A communication management optimization method and system based on vehicle networking

Abstract

The application relates to a communication management optimization method based on Internet of Vehicles, which comprises the following steps: step one, terminal layer multi-modal perception and dynamic watermark embedding: real-time collection of positioning data, channel state information, neighbor node distribution and service type; step two, edge layer space-time field dynamic modeling and prediction: an electromagnetic field dynamic propagation model is constructed based on Maxwell equation, historical field intensity distribution is input into a mixed network of LSTM-CNN, and a future electromagnetic field thermal map is output, meanwhile, the vehicle position is mapped to SE(3) Lie group space, a tangent vector of the adjacent time instant position and posture is calculated, and a discrete channel quality matrix and a sudden vehicle topology are converted into a continuous space-time field and differential motion; step three, cross-layer interference prediction and quantum annealing spectrum allocation: a time-frequency spectrum diagram generated by an IQ signal is analyzed through a 1D-CNN+LSTM double-path network, and interference intensity and frequency point occupation probability are predicted in advance; and step four, service-driven protocol adaptation and light field cooperation: services are classified according to delay and reliability requirements.

Description

Technical Field

[0001] This invention relates to the field of communication management technology, and in particular to a communication management optimization method and system based on the Internet of Vehicles. Background Technology

[0002] The optimization of communication management based on vehicle-to-everything (V2X) aims to improve the efficiency and accuracy of data transmission between vehicles and road infrastructure, other vehicles, pedestrians, and cloud platforms through an efficient, reliable, and low-latency information exchange mechanism. Its core objective is to build a real-time perception, dynamic coordination, and intelligent decision-making transportation ecosystem, thereby significantly enhancing road safety (e.g., early warning of potential collision risks through V2X communication), alleviating traffic congestion (optimizing route planning using real-time traffic data), improving energy efficiency (coordinated control of vehicle speed to reduce energy consumption from sudden braking and acceleration), and promoting the reliable implementation of autonomous driving technology. Its impact extends from both individual travel experience and macro-social benefits. It allows drivers to avoid congested areas and receive traffic signal priority information through real-time navigation, while enabling traffic management departments to precisely control traffic lights and deploy emergency resources based on aggregated data. It even facilitates customized services such as public transport priority and dedicated freight lines through vehicle-road collaboration, ultimately forming a new transportation model with high collaboration between people, vehicles, roads, and the cloud, laying a solid foundation for low-carbon travel and efficient logistics under the sustainable development goals.

[0003] In existing technologies, the state space (such as the channel quality matrix and vehicle location topology) assumes a static environment. However, in actual vehicle networks, the high-speed movement of vehicles leads to frequent changes in the topology, and the channel state (such as CSI) fluctuates rapidly due to multipath effects, occlusion, and other factors. Static state representation may cause model decision lag and fail to respond in a timely manner to sudden channel interference or dense vehicle scenarios. Therefore, it is particularly important to propose a communication management optimization method and system based on vehicle networks. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a communication management optimization method and system based on vehicle networking.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: A communication management optimization method based on vehicle-to-everything (V2X) communication includes the following steps: Step 1: Terminal Layer Multimodal Perception and Dynamic Watermark Embedding: Integrating 5G NR millimeter-wave radar, lidar and optical field communication receiving module, real-time collection of positioning data, channel status information, neighbor node distribution and service type, and embedding a three-dimensional coded spatiotemporal stamp of location-time-velocity in the header of the data packet, combined with the quantum computing-resistant SM9 national cryptographic algorithm for identity authentication, and using camera + CMOS sensor to decode the red and blue light pulse signals emitted by the roadside unit, the decoding result is sent to the edge layer MEC through CAN bus; Step 2: Dynamic modeling and prediction of spatiotemporal field at the edge layer: Based on Maxwell's equations, a dynamic propagation model of electromagnetic field is constructed. The historical field strength distribution is input into the LSTM-CNN hybrid network and the future electromagnetic field heat map is output. At the same time, the vehicle position is mapped to the SE(3) Lie group space and the tangent vector of the pose at adjacent time moments is calculated. The discrete channel quality matrix and the abrupt vehicle topology are transformed into a continuous spatiotemporal field and differential motion. Step 3: Cross-layer interference prediction and quantum annealing spectrum allocation: The time spectrum of the IQ signal is analyzed by 1D-CNN+LSTM dual-path network to predict the interference intensity and frequency occupancy probability in advance to switch communication frequency bands. At the same time, the spectrum allocation is modeled as a QUBO model and the optimal allocation scheme is solved by FPGA-accelerated quantum annealing. Step 4: Service-driven protocol adaptation and optical field coordination: Services are classified according to latency and reliability requirements, and communication modes are automatically switched. At the same time, optical pulse coded signals are emitted through the LED array of the RSU, and mechanical low-latency coordination is achieved after vehicle decoding. Step 5: Meta-co-distillation learning and policy optimization: The original state space is compressed using an autoencoder, and key dimensions with variance greater than the threshold are retained. The regional coordinator calculates the policy entropy difference. When the difference is greater than the threshold, the cloud-based meta-teacher network is triggered to correct the edge policy through policy distillation.

