusters, and the network structure of the unmanned vehicle cluster adopts a mathematical expression of (4):
Figure GDA0003398469070000042
forming a multi-hop cluster topology for unmanned vehicles in each direction of the highway road, as shown in fig. 10, is a three-hop cluster network topology, in which: the radius of the vehicle which can be communicated is represented by R in the figure, the dotted line among the vehicle nodes represents that the vehicles can be communicated with each other, and the vehicle nodes which are not in the communication range can be used for relaying communication by other vehicles;
thirdly, defining the vehicle state
Definition 6 to describe the formation process of the unmanned vehicle group in the highway scenario, the following five unmanned vehicle states are defined:
(1) initialization State IN (initialization)
The initialization state is the starting state of the unmanned vehicle node.
During initialization, each unmanned vehicle node maintains a vehicle Basic Information table VIBT (vehicle Information Basic Table) including vehicle Information of the node itself and its neighbor nodes, wherein: the vehicle information includes its vehicle ID, direction, speed, position coordinates, current vehicle state,
the neighbor vehicle information includes vehicle state information of each neighbor node around the vehicle,
in addition, the vehicle also needs to record the ID of the vehicle group to which the vehicle belongs, if the vehicle is a common node CN, the hop count of the leading node LN needs to be saved, and the vehicle ID is connected to the leading node LN and passes through the common node CN; if the node is a leading node LN, the ID of the vehicle group member is also stored.
By way of example and not limitation, tables 1-1 show the following vehicle basic information:
Figure GDA0003398469070000051
(2) election State SE (select State)
After each unmanned vehicle node is in the initialized state, the unmanned vehicle node comprehensively senses the information of the neighbor nodes, simultaneously sends the information of the unmanned vehicle node to the neighbor nodes around, updates the basic information table VIBT of the vehicle to the latest state, and at the moment, switches the state of the unmanned vehicle node into the election state.
(3) /(4) leading node State and ordinary node State
In order to characterize the position of the unmanned vehicle nodes in the vehicle cluster and the degree of contribution thereof in the maintenance of the vehicle cluster, after an election state, one unmanned vehicle cluster generates more than one Leading Node (LN) and a plurality of Common member nodes (CN).
Each leading node is responsible for managing various information of the vehicle group, such as a vehicle group node set, the position and the speed of each unmanned vehicle node in the vehicle group and the like; meanwhile, in the movement process of the vehicle group, the leading node often determines whether a node which is not the vehicle group can join the vehicle group.
Meanwhile, node set N of vehicle groupiExcept for the leading node, the remaining unmanned vehicle nodes are in a common node state.
(5) Free node FN (free node) state
In the movement process of the unmanned vehicle node, if the node cannot be connected to any existing vehicle cluster and no nodes capable of being communicated exist around the node, the node is in a free node state.
Fourth, unmanned vehicle state transition process
Giving a state conversion process of the unmanned vehicle according to the motion characteristics and the five states of the unmanned vehicle node in the expressway scene,the frame diagram is shown in FIG. 11;
the specific conversion process is as follows:
1) the unmanned vehicle node starts to be in an initialization state, and in the state, the vehicle periodically exchanges HELLO data packets to construct a vehicle basic information table VIBT of the vehicle;
the vehicle then transitions to an election state SE in which the vehicle makes its next state decision, step 2);
2) when no adjacent node exists near the unmanned vehicle node in the election state SE, the vehicle is converted into a free node FN state, and the step 3) is carried out;
3) when the unmanned vehicle node in the free node FN state finds other free nodes FN which can be directly connected, the vehicle is converted into an election state SE;
when the vehicle group node of the leading node LN or the common node CN exists near the unmanned vehicle node in the free node state, the node is converted into the common node, or the vehicle is converted into the election state SE;
4) if the relative attribute measurement of the node is the best, the vehicle state is converted into a leading node LN state; entering step 5) or step 6);
5) in an election state SE, if the node metric of the piloting node LN of the unmanned vehicle is not optimal, the piloting node LN is converted into a common node CN state; otherwise, converting into a leading node LN state;
6) when no common node CN exists near the leading node LN, the node is converted into an election state SE;
7) and when the vehicle group leading node LN to which the common node CN belongs does not exist, the state is converted into an election state SE.
Further, in said step 5), node relative mobility is defined for characterizing said metric to determine whether it is optimal to elect a lead node LN.
Selecting a stable lead node using the node relative mobility, node i relative mobility MobiThe mathematical expression is (5):
Figure GDA0003398469070000071
wherein the content of the first and second substances,
Nirepresenting unmanned vehicle node viSet of nodes of the vehicle group, ijIndicating a vehicle group NiJ-th node in, SiRepresents the velocity of node i;
