Pinning control method for multi-network heterogeneous vehicle-mounted system

Abstract

The invention provides a pinning control method for a multi-network heterogeneous vehicle-mounted system, which belongs to the technical field of intelligent transportation, is implemented through a distributed control system, and comprises the following steps: a road side unit acquires state information of a vehicle in real time based on a vehicle-mounted unit and sequentially constructs an asymmetric adjacent matrix, a Laplacian matrix and a symmetric part thereof; the road side unit defines a target function by taking the minimum number of unique pinning nodes in the whole network range as an optimization target, adopts stability constraint LMI and executes an algorithm to solve the target function, and outputs an optimal pinning node set; and the distributed control system sends the output to the corresponding road side unit and the vehicle. For overlapped nodes, a priority weighting formula mechanism is adopted; and the cloud platform maintains stable operation of the distributed control system through dynamic adjustment and monitoring. According to the method disclosed by the invention, the stable control on the large-scale vehicle network is realized by using the least pinning nodes, and the calculation and communication burden on the roadside unit and the cloud platform is greatly reduced.

Description

Technical Field

[0001] This invention belongs to the field of intelligent transportation technology, and in particular relates to a pinning control method for a multi-network heterogeneous vehicle system. Background Technology

[0002] Intelligent Transportation Systems (ITS) are the core of modern traffic management, relying on Vehicular Ad-hoc Networks (VANETs) to enable real-time data interaction between vehicles (V2V) and between vehicles and infrastructure (V2I). As the backbone of ITS, VANETs are crucial for traffic coordination, safety, and efficiency. With the increasing scale of vehicles and the growing complexity of applications, VANETs face highly dynamic challenges, including frequent network topology changes, differences in node movement speeds, and the coupling characteristics of heterogeneous networks (such as subnets for private vehicles, public transportation, and freight vehicles). These characteristics make traditional centralized control methods computationally and communicationally too expensive to be suitable for large-scale deployment.

[0003] In the field of VANETs control, to reduce dependence on roadside units (RSUs) and cloud platforms, the pinning control strategy has been proposed. This strategy guides the entire network to a desired state (such as consensus speed or position) by controlling a few key nodes (such as vehicles). Existing research has applied consensus algorithms (such as stability analysis based on Lyapunov theory and the Linear Matrix Inequality (LMI)) and optimization methods (such as genetic algorithms) to select pinning control nodes. However, real-world VANETs often involve multi-task networks (such as congestion management and collision avoidance), where vehicles may act as overlapping nodes belonging to multiple subnets simultaneously, resulting in asymmetric coupling (such as the unidirectional influence of the preceding vehicle on the following vehicle). Existing technologies are mostly based on static routing protocols or simple distributed strategies, without systematically considering the cooperative mechanisms of heterogeneous networks, leading to limited control efficiency.

[0004] The existing technology has the following defects and shortcomings: 1) Insufficient adaptability to dynamic topology and asymmetric coupling.

[0005] Most existing pinning control methods assume that the network topology is symmetrical or static, failing to adequately consider the asymmetric coupling (such as the unidirectional influence of the preceding vehicle on the following vehicle) and frequent topology changes prevalent in VANETs. This leads to the control strategy easily failing in real high-mobility environments, making it difficult to guarantee system stability.

[0006] 2) The coordination mechanism for overlapping nodes in multiple networks is imperfect.

[0007] In heterogeneous VANETs with multiple networks, existing methods have failed to effectively coordinate conflicting objectives of overlapping nodes in different networks (e.g., emergency vehicles participating in both emergency and regular traffic networks simultaneously). The lack of priority weights and composite objective mechanisms leads to low control efficiency and may even cause system oscillations.

[0008] 3) Limitations of the control node selection optimization method.

[0009] Existing optimization methods (such as random search, high centrality selection, and greedy algorithms) often get stuck in local optima or select all nodes as control nodes when dealing with large-scale multi-network problems, which violates the original intention of pinned control to "minimize control nodes". In addition, these methods do not effectively integrate the stability constraint of linear matrix inequality (LMI) into the optimization process, resulting in low control accuracy and large errors.

[0010] 4) High consumption of computing and communication resources.

[0011] Many existing control strategies require the participation of the entire network or a large number of control nodes, which significantly increases the computing load and communication bandwidth requirements of RSUs and cloud platforms, limiting their practical deployment in resource-constrained scenarios.

