Unmanned vehicle group dynamic evolution and collaboration method based on process mining in highway scene

By considering the historical driving state and structural changes in the dynamic evolution method of driverless vehicle groups, combined with process mining and collaborative algorithms, the problem of poor stability of vehicle groups under highway scenarios is solved, and vehicles in the vehicle group and environmental perception are realized Improvement has promoted the healthy and rapid development of unmanned driving.

CN118968752APending Publication Date: 2024-11-15TONGJI UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411025145.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-29
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

The existing dynamic evolution method of driverless vehicle groups does not take into account the historical driving state and structural changes of the vehicle groups, resulting in poor stability of the vehicle groups and cannot meet the dynamic evolution needs in highway scenarios.

Method used

A dynamic evolution method (AMDEM) based on process mining is proposed. Combining node joining, leaving, merging, splitting and reconstruction algorithms, defining the state of the vehicle group and the triggering conditions of the dynamic evolution event, and establishing a coordinated model of the driverless vehicle group and collaborative algorithms to evaluate vehicle group stability through simulation experiments.

Benefits of technology

Improve the collaboration time and structural stability between vehicle nodes in the vehicle group, realize the perception of beyond the visual range and the understanding of the surrounding environment of autonomous driving vehicles, ensure the stable and orderly driving of unmanned driving movement behavior, and promote unmanned Healthy and rapid development of driving.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118968752A_ABST
    Figure CN118968752A_ABST
Patent Text Reader

Abstract

The invention discloses an unmanned vehicle group dynamic evolution and cooperation method based on process mining in an expressway scene. The method comprises the following steps: step 1, related definition; step 2, a dynamic evolution event of the unmanned vehicle group; step 3, a dynamic evolution method of an unmanned vehicle group; and step 4, an unmanned vehicle group cooperation method. A node adding algorithm, a node leaving algorithm, a vehicle group merging algorithm, a vehicle group splitting algorithm and a vehicle group reconstruction algorithm are further given by the unmanned vehicle group construction method based on flow mining in the highway scene. Defining a vehicle group state, a dynamic evolution event triggering condition and a state transfer function, and giving an unmanned vehicle group dynamic evolution algorithm; an unmanned vehicle group cooperation model is given, and an unmanned vehicle group cooperation algorithm is given on the basis of the unmanned vehicle group cooperation model; evaluation indexes such as vehicle group extinction frequency, node participation rate, node dependency degree, vehicle group average survival duration and average cooperation duration proportion are given, and a simulation experiment is constructed to perform stability verification on the proposed unmanned vehicle group.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of unmanned driving, and in particular to a dynamic evolution and coordination method of an unmanned vehicle group based on process mining in a highway scenario. Background Art

[0002] (1) Current status of research on community dynamics and evolution

[0003] There are many studies on the dynamic evolution methods of complex community networks. Palla et al. proposed an algorithm based on community permeation, which can study the time dependence of overlapping communities on a large scale, thereby revealing the basic relationship of community evolution. Asur et al. proposed an event-based representation of the key behavior patterns of time-varying interaction graphs. Using non-overlapping snapshots of interaction graphs, a framework for obtaining events from them was developed to describe the complex behavior patterns of individuals and communities over time, and to incorporate semantic information into the evolution of community behavior events. Backstrom et al. used decision tree technology to determine the determinants of community evolution and proposed a method to measure the flow of individuals between communities to reflect the dynamic evolution of communities. Berger et al. proposed a computational framework to analyze dynamic social networks, explicitly utilizing information about when social interactions occur, and demonstrated the effectiveness of the framework in some practical applications. Tantipathananandh et al. proposed a framework and algorithm for identifying communities that change over time in social networks. The algorithm is an optimization-based dynamic community structure modeling method. It is proved that finding the most explanatory community structure is an NP-hard and APX-hard problem, and an algorithm based on dynamic programming, exhaustive search, maximum matching, and greedy heuristics is proposed. Leskovec et al. studied the evolutionary behavior of communities where the node distance and the number of edges gradually shorten and become dense over time, respectively, and proposed a graph generator based on the "forest fire" spreading process to discover communities and study their dynamic evolution. Falkowski et al. proposed two methods to analyze the evolution of two different types of online communities at the subgroup level. The first method includes statistical analysis and visualization, allowing interactive analysis of subgroup evolution in communities that exhibit considerable membership structure; the second method is designed to detect communities in environments with highly fluctuating membership. Wang et al. proposed a core-based community evolution tracking algorithm that relies on core nodes to establish evolutionary relationships between communities under different snapshots. Kumar et al. proposed a network growth model to analyze three types of nodes in a community and the evolution of the community. Greene et al. proposed a model for tracking the progression of communities over time in a dynamic network, in which each community is characterized by a series of important evolutionary events.

[0004] (2) Current status of research on the dynamic evolution of driverless vehicle swarms

[0005] After the formation of the unmanned vehicle group, a dynamic evolution method is needed to cope with changes in external conditions or changes in the scale and structure of the vehicle group. When the unmanned vehicle group undergoes dynamic evolution, the structure of the vehicle group is updated through corresponding evolution steps according to the type of event that occurs. The existing unmanned driving dynamic evolution methods are designed according to the corresponding vehicle group formation method, so the existing dynamic evolution methods are also divided into two categories: centralized and distributed. Regardless of the dynamic evolution method, its basis is the change in the scale of the vehicle group, such as nodes joining and leaving the vehicle group, and the merger, splitting, and extinction of the vehicle group. Cheng et al. proposed a centralized unmanned vehicle group dynamic evolution method, and in its evolution process, special consideration needs to be given to the changes in the leading node of the vehicle group. When the merger or split event of the vehicle group occurs, the leading node needs to be re-selected. The distributed unmanned vehicle group dynamic evolution method does not need to consider the problem of the leading node. Cheng et al. proposed SCCDEM, and the evolution of the vehicle group depends on the changes in the side chain consensus. When the side chain consensus increases or decreases, the merger or split event of the vehicle group occurs accordingly.

[0006] In summary, from the current status of community evolution research, it can be seen that most methods generally consider the evolution behavior of static networks. For the highway scene where nodes are traveling at high speed, they cannot meet the dynamic evolution behavior of the vehicle group. From the current status of the research on the dynamic evolution of the unmanned vehicle group, it can be seen that the existing methods only consider the current driving state of the vehicle group, and only consider the change of the size of the vehicle group when the vehicle group evolves. The existing methods have the following defects:

[0007] (1) Not considering the historical driving status of the vehicle group: The existing methods only consider the instantaneous driving status of the vehicle group for dynamic evolution, but do not consider its historical driving status. This may affect the stability of the vehicle group due to differences in driving status after the evolution of the vehicle group.

[0008] (2) Failure to consider changes in the structure of the vehicle group: Existing methods only consider changes in the size of the vehicle group, but do not consider changes in the structure of the vehicle group. Summary of the invention

[0009] Purpose of the invention:

[0010] In response to the current problems that unmanned vehicle swarms do not take into account the historical driving status of the vehicle swarm and the changes in the vehicle swarm structure, the present invention is based on ("Unmanned Vehicle Swarm Construction Method Based on Process Mining in Highway Scenarios" applied by inventors Cheng Jiujun and others on January 5, 2024 (Applicant: Cheng Jiujun Tongji University, Patent Application Number: 2024100138829)) and provides a node joining algorithm, a node leaving algorithm, a vehicle swarm merging algorithm, a vehicle swarm splitting algorithm and a vehicle swarm reconstruction algorithm; defines the vehicle swarm state, dynamic evolution event trigger conditions and state transition functions, and provides an unmanned vehicle swarm dynamic evolution algorithm; provides an unmanned vehicle swarm collaboration model, and on this basis, provides an unmanned vehicle swarm collaboration algorithm; provides evaluation indicators such as the number of vehicle swarm extinctions, node participation rate, node dependence, average vehicle swarm survival time and average collaboration time ratio, and constructs a simulation experiment to verify the stability of the proposed unmanned vehicle swarm. Selecting vehicle nodes with similar instantaneous and historical motion states to form a vehicle group can improve the time of mutual cooperation between vehicle nodes in the vehicle group and maintain the stability of the vehicle group structure, thereby effectively realizing beyond-visual-range perception and improving the ability of autonomous driving vehicles to understand the surrounding driving environment. It provides effective guarantee for the stable and orderly driving of unmanned driving movement behavior, and has important significance and application value for promoting the healthy and rapid development of unmanned driving.

[0011] The technical solution of the present invention specifically comprises the following steps:

[0012] Step 1. Relevant definitions;

[0013] Step 2. Dynamic evolution of driverless car group events

[0014] Step 2.1 Node joins the event;

[0015] Step 2.2 Node leaving event;

[0016] Step 2.3: Vehicle group merging event;

[0017] Step 2.4: Vehicle group split event;

[0018] Step 2.5: Vehicle group reconstruction event.

[0019] Step 3: Dynamic evolution method of unmanned vehicle group;

[0020] Step 3.1: Vehicle group status;

[0021] Step 3.2 Dynamic evolution event triggering conditions and state transition functions;

[0022] Step 3.3 Dynamic evolution method.

[0023] Step 4: Unmanned vehicle group collaboration method;

[0024] Step 4.1 Unmanned vehicle group collaboration model;

[0025] Step 4.2 Unmanned vehicle group collaboration algorithm.

[0026] Compared with the prior art, the present invention has the following beneficial effects:

[0027] The purpose of the present invention is to disclose a dynamic evolution and coordination method of a swarm of unmanned vehicles based on process mining in a highway scenario, which can effectively realize beyond-visual-range perception and enhance the ability of autonomous driving vehicles to understand the surrounding driving environment, thereby effectively helping to solve current unmanned driving single intelligent body problems such as lane changing, avoidance, safety distance adjustment, and road traffic efficiency. It has important significance and application value for promoting the healthy and rapid development of unmanned driving in highway scenarios.

[0028] The present invention provides an Alpha-Miner-based Dynamic Evolution Method (AMDEM) for unmanned vehicle swarms based on process mining in highway scenarios. Starting from the changes in vehicle driving logs, adjacent logs and feature matrices, the instantaneous driving state and historical driving state of nodes are comprehensively considered, and events such as node joining, node leaving, vehicle group merging, vehicle group splitting, and vehicle group reconstruction as well as specific steps are given; on this basis, a dynamic evolution and coordination method for unmanned vehicle swarms in highway scenarios is given. Based on the unmanned vehicle swarm construction method based on process mining in highway scenarios, a node joining algorithm, a node leaving algorithm, a vehicle group merging algorithm, a vehicle group splitting algorithm, and a vehicle group reconstruction algorithm are further given; the vehicle group state, dynamic evolution event triggering conditions, and state transition functions are defined, and an unmanned vehicle group dynamic evolution algorithm is given; an unmanned vehicle group coordination model is given, and on this basis, an unmanned vehicle group coordination algorithm is given; evaluation indicators such as the number of vehicle group extinctions, node participation rate, node dependence, average vehicle group survival time, and average coordination time ratio are given, and a simulation experiment is constructed to verify the stability of the proposed unmanned vehicle group.

[0029] Selecting vehicle nodes with similar instantaneous and historical motion states to form a vehicle group can improve the time of mutual cooperation between vehicle nodes in the vehicle group and maintain the stability of the vehicle group structure, thereby effectively realizing beyond-visual-range perception and improving the ability of autonomous driving vehicles to understand the surrounding driving environment. It provides effective guarantee for the stable and orderly driving of unmanned driving movement behavior, and has important significance and application value for promoting the healthy and rapid development of unmanned driving.