[0006] The above plan further includes: Furthermore, the positioning data is obtained by calculating the vehicle's position coordinates using TOF ranging from a 5G NR millimeter-wave radar and SLAM algorithms from a lidar system. ; The channel state information is obtained by estimating the channel matrix using 5G NR pilot signals. Where Y is the received signal, X is the transmitted pilot, and N is noise; The neighbor node distribution is as follows: the BSM broadcast by V2X obtains the IDs, locations, and speeds of surrounding vehicles. Service type classification: Based on service QoS requirements, the SVM classifier is used to classify services into security, control, and entertainment.

[0007] Furthermore, the specific steps of the dynamic watermark embedding are as follows: Spacetime stamp encoding: The timestamp T and velocity v are mapped to a binary sequence S, represented as follows: , where ⊕ is the XOR operation and Hash is SHA-256; Watermark embedding: Insert 'S' into a reserved field in the header of the data packet to form a watermarked data packet. ; SM9 Signature: Using terminal private key d Signature, as ,in It serves as a unique identifier for the terminal.

[0008] Furthermore, the specific steps for the dynamic modeling and prediction of the spatiotemporal field of the edge layer are as follows: Step 1: Construction of Electromagnetic Field Dynamic Propagation Model: Based on Maxwell's equations, an electromagnetic field propagation model is constructed. Combining the multipath effect of vehicle-to-everything (V2X) networks and the characteristics of obstacle occupancy, a spatiotemporal evolution equation for the field strength distribution is established: In free space, the propagation of electric field E and magnetic field H satisfies the wave equation: ,in, Permeability, Given the dielectric constant, a path loss model is introduced to correct the field strength in the complex environment of vehicle-to-everything (V2X) networks: ,in, For transmission power, and For antenna gain, Let λ be the wavelength, d be the distance, and L(d) be the obstacle penetration loss. The field strength E(x,y,z,t) is considered as a spatiotemporal field, and its rate of change is determined by both the propagation model and the vehicle's motion. ,in, For vehicle speed, The field strength recovery coefficient, The interference attenuation coefficient is... This represents the number of neighboring nodes; Step 2: Spatiotemporal field prediction using a hybrid LSTM-CNN network: Input historical field strength distribution, extract temporal features using LSTM, extract spatial features using CNN, and output a future electromagnetic field heatmap: Historical field strength data is in 4D tensor Where T is the time step, H=W is the spatial resolution, C is the electric / magnetic field strength, and the LSTM layer extracts temporal features and outputs the hidden state. ,in It is the Sigmoid activation function. and This is the weight matrix. This is the hidden state from the previous moment. For the current input, For the bias term, the CNN layer extracts spatial features through convolutional kernels. Calculate the feature map: ,in, Convolution kernel weights, Input feature map values, output thermal field intensity Y for future Δt ; Step 3: Vehicle pose SE(3) Lie group mapping and differential motion modeling: Map the vehicle position to the SE(3) Lie group space, calculate the tangent vector of the pose at adjacent time steps, and transform the discrete pose into continuous motion: The vehicle pose g∈SE(3) is a 4×4 homogeneous transformation matrix: ,in, Let g(t) be the rotation matrix, t be the translation vector, and g(t+Δt) be the tangent vector of the poses at adjacent time points. Calculated via logarithmic mapping: The ∨ operation transforms the Lie algebra matrix into a 6-dimensional vector, and the rate of pose change is determined by the tangent vector: ,in ∈se(3) is the matrix form corresponding to the Lie algebra; Step 4: Conversion of Discrete Channel Quality Matrix and Topology to Continuous Spatiotemporal Field: The discrete CSI matrix and vehicle topology uploaded by the terminal are converted into a continuous spatiotemporal field to avoid state lag. Discrete CSI matrix Transformed into a continuous function through cubic spline interpolation. : ,in, For spline basis functions, the vehicle topology is represented by a graph. This represents a graph where V is a vehicle node and E(t) is an edge. The topological rate of change is transformed into a continuous field using a graph Fourier transform. Where F is the graph Fourier transform; Step 5: Generation of Local Spatiotemporal Field Prediction Results: Combining the sensing data uploaded by the terminal and the global electromagnetic field twin parameters sent from the cloud, local spatiotemporal field prediction results are generated. By fusing terminal data and cloud parameters using Kalman filtering, a thermal map of field strength Y and the rate of topological change for future Δt are generated. Input spectrum allocation engine.