[0012] 5) Lack of a unified control framework for multiple networks.

[0013] Existing research focuses on single-network control and lacks a unified control framework applicable to multi-network heterogeneous systems, making it impossible to achieve system-level coordination while ensuring the independent objectives of each network. Summary of the Invention

[0014] The purpose of this invention is to address the problems of insufficient adaptability to dynamic topology and asymmetric coupling, imperfect coordination mechanisms for overlapping nodes in multiple networks, limitations of control node selection optimization methods, high consumption of computational and communication resources, and lack of a unified control framework for multiple networks. It provides a pinning control method for heterogeneous multi-network vehicle systems, implemented through a distributed control system. The roadside unit distributed control system includes vehicle units, roadside units, and a cloud platform. The method includes the following steps: Step 1: The onboard unit's sensors collect vehicle status information in real time and transmit it to the roadside unit. The roadside unit then constructs an asymmetric adjacency matrix based on this status information. Each subnet Laplace matrix and its symmetrical parts Used for stability analysis; Step 2: The roadside unit optimizes by minimizing the number of unique pinned nodes across the entire network. An objective function is defined, and a stability-constrained LMI (Low Mobility Index) is applied. An adaptive genetic algorithm is then executed to solve the objective function, outputting the globally optimal set of pinned nodes. ; Step 3: The distributed control system outputs the optimal solution. The information is distributed to the corresponding roadside units and vehicles. For overlapping nodes, a priority weighting formula mechanism is adopted to ensure that safety-critical objectives dominate the control process. Step 4: The cloud platform maintains the stable operation of the distributed control system through dynamic adjustments and monitoring.

[0015] Furthermore, in step 1, the roadside unit state information constitutes a state vector. Including position coordinates and velocity vector It is transmitted to the roadside unit via V2V communication broadcast, whereby... .

[0016] Furthermore, in step 1, the roadside unit calculates communication quality indicators between vehicles based on the received status information, including received signal strength indication, packet delivery rate, and distance between vehicles. The roadside units dynamically construct an asymmetric adjacency matrix according to rules. , used for each subnet k, Elements of the network adjacency matrix The rule is expressed as: ,if ; in, For receiving signal strength indication; Package delivery rate; The distance between vehicles; This is a function that maps signal quality metrics to link weights. For communication range; It is a time-varying channel quality factor used to characterize environmental disturbances.

[0017] Furthermore, in step 1, the Laplace matrix of the roadside unit... Represented as: - ; in, It is a degree matrix; Laplace matrix The elements are represented as: ; The symmetric part of the Laplace matrix is ​​expressed as: .

[0018] Furthermore, in step 2, the objective function of the roadside unit is defined as: ,in ; in, For vehicles In subnet Whether a binary variable is pinned; The stability constraint LMI means that the objective function must satisfy a linear matrix inequality to guarantee system stability. This linear matrix inequality is expressed as: ; in, Let Q be the pinning control matrix for subnet k; Q = >0, q>0, there exists a sufficiently large constant. >0、 >0, and .

[0019] Furthermore, in step 2, the execution of the roadside unit adaptive genetic algorithm includes the following steps: Step 2.1: Generate Population Size Each individual is an N×K binary matrix representing the pinning state of all vehicles across all networks, with crossover probabilities set. =0.8, Probability of Mutation =0.05 and initial pinning probability =0.2; Step 2.2: For each individual in the population, use the LMI solver to find the feasible control gain and weighting matrix based on the current solution. When LMI is feasible, the fitness is expressed as: ; If LMI is not feasible, a penalty is imposed, expressed as: ; in, The Frobenius norm for LMI constraint violation; For adaptive penalty coefficients; m represents the "multi-network" scenario; Step 2.3: Repeat the selection, crossover, mutation, and evaluation process until the maximum number of generations is reached. =20 or convergence, output the globally optimal set of pinned nodes. .

[0020] Furthermore, in step 3, the composite target state of the overlapping nodes. The calculation formula is expressed as: ; in, [0,1] represents the priority of subnet k; In a multi-network system, the target consistent state of each subnet k is the desired state that all vehicles in subnet k should converge to.

[0021] Furthermore, in step 3, for each vehicle that was nailed... The instructions for its control input are determined jointly by all the networks to which it belongs, and are expressed as: ; in, For subnet The internal coupling matrix is ​​a positive definite matrix used to describe the internal coupling relationship between vehicle state variables in subnet k; The roadside unit sends the control input command to the vehicle's onboard unit via the V2I link. The pinned node executes the command and is influenced by neighboring vehicles, ultimately causing the entire network to converge to the desired state within 1-2 seconds.

[0022] Furthermore, in step 4, dynamic adjustments and monitoring are performed through topology updates, parameter reconfiguration, and performance monitoring. The roadside unit topology update is as follows: the roadside unit periodically updates the adjacency matrix and Laplace matrix based on the latest vehicle status and communication quality; Specifically, parameter reconfiguration involves re-triggering the genetic algorithm to optimize the network topology when the changes exceed a threshold, updating the pinned node selection and control gain. Performance monitoring specifically involves: real-time monitoring of the system state error ||e(t)|| by the cloud platform and roadside units, ensuring its convergence to... Magnitude.

[0023] Furthermore, the on-board unit is integrated into an embedded system inside the vehicle, connected to the vehicle's control system via a CAN bus, communicating with neighboring vehicles via a V2V antenna, and connected to roadside units via a V2I antenna. The roadside unit is a chassis deployed on the roadside. Each roadside unit communicates with each other and with the cloud platform through the backhaul network. The roadside unit cloud platform is a server cluster located in the data center, which is connected to all roadside units through the core network; The roadside unit distributed control system uses a cloud platform to specify and monitor global strategies, receives status information from each vehicle through roadside units and generates optimal control commands, and collects status information from each vehicle through on-board units and executes commands.

[0024] Compared with the prior art, the beneficial effects of the present invention are mainly reflected in:

[0025] Compared with the prior art, the beneficial effects of the present invention are mainly reflected in: 1. This invention establishes for the first time a unified Lyapunov stability theory framework for pinning control of multi-network vehicle-to-everything (V2X) systems with asymmetric coupling and overlapping nodes, solving the core problem of multi-network target conflict, ensuring that the stability of safety-critical applications (such as collision avoidance) takes precedence over general applications (such as traffic coordination), and providing a rigid theoretical basis for the controllability of complex V2X systems.

[0026] 2. This invention introduces a multi-network overlapping node coordination mechanism, including priority weights and composite target states, which naturally guides the algorithm to prioritize the selection of overlapping nodes that can simultaneously affect multiple networks as pinning points, thereby achieving a "killing multiple birds with one stone" control effect.

[0027] 3. This invention transforms the complex stability LMI conditions into a penalty function in a genetic algorithm, guiding the search process to automatically tend toward a solution that is both stable and efficient.

[0028] 4. The method of the present invention can achieve stable control of large-scale vehicle-to-everything (V2X) networks with the fewest direct control nodes (pinned nodes), which greatly reduces the computational and communication burden on roadside units and cloud platforms.

[0029] 5. Penalty coefficient in the algorithm of this invention The algorithm can be dynamically adjusted according to the proportion of infeasible solutions in the population, which enhances the convergence and robustness of the algorithm under complex constraints. Attached Figure Description

[0030] Figure 1 This is a flowchart illustrating the pinning control method for a multi-network heterogeneous vehicle system according to the present invention. Detailed Implementation

[0031] The following will describe in more detail a pinning control method for a multi-network heterogeneous vehicle system according to the present invention with reference to the schematic diagram, which illustrates the preferred embodiments of the present invention. It should be understood that those skilled in the art can modify the present invention described herein while still achieving the advantageous effects of the present invention. Therefore, the following description should be understood as being of general knowledge to those skilled in the art and is not intended to limit the present invention.

[0032] A distributed control system mainly includes an on-board unit, a roadside unit, a cloud platform, and communication network components.

[0033] 1. On-Board Unit (OBU) 1) Shape and structure: Embedded systems integrated into vehicles include multi-core processors (such as the ARM Cortex-A series), memory (RAM / Flash), V2X communication modules (supporting IEEE 802.11p / DSRC), and sensor interfaces (GPS receiver, IMU).

[0034] 2) Connection relationship: It connects to the vehicle control system via CAN bus; communicates with surrounding vehicles via V2V antenna; and connects to the RSU via V2I antenna.

[0035] 3) Working principle and function: The OBU periodically collects its own status, encapsulates it into a BSM message, and broadcasts it via V2V. Simultaneously, it receives control commands from the RSU. After being processed by the processor, it executes specific control actions through the vehicle control system (such as accelerator, brake, and steering). It is the controlled node and state-aware node in the network.

[0036] 2. Road-Side Unit (RSU) 1) Shape and structure: The robust chassis, deployed on the roadside (such as traffic light poles), contains an industrial-grade multi-core CPU (such as Intel Xeon), large-capacity memory and solid-state drives, a V2I communication module (DSRC / C-V2X), and a fiber optic or 5G backhaul network interface.

[0037] 2) Connection relationship: It connects to the vehicle OBU within the coverage area via V2I communication; and communicates with the cloud platform and other RSUs via the high-speed backhaul network.

[0038] 3) Working principle and function: The RSU acts as the "brain" of its coverage area. It collects BSM messages from all vehicles, constructing a dynamic adjacency matrix and Laplace matrix. It runs an adaptive genetic algorithm to solve for the optimal set of pinned nodes. Finally, it distributes the calculated control commands to the selected pinned nodes. Its role is to perform optimization and control locally, significantly reducing the burden on the cloud platform.

[0039] 3. Cloud Platform 1) Shape and structure: The server cluster is located in the data center and is equipped with high-performance CPUs / GPUs, massive distributed storage systems, and high-speed network switching equipment.

[0040] 2) Connectivity: Connects to all RSUs through the core network to aggregate global data.

[0041] 3) Working principle and function: The cloud platform is responsible for overall monitoring and high-level coordination. It analyzes cross-regional traffic conditions and sets unified target states for each RSU. and priority weight In extreme cases, it can take over and perform global optimization. It is the highest decision-making level of the system.

[0042] 4. Communication network components 1) V2V communication module: Based on the IEEE 802.11p protocol, it operates in the 5.9GHz frequency band, with a transmission range of 300-1000 meters and a data rate of 3-27Mbps. It is responsible for low-latency status information exchange between vehicles.

[0043] 2) V2I communication module: Supports DSRC and C-V2X standards, providing low-latency (<100ms) and high-reliability (>99%) control commands and aggregated data transmission between the vehicle and RSU.

[0044] 3) Backhaul network: It adopts a hybrid network of fiber optic and 5G to provide a high-speed (>1Gbps) and low-latency (<10ms) data transmission channel between RSU and cloud platform.

[0045] Summary of Distributed Control Systems: All vehicles form a multi-hop self-organizing network through V2V communication. The OBU (On-Board Unit) acts as the end node, responsible for perception and execution. The RSU (Remote Utility Unit) acts as the edge computing node, responsible for regional real-time optimization and control. The cloud platform acts as the central hub, responsible for global policy formulation and monitoring. The typical control flow path is: Vehicle status → OBU → (V2V / V2I) → RSU → (Genetic algorithm optimization) → Control command → (V2I) → Pinned node OBU → Vehicle execution. This layered distributed architecture enables efficient and reliable large-scale vehicular network pinning control.

[0046] A pinning control method for a multi-network heterogeneous vehicle system is implemented through the aforementioned distributed control system, such as... Figure 1 As shown, the specific steps include: Step 1: System parameter acquisition and network topology construction.

[0047] 1. Vehicle status information collection: The vehicle's status information, including location coordinates, is collected in real time by sensors (such as GPS and IMU) in the on-board unit (OBU). velocity vector Together they constitute the state vector This data is broadcast at a frequency of 10 Hz via V2V communication (based on the IEEE 802.11p protocol) to form basic BSM messages.

[0048] 2. Adjacency matrix construction: The roadside unit (RSU) calculates inter-vehicle communication quality metrics based on BSM messages reported by vehicles, including Received Signal Strength Indicator (RSSI), Packet Delivery Rate (PDR), and inter-vehicle distance. The roadside units dynamically construct an asymmetric adjacency matrix according to rules. For each subnet k, the network adjacency matrix elements Defined as follows: ,like ; in, For receiving signal strength indication; Package delivery rate; The distance between vehicles; This is a function that maps signal quality metrics to link weights. The communication range is typically 300-1000 meters. It is a time-varying channel quality factor used to characterize environmental disturbances.