[0030] Explanation of the attached table

[0031] Table 1 Symbols

[0032] Table 2 State transition table of finite state automaton for dynamic evolution of unmanned vehicle group

[0033] Table 3 Simulation experiment parameter settings BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 Node joining diagram

[0035] Figure 2 Schematic diagram of the change in the group structure after the node leaves

[0036] Figure 3 Diagram of the split of the vehicle group caused by the node leaving

[0037] Figure 4 Schematic diagram of vehicle group merging

[0038] Figure 5 Schematic diagram of car group splitting

[0039] Figure 6 Schematic diagram of vehicle group reconstruction (without causing vehicle group splitting)

[0040] Figure 7 Schematic diagram of vehicle group reconstruction (causing the vehicle group to split)

[0041] Figure 8 State transition diagram of finite state automaton for dynamic evolution of driverless car group

[0042] Fig. 9 Sequence diagram of unmanned vehicle group collaboration method

[0043] Fig.10 Number of car group deaths

[0044] Fig.11 Node participation rate

[0045] Fig.12 Changes in node dependence when the maximum speed of the node is 65km / h, 70km / h, and 75km / h Fig.13 Changes in node dependence when the maximum speed of the node is 80km / h, 85km / h, and 90km / h Fig.14 Changes in node dependence when the maximum speed of the node is 95km / h, 100km / h, and 105km / h

[0046] Fig.15 Changes in node dependence when the maximum speed of the node is 110km / h, 115km / h, and 120km / h

[0047] Fig.16 Average survival time of a group of vehicles

[0048] Fig.17 Average collaborative time ratio

[0049] Fig.18 Flowchart of the invention DETAILED DESCRIPTION

[0050] The technical solution provided by the present application will be further described below in conjunction with specific embodiments and accompanying drawings. The advantages and features of the present application will become more apparent with the following description.

[0051] The present invention specifically includes the following five aspects:

[0052] Step 1. Relevant definitions;

[0053] Step 2. Dynamic evolution of driverless car group events

[0054] Step 2.1 Node joins the event;

[0055] Step 2.2 Node leaving event;

[0056] Step 2.3: Vehicle group merging event;

[0057] Step 2.4: Vehicle group split event;

[0058] Step 2.5: Vehicle group reconstruction event.

[0059] Step 3: Dynamic evolution method of unmanned vehicle group;

[0060] Step 3.1: Vehicle group status;

[0061] Step 3.2 Dynamic evolution event triggering conditions and state transition functions;

[0062] Step 3.3 Dynamic evolution method.

[0063] Step 4: Unmanned vehicle group collaboration method;

[0064] Step 4.1 Unmanned vehicle group collaboration model;

[0065] Step 4.2 Unmanned vehicle group collaboration algorithm.

[0066] Step 5. Simulation experiment and result analysis

[0067] Step 4.1 Simulation experiment parameter setting and method description

[0068] Step 4.2 Analysis of simulation experiment results

[0069] Details are as follows:

[0070] The symbols used in the present invention are shown in Table 1.

[0071] Step 1: Definitions

[0072] Definition 1 Node Join: The unmanned vehicle node that has not joined the swarm has a new neighbor relationship with the nodes in the swarm, and sends a request message to join the swarm. After the conditions of the nodes in the swarm are judged, it agrees to join the swarm. Its formal representation is as follows:

[0073]

[0074] Among them, VG=(V,L,M) represents the unmanned vehicle group, v q It is a node in the car group VG.

[0075] Definition 2: Node Leave: A node in a driverless vehicle group loses its communication connection with other nodes in the group due to communication failure, vehicle failure, excessive distance between vehicles, or leaving the highway, and loses its neighbor relationship with other nodes. Its formal representation is as follows:

[0076]

[0077] Among them, node v p It is a node in the vehicle group VG = (C, L, M), and has no adjacent relationship with any other node in the vehicle group except itself.

[0078] Definition 3: Vehicle group merging (Merge): Two vehicle nodes from different driverless vehicle groups have a new adjacent relationship and are merged into a new vehicle group after negotiation. Its formal representation is as follows:

[0079]

[0080] Among them, VG1 = (V1, L1, M1) and VG2 = (V2, L2, M2) represent two driverless car groups, and node v p and v q They are from car groups VG1 and VG2 respectively.

[0081] Definition 4: Split: When the distance between the two nodes in the following relationship in the unmanned vehicle group is too large, the group will be split into two new groups after negotiation. The formal representation is as follows:

[0082]

[0083] in Indicates v p and v q The distance between nodes; θ represents the threshold of the distance between nodes, and satisfies δ<COMM, which represents the maximum communication distance between vehicles.

[0084] Definition 5: Change: A vehicle node in the driverless vehicle group changes lanes, which changes the neighbor node set of the node and updates the adjacent set, resulting in a change in the structure of the vehicle group. However, no node joins or leaves the vehicle group itself. Its formal representation is as follows:

[0085]

[0086] Where VG = (V, L, M) represents the group of unmanned vehicles; L t ∈L represents the latest adjacent set before the group is reconstructed; L t+1 represents the adjacent set after the vehicle group is reconstructed; V and V′ represent the unmanned vehicle node sets before and after the vehicle group is reconstructed, respectively.

[0087] Definition 6 Dissolution: When a node leaves or the group splits, the number of vehicles in the group is reduced to less than two and the group size cannot be maintained, which means the group is dissoluted. Its formal representation is as follows:

[0088]

[0089] Where VG = (V, L, M) represents the unmanned vehicle group, and |V| represents the number of nodes in the unmanned vehicle set.

[0090] Definition 7 Vehicle Group Component (VGC): When the structure of the unmanned vehicle group evolves dynamically, the vehicle group structure changes. All nodes that a node can access through the unmanned vehicle group maintenance method constitute the vehicle group component where the node is located, that is, the set of nodes that have a communication path with the node constitutes a vehicle group component. Its formal representation is as follows:

[0091] VGC(v p )={v|Path(v,v p )>0,v∈V-{v p}} (7)

[0092] Among them, Path(v,v p ) represents node v and node v p The number of communication connections between them; V represents the set of unmanned vehicles.

[0093] Definition 8 Vehicle Group Component Count (VGCC): The number of vehicle group components contained in the dynamically evolving unmanned vehicle group. Its formal representation is as follows:

[0094] VGCC(M)=|{VGC(v i )|vi ∈V, i=1,2,...,N}| (8)

[0095] Definition 9 Speed ​​Confusion (SC): The degree of inconsistency between the instantaneous speed of all nodes in the vehicle group and the historical average speed, that is,

[0096]

[0097] Where N is the number of vehicles in the driverless car group; α and β are two coefficients; Represents node v i The instantaneous velocity vector of Indicates the average instantaneous speed of all nodes in the vehicle group; Avgspeed i Represents node v i The historical average speed of It represents the average of the historical average speeds of all nodes in the vehicle group. and The formal expression of is shown in formula (10) and formula (11):

[0098]

[0099] Step 2: Dynamic evolution of driverless vehicle swarm events

[0100] (1) Node joining event

[0101] When an unmanned vehicle node that has not yet joined the swarm encounters a node that has already joined the swarm, it sends a message to the node in the swarm requesting to join the swarm. After receiving the message, the node in the swarm uses a series of judgment conditions to decide whether to agree to the node joining the swarm. If not, it replies with a rejection message. If yes, it updates the adjacent set and feature matrix, and sends the updated swarm information to the node, adding the node to the swarm. The schematic diagram of node joining is shown in the figure below. Figure 1 shown.

[0102] The specific steps of the node joining event are as follows:

[0103] (1) Node v that has not yet joined the swarm p Broadcast a HELLO message to the surroundings and attach the latest driving log, intending to join the car group or form a car group with other nodes that have not yet joined the car group.

[0104] (2) Node v that is already in the vehicle group VG = (V, L, M) q After receiving the HELLO message, reply with a BUILT message to indicate that you have joined the car group.

[0105] (3) If at this time node v pIf it is still in the state of not joining the car group, then it will send a q Issue a join request JOIN.

[0106] (4) Node v q After receiving the JOIN message, the following three conditions are used to determine whether to allow the node to join:

[0107] ①

[0108] ②NSS(v p , v q )≥ε(ε is the node speed similarity threshold)

[0109] ③SL p =SL q

[0110] If any of the conditions are not met, then v q To v p Send a DENY_JOIN message to reject v p If all three conditions above are met, then v q To v p Send AGREE_JOIN message to indicate agreement p The joining request is made and in its neighbor node set Neighbor p Add node v q . Node v p After receiving the AGREE_JOIN message, build your own adjacent set and add element v to it q and v p .

[0111] (5) Construct a new adjacent set L t+1 , node v q According to Adjacent(v p , v q ) and Adjacent(v q , v p ) is true or false to decide whether to p , v q ) and (v q , v p ) add a new adjacent set, and calculate a new feature matrix based on the new adjacent set through the Alpha Miner algorithm. Let the new adjacent log be L′=LU{L t+1}, and get the new vehicle group VG′=(V′, L′, M′).

[0112] (6) Send the new group structure to other nodes in the group (including the node v that has just joined the group). p ).

[0113] The node joining algorithm is shown in Algorithm 1.

[0114]

[0115]

[0116]

[0117] Algorithm 1 can be divided into three parts: The first part is for the node v that has not yet joined the car group p and the node v in the group q The process of establishing contact and sending a request to join the car group (lines 1-15); the second part is the node v q After conditional judgment, agree to node v p The third part is the node v p Build your own neighbor node set (Lines 38-44).

[0118] (2) Node leaving event

[0119] When a node in a swarm of unmanned vehicles detects that it has lost communication with another node, it may be caused by a vehicle failure, communication failure, or leaving the highway, so the node needs to leave the swarm; it may also be simply because the distance between the two nodes exceeds the maximum range of the communication connection, but the node still maintains communication with other nodes and does not leave the swarm.

[0120] Regarding the former, when a node leaves the swarm, the nodes in its neighboring set will lose the communication connection with the node that left the swarm. The node's departure can be divided into the following two cases:

[0121] (1) Delete the node from the vehicle group structure. Figure 2 As shown, in the topological structure of the vehicle group, a node at the edge leaves the vehicle group for some reason, but the structure of the vehicle group does not change significantly due to the departure of the node. It only needs to delete the node based on the original vehicle group structure.

[0122] (2) The vehicle group is split into two or more sub-vehicle groups. Figure 3 As shown, in the topological structure of the vehicle group, the middle node acts as a medium to connect the two parts of the vehicle group. When the node leaves the vehicle group, the topological structure of the original vehicle group is destroyed and is no longer a connected structure, so it must be split into two vehicle groups.

[0123] In order to determine whether the departure of the node will cause the group to split, the concept of vehicle group component (VGC) is introduced for judgment. The concept of vehicle group component is borrowed from the connected component (CC) in graph theory. In graph theory, the maximal connected subgraph of an undirected graph is called the connected component of the graph. In a group, all nodes that can be accessed by the unmanned vehicle group maintenance method constitute a vehicle group component. When a node leaves the group, the number of vehicle group components (Vehicle Group Component Count, VGCC) of the group composed of the remaining nodes is calculated. If the calculated number of vehicle group components is equal to 1, it means that the group has not split due to the departure of the node, and only the node needs to be deleted. If the calculated number of vehicle group components is greater than 1, it means that the group has split into multiple sub-groups due to the departure of the node. However, in either case, only the nodes that are still in the same vehicle group component as itself need to be retained.

[0124] The specific steps of the node leaving event are as follows:

[0125] (1) Vehicle group VG = node v in (V, L, M) p Detects that node v cannot be found in the set of neighbor nodes q Continue to communicate, then (v p , v q ) and (v q , v p ) is removed from the adjacent set, and a new feature matrix M′ is calculated, and m′ in the feature matrix is ​​replaced pq and m′ qp Change it to Unrelated. Then calculate the number of vehicle group components at this time.

[0126] (2) If the calculated number of vehicle group components is still equal to 1, it means that node v q Still communicating with other nodes and not leaving the group. Then node v p The node v q Delete it from the neighbor node set and update the vehicle group structure VG′ according to the adjacent set and feature matrix of step (1).

[0127] (3) If the calculated number of vehicle group components is greater than 1, the nodes in the unmanned vehicle set V that are not in the vehicle group component to which it belongs are removed, and the elements related to these nodes in the neighboring set are deleted. At the same time, the rows and columns where these elements are located in the feature matrix are also deleted to obtain the updated vehicle group VG′. In addition, these nodes in its neighboring node set are also deleted.