[0009] Furthermore, the specific steps for generating the temporal spectrogram using the 1D-CNN+LSTM dual-path network are described below; IQ signal representation: The IQ signal is in complex form. , where I(t) and Q(t) are real components; Time-frequency spectrum generation: The signal is divided into frames using the Hamming window function w(n), and the spectrum of each frame is calculated. Where t is the time index and f is the frequency index; 1D-CNN Feature Extraction: One-dimensional convolution is performed on each frame of the temporal spectrogram using convolution kernels to extract local time-frequency features. ,in, For convolution kernel; LSTM Temporal Modeling: Input Feature Sequence Long-term dependencies are captured through LSTM gating mechanism; Interference prediction output: Outputs the interference intensity in the future Δt through a fully connected layer. Frequency occupancy probability : ,in, This represents the hidden state of the last frame of the LSTM. This is the weight matrix.

[0010] Furthermore, the specific steps for modeling spectrum allocation as a QUBO model and solving for the optimal allocation scheme using an FPGA-accelerated quantum annealer are as follows: QUBO objective function: Let the spectrum allocation variable be... ( =1 indicates that frequency band i is allocated, and the objective function is: ,in, For the interference matrix ( =1 indicates that frequency bands i and j interfere with each other. Let i be the spectral efficiency of frequency band i; Constraint transformation: Delay constraints ≤ : Through penalty items Add the objective function; Reliability constraints ≥ : Through penalty items Add the objective function; Quantum annealing energy function: Mapping the QUBO objective function to the quantum annealing energy function: ,in, As constraints, This is the constraint threshold; Annealing process: The energy function is calculated in parallel using FPGA, and the annealing path is dynamically adjusted. ,in The initial temperature. This is the decay time constant.

[0011] According to claim 1, the communication management optimization method based on vehicle-to-everything (V2X) is characterized in that, in the service-driven protocol adaptation and optical field coordination, V2X services are divided into 5 levels according to latency and reliability requirements, and the service priority score S is determined by latency weight. and reliability weight Weighted summation yields: Where D is the actual time delay. R is the maximum allowable delay; R is the actual reliability. This represents the minimum reliability requirement.

[0012] A communication management optimization system used in a vehicle-to-everything (V2X) communication management optimization method includes: Spatiotemporal perception layer: By integrating a multimodal sensor array, a spatiotemporal field theory modeling engine, and edge computing nodes, it constructs a dynamic spatiotemporal field model of the vehicle and extracts spatiotemporal feature parameters in real time; Meta-collaborative decision-making layer: Based on the meta-learning framework, collaborative distillation network and digital twin engine, a joint decision-making model of vehicle-road-network is constructed and a dynamic resource allocation scheme is generated; Communication layer: Utilizing terahertz communication modules, intelligent metasurface arrays, and orbital angular momentum multiplexing technology, it enables vehicle-to-vehicle, vehicle-to-infrastructure, and vehicle-to-cloud converged communication, constructs multi-path redundant transmission channels, and dynamically adjusts communication parameters; Intelligent Application Layer: Through blockchain security protocols, federated learning platforms, and AR augmented reality interfaces, it provides vehicle-to-everything (V2X) services, supports cross-platform data value sharing, and builds human-machine interaction interfaces.