[0049] 3. Calculation of the Laplace matrix: RSU is based on the adjacency matrix. Calculate the Laplacian matrix for each subnet k. First, calculate the degree matrix. Then the Laplace matrix is ​​calculated. - The Laplace matrix The element is represented as : ; And further calculate its symmetrical part Used for stability analysis.

[0050] Step 2: Optimize problem construction and algorithm solution.

[0051] The pinning control problem in multi-network heterogeneous vehicular systems represents a significant extension and deepening of the single-network model. In real-world VANET environments, vehicles often simultaneously participate in multiple task-specific networks (such as congestion management, emergency response, and fleet collaboration), each with its own independent topology, coupling strength, and control objective. Therefore, this paper first constructs a stability theoretical framework for single-network pinning control. Based on this, by introducing the concept of overlapping nodes (i.e., vehicles simultaneously belonging to multiple networks) and extending the vehicle dynamics equations to a superposition of the coupling effects of multiple networks, a more realistic multi-network system model is established. 1. Single-network vehicle system pinning control model.

[0052] For dynamic topology and asymmetric coupling problems, a stability theoretical framework for pinning control is established to ensure system convergence under various conditions. Consider a vehicular ad hoc network (VANET) consisting of N vehicles, modeling VANET as a continuous-time multi-agent system. The state of each vehicle i is represented by a vector... This indicates that physical quantities, including position and velocity, are considered. The dynamic equations of the system are described as follows: ; in, Let be the state vector of vehicle i at time t in a single network; C>0: the coupling strength between vehicles determines the degree of influence between connected vehicles; The internal coupling matrix satisfies Γ≻0 (positive definite); G is an asymmetric adjacency matrix The elements describe the communication links between vehicles; c>0 is the pinning control gain constant, which adjusts the strength of the pinning control. ∈{0,1} is the pinning indicator variable. =1 indicates that vehicle i is nailed (directly controlled); The desired state that all vehicles should converge to, representing a consistent objective state.

[0053] Define error The error dynamics equation is derived, and the conditions for achieving pinned consensus are obtained as follows: .

[0054] 2. Stability analysis of a single network.

[0055] Consider a continuous-time vehicular network system consisting of N vehicles. The Laplace matrix L corresponding to the adjacency matrix G is stable (all non-zero eigenvalues ​​have positive real parts). Let Q = >0, where q>0, and Γ= Using Lyapunov functions Prove its stability.

[0056] ; If there exists a sufficiently large constant c > 0 such that the following matrix inequality holds, the stability condition is given by LMI: ; in Laplace matrix The symmetrical part, For pinning matrix, It is a constant; These are the identity matrices, which are N×N and m×m respectively.

[0057] The system can then asymptotically achieve pinned consensus, that is: ; 3. Extended model of multi-network heterogeneous vehicle system.

[0058] Consider K consecutive-time vehicular networks, each denoted as . ,in A vehicle may belong to multiple networks simultaneously, forming an overlapping set of nodes: ; For multi-network systems, the vehicle dynamics equations are extended to: ; The newly added parameters are: Let i be the set of networks to which vehicle i belongs; Let i be the set of neighbors of vehicle i in subnet k; >0 represents the coupling strength of subnet k; Let be the internal coupling matrix of subnet k; Let k be the adjacency matrix of subnet k. Element; >0 represents the pinning control gain of subnet k; ∈{0,1} is the pinning indicator variable for vehicle i in subnet k; This indicates the goal-consistent state of subnet k.

[0059] 4. Multi-network stability analysis.

[0060] Consider the aforementioned multi-network heterogeneous in-vehicle system. Assume the internal coupling matrix... ,in Let Q = >0, where q>0, if there exists a sufficiently large constant. >0、 A value greater than 0 makes the following matrix inequalities hold: ; in, For the symmetric part of the Laplace matrix of subnet k; Let k be the pinning control matrix for subnet k; The system can then asymptotically achieve pinned consensus, that is: ; By formalizing the stability conditions through LMI, the control gain c and pinning configuration can be dynamically adjusted to adapt to topology changes and asymmetric coupling, ensuring asymptotic convergence of the system.

[0061] 5. Execution of Adaptive Genetic Algorithm (GA).

[0062] 1) Genetic manipulation parameters Population size generated , Adjusted according to network size (50 vehicles: 100, 100 vehicles: 200, 200 vehicles: 400), each individual is an N×K binary matrix representing the pinning state of all vehicles in all networks.

[0063] Crossover probability: ; Mutation probability: ; Initial pinning probability: Encourage sparse selection.

[0064] Fitness assessment: for each individual in the population (i.e., a pinned configuration): (1) LMI solution: Use an LMI solver (such as MATLAB YALMIP / SeDuMi) based on the current Solve for feasible control gain And the weighted matrix Q.

[0065] (2) Calculate fitness: If LMI is feasible, the fitness is If this is not feasible, then a penalty will be imposed: .in, The sum of all network LMI constraint violations. For subnet The constraint is violated (Frobenius norm). It is an adaptive penalty coefficient that dynamically adjusts with each generation: ; in, The proportion of infeasible solutions in the g-th generation; For the fitness rate.

[0066] (3) Iteration and output: Repeat the selection, crossover, mutation and evaluation process to select the best individual to enter the next generation until the maximum number of generations is reached. =20 or convergence. Output the globally optimal set of pinned nodes. .

[0067] This invention achieves stability under the minimum control node by using a genetic algorithm to globally search for pinned configurations, a penalty function to guide the search for feasible solutions that satisfy LMI, and an adaptive mechanism to balance exploration and utilization.

[0068] Step 3: Deployment and execution of control strategies.

[0069] 1. Selection and coordination of pinning nodes: The optimal solution output by the genetic algorithm The data is distributed to the corresponding RSUs and vehicles. For overlapping nodes (vehicles belonging to multiple networks), a priority-weighted consensus mechanism is used. Its composite target state... The calculation is as follows: ; in, In a multi-network system, the target consistent state of each subnet k is the desired state that all vehicles in subnet k should converge to. [0,1] represents the priority of subnet k (safety-related subnets have higher weights). For example, collision avoidance network ρ=0.9, traffic flow network ρ=0.5). This ensures that safety-critical objectives dominate control.

[0070] To handle target conflicts between networks, slack variables are introduced. In the LMI constraint, a moderate degradation in the performance of non-critical networks is allowed: ;

[0071] in, Let be the relaxation parameters of subnet k; These are the baseline relaxation parameters. This framework allows for moderate performance degradation in non-critical networks while ensuring stringent performance for critical networks.

[0072] 2. Control input calculation and application: For each nailed vehicle i, the control input instructions are determined jointly by all the networks to which it belongs: ; in, For subnet The internal coupling matrix is ​​a positive definite matrix used to describe the subnet. The internal coupling relationship between vehicle state variables; The control input command is sent to the vehicle's OBU via the RSU through the V2I link. Pinned nodes execute the control law, while non-pinned nodes communicate via V2V, influenced by the status of neighboring vehicles, ultimately causing the entire network system to converge to the desired state within 1-2 seconds.

[0073] Step 4: Dynamic adjustment and monitoring.

[0074] 1. Topology update: RSU periodically (e.g. every 10 seconds) updates the adjacency matrix and Laplace matrix based on the latest vehicle status and communication quality.

[0075] 2. Parameter reconfiguration: When the network topology changes beyond a certain threshold, the genetic algorithm is re-triggered for optimization, updating the pinned node selection and control gain.

[0076] 3. Performance Monitoring: The cloud platform and RSU monitor the system state error ||e(t)|| in real time to ensure that it converges to The scale is sufficient to maintain stable system operation.

[0077] Computational complexity analysis: The time complexity of a single-generation genetic algorithm is: ; in, To reduce the complexity of solving LMI using the interior point method; The complexity of fitness evaluation is reduced. Through distributed LMI solving, the LMI of each subnet k is solved independently by the local controller, reducing the complexity from... Down to Only overlapping nodes need to be coordinated.

[0078] The total complexity in the case of multiple networks is: ; This study reduces the burden on central processing by using parallelization and localized computing, making it suitable for large-scale VANET deployments.

[0079] This invention, through extensive simulation experiments and comparison with various mainstream optimization methods, and in-depth parameter sensitivity analysis, fully demonstrates its significant beneficial effects in the following aspects: 1. Significantly reduces control costs and communication load, achieving efficient control.

[0080] This invention enables stable control of large-scale vehicle-to-everything (V2X) networks with a minimum number of direct control nodes (pinned nodes), greatly reducing the computational and communication burden on roadside units and cloud platforms.