[0128] (4) Call the unmanned vehicle group maintenance method to send a LEAVE message to other nodes in the vehicle group, and attach the changed vehicle group VG′. After receiving the LEAVE message, other nodes update the vehicle group structure and their own neighbor node set.

[0129] The node leaving algorithm is shown in Algorithm 2.

[0130]

[0131]

[0132] Algorithm 2 can be divided into four parts: The first part is node v p Found lost with node v q The communication connection is established and the number of vehicle group components is calculated (lines 1-12). The number of vehicle group components calculated is divided into two cases. The second part is the scenario where the number of vehicle group components is equal to 1. q The position feature relationship between them can be deleted and modified from the adjacent set and feature matrix (lines 13-15). The third part is the scenario where the number of vehicle group components is greater than 1, and all nodes that are not in the same vehicle group component need to be deleted (lines 16-32). The fourth part is to update the vehicle group (lines 33-37).

[0133] (3) Vehicle group merging event

[0134] When a vehicle node in an unmanned vehicle group encounters a vehicle node in another unmanned vehicle group, the node can determine whether to merge with the vehicle group based on the known information of the vehicle group. At this time, it is not reasonable to rely solely on speed similarity to determine whether the vehicle group can be merged, because there are also differences in the speed of the nodes in the vehicle group. It is not possible to rely solely on the speed of the node itself to determine whether the two vehicle groups can be merged. Instead, the consistency of the speed of all nodes in the two vehicle groups must be considered. This section uses speed chaos to measure the consistency of the speed of the nodes in the vehicle group. The vehicle group chaos degree consists of two parts. The first part is the variance of the current instantaneous speed of the node in the vehicle group, and the second part is the variance of the average speed of the node in the vehicle group from entering the highway and starting to the current moment. The two are weighted by a pre-set coefficient to obtain the degree of chaos of the speed of all vehicles in the current vehicle group. From the concept of variance, it can be seen that the larger the value of the vehicle group chaos degree, the greater the speed difference of the vehicle segments in the vehicle group, and the more difficult it is to maintain the structure of the vehicle group for a long time. Therefore, only when the value of the vehicle group chaos degree is less than a certain threshold, the vehicle group can be merged. The schematic diagram of the vehicle group merger is shown in Figure 4 shown.

[0135] The specific steps of the vehicle group merging event are as follows:

[0136] (1) Node v in vehicle group VG1 pDuring the driving process, the node v in the vehicle group VG2 is detected. q . Node v p Calculate the speed disorder SC(VG1) of vehicle group VG1. If the value of speed disorder SC(VG1) is lower than the threshold ε, then send a signal to node v in vehicle group VG2. q Send a MERGE message to indicate a request to merge vehicle groups.

[0137] (2) When node v q After receiving the MERGE message, the vehicle groups are temporarily merged. The specific steps are:

[0138] ① Take the union of the unmanned vehicle sets of the two vehicle groups to obtain a temporary unmanned vehicle group set V′.

[0139] ② Calculate the adjacent set, that is, according to Adjacent(v p , v q ) and Adjacent(v q , v p ) is true or false to decide whether to p , v q ) and (v q , v p ) is added to the adjacent set to obtain a new adjacent log L′.

[0140] ③ Calculate the new feature matrix M′ through the Alpha Miner algorithm.

[0141] Through the above three steps, a temporary vehicle group VG′=(V′, L′, M′) is obtained, and the speed disorder degree SC(VG′) of the temporary vehicle group is calculated.

[0142] (3) Determine whether to agree to the vehicle group merger based on the following three conditions:

[0143] ①

[0144] ②SC(VG′)≤γ (γ is the velocity chaos threshold)

[0145] ③SL p =SL q

[0146] If any of the conditions are not met, then v q To v p Send a DENY_MERGE message to reject the group merge request; if the above three conditions are met, then v q To v p Send AGREE_MERGE message to indicate the group merge request and v p Add to your own neighbor node set. pAfter receiving the AGREE_MERGE message, v q Add your own neighbor node set.

[0147] (4) Send the new swarm structure to other nodes in the swarm.

[0148] The vehicle group merging algorithm is shown in Algorithm 3.

[0149]

[0150]

[0151] Algorithm 3 can be divided into three parts: The first part is the node v in the driverless car group VG1 p and node v in the driverless car group VG2 q The process of meeting and calculating the speed disorder degree SC(VG1) of the vehicle group in which they are located. If the conditions are met, a request for merging the vehicle groups is issued (lines 1-6); the second part is the process of ... q After receiving the request to merge the vehicle group, the speed disorder degree SC(VG′) of the merged vehicle group is calculated, and the conditions are used to determine whether to agree to the merger. If the merger is agreed, the vehicle group is updated (lines 7-21); the third part node v p Update its own neighbor node set (lines 22-27).

[0152] (4) Vehicle Split Incident

[0153] When a group of unmanned vehicles is driving on the highway, the speeds of the vehicles in the group vary, which in turn affects the distance between the vehicles. When the distance between vehicles is too large, the communication connection between the vehicles may be interrupted. In order to avoid this situation as much as possible, the group needs to maintain the stability of the group structure by actively splitting.

[0154] In a following relationship, if the speed of the leading vehicle is greater than that of the trailing vehicle, and the distance between the two vehicles is greater than a certain threshold (which is less than the communication range between the vehicles), then there is a risk of communication interruption between the two vehicle nodes. Therefore, the vehicle group needs to be split into two smaller vehicle groups through a split event. Figure 5 shown.

[0155] The specific steps of the car group split event are as follows:

[0156] (1) There is a node v in the driverless car group VG p Follow node v q When node v p Detected with node v q Distance(v p , v q) is greater than the threshold θ (θ<COMM, COMM is the communication range of the unmanned vehicle), the following vehicle node v p To the front car node v q Send a QUERY_SPILT message to query node v q Whether the group of vehicles needs to split and attach their own speed.

[0157] (2) Current car node v q After receiving the QUERY_SPILT message, determine the speed of the two vehicles. q The velocity is greater than that of node v p The distance between the two vehicles will gradually increase over time and may exceed the communication range COMM at some point in the future. In addition, it is necessary to determine v p With v q Is the following relationship between them the only medium connecting the front and rear parts of the group? The element m in the original feature matrix pq and m qp The values ​​of are all FOLLOW. Now let’s assume that m pq and m qp The value of is changed to Unrelated, and the new feature matrix is ​​set as M′, and then the number of vehicle group components of the vehicle group is calculated. If the calculated number of vehicle group components is still 1, it means that even if the follow-up relationship between the two nodes is disconnected, there are other follow-up relationships in the vehicle group that make the vehicle group still connected. If the calculated number of vehicle group components is greater than 1, it means that if the node v is disconnected p and node v q If there is no following relationship between nodes, the group will be split into two parts. q Whether the group needs to be split needs to be determined based on the following two conditions:

[0158] ①

[0159] ②VGCC(M′)>1

[0160] If any of the conditions are not met, the preceding node v q Backward car node v p Send a DENY_SPILT message to indicate that it disagrees with the group split. Otherwise, send a DENY_SPILT message to the following node v p Send an AGREE_SPILT message to indicate that the group agrees to split.

[0161] (3) At the front vehicle node v q Backward car node v p After sending the AGREE_SPILT message, the new unmanned vehicle group structure needs to be calculated. The specific steps are as follows:

[0162] ① Remove the nodes in the second half of the original vehicle group from the set of unmanned vehicles contained in the original vehicle group.

[0163] ② Construct a new adjacent log, in which the adjacent set is based on the last adjacent set of the original vehicle group, and deletes the tuple containing the removed node.

[0164] ③Build a new feature matrix and delete the rows and columns related to the removed nodes.

[0165] (4)v q The unmanned vehicle group maintenance algorithm is called to send a SPILT message to the nodes in the first half of the vehicle group, notifying them of the vehicle group split and attaching the new vehicle group structure.

[0166] (5) The following node v p After receiving the AGREE_SPILT message, it is also necessary to build a new swarm structure in the same way as step (3) and delete the relevant elements of the nodes in the first half of the original swarm. Call the unmanned swarm maintenance algorithm to send a SPILT message to the nodes in the second half of the swarm. (Note: The unmanned swarm maintenance algorithm is disclosed in patent application: 2024100138829.)

[0167] The group splitting algorithm is shown in Algorithm 4.

[0168]

[0169]

[0170] Algorithm 4 can be divided into three parts: the first part is node v p Detected with node v q The distance between them exceeds the threshold δ, and the node v q Send a group split request message QUERY_SPILT (lines 1-5). The second part is node v q After receiving the QUERY_SPILT message, the condition is judged. If the condition is met, the AGREE_SPILT message is replied and the first half of the nodes in the swarm are notified of the change in the swarm structure. Otherwise, the DENY_SPILT message is replied (lines 6-25). The third part is node v p After receiving the AGREE_SPILT message, the nodes in the second half of the group are notified of the change in the group structure (lines 26-38).

[0171] (5) Vehicle group reconstruction event

[0172] When the vehicles in the driverless car group are driving on the highway, they sometimes change lanes or overtake according to the actual situation or their own needs. However, the behavior of this node causes the structure of the car group to change, and the original adjacent sets and feature matrices are destroyed. The new structure of the car group needs to be analyzed according to the actual situation. The car group reconstruction event can be divided into the following two situations:

[0173] (1) The reconstruction of the vehicle group does not lead to the split of the vehicle group. Figure 6 As shown, an unmanned vehicle node changes lanes and overtakes in sequence, causing the structure of the vehicle group to change, and the vehicle group remains a connected structure without any split.

[0174] (2) The reconstruction of the vehicle group leads to the splitting of the vehicle group. Figure 7 As shown in FIG. 1 , after changing lanes, an unmanned vehicle node loses the following relationship with the node behind it, causing the unmanned vehicle group to split.

[0175] For the situation in (2), after the group reconstruction event occurs, the subsequent group can be applied to the situation (2) of the node departure event, that is, the connection with the node is lost, resulting in the split of the group structure. Therefore, the subsequent group evolution event is no longer discussed in detail. As for the previous group, it is only necessary to calculate the number of group components of the group after the group reconstruction event occurs. If the result is greater than 1, it means that the group reconstruction leads to the split of the group. Therefore, the above two types of group reconstruction situations can be combined. It is only necessary to calculate the number of group components of the group at the end and classify and discuss them according to the calculation results.

[0176] The specific steps of the vehicle group reconstruction event are as follows:

[0177] (1) Autonomous vehicle group VG = node v in (V, L, M) p The lane is changed and a CHANGE message is broadcast to the surrounding nodes in the vehicle group. When the node receives the CHANGE message, it sends a p Reply RESP_CHANGE with your location.

[0178] (2) After time τ, node v p Construct a new neighbor node set Neighbor′ based on the received RESP_CHANGE message p . Create a new adjacent set about node v i ∈Neighbor p , delete the newly created adjacent set (v i , v p ) and (v p , v i ) element; about node v i ∈Neighbor′p , if Adjacent(v p , v i )=true, then add (v p , v i ) element; if Adjacent(v i , v p )=true, then add (v i , v p )element.

[0179] (3) Construct a new feature matrix M′ based on the new adjacent set and calculate the number of group components. If the number of group components is equal to 1, it means that the group has not split. Add the newly created adjacent set to the adjacent log to obtain a new adjacent log L′.

[0180] (4) If the number of vehicle group components is greater than 1, it means that the vehicle group has split due to the reconstruction event. Delete the vehicle nodes in the unmanned vehicle set that are not in the same vehicle group component as the self, and delete the tuples related to these nodes in the new adjacent set. For the new feature matrix, it is also necessary to delete the rows and columns related to these nodes.

[0181] (5) Node v p The new vehicle group structure VG′=(V′, L′, M′) is sent to other nodes via the UPDATE_CHANGE message.

[0182] The vehicle group reconstruction algorithm is shown in Algorithm 5.