Mobithe smaller the value, the smaller the difference in relative speed between the node i and other nodes in the vehicle group, and the more stable the relative mobility.
If v isiIn the leading node state, the mathematical expression is (6):
Figure GDA0003398469070000073
wherein the content of the first and second substances,
Mobiindicating the relative mobility of the nodes of the unmanned vehicle i, i belonging to the unmanned vehicle node in the election state.
As an embodiment, the main symbols required in the formation process of the vehicle group in the expressway scene are given, and the meaning description is shown in tables 1-2.
Tables 1 to 2
Figure GDA0003398469070000072
Figure GDA0003398469070000081
In conclusion, the invention provides the concept of open type unmanned vehicle group for the first time, supposes that the unmanned vehicles form a multi-hop vehicle group in each direction of the highway road, designs and constructs an unmanned vehicle group model which is considered under the highway scene, faces the driving environment of the open type unmanned vehicle group, can always keep interconnection and intercommunication among the vehicle groups and effectively meets the requirement of intellectualization of future unmanned movement behaviors, meanwhile, the conversion processes of the initialization state, the election state, the leading node state, the common node state, the free node state and the five states of the unmanned vehicle node are researched, the prototype of the unmanned vehicle group forming method is provided, therefore, the theory and the method needed by the future unmanned motion behavior intellectualization are provided, and the practical application of the unmanned vehicle in the expressway scene becomes possible.
The unmanned vehicle serves as a terminal node of a vehicle group, serves as an intelligent agent, and is internally provided with a plurality of devices for sensing, data processing, data storage, communication transmission and the like, so that the unmanned vehicle can acquire information in the vehicle and real-time information of adjacent vehicles, and can effectively keep interconnection and intercommunication among the vehicle group nodes. These supporting devices are not an inventive task of the present invention.
Various networks in the physical environment, including different types of roadside infrastructure networks, mobile communication networks, etc., are considered prior art, and are considered road network space-time resources that can be perceived by the unmanned single intelligent vehicle nodes of the present invention, but are not the inventive task of the present invention.
The invention is used as an original technical scheme. The type of the access network, the service quality of the network, the protocol type, the network bandwidth, the terminal capability and the like are not the invention tasks of the invention, and other subsequent patents further disclose and perfect. The unmanned vehicle has application value in highway scenes.
Example 2
In a highway scene, unmanned driving and manned driving coexist, an unmanned vehicle group can be interfered by manned vehicle nodes, and various evolution events exist in the life cycle of the unmanned vehicle group, including leaving and joining of the unmanned vehicle nodes on a microscopic level, formation, merging, splitting, extinction and other events of the unmanned vehicle group on a macroscopic level. Therefore, how to accurately predict the behavior of the unmanned vehicle group after dynamic evolution can reasonably guide the establishment of a new vehicle group or a vehicle group, so that the stability and the order of the motion behavior of the unmanned vehicle group are ensured.
At present, the prediction research of the unmanned vehicle group in the expressway scene after dynamic evolution does not exist. For this reason, based on example 1, further disclosed is example 2
The specific implementation process of example 2 is shown in fig. 9, and includes the following 6 aspects:
(ii) associated definitions
② extraction of vehicle group characteristics
Third, vehicle group sample generation
Vehicle group dynamic evolution prediction algorithm
Sixth, simulation experiment verification
①
Correlation definition
The method is based on main symbols required in the dynamic evolution detection process of the open unmanned vehicle group in the expressway scene, and the specific meaning description is shown in table 1.
In order to research a dynamic evolution detection method of an open unmanned vehicle group based on a highway scene, the invention provides a definition of vehicle group similarity, which is as follows:
defining vehicle group similarity
Figure GDA0003398469070000091
And
Figure GDA0003398469070000092
the vehicle groups detected at times i and j, respectively. If the proportion of the same unmanned vehicle node in the two vehicle groups to the sum of the nodes of the two vehicle groups exceeds a set threshold value k, the two vehicle groups are called
Figure GDA0003398469070000093
And
Figure GDA0003398469070000094
is similar to that of
Figure GDA0003398469070000095
Expressed, the mathematical expression is (1):
Figure GDA0003398469070000101
wherein the content of the first and second substances,
Figure GDA0003398469070000102
indicates the group of vehicles detected at time i,
Figure GDA0003398469070000103
representing the number of unmanned vehicle nodes in the unmanned vehicle clusters p and q at times i and j, respectively.
②
Vehicle group feature extraction