[0013] The present invention has the following beneficial effects: In this invention, a dynamic electromagnetic field model is reconstructed using Maxwell's equations, and a hybrid LSTM-CNN network is used to predict future field strength distribution. This transforms the discrete channel quality matrix into a spatiotemporally continuous field, avoiding the decision lag caused by sampling intervals in traditional static models. This achieves millisecond-level prediction of field strength evolution, accurately capturing dynamic interference such as multipath effects and vehicle obstruction. Based on 1D-CNN, the time-frequency features of the IQ signal are extracted, and combined with LSTM to predict the interference intensity and frequency occupancy probability in the next 100ms, allowing for early detection of spectrum holes and reducing the decision lag for dynamic spectrum access. This reduces the probability of interference in the 5.9GHz CBRS band. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating the method steps of a communication management optimization method based on vehicle networking proposed in this invention. Figure 2 This is a system block diagram of a communication management optimization method based on vehicle networking proposed in this invention. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] Please see Figures 1-2 As shown, this invention is a communication management optimization method based on vehicle-to-everything (V2X) communication, comprising the following steps: Step 1: Terminal Layer Multimodal Sensing and Dynamic Watermark Embedding: Integrating 5G NR millimeter-wave radar, lidar, and optical field communication receiving modules, it collects location data, channel state information (CSI), neighbor node distribution, and service type (security / control / entertainment) in real time. A three-dimensional spatiotemporal stamp with location-time-velocity encoding is embedded in the data packet header using Dynamic Watermark 2.0. This is combined with the quantum-resistant SM9 national cryptographic algorithm for identity authentication. Simultaneously, it uses a camera and CMOS sensor to decode the red and blue light pulse signals emitted by the roadside unit (RSU) (short flashes indicate limited spectrum resources, continuous light indicates secure access, and alternating red and blue indicate a conflict alarm). The decoding results are sent to the edge layer MEC via the CAN bus, triggering local protocol adjustments in advance. The terminal uploads the sensing data (CSI, location, service type) to the edge layer MEC server via a V2X link. The optical field communication signal is broadcast unidirectionally from the RSU to the vehicle, without occupying wireless spectrum resources; the vehicle only receives and does not transmit. Step 2: Dynamic modeling and prediction of spatiotemporal field at the edge layer: Based on Maxwell's equations, a dynamic propagation model of electromagnetic field is constructed. The historical field strength distribution is input into the LSTM-CNN hybrid network and the future electromagnetic field heat map is output. At the same time, the vehicle position is mapped to the SE(3) Lie group space and the tangent vector of the pose at adjacent time moments is calculated. The discrete channel quality matrix and the abrupt vehicle topology are transformed into a continuous spatiotemporal field and differential motion, avoiding state lag and improving the accuracy of motion trend perception. The edge layer receives the perception data uploaded by the terminal and combines it with the global electromagnetic field twin parameters sent by the cloud to generate local spatiotemporal field prediction results. The prediction results (field strength heat map, topology change rate) are input to the spectrum allocation engine. Step 3: Cross-layer interference prediction and quantum annealing spectrum allocation: The time-frequency spectrum generated by the IQ signal (baseband complex signal in communication, where I is the in-phase component and Q is the quadrature component) is analyzed by a 1D-CNN+LSTM dual-path network (1D-CNN: one-dimensional convolutional neural network to extract local time-frequency features, LSTM: long short-term memory network to capture long-term time dependencies). This predicts the interference intensity and frequency occupancy probability in advance to switch communication frequency bands. At the same time, the spectrum allocation is modeled as a QUBO model (quadratic unconstrained binary optimization) and the optimal allocation scheme is solved using an FPGA-accelerated quantum annealer to pre-cancel radio frequency interference and solve resource competition problems. The interference prediction results are input into the quantum annealer to generate spectrum allocation instructions. The allocation instructions are sent to the terminal through the V2X link, and the edge layer resource pool status is updated simultaneously. Step 4: Service-Driven Protocol Adaptation and Optical Field Coordination: Services are classified into five levels based on latency and reliability requirements, and communication modes are automatically switched (DSRC / 802.11p / NR-V2X). At the same time, optical pulse coded signals are emitted through the LED array of the RSU (Roadside Unit, the roadside infrastructure in the vehicle-to-everything (V2X) network deployed beside roads, intersections or highways, and interact with the On-Board Unit (OBU), other RSUs, and cloud platforms through wireless communication (such as C-V2X, DSRC). After vehicle decoding, mechanical low-latency coordination is performed to ensure that high-priority services have exclusive access to resources, low-priority services dynamically access resources, and policy conflict alarms are resolved. Step 5: Meta-cooperative distillation learning and policy optimization: The original state space is compressed into a 32-dimensional latent space using an autoencoder (AE) while retaining key dimensions with variance greater than a threshold. The regional coordinator calculates the policy entropy difference. When the difference exceeds the threshold, the cloud-based meta-teacher network is triggered to correct the edge policy through policy distillation, solving the problems of dimensional explosion and local optima in multi-agent reinforcement learning. The edge layer uploads local experiences (state-action pairs) to the regional coordinator, and the cloud distributes the distilled meta-policy to update the edge layer's execution network parameters.

[0017] In one embodiment, the positioning data is obtained by calculating the vehicle's position coordinates using TOF (Time-of-Flight) ranging from a 5G NR millimeter-wave radar and SLAM (Simultaneous Localization and Mapping) algorithms from a LiDAR. ; The Channel State Information (CSI) is estimated using 5G NR pilot signals to determine the channel matrix. Where Y is the received signal, X is the transmitted pilot, and N is noise; The neighbor node distribution is as follows: the surrounding vehicle IDs, locations, and speeds are obtained through BSM (Basic Safety Message) broadcast via V2X; Service type classification: Based on service QoS requirements (latency, reliability), the SVM classifier is used to classify services into security (latency <10ms), control (latency <50ms), and entertainment (latency >100ms).

[0018] In one embodiment, the specific steps of the dynamic watermark embedding are as follows: Spacetime stamp encoding: The timestamp T and velocity v are mapped to a binary sequence S, represented as follows: , where ⊕ is the XOR operation and Hash is SHA-256; Watermark embedding: The 'S' is inserted into a reserved field in the header of the data packet (such as the "reserved bit" in IEEE 802.11p) to form a watermarked data packet. ; SM9 Signature: Using terminal private key d Signature, as ,in It serves as a unique identifier for the terminal.

[0019] In one embodiment, the specific steps for dynamic modeling and prediction of the spatiotemporal field of the edge layer are as follows: Step 1: Construction of Electromagnetic Field Dynamic Propagation Model: Based on Maxwell's equations, an electromagnetic field propagation model is constructed. Combining the multipath effect of vehicle-to-everything (V2X) networks and the characteristics of obstacle occupancy, a spatiotemporal evolution equation for the field strength distribution is established: In free space, the propagation of electric field E and magnetic field H satisfies the wave equation: ,in, Permeability, Given the dielectric constant, in complex environments of vehicle-to-everything (V2X) networks (such as urban canyons), a path loss model is introduced to correct the field strength: ,in, For transmission power, and For antenna gain, Let λ be the wavelength, d be the distance, and L(d) be the obstacle penetration loss (e.g., 20dB attenuation due to a concrete wall). The field strength E(x,y,z,t) is considered as a spatiotemporal field, and its rate of change is determined by both the propagation model and the vehicle's motion. ,in, For vehicle speed, This is the field strength recovery coefficient (such as RSU transmit power adjustment). The interference attenuation coefficient is... This represents the number of neighboring nodes; Step 2: LSTM-CNN hybrid network for spatiotemporal field prediction: Input historical field strength distribution (spatiotemporal sequence), extract temporal features through LSTM, extract spatial features through CNN, and output a future electromagnetic field heatmap: Historical field strength data is in 4D tensor Where T=10 is the time step, H=W=50 is the spatial resolution, C=2 is the electric / magnetic field strength, and the LSTM layer extracts temporal features and outputs the hidden state. ,in It is the Sigmoid activation function. and This is the weight matrix. This is the hidden state from the previous moment. For the current input, For the bias term, the CNN layer extracts spatial features through convolutional kernels. Calculate the feature map: ,in, Convolution kernel weights, Input feature map values, output thermal field intensity Y for future Δt ; Step 3: Vehicle pose SE(3) Lie group mapping and differential motion modeling: Map the vehicle position to the SE(3) Lie group space, calculate the tangent vector of the pose at adjacent time steps, and transform the discrete pose into continuous motion: The vehicle pose g∈SE(3) is a 4×4 homogeneous transformation matrix: ,in, Let g(t) be the rotation matrix, t be the translation vector, and g(t+Δt) be the tangent vector of the poses at adjacent time points. Calculated via logarithmic mapping: The ∨ operation transforms the Lie algebra matrix into a 6-dimensional vector (3-dimensional translational velocity + 3-dimensional angular velocity), and the pose change rate is determined by the tangent vector. ,in ∈se(3) is the matrix form corresponding to the Lie algebra; Step 4: Conversion of Discrete Channel Quality Matrix and Topology to Continuous Spatiotemporal Field: The discrete CSI matrix and vehicle topology uploaded by the terminal are converted into a continuous spatiotemporal field to avoid state lag. Discrete CSI matrix (where n is the time index) is transformed into a continuous function using cubic spline interpolation. : ,in, For spline basis functions, the vehicle topology (neighbor relationships) is represented by a graph. This represents the topology, where V is the vehicle node and E(t) is the edge (communication link). The rate of change of the topology is transformed into a continuous field through graph Fourier transform: Where F is the graph Fourier transform; Step 5: Generation of Local Spatiotemporal Field Prediction Results: Combining the sensing data uploaded by the terminal (CSI, location, service type) and the global electromagnetic field twin parameters (such as RSU coverage area, spectrum usage history) distributed from the cloud, the local spatiotemporal field prediction results are generated. By fusing terminal data and cloud parameters using Kalman filtering, a thermal map of field strength Y and the rate of topological change for future Δt are generated. Input spectrum allocation engine.

[0020] In one embodiment, the specific steps for generating the temporal spectrogram using a 1D-CNN+LSTM dual-path network are as follows: IQ signal representation: The IQ signal is in complex form. , where I(t) and Q(t) are real components; Time-to-Frequency Spectrum Generation (STFT): The signal is framed using a Hamming window function w(n) (frame length N=256, frame shift M=128), and the spectrum of each frame is calculated. Where t is the time index and f is the frequency index (0 to 1). ); 1D-CNN Feature Extraction: One-dimensional convolution is performed on each frame of the temporal spectrogram using convolution kernels to extract local time-frequency features. ,in, For convolution kernel; LSTM Temporal Modeling: Input Feature Sequence Long-term dependencies are captured using the LSTM gating mechanism: Forgotten Gate: Deciding which historical information to discard. ; in, For the Sigmoid function, This is the weight matrix. This is the hidden state from the previous moment. For the current input, For the bias term of the forget gate; Input gate: Determine which new information to update: ; ; in, The output of the input gate, Here is the weight matrix of the input gate. For the bias term of the input gate, The candidate cell states are generated using the tanh function, representing new information, and their range is between -1 and 1. To generate the weight matrix for candidate cell states, Bias terms for generating candidate cell states; Cell state update: fusing historical and new information: ; in, This represents the current cell state, and ⊙ represents element-wise multiplication. This represents the cell state at the previous moment; Output gate: Generates the current hidden state: ; ; in, For the output of the output gate, This is the weight matrix of the output gate. This is the bias term for the output gate. The current hidden state; Weight mapping: mapping hidden states Mapped to a weight vector: , The weight matrix is ​​used to map the hidden states to the weight vectors. This is the bias term that maps the hidden state to the weight vector; Interference prediction output: Outputs the interference intensity in the future Δt through a fully connected layer. Frequency occupancy probability : ,in, This represents the hidden state of the last frame of the LSTM. This is the weight matrix.