[0081] In a multi-network scenario with 50 nodes, as shown in the comprehensive data in Table 1-4, the present invention only requires an average of 11-12 pinned nodes (approximately 22%-24% of the total number of nodes) to achieve system stability.

[0082] In the comparative experiments, random search, the overlapping node priority strategy, and the ant colony optimization algorithm all failed to select all 50 nodes as pinning nodes, which completely violates the original intention of pinning control as "partial control." Particle swarm optimization also requires 15-20 nodes and has a higher control error. This result strongly demonstrates the superior effectiveness of this invention in minimizing control nodes.

[0083] Table 1: Initial pinning probability involved ( The impact on the number of pinning nodes and control error.

[0084]

[0085] Table 2: Population Size Involved ( The impact on performance.

[0086]

[0087] Table 3: Variation Rates Involved The impact of this on the optimization results.

[0088]

[0089] Table 4: The role of the penalty coefficient (λ) in the tradeoff between the number of pinning nodes and control error.

[0090]

[0091] 2. Achieve an excellent balance between control precision and the number of nodes.

[0092] This invention does not simply pursue the minimum number of nodes, but rather finds the optimal balance between reducing control costs and maintaining high control accuracy.

[0093] As shown in Table 4, by adjusting the penalty coefficient This invention can precisely control this trade-off. When At that time, the algorithm achieved a low latency of only 11.7 pinning nodes on average. The magnitude of the control error.

[0094] This performance far surpasses that of the greedy sequence method (20-25 nodes, error). ) and higher-order adaptive methods (error) This demonstrates the unique advantages of the LMI penalty function mechanism employed in this invention in resolving this core trade-off problem.

[0095] 3. It possesses excellent robustness and parameter insensitivity, making it easy to deploy in practice.

[0096] The algorithm of this invention is not sensitive to the setting of key parameters and can maintain stable performance in various scenarios without cumbersome parameter tuning. This greatly enhances its practicality and ease of use in real and ever-changing vehicle connectivity environments.

[0097] Table 1 shows the initial pinning probability. When the value varied over a wide range of 0.1 to 0.5, the number of pinning nodes (11.4–12.2) and the control error remained stable.

[0098] Table 3 further shows the variability rate. Within the range of 0.05-0.10, the algorithm's performance also fluctuates very little.

[0099] This low sensitivity to parameters means that in actual deployments, engineers do not need to perform a lot of trial-and-error parameter tuning, which lowers the application threshold and maintenance costs.

[0100] The above are merely preferred embodiments of the present invention and do not constitute any limitation on the present invention. Any equivalent substitutions or modifications made by those skilled in the art to the technical solutions and content disclosed in the present invention without departing from the scope of the present invention shall be deemed to have remained within the protection scope of the present invention.

Claims

1. A pinning control method for a multi-network heterogeneous vehicle system, characterized in that, Implemented through a distributed control system, which includes an on-board unit, a roadside unit, and a cloud platform, the method includes the following steps: S1: The vehicle's status information is collected in real time by the sensors of the on-board unit and transmitted to the roadside unit. The roadside unit then constructs an asymmetric adjacency matrix based on the status information. Each subnet Laplace matrix and its symmetrical parts Used for stability analysis; S2: The roadside unit optimizes by minimizing the number of unique pinned nodes across the entire network. It defines an objective function, employs a stability-constrained LMI algorithm, and executes an adaptive genetic algorithm to solve the objective function, outputting the globally optimal set of pinned nodes. ; S3: The optimal solution output by the distributed control system The information is distributed to the corresponding roadside units and vehicles. For overlapping nodes, a priority weighting formula mechanism is adopted to ensure that safety-critical objectives dominate the control process. S4: The cloud platform maintains the stable operation of the distributed control system through dynamic adjustment and monitoring.

2. The pinning control method for a multi-network heterogeneous vehicle system according to claim 1, characterized in that, In step S1, the state information constitutes a state vector. Including position coordinates and velocity vector It is transmitted to the roadside unit via V2V communication broadcast, whereby... .