The unmanned vehicle group is a group consisting of unmanned vehicle nodes with similar positions, similar movement trends and close communication in a certain scale, can be abstracted into a complex network consisting of a plurality of nodes according to the knowledge of graph theory, and has corresponding structural characteristics, specifically comprising the number of nodes in the vehicle group, the number of edges connected by the nodes, the connection ratio between the interior and the exterior of the vehicle group, the duration of the vehicle group and the like. The details of the structural features of the unmanned vehicle cluster are shown in table 2.
③
Vehicle fleet sample generation
In order to predict the dynamic evolution trend of the unmanned vehicle group, feature extraction needs to be carried out on each vehicle group at all the time, namely, a set of feature sets is used for representing a single vehicle group from multiple dimensions. In this way, each vehicle group at each time is quantized into a sample, and the sample is used as a training sample for training a classification model. Examples of the dynamic evolution prediction samples of the unmanned vehicle group are respectively shown as follows:
an example of formation and extinction of a fleet of vehicles is shown in FIG. 1;
an example of survival and splitting of one fleet into two is shown in FIG. 2;
an example of two vehicle groups merging into one vehicle group and then splitting into two vehicle groups again is shown in fig. 3;
figure 4 is a representation of different vehicle groups and their evolution.
Wherein the content of the first and second substances,
Figure GDA0003398469070000104
represents the time t1The p-th vehicle group (2),
Figure GDA0003398469070000105
indicating vehicle group
Figure GDA0003398469070000106
The first characteristic value of the invention is 9, tar is a target variable, namely a classification result, and the method respectively assigns the values of tar by forming a train group, eliminating the train group, living the train group, splitting the train group and combining the train group in the dynamic evolution time of the train group. Each row feature represents a vehicle cluster, and these features of all vehicle clusters constitute training samples and test samples of the classification model.
⑤
Vehicle group dynamic evolution prediction algorithm
The dynamic evolution prediction of the unmanned vehicle group is a multi-classification single-label problem, namely, five evolution events can occur in each vehicle group, but only one event can occur at a specific moment. The more general algorithms for solving the problem of multi-classification single labels include a decision tree algorithm, a decision table algorithm, an adaboost.m1 multi-classification algorithm and a naive bayes classification algorithm, any one of the above classification algorithms can be directly used in WEKA data mining software, and the corresponding classifier names are shown in table 3.
Through feature extraction and data sample generation, the optimal and most suitable classifier algorithm can be screened out through the algorithm 1 by utilizing the obtained data sample, and therefore the dynamic evolution trend of the unmanned vehicle group is predicted.
Note: each classifier in table 3 is the prior art, and after the optimal and most suitable classifier algorithm is selected by using algorithm 1, the classifier is directly used and applied for prediction in the invention.
Specifically, as shown in algorithm 1, a specific algorithm flowchart is shown in fig. 5.
Figure GDA0003398469070000111
Figure GDA0003398469070000121
Firstly, calculating the characteristics of the unmanned vehicle group, and generating a data sample, wherein the data sample is specifically divided into a training sample and a testing sample (an algorithm 1-2 lines).
Then, training is carried out on different classification algorithms by using the training samples, and a classification model corresponding to each classification algorithm is obtained (algorithm 4-16 lines). The test samples are tested on different classification models to yield prediction results (rows 18-23 of the algorithm).
Then, according to the prediction result, the evolution events marked by the test samples are compared, the classification accuracy and recall rate are counted, then the f1 value of the model is calculated, the advantages and the disadvantages of different classification models are evaluated (in the lines of 24-28), and finally the most suitable dynamic evolution prediction model of the unmanned vehicle group is obtained (in the lines of 31-32). Through the steps, a classification algorithm suitable for the dynamic evolution characteristics of the unmanned vehicle group can be selected, and the dynamic evolution trend of the vehicle group can be accurately predicted by applying the model, so that corresponding measures can be accurately made in advance in real time, the establishment of a new vehicle group or a vehicle group is guided, and the motion behaviors of the vehicle group are kept stable and orderly. For example, the unmanned vehicle cluster is predicted to be subjected to a split event, if the split event is not processed, the unmanned vehicle cluster changes a part of nodes of the vehicle cluster from leaving the vehicle cluster into an isolated node state, and the isolated nodes regenerate a new vehicle cluster through a vehicle cluster forming method. If the fact that the vehicle group is to be split is predicted, corresponding leading nodes can be generated in advance, so that the vehicle group is changed into two independent vehicle groups to operate in advance, time and network cost do not need to be spent to regenerate the vehicle group, and the movement behaviors of the vehicle group are guaranteed to be stable and orderly.