[0021] In one embodiment, the specific steps for modeling spectrum allocation as a QUBO model and solving for the optimal allocation scheme using an FPGA-accelerated quantum annealer are as follows: QUBO objective function: Let the spectrum allocation variable be... ( =1 indicates that frequency band i is allocated, and the objective function is: ,in, For the interference matrix ( =1 indicates that frequency bands i and j interfere with each other. Let be the spectral efficiency of frequency band i (bps / Hz). Constraint transformation: Delay constraints ≤ : Through penalty items Add the objective function; Reliability constraints ≥ : Through penalty items Add the objective function; Quantum annealing energy function: Mapping the QUBO objective function to the quantum annealing energy function: ,in, As constraints, This is the constraint threshold; Annealing process: The energy function is calculated in parallel using FPGA, and the annealing path is dynamically adjusted. ,in The initial temperature. This is the decay time constant.

[0022] In one embodiment, in the service-driven protocol adaptation and optical field coordination, vehicle-to-everything (V2X) services are divided into 5 levels based on latency and reliability requirements, and the service priority score S is determined by latency weighting. and reliability weight Weighted summation yields: Where D is the actual time delay. R is the maximum allowable delay; R is the actual reliability. Minimum reliability requirements; business: L1 (Emergency Control): S≥0.9 L2 (Safety Control): 0.7 ≤ S < 0.9 L3 (Cooperative Control): 0.5 ≤ S < 0.7 L4 (Entertainment): 0.3 ≤ S < 0.5 L5 (Background Data): S < 0.3 Dedicated spectrum resources are reserved for L1-L2 level services; Level 3-Level 5 services dynamically access remaining resources through the contention window (CW); When resource contention is detected, the RSU transmits alternating red and blue light pulses (spectrum shortage alarm), which trigger the backoff algorithm upon being received by the vehicle. .

[0023] A communication management optimization system used in a vehicle-to-everything (V2X) communication management optimization method includes: Spatiotemporal perception layer: By integrating a multimodal sensor array (5G millimeter-wave radar, lidar and V2X camera), a spatiotemporal field theory modeling engine (dynamic environment reconstruction based on four-dimensional spatiotemporal tensor) and edge computing nodes (equipped with spatiotemporal feature extraction acceleration chip), a dynamic spatiotemporal field model of the vehicle is constructed and spatiotemporal feature parameters (road topology, obstacle movement trajectory, electromagnetic environment changes) are extracted in real time. Meta-cooperative decision layer: Based on the meta-learning framework, cooperative distillation network and digital twin engine, a joint decision model of vehicle-road-network is constructed and a dynamic resource allocation scheme is generated, including spectrum allocation (dynamic scheduling of terahertz band), power control (intelligent metasurface adjustment) and time slot allocation (dynamic frame structure based on reinforcement learning). Communication layer: Utilizing terahertz communication modules (supporting Tbps-level transmission in the 0.1-10THz frequency band), intelligent metasurface (RIS) arrays (dynamic beamforming coverage angle adjustment accuracy of 0.1°), and orbital angular momentum (OAM) multiplexing technology (spatial dimension multiplexing improves spectral efficiency by 12 times), it enables converged communication between vehicles (V2V), vehicles (V2I), and vehicles (V2N), constructs multi-path redundant transmission channels, and dynamically adjusts communication parameters to maintain stable links in mobile scenarios at 300km / h. Intelligent Application Layer: Through blockchain security protocols (privacy protection mechanism based on zero-knowledge proofs), federated learning platforms (cross-vendor model collaborative training framework) and AR augmented reality interfaces (holographic projection and spatial audio interaction), it provides vehicle networking services (cooperative adaptive cruise, intersection priority passage, platooning), supports cross-platform data value sharing (300% improvement in algorithm iteration speed) and builds human-machine interaction interfaces.

[0024] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for communication management optimization based on Internet of Vehicles, characterized in that, Includes the following steps: Step 1: Terminal Layer Multimodal Perception and Dynamic Watermark Embedding: Integrating 5G NR millimeter-wave radar, lidar and optical field communication receiving module, real-time collection of positioning data, channel status information, neighbor node distribution and service type, and embedding a three-dimensional coded spatiotemporal stamp of location-time-velocity in the header of the data packet, combined with the quantum computing-resistant SM9 national cryptographic algorithm for identity authentication, and using camera + CMOS sensor to decode the red and blue light pulse signals emitted by the roadside unit, the decoding result is sent to the edge layer MEC through CAN bus; Step 2: Dynamic modeling and prediction of spatiotemporal field at the edge layer: Based on Maxwell's equations, a dynamic propagation model of electromagnetic field is constructed. The historical field strength distribution is input into the LSTM-CNN hybrid network and the future electromagnetic field heat map is output. At the same time, the vehicle position is mapped to the SE(3) Lie group space and the tangent vector of the pose at adjacent time moments is calculated. The discrete channel quality matrix and the abrupt vehicle topology are transformed into a continuous spatiotemporal field and differential motion. Step 3: Cross-layer interference prediction and quantum annealing spectrum allocation: The time spectrum of the IQ signal is analyzed by 1D-CNN+LSTM dual-path network to predict the interference intensity and frequency occupancy probability in advance to switch communication frequency bands. At the same time, the spectrum allocation is modeled as a QUBO model and the optimal allocation scheme is solved by FPGA-accelerated quantum annealing. Step 4: Service-driven protocol adaptation and optical field coordination: Services are classified according to latency and reliability requirements, and communication modes are automatically switched. At the same time, optical pulse coded signals are emitted through the LED array of the RSU, and mechanical low-latency coordination is achieved after vehicle decoding. Step 5: Meta-co-distillation learning and policy optimization: The original state space is compressed using an autoencoder, and key dimensions with variance greater than the threshold are retained. The regional coordinator calculates the policy entropy difference. When the difference is greater than the threshold, the cloud-based meta-teacher network is triggered to correct the edge policy through policy distillation.

2. The communication management optimization method based on vehicle-to-everything (V2X) as described in claim 1, characterized in that, The positioning data is obtained by calculating the vehicle's position coordinates using TOF ranging from a 5G NR millimeter-wave radar and SLAM algorithms from a lidar system. ; The channel state information is obtained by estimating the channel matrix using 5G NR pilot signals. Where Y is the received signal, X is the transmitted pilot, and N is noise; The neighbor node distribution is as follows: the BSM broadcast by V2X obtains the IDs, locations, and speeds of surrounding vehicles. Service type classification: Based on service QoS requirements, the SVM classifier is used to classify services into security, control, and entertainment.

3. The communication management optimization method based on vehicle-to-everything (V2X) as described in claim 2, characterized in that, The specific steps of the dynamic watermark embedding are as follows: Spacetime stamp encoding: The timestamp T and velocity v are mapped to a binary sequence S, represented as follows: , where ⊕ is the XOR operation and Hash is SHA-256; Watermark embedding: Insert 'S' into a reserved field in the header of the data packet to form a watermarked data packet. ; SM9 Signature: Using terminal private key d Signature, as ,in It serves as a unique identifier for the terminal.

4. The communication management optimization method based on vehicle-to-everything (V2X) as described in claim 1, characterized in that, The specific steps for dynamic modeling and prediction of the spatiotemporal field of the edge layer are as follows: Step 1: Construction of Electromagnetic Field Dynamic Propagation Model: Based on Maxwell's equations, an electromagnetic field propagation model is constructed. Combining the multipath effect of vehicle-to-everything (V2X) networks and the characteristics of obstacle occupancy, a spatiotemporal evolution equation for the field strength distribution is established: In free space, the propagation of electric field E and magnetic field H satisfies the wave equation: ,in, Permeability, Given the dielectric constant, a path loss model is introduced to correct the field strength in the complex environment of vehicle-to-everything (V2X) networks: ,in, For transmission power, and For antenna gain, Let λ be the wavelength, d be the distance, and L(d) be the obstacle penetration loss. The field strength E(x,y,z,t) is considered as a spatiotemporal field, and its rate of change is determined by both the propagation model and the vehicle's motion. ,in, For vehicle speed, The field strength recovery coefficient, The interference attenuation coefficient is... This represents the number of neighboring nodes. Step 2: Spatiotemporal field prediction using a hybrid LSTM-CNN network: Input historical field strength distribution, extract temporal features using LSTM, extract spatial features using CNN, and output a future electromagnetic field heatmap: Historical field strength data is in 4D tensor Where T is the time step, H=W is the spatial resolution, C is the electric / magnetic field strength, and the LSTM layer extracts temporal features and outputs the hidden state. ,in It is the Sigmoid activation function. and This is the weight matrix. This is the hidden state from the previous moment. For the current input, For the bias term, the CNN layer extracts spatial features through convolutional kernels. Calculate the feature map: ,in, Convolution kernel weights, Input feature map values, output thermal field intensity Y for future Δt ; Step 3: Vehicle pose SE(3) Lie group mapping and differential motion modeling: Map the vehicle position to the SE(3) Lie group space, calculate the tangent vector of the pose at adjacent time steps, and transform the discrete pose into continuous motion: The vehicle pose g∈SE(3) is a 4×4 homogeneous transformation matrix: ,in, Let g(t) be the rotation matrix, t be the translation vector, and g(t+Δt) be the tangent vector of the poses at adjacent time points. Calculated via logarithmic mapping: The ∨ operation transforms the Lie algebra matrix into a 6-dimensional vector, and the rate of pose change is determined by the tangent vector: ,in ∈se(3) is the matrix form corresponding to the Lie algebra; Step 4: Conversion of Discrete Channel Quality Matrix and Topology to Continuous Spatiotemporal Field: The discrete CSI matrix and vehicle topology uploaded by the terminal are converted into a continuous spatiotemporal field to avoid state lag. Discrete CSI matrix Transformed into a continuous function through cubic spline interpolation. : ,in, For spline basis functions, the vehicle topology is represented by a graph. This represents a graph where V is a vehicle node and E(t) is an edge. The topological change rate is transformed into a continuous field using a graph Fourier transform. Where F is the graph Fourier transform; Step 5: Generation of Local Spatiotemporal Field Prediction Results: Combining the sensing data uploaded by the terminal and the global electromagnetic field twin parameters sent from the cloud, local spatiotemporal field prediction results are generated. By fusing terminal data and cloud parameters using Kalman filtering, a thermal map of field strength Y and the rate of topological change for future Δt are generated. Input spectrum allocation engine.

5. The communication management optimization method based on vehicle-to-everything (V2X) as described in claim 1, characterized in that, The specific steps for generating the temporal spectrogram using the 1D-CNN+LSTM dual-path network; IQ signal representation: The IQ signal is in complex form. , where I(t) and Q(t) are real components; Time-frequency spectrum generation: The signal is divided into frames using the Hamming window function w(n), and the spectrum of each frame is calculated. Where t is the time index and f is the frequency index; 1D-CNN Feature Extraction: One-dimensional convolution is performed on each frame of the temporal spectrogram using convolution kernels to extract local time-frequency features. ,in, For convolution kernel; LSTM Temporal Modeling: Input Feature Sequence Long-term dependencies are captured through LSTM gating mechanism; Interference prediction output: Outputs the interference intensity in the future Δt through a fully connected layer. Frequency occupancy probability : ,in, This represents the hidden state of the last frame of the LSTM. This is the weight matrix.

6. The communication management optimization method based on vehicle-to-everything (V2X) as described in claim 5, characterized in that, The specific steps for modeling spectrum allocation as a QUBO model and solving for the optimal allocation scheme using an FPGA-accelerated quantum annealer are as follows: QUBO objective function: Let the spectrum allocation variable be... ( =1 indicates that frequency band i is allocated, and the objective function is: ,in, For the interference matrix ( =1 indicates that frequency bands i and j interfere with each other. Let i be the spectral efficiency of frequency band i; Constraint transformation: Delay constraints ≤ : Through penalty items Add the objective function; Reliability constraints ≥ : Through penalty items Add the objective function; Quantum annealing energy function: Mapping the QUBO objective function to the quantum annealing energy function: ,in, As constraints, This is a constraint threshold; Annealing process: The energy function is calculated in parallel using FPGA, and the annealing path is dynamically adjusted. ,in The initial temperature, This is the decay time constant.

7. The communication management optimization method based on vehicle-to-everything (V2X) as described in claim 1, characterized in that, In the business-driven protocol adaptation and optical field coordination, vehicle-to-everything (V2X) services are divided into 5 levels based on latency and reliability requirements. The service priority score S is determined by latency weighting. and reliability weight Weighted summation yields: Where D is the actual time delay. R is the maximum allowable delay; R is the actual reliability. This represents the minimum reliability requirement.

8. The communication management optimization system used in the communication management optimization method based on vehicle networking according to claim 1, characterized in that, include: Spatiotemporal perception layer: By integrating a multimodal sensor array, a spatiotemporal field theory modeling engine, and edge computing nodes, it constructs a dynamic spatiotemporal field model of the vehicle and extracts spatiotemporal feature parameters in real time; Meta-collaborative decision-making layer: Based on the meta-learning framework, collaborative distillation network and digital twin engine, a joint decision-making model of vehicle-road-network is constructed and a dynamic resource allocation scheme is generated; Communication layer: Utilizing terahertz communication modules, intelligent metasurface arrays, and orbital angular momentum multiplexing technology, it enables vehicle-to-vehicle, vehicle-to-infrastructure, and vehicle-to-cloud converged communication, constructs multi-path redundant transmission channels, and dynamically adjusts communication parameters; Intelligent Application Layer: Through blockchain security protocols, federated learning platforms, and AR augmented reality interfaces, it provides vehicle-to-everything (V2X) services, supports cross-platform data value sharing, and builds human-machine interaction interfaces.

Images