[0183]

[0184]

[0185]

[0186] Algorithm 5 can be divided into four parts: the first part is node v p After changing lanes, obtain a new set of neighbor nodes (lines 1-15); the second part is node v p A new neighbor set and feature matrix are constructed based on the new neighbor node set, and the number of vehicle group components at this time is calculated (lines 16-29); the third part updates the structure of the vehicle group according to whether the number of vehicle group components is 1 (lines 30-46); the fourth part is to notify other nodes in the vehicle group of the new vehicle group structure (lines 47-50).

[0187] Step 3: Dynamic evolution method of driverless vehicle group

[0188] (1) Vehicle group status

[0189] After the formation of the unmanned vehicle group, the corresponding dynamic evolution events will occur according to the messages sent by other unmanned vehicles or the changes of its own adjacent sets and feature matrices. A non-deterministic finite state automaton (NFA) is given to describe the dynamic evolution process of the unmanned vehicle group, and the various evolution behaviors that the vehicle group needs to make are judged according to the triggering conditions of various evolution events. The dynamic evolution finite state automaton NFA M of the unmanned vehicle group is represented by a five-tuple:

[0190] NFA M=(S, ∑, Δ, s, F) (12)

[0191] Where S = (S i |I = 0, 1, ..., n} represents the set of all possible states of the vehicle group in the dynamic evolution event; ∑ = {∑ i |i=1,...,n} refers to the input alphabet of the finite state automaton and represents the conditions for triggering various dynamic evolution events; Δ represents the set of state transition functions of the finite state automaton of the dynamic evolution of the unmanned vehicle swarm; s represents the set of initial states of the finite state automaton of the dynamic evolution of the unmanned vehicle swarm; F represents the set of terminal states of the finite state automaton of the dynamic evolution of the unmanned vehicle swarm.

[0192] The status of the vehicle group is as follows:

[0193] (1) Collaborative state S0: After the formation of the unmanned vehicle group or when no dynamic evolution event occurs, the vehicle group is in a collaborative state. This state is the initial state of the dynamic evolution finite state automaton of the unmanned vehicle group. In the unmanned vehicle group in a collaborative state, the node speed similarity NSS between the nodes with adjacent relationships is higher than the threshold. In this state, the nodes in the vehicle group need to periodically perform the following actions:

[0194] ① The node sends a SYN message to the nodes in its neighbor node set to confirm whether the communication connection can be maintained, and attaches its latest driving log in the SYN message

[0195] ② After receiving the SYN message, the node replies with an ACK message to confirm that the communication connection is still stable, and also needs to attach its latest driving log.

[0196] ③ The node periodically broadcasts HELLO messages to the surrounding area, intending to find nodes of other vehicle groups within the communication range and prepare for possible vehicle group merger events.

[0197] ④After receiving the HELLO message, the node replies with a BUILT message, indicating that it is in the group of driverless cars.

[0198] In the definition of the finite state automaton for the dynamic evolution of a swarm of unmanned vehicles, the cooperative state S0 is the only element of the initial state set s, that is, s = {S0}.

[0199] When the driverless car group receives a joining request from a node, the car group is in the joining state. In this state, the node that receives the joining request needs to determine whether to agree to the node joining according to the conditions in Algorithm 1.

[0200] (3) Leaving state S2: When a node in the unmanned vehicle group loses communication with its neighboring nodes, the vehicle group is in the leaving state. In this state, the node needs to calculate the number of vehicle group components. And according to the calculation result of the number of vehicle group components, the adjacent set of the vehicle group is modified to calculate the new feature matrix.

[0201] (4) Merge state S3: When the driverless car group receives a merge request from another car group, the car group is in the merge state. In this state, the node that receives the join request needs to determine whether to agree to the car group merge according to the conditions in Algorithm 3.

[0202] (5) Split state S4: When the distance between two nodes in the following state in the unmanned vehicle group is too large, the vehicle group is in a split state. In this state, the vehicle group needs to further determine whether to split based on conditions such as speed and number of vehicle group components.

[0203] (6) Reconstruction state S5: When a vehicle node of the driverless vehicle group changes lanes, the vehicle group is in the reconstruction state. In this state, the lane-changing node needs to construct a new set of neighbor nodes and calculate the new vehicle group component it belongs to. And based on the calculation results, the adjacent set of the vehicle group is modified and a new feature matrix is ​​calculated.

[0204] (7) Update state S6: When the swarm structure needs to be updated, the swarm is in the update state. The nodes in the swarm in the update state need to update the swarm structure.

[0205] (8) Extinction state S7: When the number of vehicle nodes in the swarm is less than 2, the swarm is in the extinction state. The nodes in the swarm in the extinction state delete the adjacent set and feature matrix and become nodes that have not joined the swarm. This state is also the only element in the terminal state set of the finite state automaton of the dynamic evolution of the unmanned vehicle swarm, that is, F = {S7}.

[0206] (2) Dynamic evolution event trigger conditions and state transition functions

[0207] (1) Node joining ∑1: When a vehicle node of the unmanned vehicle group in the collaborative state S0 receives a JOIN message, a node joining event is triggered. The corresponding state transition function is δ(S0, ∑1) = S1, which means that the vehicle group in the collaborative state S0 enters the joining state S1 after receiving the node joining request.

[0208] (2) Node leaving ∑2: When a vehicle node of the unmanned vehicle group in the cooperative state S0 does not receive an ACK message within a period of time after sending a SYN message, it means that the communication connection with the neighboring node is lost, triggering a node leaving event. The corresponding state transition function is δ(S0, ∑2) = S2, indicating that the vehicle group in the cooperative state S0 enters the leaving state S2 when the node loses the communication connection.

[0209] (3) Group merging ∑3: When the vehicle node of the unmanned vehicle group in the collaborative state S0 receives the MERGE message, the group merging event is triggered. The corresponding state transition function is δ(S0, ∑3) = S3, which means that after the group in the collaborative state S0 receives the merge request from another group, it enters the merge state S1.

[0210] (4) Group split ∑4: When a vehicle node in the unmanned vehicle group in the cooperative state S0 receives an ACK message from a node in the following relationship, and the distance calculated by the Location in the driving log is greater than the threshold δ, the group split event is triggered. The corresponding state transition function is δ(S0, ∑4) = S4, indicating that the vehicle group in the cooperative state S0 enters the split state S4.

[0211] (5) Group Reconfiguration ∑5: When a vehicle node in the unmanned vehicle group in the cooperative state S0 changes lanes, a group reconfiguration event is triggered. The corresponding state transition function is δ(S0, ∑5) = S5, indicating that the vehicle group in the cooperative state S0 enters the reconfiguration state S5.

[0212] (6) Approve request∑6: When the vehicle group in the joining, leaving, merging, splitting, and reconstructing states passes the conditional judgment and finds that the conditions are met, the structure of the vehicle group is ready to be updated. The corresponding state transition function is as follows:

[0213] ①δ(S1,∑6)=S6: The vehicle group in the joining state S1 decides to agree to the node joining the vehicle group after conditional judgment, and the vehicle group enters the updating state S6;

[0214] ②δ(S2,∑6)=S6: After calculating the new adjacent set and feature matrix, the group in the leaving state S2 sends the group information to the remaining nodes in the group, and the group enters the updating state S6;

[0215] ③δ(S3,∑6)=S6: The vehicle group in the merge state S3 agrees to merge after conditional judgment, and the vehicle group enters the update state S6;

[0216] ④δ(S4,∑6)=S6: After calculating the new adjacent set and feature matrix, the group in the split state S4 sends the group information to the remaining nodes in the group, and the group enters the update state S6;

[0217] ⑤δ(S5,∑6)=S6: After calculating the new adjacent set and feature matrix, the car group in the reconstruction state S5 sends the car group information to the remaining nodes in the car group, and the car group enters the update state S6.

[0218] (7) Swarm maintenance ∑7: When the swarm is in the update state S6, the unmanned swarm maintenance method is called to update the swarm structure stored in the nodes in the swarm. When all nodes are updated, the swarm returns to the collaborative state S0. The corresponding state transition function is δ(S6, ∑7) = S0, indicating that the swarm in the collaborative state S0 enters the reconstruction state S5.

[0219] (8) Reject request∑8: When the vehicle group in the joining and merging states is found to not meet the corresponding conditions after conditional judgment, it returns to the collaborative state S0. The corresponding state transition function is as follows:

[0220] ①δ(S1,∑8)=S0: The vehicle group in joining state S1 rejects the joining of the node after conditional judgment, and the vehicle group returns to the collaborative state S0;

[0221] ②δ(S3,∑8)=S0: The vehicle group in the merge state S3 refuses to merge after conditional judgment, and the vehicle group returns to the cooperative state S0;

[0222] (9) Group extinction∑9: When the number of vehicle nodes in the group component where the node in the group is located in the leaving, splitting, or reconstructing state is less than 2, the group enters the extinction state S7. The corresponding state transition function is as follows:

[0223] ①δ(S2,∑9)=S7: The number of nodes in the group leaving state S2 is less than 2, and the group enters the extinction state S7;

[0224] ②δ(S4,∑9)=S7: The number of nodes in the car group in the split state S4 is less than 2, and the car group enters the extinction state S7;

[0225] ③δ(S5,∑9)=S7: The number of nodes in the car group in the reconstruction state S5 is less than 2, and the car group enters the extinction state S7;

[0226] The state transition table of the finite state automaton of the dynamic evolution of the unmanned vehicle group is shown in Table 2, and its state transition diagram is shown in Figure 8shown.

[0227] In summary, the detailed definition of the dynamic evolution finite state automaton NFA M of the unmanned vehicle group is as follows:

[0228] NFA M=(S, ∑, Δ, s, F) (13)

[0229] Where S = {S i |i=0, 1,...,7}, ∑={∑ i |i=1,...,9}, Δ={δ(S0,∑1)=S1, δ(S0,∑2)=S2, δ(S0,∑3)=S3, δ(S0,∑4)=S4, δ(S0,∑5)=S5, δ(S1,∑6)=S6, δ(S1,∑8)=S0, δ(S2,∑6)=S6 , δ(S2, ∑9)=S7, δ(S3, ∑6)=S6, δ(S3, ∑8)=S0, δ(S4, ∑6)=S6, δ(S4, ∑9)=S7, δ(S5, ∑6)=S6, δ(S5, ∑9)=S7, δ(S6, ∑7)=S0}, s=S0, F={S7}.

[0230] (3) Dynamic Evolution Method

[0231] According to the state transition logic of the finite state automaton of the dynamic evolution of unmanned vehicle swarms, a dynamic evolution method of unmanned vehicle swarms is given. The dynamic evolution method of unmanned vehicle swarms describes the state transition process that occurs when the swarm encounters conditions in various states. The specific steps are shown in Algorithm 6

[0232]

[0233]

[0234] Step 4: Unmanned vehicle group collaboration method

[0235] (1) Unmanned vehicle swarm collaboration model

[0236] When the unmanned vehicle group is in the cooperative state S0 during the dynamic evolution process, if the node speed similarity and node dependence between the nodes with adjacent relationships in the vehicle group are greater than the threshold, and the overall speed chaos of the vehicle group is less than the threshold, the unmanned vehicle group reaches the cooperative state. The unmanned vehicle group cooperation model (Cooperation Model, CM) is as follows:

[0237]

[0238] Among them, S nIndicates the dynamic evolution state of the unmanned vehicle group in which the vehicle group is currently located; NFA M=(S, ∑, Δ, s, F) represents the dynamic evolution finite state automaton of the unmanned vehicle group; m ij represents the element in the i-th row and j-th column of the feature matrix in the vehicle group VG = (V, L, M); NSS (v i , v j ) represents node v i and v j SC(VG) represents the speed disorder of the vehicle group VG; ND(VG) represents the node dependence of the vehicle group VG.

[0239] The meaning of the unmanned vehicle group coordination model expressed by formula (14) is explained as follows: first, it is necessary to ensure that the overall properties of the vehicle group meet certain conditions, that is, the node speed similarity, vehicle group chaos and node dependence between adjacent vehicles are not higher than or lower than the corresponding thresholds; secondly, the current dynamic evolution state of the vehicle group is the coordination state S0.