Remarking: algorithm 1 is as follows:
Figure GDA0003398469070000131
⑥
simulation experiment verification
In order to verify the prediction of the dynamic evolution of the unmanned vehicle group, the method adopts a simulation experiment mode to carry out computer simulation on the expressway scene vehicle group so as to obtain related statistical data.
(1) Experimental data
Based on the expressway scene, the invention adopts the microscopic traffic simulation software SUMO to create a one-way two-lane expressway with the length of 10 kilometers, and an entrance and an exit are arranged every 2km in the expressway. Specific simulation parameters are shown in table 4, and the simulation time is set to 310 seconds. The range of stable connection of the unmanned vehicle in network simulation is set to be 200 meters, the interference of the unmanned vehicle on the vehicle group is considered, and in addition, the collection frequency of node position information and node data of the whole unmanned vehicle group is 1 second/time.
(2) Experimental methods
The simulation experiment of the invention is to use SUMO and ns3 simulation software to simulate traffic on the basis of the simulation experiment data. In order to verify the influence of the vehicle group similarity threshold k on the vehicle group evolution, two groups of experiments with different average speeds of unmanned vehicles are set, wherein the speed interval of the vehicles is set to be 60,70 kilometers per hour in the first group, and 60,120 kilometers per hour in the second group. 100 unmanned vehicles are randomly arranged in the road, the speeds of the unmanned vehicles are uniformly distributed in the set section, and the destination randomly selects an exit in the positive driving direction to leave the road. Setting k as 0.1,0.2, … and 1.0 change, starting from the 10 th s of simulation experiment acquisition data, and then generating a network topology structure diagram of the unmanned vehicle every 15 seconds for analyzing the topology structure change of the vehicle group.
(3) Simulation experiment results and analysis
The invention extracts the train group characteristics of each second in the experimental data and marks the dynamic evolution events according to the state of the next second. And then, carrying out classifier training on the experimental data of the first 200 seconds by using weka data mining software to obtain a classification model corresponding to four algorithms listed in the table 3, and then verifying the advantages and disadvantages of the classifier by using the data of the last 100 seconds. The accuracy of each model in predicting five events, also called precision, is calculated, and the result is shown in fig. 6, the adaboost.m1 multi-classification algorithm has the best effect, and the average prediction accuracy of the five dynamic evolution events can reach over 86%. The decision table algorithm has poor effect, and the average prediction accuracy is only about 76%. For more verification of the merits of the classification models, the predicted recall rate, also called recall rate, of the four classification models is calculated, and the experimental results are shown in fig. 7. Similarly, the recall rate of the adaboost.M1 multi-classification algorithm is highest, the recall rate of predicted death events can reach 89.8%, and the average recall rate is 85.9%.
The accuracy and the recall rate mutually influence, and the optimal condition is that the values of the accuracy and the recall rate are high, and the average accuracy and the recall rate of five dynamic evolution events to be predicted are also the highest. To achieve the above goal, we calculate the average F1-measure of each classification model, and the experimental results are shown in FIG. 8. The F1 value of adaboost, m1 multi-classification algorithm is the best, about 85.88%, followed by naive bayes classification algorithm, decision tree algorithm, decision table algorithm in turn.
Innovation point
The innovation points are as follows: the invention aims at a method for predicting the motion behavior of an unmanned vehicle group after various events such as formation, combination, splitting, extinction and the like occur in a highway scene, firstly, the characteristics of the structure of the unmanned vehicle group are extracted, samples are generated according to the occurring evolution events, then, the classification method is used for predicting the motion behavior of the unmanned vehicle group after the evolution, the establishment of a new vehicle group or a vehicle group can be reasonably guided, the motion behavior of the unmanned vehicle group is ensured to be kept stable and ordered, and the unmanned vehicle can be widely applied to the highway.
In a highway scene, unmanned driving and manned driving coexist, an unmanned vehicle group can be interfered by manned vehicle nodes, and various evolution events exist in the life cycle of the unmanned vehicle group, including leaving and joining of the unmanned vehicle nodes on a microscopic level, formation, merging, splitting, extinction and other events of the unmanned vehicle group on a macroscopic level. Therefore, how to predict the motion behavior of the next vehicle group according to the dynamic evolution of the unmanned vehicle group is a problem which is urgently needed to be solved for ensuring that the motion behavior of the unmanned vehicle group is always stable and ordered.
Attached table of the specification
TABLE 1
Figure GDA0003398469070000151
TABLE 2
Figure GDA0003398469070000152
Figure GDA0003398469070000161
TABLE 3
Figure GDA0003398469070000162
TABLE 4
Figure GDA0003398469070000163

