Method for constructing model of unmanned vehicle group in urban scene

Abstract

The application provides a method for constructing an unmanned vehicle group model in a city scene, selects part of unmanned vehicles as leading nodes of the unmanned vehicle group by considering the activity, pre-sensitivity and mobility difference of the unmanned vehicles, further gives a method for constructing the unmanned vehicle group model based on the connectivity, coupling and real-time performance of the unmanned vehicle group, and finally gives a method for solving the unmanned vehicle group model based on distributed multi-objective optimization. In the simulation experiment, corresponding evaluation indexes are given to evaluate the unmanned vehicle group model and verify the effectiveness of the unmanned vehicle group model. Thus, the perception ability of the unmanned vehicle can be improved by sharing the perception information among the unmanned vehicle group members, the intelligent coordination of the motion behavior of the unmanned vehicle is realized, and the unmanned vehicle group which can be widely applied can be established.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving, specifically to a method for constructing an autonomous vehicle swarm model in urban scenarios. Background Technology

[0002] Autonomous vehicles rely on onboard sensors to perceive their surroundings and control the vehicle based on these perceptions, enabling them to operate safely and automatically without human intervention. Initially, autonomous vehicles may be expensive and have limited performance; however, starting in the 2030s or 40s, their cost will gradually decrease, making them more accessible to the general public. The 2022 Beijing Winter Olympics utilized a 5G intelligent vehicle-to-everything (V2X) system deployed in the Shougang Park to complete the torch relay in an autonomous vehicle. Autonomous vehicles are expected to achieve Level 5 fully autonomous driving (L5) without a driver and be legally used in certain areas by the end of the 2020s.

[0003] While autonomous vehicles can use onboard sensors to perceive their surroundings, blind spots and obstructed views are inherent problems during vehicle movement, especially in foggy or snowy weather. These conditions significantly reduce the sensing range of the onboard sensors, preventing autonomous vehicles from comprehensively and accurately perceiving their environment and leading to potential safety hazards. By grouping nearby and similarly mobile autonomous vehicles into a convoy, and enabling convoy members to share perception results via wireless communication, blind spots can be eliminated, allowing autonomous vehicles to better perceive their surroundings and make informed decisions.

[0004] The limitations of single-agent autonomous driving systems, such as restricted perception, hinder their ability to effectively support intelligent vehicle behavior in urban scenarios. Autonomous vehicle swarms, on the other hand, can achieve data sharing, thereby improving perception capabilities and promoting the widespread application of autonomous driving technology, which is undoubtedly of great significance. While some research has been conducted on methods for building autonomous vehicle swarms, the following issues remain before they can be applied to urban scenarios: (1) Load concentration: Existing semi-centralized autonomous vehicle swarm construction methods for closed and highway scenarios rely on the leader node to build and maintain the vehicle swarm structure. In urban scenarios, the movement of autonomous vehicles is affected by human drivers, roadside obstacles, traffic lights and pedestrians, which causes the following relationship between vehicle swarm members to change frequently, resulting in excessive computational and network load on the leader node.

[0005] (2) Lack of connectivity metrics: Existing methods for building autonomous vehicle swarms only consider attributes such as speed, location and destination between autonomous vehicles, lacking connectivity metrics between autonomous vehicles, making it difficult to guarantee the connectivity quality between swarm members. Summary of the Invention

[0006] Purpose of the invention: To address the above problems, this invention presents a method for constructing an autonomous vehicle swarm model. By considering the activity level, awareness, and mobility differences of autonomous vehicles, a subset of autonomous vehicles are selected as leading nodes in the swarm. Furthermore, based on the connectivity, coupling, and real-time performance of the swarm, a method for constructing the autonomous vehicle swarm model is provided. Finally, a solution method for the autonomous vehicle swarm model based on distributed multi-objective optimization is presented. Simulation experiments provide corresponding evaluation metrics to assess the effectiveness of the autonomous vehicle swarm model. This allows for the sharing of perception information among swarm members, improving the perception capabilities of autonomous vehicles and enabling intelligent collaborative movement behavior, thus establishing a widely applicable autonomous vehicle swarm.

[0007] The technical solution of the present invention specifically includes the following steps: Step 1. Selection of Leader Node for Autonomous Vehicle Swarm Step 1.1 Relevant Definitions Step 1.2 Activity Measurement of Autonomous Vehicles Step 1.3 Sensitivity Measurement of Autonomous Vehicles Step 1.4 Measurement of Mobility Differences of Autonomous Vehicles Step 1.5 Selection of Leading Node for Autonomous Vehicle Swarm Step 2. Construction of Autonomous Vehicle Swarm Model Step 2.1 Connectivity Measurement of Autonomous Vehicle Swarm Step 2.2 Coupling Measurement of Autonomous Vehicle Swarm Step 2.3 Real-time measurement of autonomous vehicle swarm Step 2.4 Autonomous Vehicle Swarm Model Step 3. Solving the autonomous vehicle swarm model Step 3.1 Solving the autonomous vehicle swarm model Beneficial effects: This invention discloses a method for constructing an autonomous vehicle swarm model in urban scenarios, and provides a method that enables the sharing of perception information among members of the autonomous vehicle swarm and ensures the intelligent movement behavior of the autonomous vehicle swarm, thereby obtaining a method that can be widely applied in urban scenarios for autonomous vehicle swarms. Attached Figure Description