[0240] The coordination of unmanned vehicle swarms means that the vehicle nodes in the swarm share data and exchange information, so that the nodes can obtain more information than a single vehicle, and the nodes in the swarm share information so that the swarm status stored by each node is consistent. The following is an analysis and proof of the coordinated status of unmanned vehicle swarms.

[0241] Proposition: A necessary and sufficient condition for a swarm of unmanned vehicles to achieve coordination is that the current state of the finite state automaton of the dynamic evolution of the swarm of unmanned vehicles is the coordination state S0.

[0242] prove:

[0243] (1) Sufficiency, that is, to prove that when the swarm of unmanned vehicles achieves coordination, the current state of the finite state automaton of the dynamic evolution of the swarm of unmanned vehicles is the coordinated state S0.

[0244] The following proof by contradiction is used to prove the sufficiency. It is assumed that when the unmanned vehicle group dynamically evolves and the unmanned vehicle group reaches coordination, the current state of the unmanned vehicle group dynamic evolution finite state automaton is any one of the joining state S1, leaving state S2, merging state S3, splitting state S4, reconstructing state S5, updating state S6 or extinction state S7.

[0245] ① If the swarm is in the joining state S1 and the merging state S3, a node in the swarm is calculating the node speed similarity or the speed chaos of the swarm, and cannot exchange information with other nodes in the swarm, which violates the condition of maintaining data sharing during swarm collaboration;

[0246] ② If the swarm is in the leaving state S2, splitting state S4 or reconstructing state S5, a node in the swarm is calculating the number of swarm components to determine whether the swarm will split at the next moment, and cannot exchange information with other nodes in the swarm, violating the condition of maintaining data sharing when the swarm is coordinated;

[0247] ③ If the swarm is in the updating state S6, even if all nodes are exchanging information and sharing data at this time, when the swarm status is updated, the swarm structures stored by different nodes are different, which is contrary to the swarm status stored by each node through information sharing;

[0248] ④ If the vehicle group is in the extinction state S7, that is, the number of vehicles is less than 2, it is impossible to form a vehicle group.

[0249] In summary, it is proved that when the unmanned vehicle group achieves coordination, the current state of the unmanned vehicle group dynamic evolution finite state automaton is the coordination state S0.

[0250] (2) Necessity: That is, to prove that when the current state of the finite state automaton of the dynamic evolution of the unmanned vehicle group is the cooperative state S0, the unmanned vehicle group achieves coordination.

[0251] When the swarm is in the collaborative state S0, the nodes in the swarm need to periodically send SYN messages to neighboring nodes, and reply with ACK messages after receiving the SYN messages, which satisfies the condition that the nodes maintain information exchange when the unmanned vehicle swarm reaches collaboration. In addition, the nodes in the swarm also need to periodically broadcast HELLO messages in order to sense whether there are other unmanned vehicles around, and obtain more information than a single unmanned vehicle. Therefore, when the current state of the finite state automaton of the dynamic evolution of the unmanned vehicle swarm is the collaborative state S0, the unmanned vehicle swarm reaches collaboration.

[0252] Combining sufficiency and necessity, it is proved that the necessary and sufficient condition for the unmanned vehicle group to achieve coordination is that the current state of the dynamic evolution finite state automaton of the unmanned vehicle group is the coordinated state S0.

[0253] (2) Autonomous vehicle swarm collaboration algorithm

[0254] The specific steps of the unmanned vehicle group collaboration algorithm are as follows:

[0255] (1) Nodes in the swarm broadcast HELLO packets to the surroundings and send SYN packets to neighboring nodes. After receiving a HELLO message or a SYN message, they reply with a BUILT or ACK message, respectively.

[0256] (2) When a node in the swarm receives a JOIN message, Algorithm 1 is executed and a node joining event occurs.

[0257] (3) When a node in the swarm does not receive an ACK message after sending a SYN message, Algorithm 2 is executed to calculate the swarm component. If it is greater than 1, a node departure event occurs; otherwise, the swarm structure is updated. After the node departure is completed, the swarm is determined to be extinct based on the number of remaining vehicles in the swarm.

[0258] (4) When a node in the swarm receives the MERGE message, it executes Algorithm 3 to calculate the swarm chaos degree and decides whether a swarm merge event occurs based on the result.

[0259] (5) When the distance between a node in the group and its following node exceeds the threshold, execute Algorithm 4 to calculate the group component after removing the following relationship. If it is greater than 1, a group split event occurs; otherwise, the group structure is updated. After the group splits, determine whether the group has disappeared based on the number of remaining vehicles in the group.

[0260] (6) When a node in the vehicle group changes lanes, Algorithm 5 is executed and a vehicle group reconstruction event occurs. After the event ends, the number of remaining vehicles in the vehicle group is used to determine whether the vehicle group has disappeared.

[0261] The unmanned vehicle group collaboration algorithm is shown in Algorithm 7.

[0262]

[0263]

[0264]

[0265] Algorithm 7 can be divided into 6 parts. The first part is that the swarm broadcasts HELLO messages in a collaborative state and sends SYK messages to neighbors (lines 1-6). The second to sixth parts represent node joining (lines 7-10), node leaving (lines 11-20), swarm merging (lines 21-26), swarm splitting (lines 27-36) and swarm reconstruction (lines 37-43).

[0266] The sequence diagram of the unmanned vehicle group collaboration method is as follows: Fig. 9 shown.

[0267] Step 5: Simulation experiment and result analysis

[0268] (1) Simulation experiment parameter setting and method description

[0269] The present invention uses SUMO and NS-3 to conduct simulation experiments. The various parameter settings of the simulation experiments are shown in Table 3.

[0270] The present invention uses node dependency, the number of vehicle group extinctions and the node participation rate to evaluate the effect of the dynamic evolution method AMDEM. In addition, the average cooperation time ratio (AverageCooperationTimeRate, ACTR) and the average survival time of the vehicle group are introduced to evaluate the unmanned vehicle group cooperation algorithm. The average cooperation time ratio refers to the average value of the proportion of the time that all vehicle groups are in a cooperative state in the simulation experiment, that is,

[0271]

[0272] Among them, Cooperation time i Indicates vehicle group VG i The duration of the collaborative state during the simulation experiment, Survivaltime i Indicates vehicle group VG i The survival time of n is , and n represents the number of vehicle groups formed during the simulation experiment.

[0273] The specific experimental steps are as follows:

[0274] (1) Use the netedit plug-in of the traffic simulation software SUMO to generate a 10 km one-way four-lane highway with speed limits set from left to right as 120 km / h, 120 km / h, 100 km / h, and 100 km / h. Set a node every 2 km as the exit of the highway.

[0275] (2) Place 100 vehicles in the scenario set in (1), set the maximum speed between 65km / h and 120km / h, and conduct an experiment every 5km / h for a total of 12 times.

[0276] (3) Write the logic of the unmanned vehicle group formation method AMAVGF in NS-3, use the 12 traffic simulation results generated by SUMO in (2) as the input of NS3, and obtain the behavior of each node at each moment.

[0277] (4) According to the NS-3 simulation results of (3), the average survival time of the swarm, the number of swarm deaths, the node participation rate and the average proportion of the coordination time are calculated, and the node dependency at each moment is calculated. In terms of dynamic evolution, AMDEM is compared with EDEM and SCCDEM to analyze the advantages and disadvantages of AMDEM. In terms of unmanned vehicle swarm coordination, CM is compared with DEModel to evaluate the performance of CM.

[0278] (2) Analysis of simulation experiment results

[0279] For the dynamic evolution method, the number of vehicle group deaths and node participation rates under different speed limit conditions were calculated, and AMDEM was compared with the EDEM method and the SCCDEM method. The average value of the node dependency of each vehicle group under different speed limit conditions was also calculated over time. For vehicle group collaboration, the average survival time and average collaboration time ratio of vehicle groups under different speed limit conditions were calculated, and CM was compared with DEModel.

[0280] (1) Analysis of dynamic evolution experimental results

[0281] ①Number of car group deaths

[0282] like Fig.10 As shown in the figure, the blue line, red line and green line represent the number of group deaths of AMDEM, EDEM and SCCDEM under different node maximum speed conditions. Among them, the number of group deaths of AMDEM is the lowest under various speed conditions, with an overall reduction of 42.29%, which reflects the superiority of considering the group splitting caused by node departure. As the maximum speed of the node continues to increase, the three curves generally show an increasing trend, indicating that it is indeed difficult for vehicles to maintain a stable structure of the group under high-speed conditions. However, the number of group deaths corresponding to AMDEM does not increase significantly with the increase of the maximum speed of the node, indicating that AMDEM has a certain effect in dealing with high-speed nodes. However, compared with Chapter 3, the number of group deaths of AMDEM has increased slightly. This is because when AMDEM defines group splitting, when the distance between vehicles exceeds the threshold δ, the node will start to judge the group splitting. When the number of group components of the node is less than 2 when the group splits, the group will die, so the number of group deaths of AMDEM will increase slightly compared with AMAVGF.

[0283] ②Node participation rate

[0284] like Fig.11As shown in the figure, the blue line, red line and green line represent the node participation rate of AMDEM, EDEM and SCCDEM under different node maximum speed conditions. It can be seen that the node participation rate of AMDEM is much higher than that of EDEM and SCCDEM under both high and low speed conditions, and the overall increase is 14.55%. When the maximum speed of the node is in the range of 85km / h-100km / h, the node participation rate of AMDEM is slightly higher than that of SCCDEM, but still much higher than that of EDEM. In addition, the average survival time of the same vehicle group and the number of vehicle group extinctions have similar trends, and the node participation rate of AMDEM will still decrease with the increase of the maximum speed of the node. In Chapter 3, the node participation rate of AMAVGF was once lower than that of SCCAVGF and AVGFM when the maximum speed of the node was at a high level, but after considering the dynamic evolution, the node participation rate of AMDEM has exceeded SCCDEM and EDEM under high speed conditions, which further illustrates the improvement of AMDEM on the survivability of the vehicle group.

[0285] ③Node Dependency

[0286] like Fig.12 As shown in the figure, the blue line, red line and green line respectively represent the change of the average value of node dependence over time when the maximum speed of the node is 65km / h, 70km / h and 75km / h. It can be seen that even if the curve of 70km / h and the curve of 75km / h intersect, the average value of node dependence will generally increase with the increase of the maximum speed of the node, and the change is relatively gentle. This is consistent with the conclusion drawn from the experiment in Chapter 3. In addition, compared with AMAVGF under the same speed conditions, the average value of node dependence of AMDEM is about 0.15 higher overall, which shows that the vehicle group reconstruction event plays a role, that is, when the vehicle changes lanes, the adjacent set and feature matrix are recalculated, the closeness of the connection between nodes is maintained, and the node dependence is improved.

[0287] like Fig.13 As shown in the figure, the blue line, red line and green line respectively represent the change of the average value of node dependence over time when the maximum driving speed of the node is 80km / h, 85km / h and 90km / h. As can be seen from the figure, only the average value of node dependence at 85km / h has a slightly larger change, and the average value of node dependence under other speed conditions is still relatively smooth. Therefore, the properties of node dependence under this speed condition are roughly consistent with those under low-speed driving. At the same time, compared with AMAVGF under the same speed condition, the average value of node dependence of AMDEM has also increased, reflecting the positive impact of the vehicle group reconstruction event on the closeness between the middle points of the vehicle group.

[0288] like Fig.14As shown in the figure, the blue line, red line and green line respectively represent the change of the average value of node dependence over time when the maximum driving speed of the node is 95km / h, 100km / h and 105km / h. With the increase of the maximum driving speed of the node, the average value of node dependence increases further, and has reached about 0.8 under medium and high speed conditions. This shows that compared with AMAVGF, the formed vehicle group is easier to travel on the highway in a smaller scale after the dynamic evolution event of AMDEM. The smaller the scale of the vehicle group, the smaller the proportion of Unrelated in the feature matrix, and the larger the average value of node dependence reflected. From a safety perspective, under medium and high speed conditions, vehicles need to maintain a larger safety distance, so it is not easy to form a larger vehicle group. Small-scale vehicle groups can not only enable vehicles to obtain more effective information to achieve coordination when driving, but also ensure the safety of vehicle driving.

[0289] like Fig.15 As shown in the figure, the blue line, red line and green line respectively represent the change of the average value of node dependence over time when the maximum driving speed of the node is 110km / h, 115km / h and 120km / h. Among them, the average value of node dependence of the two curves of 115km / h and 120km / h is equal to 1 at certain moments. From the calculation method of node dependence, it can be concluded that the topological structure inside the car group formed at this time is all fully connected. It further shows that under high-speed conditions, the scale of the car group maintained by dynamic evolution is further reduced. Most car groups are maintained at about 2-3 vehicles, so it is easy to form a fully connected topological structure. Under this condition, once a node leaves or a car group splits, the car group is likely to die because the number of vehicles is less than 2, and the remaining car groups are still mostly small-scale car groups, maintaining a fully connected topological structure.

[0290] In summary, AMDEM's average survival time, number of group extinctions, and node participation rate are still better than EDEM and SCCDEM. Compared with its own formation method AMAVGF, AMDEM has a higher average survival time and higher node participation rate. The slight increase in the number of group extinctions reflects that the group split caused by the node departure event has played a role, ensuring the reasonable stability of the group topology. In terms of the average proportion of collaborative time, AMDEM is better than SCCDEM under low-speed conditions and has room for improvement under medium and high-speed conditions. In terms of node dependence, the node dependence of the group under different speed conditions has been improved to a certain extent, reflecting the effectiveness of the dynamic evolution method.

[0291] (2) Analysis of vehicle group collaboration experiment results

[0292] ① Average survival time of vehicle groups

[0293] The performance of the average node survival time under different speed conditions is as follows Fig.16 As shown. The blue line and green line represent the average survival time of the vehicle group of CM and DEModel under different node maximum speed conditions. The average survival time of the vehicle group of CM under various speed conditions is higher than that of DEModel. In addition, CM has a downward trend when the maximum driving speed of the node is higher than 100km / h, indicating that under high-speed driving conditions, the average survival time of the vehicle group will become shorter. The possible reason is that when the vehicle is driving at a high speed, it changes lanes to the two lanes on the left and gradually separates from the original vehicle group, causing the vehicle group to disappear and unable to maintain a long average survival time of the vehicle group. But even so, CM performs slightly better under high-speed driving conditions. Overall, the average survival time of the vehicle group of CM has increased by 87.52%, which is a significant increase.

[0294] ②Average collaborative time ratio

[0295] like Fig.17 As shown in the figure, when the maximum speed of the node is between 65km / h and 80km / h, the average coordination time ratio of CM is slightly better than that of DEModel. This shows that CM can make the vehicle group better coordinated under low-speed conditions. In the case of 85km-105km / h, the difference between the two is not obvious; when the speed is higher than 105km / h, the average coordination time ratio of CM is lower than that of DEModel. Overall, the average coordination time ratio of CM under low and medium speed conditions is relatively stable, which can be maintained between 0.5-0.6, but the average coordination time ratio will drop to 0.4-0.5 under high speed conditions. But overall, the average coordination time ratio of CM is also 1.64% higher than that of DEModel, which has a certain improvement. Considering the average survival time of the vehicle group, the survival time of the vehicle group in the high-speed scenario is short, and the time spent by the vehicle group for evolution and subsequent maintenance is relatively fixed. Therefore, the proportion of time that the vehicle group is in a non-cooperative state increases, and the proportion of time in a cooperative state decreases.

[0296] In summary, the average group survival time of CM under various speed conditions is better than that of DEModel, indicating that CM has better group coordination effect. The average coordination time ratio of CM under low speed conditions is better than that of DEModel, and the effect is similar to that of DEModel under medium speed conditions, and is lower than that of DEModel under high speed conditions, which means there is room for improvement.

[0297] Innovation

[0298] In response to the current problems that unmanned vehicle groups do not take into account the historical driving status of the vehicle group and the changes in the vehicle group structure, the present invention is based on ("Unmanned Vehicle Group Construction Method Based on Process Mining in Highway Scenarios" applied by inventors Cheng Jiujun and others on January 5, 2024 (Applicant: Cheng Jiujun Tongji University, Patent Application Number: 2024100138829)) and provides a node joining algorithm, a node leaving algorithm, a vehicle group merging algorithm, a vehicle group splitting algorithm and a vehicle group reconstruction algorithm; defines the vehicle group state, dynamic evolution event trigger conditions and state transition functions, and provides an unmanned vehicle group dynamic evolution algorithm; provides an unmanned vehicle group collaboration model, and on this basis, provides an unmanned vehicle group collaboration algorithm; provides evaluation indicators such as the number of vehicle group extinctions, node participation rate, node dependence, average vehicle group survival time and average collaboration time ratio, and constructs a simulation experiment to verify the stability of the proposed unmanned vehicle group. Selecting vehicle nodes with similar instantaneous and historical motion states to form a vehicle group can improve the time of mutual cooperation between vehicle nodes in the vehicle group and maintain the stability of the vehicle group structure, thereby effectively realizing beyond-visual-range perception and improving the ability of autonomous driving vehicles to understand the surrounding driving environment. It provides effective guarantee for the stable and orderly driving of unmanned driving movement behavior, and has important significance and application value for promoting the healthy and rapid development of unmanned driving.

[0299] Instruction Manual

[0300] Table 1

[0301]

[0302]

[0303] Table 2

[0304]

[0305]

[0306] Table 3

[0307]

Claims

1. The dynamic evolution and coordination method of unmanned vehicle groups based on process mining in the highway scenario according to claim 1, It is characterized in that The steps include: Step 1. Relevant definitions; Step 2. Dynamic evolution events of the driverless vehicle group, including: Step 2.1 Node joins the event; Step 2.2 Node leaving event; Step 2.3: Vehicle group merging event; Step 2.4: Vehicle group split event; Step 2.5: Vehicle group reconstruction event; Step 3: The method for dynamic evolution of the unmanned vehicle group includes: Step 3.1: Vehicle group status; Step 3.2 Dynamic evolution event triggering conditions and state transition functions; Step 3.3 Dynamic evolution method; Step 4: The unmanned vehicle group collaboration method includes: Step 4.1 Unmanned vehicle group collaboration model; Step 4.2 Unmanned vehicle group collaboration algorithm.

2. The dynamic evolution and coordination method of unmanned vehicle groups based on process mining in a highway scenario as claimed in claim 1, It is characterized in that The relevant definitions of step 1 are: Definition 1 Node Join: The unmanned vehicle node that has not joined the swarm has a new neighbor relationship with the nodes in the swarm. It sends a request message to join the swarm. After the conditions of the nodes in the swarm are judged, it agrees to join the swarm. Its formal representation is as follows: Among them, VG=(V,l,M) represents the unmanned vehicle group, v q It is a node in the car group VG; Definition 2: Node Leave: A node in the driverless vehicle group loses its communication connection with other nodes in the group due to communication failure, vehicle failure, excessive distance between vehicles, or leaving the highway, and loses its neighbor relationship with other nodes. Its formal representation is as follows: Among them, node v p It is a node in the vehicle group VG = (C, L, M) and has no adjacent relationship with any other node in the vehicle group except itself; Definition 3: Vehicle group merging (Merge): Two vehicle nodes from different driverless vehicle groups have a new adjacent relationship and are merged into a new vehicle group after negotiation. Its formal representation is as follows: Among them, VG 1 =(V 1 ,L 1 ,M 1 ) and VG 2 =(V 2 ,L 2 ,M 2 ) represents two driverless car groups, node v p and v q From the car group VG 1 and VG 2 ; Definition 4: Split: When the distance between the two nodes in the following relationship is too large, the unmanned vehicle group will automatically split into two new vehicle groups after negotiation. The formal representation is as follows: Among them represents v p and v q represents the distance between; θ represents the threshold of the vehicle distance between nodes, and satisfies δ < COMM, where the latter represents the maximum communication distance between vehicles; Definition 5: Change: When a vehicle node in a driverless vehicle group changes lanes, the neighbor node set of the node changes, the adjacent set is updated, and the structure of the vehicle group changes; however, no node joins or leaves the vehicle group itself; its formal representation is as follows: Where VG = (v, L, M) represents the group of unmanned vehicles; L t ∈L represents the latest adjacent set before the group is reconstructed; L t+1 represents the adjacent set after the vehicle group is reconfigured; V and V′ represent the unmanned vehicle node sets before and after the vehicle group is reconfigured, respectively; Definition 6 Dissolution: After a node leaves or the group splits, the number of vehicles in the group decreases to less than two and the group size cannot be maintained. This indicates that the group has dissoluted. The formal representation is as follows: Where VG = (V, L, M) represents the unmanned vehicle group, |V| represents the number of nodes in the unmanned vehicle set; Definition 7 Vehicle Group Component (VGC): When the structure of the unmanned vehicle group evolves dynamically, the vehicle group structure changes. All nodes that a node can access through the unmanned vehicle group maintenance method constitute the vehicle group component where the node is located, that is, the set of nodes that have a communication path with the node constitutes a vehicle group component; its formal representation is as follows: VGC(v p )={v|Path(v,v p )>0,v∈V-{v p }} (7) Among them, Path(v,v p ) represents node v and node v p The number of communication connections between them; V represents the set of unmanned vehicles; Definition 8: Vehicle Group Component Count (VGCC): The number of vehicle group components contained in the dynamically evolving unmanned vehicle group; its formal expression is as follows: VGCC(M)=|{VGC(v i )|v i ∈V,i=1,2,…,N}| (8) Definition 9 Speed ​​Confusion (SC): The degree of inconsistency between the instantaneous speed of all nodes in the vehicle group and the historical average speed, that is, Where N is the number of vehicles in the driverless car group; α and β are two coefficients; Represents node v i The instantaneous velocity vector of Indicates the average instantaneous speed of all nodes in the vehicle group; Avgspeed i Represents node v i The historical average speed of Represents the average of the historical average speeds of all nodes in the vehicle group; and The formal expression of is shown in formula (10) and formula (11):

3. The dynamic evolution and coordination method of unmanned vehicle groups based on process mining in a highway scenario as claimed in claim 1, It is characterized in that, Step 2: Dynamic evolution of driverless car group: Step 2.1: Node joining event When an unmanned vehicle node that has not yet joined the swarm encounters a node that has already joined the swarm, it sends a message to the node in the swarm requesting it to join the swarm. After receiving the message, the node in the swarm determines whether to allow the node to join the swarm based on a series of judgment conditions. If it disagrees, it replies with a rejection message. If agreed, the adjacent set and feature matrix are updated, and the updated vehicle group information is sent to the node, and the node is added to the vehicle group; The specific steps of the node joining event are as follows: (1) Node v that has not yet joined the swarm p Broadcast a HELLO message to the surroundings, and attach the latest driving log, intending to join the swarm or form a swarm with other nodes that have not yet joined the swarm; (2) Node v that is already in the vehicle group VG = (V, L, M) q After receiving the HELLO message, reply with a BUILT message to indicate that you have joined the group. (3) If at this time node v p If it is still in the state of not joining the car group, then it will send a q Send a join request JOIN; (4) Node v q After receiving the JOIN message, the following three conditions are used to determine whether to allow the node to join: ① ②NSS(v p ,v q )≥ε, ε is the node speed similarity threshold; ③SL p =SL q If any of the conditions are not met, then v q To v p Send a DENY_JOIN message to reject v p If all three conditions above are met, then v q To v p Send AGREE_JOIN message to indicate agreement p The joining request is made and in its neighbor node set Neighbor p Add node v q ; Node v p After receiving the AGREE_JOIN message, it builds its own adjacent set and adds elements vq and vp to it; (5) Construct a new adjacent set L t+1 , node v q According to Adjacent(v p ,v q ) and Adjacent(v q ,v p ) is true or false to decide whether to p ,v q ) and (v q ,v p ) add a new adjacent set, and calculate a new feature matrix based on the new adjacent set through the Alpha Miner algorithm; let the new adjacent log be L′=L∪{L t+1 }, and obtain the new vehicle group VG′=(V′,L′,M′); (6) Send the new group structure to other nodes in the group, including the node v that has just joined the group. p ; Step 2.2 Node leaving event When a node in a swarm detects that it has lost communication with another node, it may be caused by a vehicle failure, communication failure, or leaving the highway, so the node needs to leave the swarm. It may also be simply because the distance between the two nodes exceeds the maximum range of the communication connection, but the node still maintains communication with other nodes and does not leave the swarm. Regarding the former, if a node leaves the swarm, the nodes in its neighboring set will lose the communication connection with the node that left the swarm. The node's departure can be divided into the following two cases: (1) Delete the node from the group structure; if a node on the edge leaves the group for some reason, the node is deleted based on the original group structure; (2) The group is split into two or more sub-groups. The middle node acts as a medium to connect the two parts of the group. When the node leaves the group, the topological structure of the original group is destroyed and split into two groups. In order to judge whether the departure of the node will cause the group to split, the concept of vehicle group component (VGC) is introduced for judgment; the concept of vehicle group component is borrowed from the connected component (CC) in graph theory; in graph theory, the maximal connected subgraph of an undirected graph is called the connected component of the graph; in a group, all nodes that can be accessed by the unmanned vehicle group maintenance method constitute a vehicle group component; when a node leaves the group, the number of vehicle group components (VGCC) of the group composed of the remaining nodes is calculated; if the calculated number of vehicle group components is equal to 1, it means that the group has not split due to the departure of the node, and the node is deleted; if the calculated number of vehicle group components is greater than 1, it means that the group is split into multiple sub-groups due to the departure of the node; but in either case, only the nodes that are still in the same vehicle group component as itself need to be retained; The specific steps of the node leaving event are as follows: (1) Vehicle group VG = node v in (V, L, M) p Detects that node v cannot be found in the set of neighbor nodes q Continue to communicate, then (v p ,v q ) and (v q ,v p ) is removed from the adjacent set, and a new feature matrix M′ is calculated, and m′ in the feature matrix is ​​replaced pq and m′ qp Modified to Unrelated; Then calculate the number of vehicle group components of the vehicle group at this time; (2) If the calculated number of vehicle group components is still equal to 1, it means that node v q Still communicating with other nodes and not leaving the group; then node v p The node v q Delete from the neighbor node set, and update the group structure VG′ according to the neighbor set and feature matrix in step (1); (3) If the calculated number of vehicle group components is greater than 1, remove the nodes in the unmanned vehicle set V that are not in the vehicle group component to which it belongs, and delete the elements related to these nodes in the neighboring set; at the same time, delete the rows and columns where these elements are located in the feature matrix to obtain the updated vehicle group VG′; in addition, delete these nodes in its neighboring node set; (4) Call the unmanned vehicle group maintenance method and send a LEAVE message to other nodes in the vehicle group, attaching the changed vehicle group VG′; After receiving the LEAVE message, other nodes update the group structure and their own neighbor node set; Step 2.3 Vehicle group merging event There are differences in the speed of nodes in the car group, and the consistency of the speed of all nodes in the two car groups should be considered; the speed chaos is used to measure the consistency of the speed of nodes in the car group; the car group chaos degree consists of two parts, the first part is the variance of the current instantaneous speed of the nodes in the car group, and the second part is the variance of the average speed of the nodes in the car group from entering the highway and starting to the current moment; the two are weighted by a pre-set coefficient to obtain the chaos degree of the speed of all vehicles in the current car group; only when the value of the car group chaos degree is less than a certain threshold, can the car group be merged; The specific steps of the vehicle group merging event are as follows: (1) Car Group VG 1 Node v in p During driving, a group of vehicles VG is detected 2 Node v in q ; Node v p Calculate the vehicle group VG 1 The speed disorder SC(VG 1 ), if the speed disorder S(VG 1 ) is lower than the threshold ε, then the vehicle group VG 2 Node v in q Send a MERGE message to indicate a request to merge the vehicle groups; (2) When node v q After receiving the MERGE message, the vehicle groups are temporarily merged first; the specific steps are as follows: ① Take the union of the unmanned vehicle sets of the two vehicle groups to obtain the temporary unmanned vehicle group set V′; ② Calculate the adjacent set, that is, according to Adjacent(v p ,v q ) and Adjacent(v q ,v p ) is true or false to decide whether to p ,v q ) and (v q ,v p ) Add to the adjacent set and get the new adjacent log L′; ③ Calculate the new feature matrix M′ through the Alpha Miner algorithm; Through the above three steps, a temporary vehicle group VG′=(V′,L′,M′) is obtained, and the speed disorder degree SV(VG′) of the temporary vehicle group is calculated; (3) Determine whether to agree to the vehicle group merger based on the following three conditions: ① ②SC(VG′)≤γ (γ is the velocity chaos threshold) ③SL p =SL q If any of the conditions are not met, then v q To v p Send a DENY_MERGE message to reject the group merge request; if the above three conditions are met, then v q To v p Send AGREE_MERGE message to indicate the group merge request and v p Add your own neighbor node set; v p After receiving the AGREE_MERGE message, v q Join your own set of neighbor nodes; (4) Send the new swarm structure to other nodes in the swarm; Step 2.4: Group Split Event In a following relationship, if the speed of the leading vehicle is greater than that of the trailing vehicle, and the distance between the two vehicles is greater than a certain threshold (which is less than the communication range between the vehicles), then there is a risk of communication interruption between the two vehicle nodes; therefore, the vehicle group needs to be split into two smaller vehicle groups through a split event; The specific steps of the car group split event are as follows: (1) There is a node v in the driverless vehicle group VG p Following node v q ; When node v p Detects that the vehicle distance Distance(v q , v p , v q ) between is greater than the threshold θ (θ < COMM, COMM is the communication range of the driverless vehicle), the following vehicle node v p Sends a QUERY_SPILT message to the preceding vehicle node v q , indicating an inquiry about whether node v q Needs vehicle group splitting and attaches its own speed magnitude; (2) Current car node v q After receiving the QUERY_SPILT message, determine the speed of the two vehicles; if point v q The velocity is greater than that of node v p The distance between the two vehicles will gradually increase over time and may exceed the communication range COMM at some point in the future. In addition, it is necessary to determine v p With v q Is the following relationship between them the only medium connecting the front and rear parts of the group? The element m in the original feature matrix pq and m qp The values ​​of are all FOLLOW; at this time, let’s assume that m pq and m qp The value of is changed to Unrelated, and the new characteristic matrix is ​​set as M′. Then the number of vehicle group components of the vehicle group is calculated. If the calculated number of vehicle group components is still 1, it means that even if the follow-up relationship between the two nodes is disconnected, there are other follow-up relationships in the vehicle group so that the vehicle group is still connected. If the calculated number of vehicle group components is greater than 1, it means that if the node v is disconnected p and node v q If there is no following relationship between nodes, the group will be split into two parts; in summary, node v q Whether the group needs to be split needs to be determined based on the following two conditions: ① ②VGCC(M′)>1 If any of the conditions are not met, the preceding node v q Backward car node v p Send a DENY_SPILT message to indicate that it disagrees with the group split; otherwise, send a DENY_SPILT message to the following node v p Send AGREE_SPILT message to indicate agreement to split the group; (3) At the front vehicle node v q Backward car node v p After sending the AGREE_SPILT message, the new swarm structure needs to be calculated; the specific steps are as follows: ① Remove the nodes in the second half of the original vehicle group from the set of unmanned vehicles contained in the original vehicle group; ② Construct a new adjacent log, in which the adjacent set is based on the last adjacent set of the original vehicle group, and deletes the tuple containing the removed node; ③Build a new feature matrix and delete the rows and columns related to the removed nodes; (4)v q Call the unmanned vehicle group maintenance algorithm to send a SPILT message to the nodes in the first half of the vehicle group, notifying them of the vehicle group split and attaching the new vehicle group structure; (5) The following node v p After receiving the AGREE_SPILT message, it is also necessary to build a new car group structure in the same way as step (3) and delete the relevant elements of the first half of the nodes in the original car group; Call the unmanned vehicle group maintenance algorithm to send SPILT messages to the nodes in the second half of the vehicle group; Step 2.5 Vehicle group reconstruction event Vehicle group reconstruction events can be divided into the following two situations: (1) The reconstruction of the vehicle group did not lead to the split of the vehicle group. An unmanned vehicle node changed lanes and overtook other vehicles in turn, causing the structure of the vehicle group to change. However, the vehicle group remained a connected structure and did not split. (2) The reconstruction of the vehicle group leads to the splitting of the vehicle group; After changing lanes, an unmanned vehicle node loses the following relationship with the node behind it, causing the unmanned vehicle group to split; The above two types of vehicle group reconstruction are combined, and finally the number of vehicle group components of the vehicle group is calculated, and classification processing is performed according to the calculation results; The specific steps of the vehicle group reconstruction event are as follows: (1) Autonomous vehicle group VG = node v in (V, L, M) p Changing lanes, broadcasting CHANGE messages to surrounding nodes in the vehicle group; when a node receives the CHANGE message, it sends a p Reply to the RESP_CHANGE message and include your location information; (2) After time τ, node v p Construct a new neighbor node set Neighbor′ based on the received RESP_CHANGE message p ; Create a new adjacent set about node v i ∈Neighbor p , delete the newly created adjacent set (v i ,v p ) and (v p ,v i ) element; about node v i ∈Neighbor′ p , if Adjacent(v p ,v i )=true, then add (v p ,v i ) element; if Adjacent(v i ,v p )=true, then add (v i ,v p )element; (3) Construct a new feature matrix M′ based on the new adjacent set and calculate the number of vehicle group components; If the number of group components is equal to 1, it means that the group has not split; add the newly created adjacent set to the adjacent log to obtain a new adjacent log L′; (4) If the number of vehicle group components is greater than 1, it means that the vehicle group has split due to the reconstruction event; delete the vehicle nodes in the unmanned vehicle set that are not in the same vehicle group component as the self, and delete the tuples related to these nodes in the new adjacent set; for the new feature matrix, it is also necessary to delete the rows and columns related to these nodes; (5) Node v p The new vehicle group structure VG′=(V′, L′, M′) is sent to other nodes via the UPDATE_CHANGE message.

4. The dynamic evolution and coordination method of unmanned vehicle groups based on process mining in a highway scenario as claimed in claim 1, It is characterized in that, The step 3 is a method for the dynamic evolution of the unmanned vehicle group: Step 3.1 Group status After the formation of a swarm of unmanned vehicles, corresponding dynamic evolution events will occur according to the messages sent by other unmanned vehicles or the changes in its own adjacent sets and feature matrices. A non-deterministic finite automation (NFA) is given to describe the dynamic evolution process of the swarm of unmanned vehicles, and the various evolutionary behaviors that the swarm needs to make are judged according to the triggering conditions of various evolutionary events. The dynamic evolution finite automation NFAM of the unmanned vehicle swarm is represented by a five-tuple: NFAM=(S,Σ,Δ,s,F) (12) Where S = {S i |i=0,1,…,n} represents the set of all possible states of the vehicle group in the dynamic evolution event; Σ={Σ i |i=1,…,n} refers to the input alphabet of the finite state automaton, indicating the conditions for triggering various dynamic evolution events; Δ indicates the set of state transition functions of the finite state automaton of the dynamic evolution of the unmanned vehicle group; s indicates the set of initial states of the finite state automaton of the dynamic evolution of the unmanned vehicle group; F indicates the set of terminal states of the finite state automaton of the dynamic evolution of the unmanned vehicle group; The status of the vehicle group is as follows: (1) Collaborative state S 0 :After the formation of the unmanned vehicle group or when no dynamic evolution event occurs, the vehicle group is in a coordinated state; this state is the initial state of the dynamic evolution finite state automaton of the unmanned vehicle group; in the unmanned vehicle group in a coordinated state, the node speed similarity NSS between the nodes with adjacent relationships is higher than the threshold; in this state, the nodes in the vehicle group need to periodically perform the following behaviors: ① The node sends a SYN message to the nodes in its neighbor node set to confirm whether the communication connection can be maintained, and attaches its latest driving log in the SYN message ②After receiving the SYN message, the node replies with an ACK message to confirm that the communication connection is still stable, and also needs to attach its latest driving log; ③ The node periodically broadcasts HELLO messages to the surrounding area, intending to find nodes of other vehicle groups within the communication range to prepare for possible vehicle group merger events; ④After receiving the HELLO message, the node replies with a BUILT message, indicating that it is in the group of driverless cars; In the definition of the finite state automaton for the dynamic evolution of the unmanned vehicle group, the collaborative state S 0 is the only element of the initial state set s, that is, s = {S 0 }; When the unmanned vehicle group receives a joining request from a node, the group is in the joining state. In this state, the node that receives the joining request needs to determine whether to agree to the node joining according to the conditions in the node joining algorithm (Algorithm 1). (3) Leave state S 2 :When a node in the unmanned vehicle group loses communication with its neighboring nodes, the vehicle group is in a departure state; in this state, the node needs to calculate the number of vehicle group components; and according to the calculation result of the number of vehicle group components, the adjacent set of the vehicle group is modified to calculate the new feature matrix; (4) Merge state S 3 : When an unmanned vehicle group receives a merging request from another vehicle group, the vehicle group is in a merging state; in this state, the node that receives the joining request needs to determine whether to agree to the vehicle group merging according to the conditions in the vehicle group merging algorithm (Algorithm 3); (5) Split state S 4 : When the distance between two nodes in the following state in the unmanned vehicle group is too large, the vehicle group is in a split state; in this state, the vehicle group needs to further determine whether to split based on conditions such as speed and number of vehicle group components; (6) Reconstruction state S 5 :When a vehicle node of an unmanned vehicle group changes lanes, the vehicle group is in a reconstruction state; in this state, the lane-changing node needs to construct a new set of neighbor nodes and calculate the new vehicle group component to which it belongs; and based on the calculation results, the adjacent set of the vehicle group is modified to calculate a new feature matrix; (7) Update status S 6 : When the swarm structure needs to be updated, the swarm is in an updating state; the nodes in the swarm in the updating state need to update the swarm structure; (8) Extinction state S 7 :When the number of vehicle nodes in the swarm is less than 2, the swarm is in the extinction state; the nodes in the swarm in the extinction state delete the adjacent set and the characteristic matrix and become nodes that have not joined the swarm; this state is also the only element in the terminal state set of the finite state automaton of the dynamic evolution of the unmanned vehicle swarm, that is, F = {S 7 }; Step 3.2 Dynamic evolution event triggering conditions and state transition functions (1) Node joins Σ 1 :When in collaborative state S 0 After the vehicle node of the driverless car group receives the JOIN message, the node joining event is triggered; the corresponding state transfer function is δ(S 0 ,Σ 1 )=S 1 , indicating that it is in the collaborative state S 0 After receiving the joining request from the node, the vehicle group enters the joining state S 1 ; (2) The node leaves Σ 2 :When in collaborative state S 0 If a vehicle node in the unmanned vehicle group does not receive an ACK message within a period of time after sending a SYN message, it means that the communication connection with the neighboring node is lost, triggering a node leaving event; the corresponding state transition function is δ(S 0 ,Σ 2 )=S 2 , indicating that it is in the collaborative state S 0 The vehicle group enters the departure state S when the node loses the communication connection 2 ; (3) Vehicle group mergingΣ 3 :When in collaborative state S 0 After the vehicle node of the unmanned vehicle group receives the MERGE message, the vehicle group merge event is triggered; the corresponding state transfer function is δ(S 0 ,Σ 3 )=S 3 , indicating that it is in the collaborative state S 0 After receiving a merge request from another car group, the car group enters the merge state S 1 ; (4) Group splitΣ 4 :When in collaborative state S 0 After receiving the ACK message from the node in the follow-up relationship, if the distance calculated by the Location in the driving log is greater than the threshold δ, the group split event is triggered; the corresponding state transfer function is δ(S 0 ,Σ 4 )=S 4 , indicating that it is in the collaborative state S 0 The car group enters the split state S 4 ; (5) Vehicle group reconstructionΣ 5 :When in collaborative state S 0 When a vehicle node of the unmanned vehicle group changes lanes, a vehicle group reconstruction event is triggered; the corresponding state transfer function is δ(S 0 ,Σ 5 )=S 5 , indicating that it is in the collaborative state S 0 The vehicle group enters the reconstruction state S 5 ; (6) Approval requestΣ 6 :When the vehicle group in the joining, leaving, merging, splitting, and reconstructing states passes the conditional judgment and finds that the conditions are met, the structure of the vehicle group is ready to be updated; the corresponding state transition function is as follows: ①δ(S 1 ,Σ 6 )=S 6 : In joining state S 1 The group of vehicles decides to agree to the node joining the group after conditional judgment, and the group enters the update state S 6 ; ②δ(S 2 ,Σ 6 )=S 6 : In the leaving state S 2 After calculating the new adjacent set and feature matrix, the car group sends the car group information to the remaining nodes in the car group, and the car group enters the update state S 6 ; ③δ(S 3 ,Σ 6 )=S 6 : In merge state S 3 The vehicle group decides to agree to merge the vehicle group after conditional judgment, and the vehicle group enters the update state S 6 ; ④δ(S 4 ,Σ 6 )=S 6 : In split state S 4 After calculating the new adjacent set and feature matrix, the car group sends the car group information to the remaining nodes in the car group, and the car group enters the update state S 6 ; ⑤δ(S 5 ,Σ 6 )=S 6 : In reconstruction state S 5 After calculating the new adjacent set and feature matrix, the car group sends the car group information to the remaining nodes in the car group, and the car group enters the update state S 6 ; (7) Vehicle group maintenanceΣ 7 :When the vehicle group is in the updating state S 6 When the unmanned vehicle group maintenance method is called to update the vehicle group structure stored in the nodes in the vehicle group; when all nodes are updated, the vehicle group returns to the collaborative state S 0 ; The corresponding state transfer function is δ(S 6 ,Σ 7 )=S 0 , indicating that it is in the collaborative state S 0 The vehicle group enters the reconstruction state S 5 ; (8) Request rejectionΣ 8 :When the vehicle group in the joining and merging state is found to not meet the corresponding conditions after conditional judgment, it returns to the collaborative state S 0 ; The corresponding state transfer function is as follows: ①δ(S 1 ,Σ 8 )=S 0 : In joining state S 1 The vehicle group rejects the joining of the node after conditional judgment, and the vehicle group returns to the collaborative state S 0 ; ②δ(S 3 ,Σ 8 )=S 0 : In merge state S 3 The vehicle group refuses to merge after conditional judgment, and the vehicle group returns to the cooperative state S 0 ; (9) The car group disappearsΣ 9 :When the number of vehicle nodes in the vehicle group component where the node in the vehicle group in the leaving, splitting, and reconstructing state is less than 2, the vehicle group enters the extinction state S 7 ; The corresponding state transfer function is as follows: ①δ(S 2 ,Σ 9 )=S 7 : In the leaving state S 2 The number of nodes in the car group is less than 2, and the car group enters the extinction state S 7 ; ②δ(S 4 ,Σ 9 )=S 7 : In split state S 4 The number of nodes in the car group is less than 2, and the car group enters the extinction state S 7 ; ③δ(S 5 ,Σ 9 )=S 7 : In reconstruction state S 5 The number of nodes in the car group is less than 2, and the car group enters the extinction state S 7 ; In summary, the detailed definition of the finite state automaton NFAM for the dynamic evolution of the unmanned vehicle group is as follows: NFAM=(S,Σ,Δ,s,F) (13) where \(S = \{S i | i = 0, 1, \ldots, 7\}\), \(\Sigma=\{\Sigma i | i = 1, \ldots, 9\}\), \(\Delta=\{\delta(S 0 , \Sigma 1 ) = S 1 , \delta(S 0 , \Sigma 2 ) = S 2 , \delta(S 0 , \Sigma 3 ) = S 3 , \delta(S 0 , \Sigma 4 ) = S 4 , \delta(S 0 , \Sigma 5 ) = S 5 , \delta(S 1 , \Sigma 6 ) = S 6 , \delta(S 1 , \Sigma 8 ) = S 0 , \delta(S 2 , \Sigma 6 ) = S 6 , \delta(S 2 , \Sigma 9 ) = S 7 , \delta(S 3 , \Sigma 6 ) = S 6 , \delta(S 3 , \Sigma 8 ) = S 0 , \delta(S 4 , \Sigma 6 ) = S 6 , \delta(S 4 , \Sigma 9 ) = S 7 , \delta(S 5 , \Sigma 6 ) = S 6 , \delta(S 5 , \Sigma 9 ) = S 7 , \delta(S 6 , \Sigma 7 ) = S 0 \}, s = S 0 , F = \{S 7 \}; Step 3.3 Dynamic Evolution Method According to the state transition logic of the finite state automaton of the dynamic evolution of the unmanned vehicle group, the dynamic evolution method of the unmanned vehicle group is given; the dynamic evolution method of the unmanned vehicle group describes the state transition process that occurs when the vehicle group encounters conditions in various states. The specific steps are shown in Algorithm 6 5. The dynamic evolution and coordination method of unmanned vehicle groups based on process mining in a highway scenario as claimed in claim 1, It is characterized in that The step 4 unmanned vehicle group collaboration method: Step 4.1 Unmanned vehicle group collaboration model When the driverless car group is in a coordinated state S during the dynamic evolution process 0 When , if the node speed similarity and node dependence between adjacent nodes in the vehicle group are greater than the threshold, and the overall speed chaos of the vehicle group is less than the threshold, the unmanned vehicle group reaches a coordinated state; The specific cooperation model (CM) of the unmanned vehicle group is as follows: Among them, S n represents the dynamic evolution state of the unmanned vehicle group in which the vehicle group is currently located; NFAM=(S,Σ,Δ,s,F) represents the dynamic evolution finite state automaton of the unmanned vehicle group; m ij represents the element in the i-th row and j-th column of the feature matrix in the vehicle group VG=(V,L,M); NSS(v i ,v j ) represents node v i and v j SC(VG) represents the speed disorder of the vehicle group VG; ND(VG) represents the node dependence of the vehicle group VG; Step 4.2 Autonomous Vehicle Group Collaboration Algorithm The specific steps of the unmanned vehicle group collaboration algorithm are as follows: (1) Nodes in the swarm broadcast HELLO packets to the surroundings and send SYN packets to neighboring nodes. After receiving a HELLO message or a SYN message, they reply with a BUILT or ACK message, respectively. (2) When a node in the swarm receives a JOIN message, it executes the node joining algorithm (Algorithm 1), and a node joining event occurs; (3) When a node in the swarm does not receive an ACK message after sending a SYN message, the node departure algorithm (Algorithm 2) is executed to calculate the swarm component. If it is greater than 1, a node departure event occurs. Otherwise, the swarm structure is updated. After the node departure is completed, the swarm is judged to be extinct based on the number of remaining vehicles in the swarm. (4) When a node in a swarm receives a MERGE message, it executes the swarm merging algorithm (Algorithm 3), calculates the swarm chaos degree, and decides whether a swarm merging event occurs based on the result; (5) When the distance between a node in the group and its following node exceeds the threshold, the group splitting algorithm (Algorithm 4) is executed to calculate the group component after removing the following relationship. If it is greater than 1, a group splitting event occurs. Otherwise, the group structure is updated. After the group splits, the number of remaining vehicles in the group is used to determine whether the group has disappeared. (6) When a node in the swarm changes lanes, the swarm reconstruction algorithm (Algorithm 5) is executed, and a swarm reconstruction event occurs. After the event, the number of remaining vehicles in the swarm is used to determine whether the swarm has disappeared.