Claims (1)

1. A dynamic evolution prediction method for an open unmanned vehicle group in a highway scene is characterized by comprising the following steps:
step 1. correlation definition
Defining vehicle group similarity
Figure FDA0003398469060000011
And
Figure FDA0003398469060000012
vehicle groups detected at times i and j, respectively; if the proportion of the same unmanned vehicle node in the two vehicle groups to the sum of the nodes of the two vehicle groups exceeds a set threshold value k, the two vehicle groups are called
Figure FDA0003398469060000013
And
Figure FDA0003398469060000014
is similar to that of
Figure FDA0003398469060000015
Expressed, the mathematical expression is (1):
Figure FDA0003398469060000016
wherein the content of the first and second substances,
Figure FDA0003398469060000017
indicates the group of vehicles detected at time i,
Figure FDA0003398469060000018
indicates that no person is present in the unmanned vehicle groups p and q at times i and j, respectivelyThe number of driven vehicle nodes;
step 2, predicting the dynamic evolution of the unmanned vehicle group
Step 2.1 vehicle group feature extraction
The unmanned vehicle group is a group consisting of unmanned vehicle nodes with similar positions, similar movement trends and close communication in a certain scale, is abstracted into a complex network consisting of a plurality of nodes according to the knowledge of graph theory, and has corresponding structural characteristics, specifically comprises the number of nodes in the vehicle group, the number of edges connected by the nodes, the connection ratio between the interior and the exterior of the vehicle group and the duration of the vehicle group; the structural features of the unmanned vehicle group are described in detail in the following table;
Figure FDA0003398469060000019
Figure FDA0003398469060000021
step 2.2 fleet sample Generation
In order to predict the dynamic evolution trend of the unmanned vehicle group, extracting the characteristics of each vehicle group at all times, namely representing a single vehicle group from multiple dimensions by using a group of characteristic sets; each vehicle group at each moment can be quantized into a sample, and the sample is used as a training sample for training a classification model;
step 2.3 vehicle group dynamic evolution prediction
The dynamic evolution prediction of the unmanned vehicle group is a multi-classification single-label problem, namely, each vehicle group can generate five evolution events, and only one event can be generated at a specific moment;
performing feature extraction and generating data samples, and screening out the optimal and most suitable classifier algorithm by using the obtained data samples through the algorithm 1, so as to predict the dynamic evolution trend of the unmanned vehicle group;
algorithm 1 is as follows:
Figure FDA0003398469060000022
Figure FDA0003398469060000031
wherein J48 is a classifier used by a C4.5 decision tree algorithm, DT is a classifier used by a decision table algorithm, ADA is a classifier used by an adaboost.M1 multi-classification algorithm, and NB is a classifier used by a naive Bayes classification algorithm; VGS represents the set of unmanned vehicle groups, and best Classification model represents the optimal Classification model.
CN202110195975.4A 2021-02-22 2021-02-22 Dynamic evolution prediction method for open type unmanned vehicle group in expressway scene Active CN113012424B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110195975.4A CN113012424B (en) 2021-02-22 2021-02-22 Dynamic evolution prediction method for open type unmanned vehicle group in expressway scene