3. The pinning control method for a multi-network heterogeneous vehicle system according to claim 2, characterized in that, In step S1, the roadside unit calculates communication quality indicators between vehicles based on the received status information, including received signal strength indication, packet delivery rate, and distance between vehicles. The roadside units dynamically construct an asymmetric adjacency matrix according to rules. For each subnet k, the elements of the network adjacency matrix The rule is expressed as: ,like ; in, For receiving signal strength indication; Package delivery rate; The distance between vehicles; This is a function that maps signal quality metrics to link weights. For communication range; It is a time-varying channel quality factor used to characterize environmental disturbances.

4. The pinning control method for a multi-network heterogeneous vehicle system according to claim 3, characterized in that, In S1, the Laplace matrix , It is a degree matrix; Laplace matrix The elements are represented as: ； The symmetric part of the Laplace matrix is ​​expressed as: 。 5. The pinning control method for a multi-network heterogeneous vehicle system according to claim 1, characterized in that, In S2, the objective function is defined as: ,in ; in, For vehicles In subnet Whether a binary variable is pinned; The stability constraint LMI means that the objective function must satisfy a linear matrix inequality to guarantee system stability. This linear matrix inequality is expressed as: ； in, Let Q be the pinning control matrix for subnet k; Q = >0, q>0, there exists a sufficiently large constant. >0、 >0, and .

6. The pinning control method for a multi-network heterogeneous vehicle system according to claim 5, characterized in that, In step S2, the execution of the adaptive genetic algorithm includes the following steps: S21: Population size generated Each individual is an N×K binary matrix representing the pinning state of all vehicles across all networks, with crossover probabilities set. =0.8, Probability of Mutation =0.05 and initial pinning probability =0.2; S22: For each individual in the population, the LMI solver is used to find feasible control gains and weighting matrices based on the current solution. When LMI is feasible, the fitness is expressed as: ； If LMI is not feasible, a penalty is imposed, expressed as: ； in, The Frobenius norm for LMI constraint violation; The adaptive penalty coefficient; m represents the multi-network scenario; S23: Repeat the selection, crossover, mutation, and evaluation process until the maximum number of generations is reached. =20 or convergence, output the globally optimal set of pinned nodes. .

7. The pinning control method for a multi-network heterogeneous vehicle system according to claim 1, characterized in that, In S3, the composite target state of the overlapping nodes The calculation formula is expressed as: ； in, [0,1] represents the priority weight of subnet k; In a multi-network system, the target consistent state of each subnet k is the desired state that all vehicles in subnet k should converge to.

8. The pinning control method for a multi-network heterogeneous vehicle system according to claim 7, characterized in that, In S3, for each vehicle that is nailed... The instructions for its control input are determined jointly by all the networks to which it belongs, and are expressed as: ； in, For subnet The internal coupling matrix is ​​used to describe the internal coupling relationships between vehicle state variables in subnet k; The roadside unit sends the control input command to the vehicle's onboard unit via the V2I link. The pinned node executes the command and is influenced by neighboring vehicles, ultimately causing the entire network to converge to the desired state within 1-2 seconds.

9. The pinning control method for a multi-network heterogeneous vehicle system according to claim 1, characterized in that, In S4, dynamic adjustment and monitoring are performed through topology updates, parameter reconfiguration, and performance monitoring. The topology update specifically involves the roadside unit periodically updating the adjacency matrix and Laplace matrix based on the latest vehicle status and communication quality. The parameter reconfiguration specifically involves: when the network topology change exceeds a threshold, the genetic algorithm is re-triggered for optimization, updating the pinned node selection and control gain; The performance monitoring specifically involves: real-time monitoring of the system state error ||e(t)|| by the cloud platform and roadside units, ensuring that it converges to the specified value. Magnitude.

10. The pinning control method for a multi-network heterogeneous vehicle system according to claim 1, characterized in that, The vehicle-mounted unit is integrated into an embedded system inside the vehicle. It is connected to the vehicle's control system via a CAN bus, communicates with neighboring vehicles via a V2V antenna, and is connected to the roadside unit via a V2I antenna. The roadside unit is a chassis deployed on the roadside, and each roadside unit communicates with each other and with the cloud platform through the backhaul network; The cloud platform is a server cluster located in a data center, which is connected to all roadside units through the core network; The distributed control system uses a cloud platform to specify and monitor global strategies, receives status information from each vehicle through roadside units and generates optimal control commands, and collects status information from each vehicle through on-board units and executes commands.

Images