[0008] Figure 1 Autonomous vehicle swarm leadership node selection process Figure 2 Solution process for autonomous vehicle swarm model Figure 3 SUMO simulation software and manually designed simulation scene diagrams Figure 4 SUMO simulation software and simulation scenario diagram of actual road network Figure 5 Average connectivity of a vehicle group at different speeds in a manually designed scenario Figure 6 Average coupling of vehicle groups at different speeds in artificially designed scenarios Figure 7 Average real-time performance of vehicle groups at different speeds in a manually designed scenario Figure 8 Average survival time of the leading node of a vehicle group at different speeds in a manually designed scenario Figure 9 Average survival time of secondary leader nodes in a group of vehicles disembarking at different speeds in an artificially designed scenario Figure 10 Average survival time of ordinary nodes in a vehicle swarm disembarking at different speeds in a manually designed scenario Figure 11 Average connectivity of vehicle groups with different numbers of vehicles in real-world road network scenarios Figure 12 Average coupling of vehicle groups with different numbers of vehicles in real-world road network scenarios Figure 13 Average real-time performance of vehicle groups with different numbers of vehicles in real-world road network scenarios Figure 14 Average survival time of vehicle group leading nodes under different numbers of vehicles in real-world road network scenarios Figure 15 Average survival time of secondary leader nodes in real-world road network scenarios with different numbers of vehicles Figure 16 Average survival time of ordinary nodes in a vehicle group under different numbers of vehicles in real-world road network scenarios Figure 17 Average connectivity of vehicle groups at different speeds in real-world road network scenarios Figure 18 Average coupling of vehicle groups at different speeds in real-world road network scenarios Figure 19 Average real-time performance of vehicle groups at different speeds in real-world road network scenarios Figure 20 Average survival time of leading nodes of vehicle groups at different speeds in real-world road network scenarios Figure 21 Average survival time of secondary leader nodes in real-world road network scenarios at different vehicle speeds Figure 22 Average survival time of ordinary nodes in a vehicle group at different speeds in a real-world road network scenario Figure 23 Table 1. Simulation Experiment Configuration in Artificially Designed Scenarios Figure 24Table 2 Simulation Experiment Configuration in Actual Road Network Scenarios Figure 25 Flowchart of the present invention Detailed Implementation The technical solutions provided in this application will be further described below with reference to specific embodiments and accompanying drawings. The advantages and features of this application will become clearer from the following description.

[0009] This invention specifically includes the following four aspects: Step 1: Selection of Leading Node for Autonomous Vehicle Swarm Step 2: Construction of the Autonomous Vehicle Swarm Model Step 3 Solving the autonomous vehicle swarm model Step 4: Simulation Experiment Verification Details are as follows: Step 1: Selection of Leading Node for Autonomous Vehicle Swarm This step involves selecting a leader node for the autonomous vehicle swarm. The leader node is located at the very front of the swarm and possesses superior activity, awareness, and mobility differentiation. The selection of a leader node for an autonomous vehicle swarm specifically includes the following five sub-steps: (1) Relevant definitions Definition 1: In urban scenarios, autonomous vehicles and external disturbances are defined as follows: in, It represents a collection of driverless vehicles. This refers to an unmanned vehicle. Represents the set of positive integers. This indicates a group of manned vehicles. This indicates a manned vehicle. This represents a collection of roadside obstacles. This represents a roadside obstacle. Indicates the collection of traffic lights. This represents a traffic light. Indicates a group of pedestrians. It represents a pedestrian.

[0010] Definition 2: In urban scenarios, a swarm of autonomous vehicles is represented as: in, It represents a collection of driverless vehicles. Represents the set of types of autonomous vehicles, where 1 and 2 represent the autonomous vehicle swarm leader, secondary leader, and ordinary node, respectively. Indicates a free node. This indicates a swarm of driverless vehicles. This represents a swarm of driverless vehicles, in which It is a group of cars The leading node, For car groups The set of members, Indicates a group of vehicles The set of follower relationships among members, This represents a mapping function for autonomous vehicle types.

[0011] Definition 3: Autonomous vehicle swarm The set of leading nodes and secondary leading nodes is as follows: in, Indicates driverless vehicles For car groups Members, express For car groups The leading or secondary leading node.

[0012] Definition 4: Autonomous vehicle swarm Middle leading node or secondary leading node The set of following nodes: in, Indicates driverless vehicles pass Join the car group .

[0013] (2) Activity measurement of autonomous vehicles Activity level measures the level of activity of autonomous vehicles. The activity level is: in, express A collection of driverless vehicles from neighboring areas. express and The distance between them This indicates the maximum communication range of the autonomous vehicle. Represents a set Size, express and The probability of communication interruption between them express and Does a communication link exist between them? (3) Priority measurement of autonomous vehicles Priority sensing measures the ability of autonomous vehicles to perceive obstacles ahead of time. The degree of prior awareness is: in, , , and They represent The number of people driving, obstacles, traffic lights, and pedestrians detected.

[0014] (4) Measurement of mobility differences of unmanned vehicles Mobility dissimilarity measures the degree of difference in mobility between an autonomous vehicle and its neighboring autonomous vehicles. The mobility disparity is: in, This indicates the maximum speed of the driverless vehicle. express and The relative speed between them.

[0015] (5) Selection of leader node for autonomous vehicle swarm Based on the meanings of autonomous vehicle activity, awareness, and mobility differences, a leader selection function for an autonomous vehicle swarm is constructed: in, , and The weighting coefficient is used to... , and Normalization.

[0016] The specific process of the autonomous vehicle swarm leader selection method (Algorithm 1) is as follows: Figure 1 As shown.

[0017] S1) Autonomous vehicles exchange vehicle attribute information such as speed and location with neighboring autonomous vehicles (lines 1-5). S2) Calculate the activity level, awareness, and mobility difference of autonomous vehicles (line 7).

[0018] S3) The autonomous vehicle calculates the leader node selection function (line 8).

[0019] S4) The autonomous vehicle will send the leader node selection function value to the neighboring autonomous vehicles (lines 9-11).

[0020] S5) The autonomous vehicle compares its own leader node selection function value with that of its neighboring autonomous vehicles (line 14).

[0021] S6) If the driverless vehicle node If the leader node's selection function value is greater than or equal to that of all neighboring autonomous vehicles, then... Convert to a vehicle group leading node (lines 15-16).

[0022] S7) If the driverless vehicle node The leader node's selection function value is less than that of its neighboring autonomous vehicles. Send a swarm join request to neighboring autonomous vehicles (lines 17-20).

[0023] Step 2: Construction of the Autonomous Vehicle Swarm Model This step addresses the lack of connectivity metrics in existing autonomous vehicle swarm construction methods by providing methods for measuring the connectivity, coupling, and real-time performance of autonomous vehicle swarms. Based on these methods, an autonomous vehicle swarm model is constructed. The construction of the autonomous vehicle swarm model specifically includes the following four sub-steps: (1) Connectivity measurement of autonomous vehicle swarms The connectivity of an autonomous vehicle swarm measures the quality of communication links between leader and secondary leader nodes, and between secondary leader nodes and ordinary nodes within the swarm. The connectivity is: in, Indicating a fleet of self-driving cars connectivity, Indicates driverless vehicles For car groups The leading or secondary leading node, Indicates driverless vehicles for The following node, , and They represent and The probability of communication interruption, data correct delivery rate, and link capacity between them.

[0024] Property 1: The connectivity of an autonomous vehicle swarm increases with the increase of the correct data delivery rate and link capacity among swarm members, and decreases with the increase of the probability of communication interruption among swarm members.

[0025] Proof: by a swarm of autonomous vehicles The process of calculating connectivity shows that for any and satisfy: in, Indicates a group of vehicles Connectivity about and The partial derivative of the probability of communication interruption between them.

[0026] because and ,so: Therefore, it can be concluded that the group of vehicles Connectivity about and The partial derivative of the probability of communication interruption between vehicles is less than or equal to 0, i.e., the vehicle group Connectivity with and The probability of communication interruption decreases as the probability of communication interruption increases.

[0027] Similarly, we have: and Therefore, the connectivity of an autonomous vehicle swarm increases with the increase in the correct data delivery rate and link capacity among swarm members, and decreases with the increase in the probability of communication interruption among swarm members.

[0028] (2) Coupling measurement of autonomous vehicle swarms The coupling property of an autonomous vehicle swarm measures the degree of interdependence between leader and secondary leader nodes, and between secondary leader nodes and ordinary nodes within the swarm. The coupling is: in, Indicates a group of vehicles Coupling, express and The connection duration between them.

[0029] Property 2: The coupling of autonomous vehicle swarms increases as the connection time between swarm members increases.

[0030] Proof: by a swarm of autonomous vehicles From the calculation process of coupling, it can be seen that for any and satisfy: in, Indicates a group of vehicles Coupling about and The partial derivative of the duration of connectivity between them.

[0031] Therefore, we can conclude that: Therefore, it can be concluded that the group of vehicles Coupling about and The partial derivative of the connectivity duration between vehicles is greater than zero, i.e., the vehicle group The coupling with and It increases as the connection time between them increases.

[0032] (3) Real-time measurement of autonomous vehicle fleet The real-time performance of an autonomous vehicle swarm is measured by the time overhead of communication between the lead and secondary lead nodes, and between the secondary lead and ordinary nodes within the swarm. The real-time performance is: in, Indicates a group of vehicles Real-time performance and They represent and Data transmission delays and communication round-trip times between the two.

[0033] Property 3: The real-time performance of autonomous vehicle swarms decreases as the data transmission delay and communication round-trip time between swarm members increase.

[0034] Proof: by a swarm of autonomous vehicles As can be seen from the real-time calculation process, for any and satisfy: in, Indicates a group of vehicles Real-time performance and The partial derivative of the data transmission delay between them.

[0035] because Therefore, we have: Therefore, the car group Real-time performance and The partial derivative of the data transmission delay between vehicles is less than zero, meaning the vehicle group... Real-time performance and It decreases as the data transmission delay between them increases.

[0036] Similarly, we can obtain: Therefore, the car group Real-time performance and The partial derivative of the round-trip time for communication between vehicles is less than zero, meaning the vehicle group... Real-time performance with and The time difference between communication round trips increases and decreases.

[0037] (4) Unmanned vehicle swarm model The autonomous vehicle swarm model is as follows: in, Indicating a fleet of self-driving cars The car group model, Indicates driverless vehicles For car groups The leading or secondary leading node, express The set of candidate follower nodes, After receiving a vehicle group join request from a free node, add it to the group. , Indicates a group of vehicles The set of candidate members Indicates Agree to join The set of free nodes, Indicates all passing train groups The leading and secondary leading nodes agreed to join the vehicle swarm. The set of free nodes, Indicates a group of vehicles Not reached maximum scale This indicates the data flow congestion density. express and The directions of movement are the same. express lie in Ahead.

[0038] Theorem 1: The autonomous vehicle swarm model is based on the vehicle swarm network capacity limit to the maximum size of the autonomous vehicle swarm.

[0039] Proof: As shown in [1], the swarm of driverless vehicles The network capacity is: in, and These represent the maximum density and transmission speed of the data stream in free-flow state, respectively.

[0040] According to the autonomous vehicle swarm model, free nodes can join the swarm. If and only if .when At that time, the autonomous vehicle swarm refuses to allow stray nodes to join the swarm. Therefore, the autonomous vehicle swarm model is based on the network capacity of the vehicle swarm to limit the maximum size of the autonomous vehicle swarm.

[0041] Step 3 Solving the autonomous vehicle swarm model This step presents a solution method for a distributed autonomous vehicle swarm model based on the multi-objective optimization method NSGA-II (Non-dominated Sorting Genetic Algorithm-II). The solution of the autonomous vehicle swarm model specifically includes the following two sub-steps: (1) Solving the vehicle group model The specific process of the algorithm for solving the autonomous vehicle swarm model based on multi-objective optimization (hereinafter referred to as Algorithm 2) is as follows: Figure 2 The flowchart shown.

[0042] S1) Leading or secondary leading node Inspecting a fleet of self-driving cars Has the maximum size been reached (line 1)?

[0043] S2) If a fleet of driverless vehicles When the maximum scale is reached, the vehicle group Reject stray nodes from joining the vehicle group (lines 2-3).

[0044] S3) If the car group Not reached maximum scale Check the set of candidate follower nodes The state of the free nodes in the data (line 4).

[0045] S4) Check whether the free node has the same direction of relative motion as the members of the autonomous vehicle swarm and is located in Behind (line 5).

[0046] S5) If A certain free node in the system has a different direction of relative motion from other members of the autonomous vehicle swarm, or is located in... Ahead, Take it from Remove from (lines 6-8).

[0047] S6) Initialize the length of the solution, the number of iterations, and the solution set (lines 9-11).

[0048] S7) The solution set is obtained through crossover and mutation operations. Generate solution set (Line 13).

[0049] S8) will and After merging, the solutions are sorted using the non-dominated sorting method of NSGA-II, and selected from the sorted solutions. Each solution generates a new set of solutions (lines 14-22).

[0050] S9) Check if the maximum number of iterations has been reached. (Line 12).

[0051] S10) If the maximum number of iterations is reached ,from Randomly select a solution As a solution to the autonomous vehicle swarm model (lines 23-25).

[0052] The solution method for the autonomous vehicle swarm model based on distributed multi-objective optimization uses binary encoding to represent the solution of the autonomous vehicle swarm model. Leading or secondary leading nodes The solution of the vehicle group model is obtained using Algorithm 2. back, examine Each of its digits can take a value. If The Middle Each element is 1. agree The Middle One driverless vehicle join in And send it a message allowing it to join the car group. Otherwise, Reject their membership .

[0053] (2) Proof of relevant properties for solving the autonomous vehicle swarm model Definition 5: The Pareto optimal solution set of the autonomous vehicle swarm model is: in, It is a Pareto optimal solution for the autonomous vehicle swarm model. This represents the solution space of the autonomous vehicle swarm model. Representing the solution Dominant Solution That is, the solution The connectivity, coupling, and real-time performance of the corresponding autonomous vehicle swarm model are all lower than those of the solution. The connectivity, coupling, and real-time performance of the corresponding autonomous vehicle swarm model.

[0054] Theorem 2: The solutions generated by Algorithm 2 converge to the Pareto optimal solution set, that is: in, This represents the solution generated by Algorithm 2. Representing the solution The probability of belonging to the Pareto optimal solution set.

[0055] Proof: Since the number of autonomous vehicles is finite, the set of candidate follower nodes for the leading and secondary leading nodes is also finite. Algorithm 2 can traverse the solution space of the autonomous vehicle swarm model in a finite number of iterations, i.e.: Algorithm 2 uses the non-dominated sorting method of NSGA-II to sort the solution set, from which we can know: If the solution generated by algorithm 2 Non-convergence Then it exists belong and Not belonging to satisfy: Therefore, we can conclude that: However, Include and Therefore, all the optimal dominant solutions are found in the set of solutions. This contradicts the non-dominated sorting method of NSGA-II. Therefore, the solution generated by Algorithm 2 converges to the Pareto optimal solution set.

[0056] Step 4: Simulation Experiment Verification This step verifies the effectiveness of the autonomous vehicle swarm model. The present invention uses simulation experiments to simulate the behavior of the vehicle swarm, including vehicle movement and inter-vehicle communication, and performs statistical analysis on the experimental results.

[0057] (1) Simulation experimental data and methods NS3 is an open-source discrete event network simulator. SUMO is a tool capable of simulating microscopic and continuous traffic flow. This invention uses SUMO and NS3 respectively to simulate the movement and communication behavior of autonomous vehicles. This invention uses two scenarios—manually designed and real-world road networks—to verify the effectiveness of the autonomous vehicle swarm model. The manually designed scenario is as follows: Figure 3 As shown, the road network structure is a crossroads with a horizontal and vertical span of 10.2 km. The simulation included 356 driverless vehicles, 44 manned vehicles, 41 traffic lights, 40 roadside obstacles, and 400 pedestrians. A 200-second segment of the simulation was captured. To cover various driving scenarios from low to high speeds, the simulation set the maximum vehicle speeds to 5, 10, 15, 20, 25, and 30 m / s. Other parameters in the manually designed scenarios are shown in Table 1. The actual road network scenario is as follows... Figure 4 As shown, its road network structure is a 3 km map generated using OpenStreetMap. A 2 km section of the urban road network in Shanghai was simulated, comprising 3469 road segments and 94 traffic lights. Sixty roadside obstacles and 600 pedestrians were simulated, with a 200-second segment captured from the simulation. Other parameters of the actual road network scenario are shown in Table 2.

[0058] To verify the effectiveness of the autonomous vehicle swarm model of this invention, the simulation experiment is designed as follows: 1) Import the manually designed and actual road network scenarios into SUMO and configure the relevant parameters, then export the vehicle movement models in the two scenarios.

[0059] 2) Use NS3 to realize communication between autonomous vehicles and implement the autonomous vehicle swarm model and its solution method of the present invention.

[0060] 3) Compare the Autonomous Vehicle Group Model (AVGM) proposed in this invention with the Autonomous Vehicle Group Formation (AVGF) [2] method proposed by Cheng et al. and the Dynamic Clustering (DC) [3] method for vehicle self-organizing networks, and analyze the advantages and disadvantages of AVGM.

[0061] (2) Analysis of simulation experiment results 1) Simulation results in artificially designed scenarios The average connectivity of an autonomous vehicle swarm measures the quality of communication links between leader and secondary leader nodes, and between secondary leader nodes and ordinary nodes. A higher average connectivity indicates better communication link quality among swarm members. In a manually designed scenario, the average connectivity of AVGM, AVGF, and DC at different maximum vehicle speeds is as follows: Figure 5 As shown, the average connectivity of the AVGM vehicle group is higher than that of AVGF and DC. This is because AVGM considers the connectivity increment generated after a free node joins the vehicle group when selecting free nodes, and uses a multi-objective optimization method to sort the free nodes.

[0062] The average coupling of an autonomous vehicle swarm measures the degree of interdependence between leader and secondary leader nodes, and between secondary leader nodes and ordinary nodes within the swarm. A higher average coupling indicates a greater degree of interdependence among swarm members. In manually designed scenarios, the average coupling of AVGM, AVGF, and DC at different maximum vehicle speeds is as follows: Figure 6 As shown, the average coupling of the AVGM group is significantly higher than that of the AVGF and DC groups. The average coupling of the AVGM group increases with increasing maximum vehicle speed because, with increased vehicle mobility, detached nodes can detect other detached nodes and the group more quickly. Then, the average coupling of the AVGM group remains at a high level because the increase in coupling caused by the addition of detached nodes is offset by the decrease in coupling caused by the departure of group members. Finally, the average coupling of the AVGM group decreases with increasing maximum speed because the relative speed differences among group members increase.

[0063] The average real-time performance of an autonomous vehicle swarm measures the time overhead of communication between the lead and secondary lead nodes, and between the secondary lead and ordinary nodes. A higher average real-time performance indicates lower communication time overhead among swarm members. In manually designed scenarios, the average real-time performance of AVGM, AVGF, and DC at different maximum vehicle speeds is as follows: Figure 7 As shown, AVGM has higher real-time performance than AVGF and DC. The average real-time performance of AVGM increases with the maximum vehicle speed. This is because the mobility of vehicle nodes increases with the maximum speed, increasing the relative distance between vehicles and thus reducing communication interference between vehicle nodes.

[0064] The average survival time of a vehicle swarm member represents the time from when an autonomous vehicle assumes that role to when it relinquishes it. The average survival times of leader, secondary leader, and ordinary nodes are as follows: Figure 8 , Figure 9 and Figure 10As shown, the average lifetime of the lead and secondary lead nodes in AVGM is greater than that in AVGF and DC because AVGM takes into account the connectivity time between vehicle group members. The lifetime of ordinary nodes in AVGM is slightly lower than that in AVGF because AVGM requires secondary lead nodes to be located in front of ordinary nodes, while AVGF does not have this restriction.

[0065] 2) Simulation results in actual road network scenarios In a real-world road network scenario, the maximum speed of autonomous vehicles was fixed at 60 km / h, and the number of vehicles was set to 600, 700, 800, 900, and 1000. 10% of the vehicles were randomly selected as manned vehicles, and the remaining vehicles were designated as autonomous vehicles. The average connectivity, coupling, and real-time performance of the autonomous vehicle group under different vehicle densities were statistically analyzed using AVGM, AVGF, and DC. The results are as follows: Figure 11 , Figure 12 and Figure 13 As shown, the average connectivity of the AVGM vehicle group is higher than that of AVGF and DC across different numbers of vehicles. Although the group size increases with vehicle density, the amount of information that needs to be exchanged between group members increases, leading to a decrease in the link capacity of group members as the group size increases. Therefore, the average connectivity of the AVGM vehicle group is relatively stable across different numbers of vehicles. The connection time between autonomous vehicle nodes is mainly affected by the relative speed between vehicle grounding points; therefore, the average coupling of the AVGM vehicle group is relatively stable across different numbers of vehicles. As the density of autonomous vehicles increases, communication interference between vehicle nodes increases; therefore, the average real-time performance of the AVGM vehicle group increases with the number of vehicles.

[0066] In real-world road network scenarios, the average survival times of AVGM, AVGF, and DC for leading, secondary leading, and ordinary nodes under different vehicle densities are as follows: Figure 14 , Figure 15 and Figure 16 As shown, the average lifetime of the lead and secondary lead nodes in AVGM is greater than that of AVGF and DC. The lifetime of ordinary nodes in AVGM is slightly lower than that of AVGF. The average lifetime of the lead and secondary lead nodes in AVGM increases with increasing vehicle density because, as vehicle density increases, lead and secondary lead nodes can choose to join the vehicle group from detached nodes with longer connectivity durations.

[0067] In a real-world road network scenario, with a fixed number of 800 vehicles, 10% are randomly selected as manned vehicles, and the remaining vehicles are set as driverless vehicles. Maximum vehicle speeds are set at 30, 40, 50, 60, and 70 km / h. The average connectivity, coupling, and real-time performance of the driverless vehicle group under different maximum vehicle speeds are statistically analyzed using the three methods. The results are as follows: Figure 17 , Figure 18 and Figure 19 As shown, AVGM exhibits higher average connectivity, coupling, and real-time performance than AVGF and DC. Compared to manually designed scenarios, AVGM shows reduced connectivity, coupling, and real-time performance in real-world road network scenarios. This is because real-world road network scenarios are more complex, leading to more variable connectivity relationships among vehicle group members.

[0068] In real-world road network scenarios, the average survival times of AVGM, AVGF, and DC at different maximum speeds for leading, secondary leading, and ordinary nodes are as follows: Figure 20 , Figure 21 and Figure 22 As shown, the average survival time of the lead and secondary lead nodes of the AVGM is greater than that of the AVGF and DC, while the survival time of the ordinary nodes of the AVGM is lower than that of the AVGF and DC. The average survival time of the lead and secondary lead nodes of the AVGM decreases as the maximum vehicle speed increases, because the increased maximum speed makes the relationships between vehicle group members more volatile.

[0069] This demonstrates that, in both artificially designed and real-world road network scenarios, AVGM can achieve high connectivity, coupling, and real-time performance for autonomous vehicle swarms regardless of vehicle speed and density, proving the effectiveness of AVGM.

[0070] Innovation This invention addresses the problems of concentrated load and lack of connectivity metrics in existing autonomous vehicle swarm construction methods for closed and highway scenarios. It presents a method for constructing autonomous vehicle swarm models in urban scenarios. Considering the activity, awareness, and mobility differences of autonomous vehicles, a method for selecting the leading node of the autonomous vehicle swarm is given. Based on the connectivity, coupling, and real-time performance of the autonomous vehicle swarm, a method for constructing the autonomous vehicle swarm model is presented. Finally, a solution method for the vehicle swarm model based on distributed multi-objective optimization is given, and its convergence is proven. Simulation experiments verify the effectiveness of the autonomous vehicle swarm model according to relevant evaluation metrics. This provides a model for autonomous vehicle swarms that maintains intelligent movement behavior in urban scenarios, making its application in urban environments possible.

[0071] [1] Cheng J, Yuan G, Zhou MC, et al. A fluid mechanics-based dataflow model to estimate VANET capacity[J]. IEEE Transactions on IntelligentTransportation Systems, 2019, 21(6): 2603-2614. [2] Cheng J, Ju M, Zhou M, et al. A dynamic evolution method for autonomous vehicle groups in a highway scene[J]. IEEE Internet of ThingsJournal, 2021, 9(2): 1445-1457. [3] Cheng J, Yuan G, Zhou MC, et al. A connectivity-prediction-based dynamic clustering model for VANET in an urban scene[J]. IEEE Internet ofThings Journal, 2020, 7(9): 8410-8418. The following content is also included in this invention application: To address the shortcomings of existing methods for analyzing and predicting the quality of vehicle-to-everything (V2X) wireless networks, such as neglecting the impact of various external interference factors on vehicle communication and the incompleteness of factors influencing wireless communication network performance, this paper presents an LSTM-based prediction model for the performance influencing factors of autonomous driving networks. This model includes steps such as feature fusion, identification of factors influencing autonomous driving network performance, neural network structure construction, loss function construction, and neural network training. This model can filter out loosely coupled and high-latency neighbor nodes, ensuring stable information sharing among autonomous vehicles and achieving stable and orderly vehicle movement.

[0072] In this article, "network performance" is short for "autonomous driving network performance".

[0073] In the text, "NPIFP" is short for "Network Performance Influencing Factors Prediction".

[0074] Technical solutions described in the text: Step 1 Feature Fusion Step This step involves fusing features from the autonomous vehicle and external disturbances to generate the input features for the LSTM. The relevant definitions of autonomous vehicles and external disturbances in urban scenarios are as follows: Definition 1: In urban scenarios, autonomous vehicles and external interference are... in, It represents a collection of driverless vehicles. This refers to an unmanned vehicle. Represents the set of positive integers. This indicates a group of manned vehicles. This indicates a manned vehicle. This represents a collection of roadside obstacles. This represents a roadside obstacle. Indicates the collection of traffic lights. This represents a traffic light. Indicates a group of pedestrians. It represents a pedestrian.

[0075] Definition 2: Driverless vehicles exist The characteristics of time are in, yes exist The speed of time, yes exist acceleration at any moment and They represent exist The horizontal and vertical coordinates of the current position at any given time. yes exist The direction of travel at any given moment.

[0076] Definition 3: Manned vehicle exist The characteristics of time are in, yes exist The speed of time, yes exist acceleration at any moment and They represent exist The x and y coordinates of the current position at any given time. yes exist The direction of travel at any given moment.

[0077] Definition 4: Roadside obstacles exist The characteristics of time are in, and They represent The x and y coordinates of the location, and They represent Length and width.

[0078] Definition 5: Traffic lights exist The characteristics of time are in, and They represent The x and y coordinates of the location, yes exist The state at any given moment, among which, 1 and 2 represent red light, green light and yellow light respectively.

[0079] Definition 6: Pedestrian exist The characteristics of time are in, yes exist The speed of time and They represent exist The x and y coordinates of the current position at any given time. express exist The direction of travel at any given moment.

[0080] Definition 7: Driverless vehicles With self-driving vehicles or With manned vehicles The relative characteristics between them are in, , , and They represent and or and Between The relative velocity, acceleration, distance, and direction of motion at any given moment.

[0081] This paper presents an LSTM-based prediction model for factors influencing the performance of autonomous driving networks. The method comprises two parts: feature fusion and prediction of these factors. Feature fusion combines features from the autonomous vehicle and external interference to generate the input features for the LSTM. In predicting the performance of autonomous vehicles... and When considering factors affecting network performance, the input characteristics of LSTM are: in, express and external interference Characteristics of time, for time A group of driverless vehicles gathered around. , , and They represent time There are drivers, roadside obstacles, traffic lights, and pedestrians gathering nearby. Similarly, express and external interference Characteristics of a moment.

[0082] Step 2: Identifying Factors Affecting Autonomous Driving Network Performance This step analyzes the metrics affecting connectivity quality between autonomous vehicles and determines the LSTM output. The main factors influencing autonomous vehicle network performance include the probability of communication interruption between autonomous vehicles, data correct delivery rate, link capacity, connectivity duration, data transmission latency, and round-trip time, with the following specific meanings: 1) Probability of communication interruption : Indicates driverless vehicles and The probability that the communication link between them may be interrupted at some point in the future.

[0083] 2) Data correct delivery rate : Indicates driverless vehicles and The probability of correct data delivery between them.

[0084] 3) Link capacity : Indicates driverless vehicles and The amount of data successfully transmitted per unit time in the communication link between them.

[0085] 4) Connection duration : Indicates an autonomous vehicle that is currently communicating. and The longest time that the communication link between them can remain connected.

[0086] 5) Data transmission delay : Towards Sending data indicates The time required to send from the first bit to the last bit.

[0087] 6) Communication round-trip time : Towards Sending data indicates from Data sent received The confirmation time.

[0088] Step 3: Neural Network Structure Construction This step constructs an LSTM-based neural network structure. Because the motion of autonomous vehicles is time-dependent—meaning the current communication state between autonomous vehicles is closely related to the previous state—and LSTM is suitable for processing and predicting time-series data, this paper inputs the fused features into the LSTM neural network, processing them through the LSTM input layer, hidden layer, and output layer. Then, the features from the LSTM output layer are input into a fully connected layer for linear transformation. Finally, the fully connected layer outputs the predicted values ​​of the factors influencing the performance of the autonomous driving network.

[0089] Step 4: Loss Function Construction As shown in step 2, predicting the factors influencing the performance of autonomous driving networks is a regression problem. Therefore, we choose the mean squared error as the loss function, and the objective function for predicting the factors influencing the performance of autonomous driving networks based on LSTM is: in, Represents the weights in a neural network. Indicates the number of training samples. Indicates the first Input features of each sample Indicates the first In the nth sample The true values ​​of factors affecting the performance of autonomous driving networks Indicates the first In the nth sample Predicted values ​​of factors affecting the performance of autonomous driving networks.

[0090] Step 5: Training the Neural Network The specific steps of the neural network training method for predicting the factors affecting the performance of autonomous driving networks based on LSTM are as follows: (1) Initialize the weight parameters of the LSTM neural network and divide the training set into several sub-training sets of the same size.

[0091] (2) For all input samples in a subset, firstly, the predicted values ​​of the factors affecting the performance of the autonomous driving network are calculated by forward propagation through the LSTM neural network, then the loss is calculated, and finally the weights of the LSTM neural network are updated by backpropagation. .

[0092] (3) Repeat step (2) for all sub-training sets to update the weight parameters of the LSTM neural network.

[0093] (4) After all sub-training sets have been iterated, save the structure of the LSTM neural network and the weights between nodes. The training of the LSTM neural network is now complete.

[0094] After the neural network is trained, a predictive model of factors affecting the performance of autonomous driving networks can be obtained, applicable to urban scenarios. This model helps autonomous vehicles filter out neighboring autonomous vehicles with low coupling and high latency, enabling them to establish data-sharing relationships with neighboring autonomous vehicles that have better connectivity (lower communication interruption probability, lower data transmission latency and round-trip time, higher data correct delivery rate, link capacity, and longer connection duration). This allows autonomous vehicles to share perception results and decision-making information with neighboring autonomous vehicles in real time and accurately during operation, improving their perception of the driving environment and driving safety, ultimately achieving stable and orderly autonomous vehicle behavior.

Claims

1. A method for constructing an autonomous vehicle swarm model in urban scenarios, characterized in that, Includes the following steps: Step 1. Selection of the leader node for the autonomous vehicle swarm: Step 1.1 Relevant Definitions; Step 1.2 Measurement of the activity level of autonomous vehicles; Activity level measures the level of activity of autonomous vehicles. The activity level is: in, express A collection of driverless vehicles from neighboring areas. Indicates driverless vehicles and The distance between them This indicates the maximum communication range of the autonomous vehicle. Represents a set Size, express and The probability of communication interruption between them express and Is there a communication link between them? Step 1.3 Sensitivity measurement of autonomous vehicles; The first sense of self-driving vehicles in, , , and These represent driverless vehicles. The number of people driving, obstacles, traffic lights, and pedestrians detected; Step 1.4 Measurement of mobility differences of autonomous vehicles; driverless vehicles Mobility differences in, This indicates the maximum speed of the driverless vehicle. express and The relative speed between them; Step 1.5 Selection of the leading node for the autonomous vehicle swarm; Step 2. Building the autonomous vehicle swarm model: Step 2.1 Connectivity measurement of autonomous vehicle swarm; Self-driving car swarm connectivity in, Indicates driverless vehicles For car groups The leading or secondary leading node, Indicates driverless vehicles for The following node, , and They represent and The probability of communication interruption, data correct delivery rate, and link capacity between them; Step 2.2 Coupling measurement of autonomous vehicle swarms; Self-driving car swarm Coupling in, express and Connection duration between them; Step 2.3 Real-time measurement of the autonomous vehicle swarm; Self-driving car swarm Real-time in, and They represent and Data transmission delay and communication round-trip time between them; Step 2.4 Autonomous vehicle swarm model, which is as follows: in, Indicating a fleet of self-driving cars The car group model, Indicates driverless vehicles For car groups The leading or secondary leading node, express The set of candidate follower nodes, After receiving a vehicle group join request from a free node, add it to the group. , Indicates a group of vehicles The set of candidate members Indicates Agree to join The set of free nodes, Indicates all passing train groups The leading and secondary leading nodes agreed to join the vehicle swarm. The set of free nodes, Indicates a group of vehicles Not reached maximum scale Indicates the data stream congestion density; express and The directions of movement are the same. express lie in Ahead; Step 3. Solve the autonomous vehicle swarm model.

2. The method for constructing an autonomous vehicle swarm model in an urban scenario as described in claim 1, characterized in that, In step 1: Definition 1: In urban scenarios, autonomous vehicles and external disturbances are defined as follows: in, It represents a collection of driverless vehicles. This refers to an unmanned vehicle. Represents the set of positive integers. This indicates a group of manned vehicles. This indicates a manned vehicle. This represents a collection of roadside obstacles. This represents a roadside obstacle. Indicates the assembly of traffic lights. This represents a traffic light. Indicates a group of pedestrians. Indicates a pedestrian; Definition 2: In urban scenarios, a swarm of autonomous vehicles is represented as: in, It represents a collection of driverless vehicles. Represents the set of types of autonomous vehicles, where 1 and 2 represent the autonomous vehicle swarm leader, secondary leader, and ordinary node, respectively. Indicates a free node. This indicates a swarm of driverless vehicles. This represents a swarm of driverless vehicles, in which It is a group of cars The leading node, For car groups The set of members, Indicates a group of vehicles The set of follower relationships among members, Represents a mapping function for autonomous vehicle types; Definition 3: Autonomous vehicle swarm The set of leading nodes and secondary leading nodes is as follows: in, Indicates driverless vehicles For car groups Members, express For car groups Leading or secondary leading nodes; Definition 4: Autonomous vehicle swarm Middle leading node or secondary leading node The set of following nodes: in, Indicates driverless vehicles pass Join the car group ; Construct a function for selecting the leader node in an autonomous vehicle swarm: in, , and The weighting coefficient is used to... , and Normalization.

3. The method for constructing an autonomous vehicle swarm model in an urban scenario as described in claim 2, characterized in that, In step 1, the specific process of selecting the leader node of the autonomous vehicle swarm is as follows: S1) Autonomous vehicles exchange speed, location, and vehicle attribute information with neighboring autonomous vehicles; S2) Calculate the activity level, awareness, and mobility differences of autonomous vehicles; S3) Calculate the leader node selection function for autonomous vehicles; S4) The autonomous vehicle will send the lead node's selected function value to its neighboring autonomous vehicles; S5) The autonomous vehicle compares its own leader node selection function value with that of its neighboring autonomous vehicles; S6) If the driverless vehicle node If the leader node's selection function value is greater than or equal to that of all neighboring autonomous vehicles, then... Transform into a vehicle convoy leading node; S7) If the driverless vehicle node The leader node's selection function value is less than that of its neighboring autonomous vehicles. Send a swarm join request to neighboring autonomous vehicles.

Images