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110195975.4A CN113012424B (en) 2021-02-22 2021-02-22 Dynamic evolution prediction method for open type unmanned vehicle group in expressway scene

Publications (2)

Publication Number Publication Date
CN113012424A CN113012424A (en) 2021-06-22
CN113012424B true CN113012424B (en) 2022-03-08

Family

ID=76405356

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110195975.4A Active CN113012424B (en) 2021-02-22 2021-02-22 Dynamic evolution prediction method for open type unmanned vehicle group in expressway scene

Country Status (1)

Country Link
CN (1) CN113012424B (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108550279A (en) * 2018-04-03 2018-09-18 同济大学 Vehicle drive behavior prediction method based on machine learning
CN109840720A (en) * 2019-03-01 2019-06-04 同济大学 Car networking community dynamic evolution method
CN111130853A (en) * 2019-12-04 2020-05-08 沈阳航空航天大学 Future route prediction method of software defined vehicle network based on time information
CN111267852A (en) * 2020-03-18 2020-06-12 清华大学苏州汽车研究院（相城） Cruise control method for motorcade cooperative driving
CN111754777A (en) * 2020-07-10 2020-10-09 清华大学 Microscopic traffic simulation method for unmanned and manned mixed traffic flow

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9459622B2 (en) * 2007-01-12 2016-10-04 Legalforce, Inc. Driverless vehicle commerce network and community
US11107031B2 (en) * 2015-02-18 2021-08-31 Ryder Integrated Logistics, Inc. Vehicle fleet control systems and methods
US20170139423A1 (en) * 2015-11-12 2017-05-18 King Fahd University Of Petroleum And Minerals Control system and method for multi-vehicle systems
US10606259B2 (en) * 2017-07-07 2020-03-31 Zoox, Inc. Interactions between vehicle and teleoperations system

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108550279A (en) * 2018-04-03 2018-09-18 同济大学 Vehicle drive behavior prediction method based on machine learning
CN109840720A (en) * 2019-03-01 2019-06-04 同济大学 Car networking community dynamic evolution method
CN111130853A (en) * 2019-12-04 2020-05-08 沈阳航空航天大学 Future route prediction method of software defined vehicle network based on time information
CN111267852A (en) * 2020-03-18 2020-06-12 清华大学苏州汽车研究院（相城） Cruise control method for motorcade cooperative driving
CN111754777A (en) * 2020-07-10 2020-10-09 清华大学 Microscopic traffic simulation method for unmanned and manned mixed traffic flow

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
在虚拟环境下分析车群行为安全的方法;纪廷婷等;《重庆交通大学学报(自然科学版)》;20100215(第01期);129-132 *
基于多人驾驶虚拟平台研究车群行为的基础模型设计;陈少军等;《交通与计算机》;20081020(第05期);19-23 *

Also Published As

Publication number Publication date
CN113012424A (en) 2021-06-22

Similar Documents

Publication Publication Date Title
EP2590151A1 (en) A framework for the systematic study of vehicular mobility and the analysis of city dynamics using public web cameras
CN110836675B (en) Decision tree-based automatic driving search decision method
CN111609864A (en) Multi-policeman cooperative trapping task allocation and path planning method under road network constraint
CN115565369B (en) Space-time hypergraph convolution traffic flow prediction method and system based on hypergraph
CN113283581B (en) Multi-fusion graph network collaborative multi-channel attention model and application method thereof
Grzybek et al. Aspects and trends in realistic VANET simulations
Zhang et al. Battery maintenance of pedelec sharing system: Big data based usage prediction and replenishment scheduling
Peixoto et al. FogJam: a fog service for detecting traffic congestion in a continuous data stream VANET
CN110472999A (en) Passenger flow pattern analysis method and device based on subway and shared bicycle data
Dai et al. Spatio-temporal deep learning framework for traffic speed forecasting in IoT
Hassan et al. Intelligent transportation systems in smart city: a systematic survey
Grzybek et al. Evaluation of dynamic communities in large-scale vehicular networks
Külzer et al. On latency prediction with deep learning and passive probing at high mobility
Thu et al. Multi-source data analysis for bike sharing systems
Manalastas et al. Where to go next?: A realistic evaluation of AI-assisted mobility predictors for HetNets
CN117521937B (en) Dynamic path induction method and system suitable for multidimensional collaborative awareness environment
Cárdenas et al. GraTree: A gradient boosting decision tree based multimetric routing protocol for vehicular ad hoc networks
Sliwa et al. Lightweight joint simulation of vehicular mobility and communication with LIMoSim
CN113159371B (en) Unknown target feature modeling and demand prediction method based on cross-modal data fusion
CN113012424B (en) Dynamic evolution prediction method for open type unmanned vehicle group in expressway scene
Jalel et al. Continuous time markov chain traffic model for urban environments
CN113015085B (en) Open type unmanned vehicle group dynamic evolution detection method in expressway scene
CN113903171B (en) Vehicle crowd sensing node optimization method based on spatial-temporal characteristics of highway network
Ramli et al. An overview of traffic congestion detection and classification techniques in VANET
Khabbaz et al. A probabilistic and traffic-aware bundle release scheme for vehicular intermittently connected networks

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant