A space-time decoupling decision method for star-ground fusion network multi-vehicle access

Abstract

The present application relates to the technical field of Internet of Vehicles automatic driving, in particular to a kind of space-time decoupling decision method for star-ground fusion network multi-vehicle access, comprising: step S1: in the star-ground fusion network of subjecting to Poisson point distribution ground road side unit and low-orbit satellite, combined with the execution environment evaluation of multiple networked automatic driving vehicles, establish multi-vehicle cooperative access problem;Step S2: through multi-vehicle access problem, construct distributed decision model based on hierarchical decoupling, including Lyapunov optimization module, mean field game approximation module and adaptive context Thompson sampling module, execute space-time decoupling decision;Step S3: the effectiveness of distributed decision model based on hierarchical decoupling is verified by simulation experiment.By Lyapunov optimization module and mean field game approximation module, time-space coupling is eliminated, scalability and stability under long-term QoS constraints are guaranteed;And cooperate with adaptive context Thompson sampling module to deal with multi-agent access problem in heterogeneous environment.

Description

Technical Field

[0001] This invention relates to the field of vehicle-to-everything (V2X) autonomous driving technology, specifically to a spatiotemporal decoupling decision-making method for multi-vehicle access in a space-ground fusion network. Background Technology

[0002] Modern connected autonomous vehicles increasingly operate as cooperative agents, including platooning, collaborative perception, and real-time map fusion. These capabilities require continuous, synchronized uplink transmissions from multiple vehicles whose data streams are temporally and semantically interdependent. To sustain this collective transmission across densely populated urban corridors and sparsely covered highways, it is necessary to integrate Low Earth Orbit (LEO) satellite constellations with ground-based roadside units into Heterogeneous Satellite-Terrestrial Integrated Networks (HSTINs). This network combines wide-area satellite coverage with low-latency terrestrial access to provide services to vehicles throughout their operating range.

[0003] However, this heterogeneous underlying layer fundamentally changes the access management problem. Multi-vehicle access in HSTINs is no longer a single-vehicle channel selection task, but a multi-agent online optimization problem constrained by two structural challenges. First, spatiotemporal decision coupling stems from vehicles' competition for shared radio and queuing resources, thus individual access decisions collectively determine congestion and bandwidth allocation and depend on the joint behavior of other vehicles. Simultaneously, cooperative applications impose long-term Quality of Service (QoS) requirements, including asynchronous latency fairness across vehicles and handover stability. These requirements couple decisions to different time slots, thus hindering stepwise greedy optimization. Second, the environment is highly non-stationary. Heterogeneous HSTINs channels evolve under dual mobility dynamics, where the movement of ground vehicles leads to the gradual fading of vehicle-to-RSU (V2RSU) links, while the movement of LEO satellites causes abrupt changes and statistical variations in the topology of vehicle-to-satellite (V2Sat) links, making fixed-model access strategies unreliable over time.

[0004] While existing research has made some progress in capturing multi-agent interactions or handling long-term QoS constraints, it remains difficult to achieve scalable distributed coordination with long-term performance guarantees under conditions of shared heterogeneous resources. On the one hand, multi-agent reinforcement learning-based methods often face the curse of dimensionality and convergence difficulties caused by channel non-stationarity as the number of vehicles increases. On the other hand, existing learning algorithms are mostly based on a single time scale, making it difficult to simultaneously capture the gradual changes in V2RSU links caused by vehicle movement and the abrupt changes in V2Sat link topology caused by LEO satellite orbital motion. Summary of the Invention

[0005] To address the technical challenges of existing technologies in simultaneously resolving non-stationarity and enforcing long-term constraints in HSTINs scenarios, this invention aims to provide a spatiotemporal decoupling decision-making method for multi-vehicle access in space-ground converged networks. The specific technical solution adopted is as follows:

[0006] Step S1: In the space-ground fusion network composed of ground roadside units that follow the Poisson point distribution and low-orbit satellites, combine multiple networked autonomous vehicles to perform environmental assessments and establish a multi-vehicle collaborative access problem;

[0007] Step S2: Construct a distributed decision model based on hierarchical decoupling for the multi-vehicle access problem, including a Lyapunov optimization module, a mean-field game approximation module, and an adaptive context Thompson sampling module, and perform spatiotemporal decoupling decision-making.

[0008] Step S3: Verify the effectiveness of the distributed decision-making model based on hierarchical decoupling using simulation experiments.

[0009] Preferably, step S1 includes:

[0010] Step S11: Define the ground roadside units and low-orbit satellites in the space-ground fusion network as candidate access nodes, and the networked autonomous driving vehicles as cooperative intelligent agents. Combine the candidate access nodes and cooperative intelligent agents to obtain V2RSU links and V2Sat links respectively.

[0011] Step S12: Define the shared unified signal-to-noise ratio model for V2RSU and V2Sat links, perform environmental assessment, define the instantaneous utility of connected autonomous vehicles, and establish the multi-vehicle collaborative access problem.

[0012] Preferably, step S12 includes:

[0013] Step S121: Obtain the transmit power and thermal noise density from the candidate access nodes and the cooperative agent respectively, and establish a signal-to-noise ratio model by combining the allocated bandwidth and the time-varying channel coefficients of each scheduling cycle;

[0014] Step S122: Solve the signal-to-noise ratio model to obtain the uplink capacity and instantaneous queuing delay for the corresponding scheduling period, and determine the end-to-end delay;

[0015] Step S123: Evaluate the switching cost of connected autonomous vehicles between consecutive scheduling cycles, define instantaneous utility by combining uplink capacity and end-to-end latency, and establish a multi-vehicle collaborative access problem.

[0016] Preferably, step S2 includes:

[0017] Step S21: Based on the multi-vehicle access problem, with the optimization objective of maximizing long-term average utility and satisfying time delay and stability constraints, construct a distributed decision model based on hierarchical decoupling;

[0018] Step S22: Decouple the Lyapunov optimization module and the mean-field game approximation module in time and space, establish the context vector, and output the exploratory sampling results;

[0019] Step S23: Utilize the adaptive context Thompson sampling module to comprehensively analyze the context vector and exploratory sampling results, select the optimal access node, and adjust the model parameters of the distributed decision-making model based on hierarchical decoupling in real time based on feedback.

[0020] Preferably, step S22 includes:

[0021] Step S221: The long-term constraints corresponding to the multi-vehicle access problem are mapped to real-time virtual queue pressure through the Lyapunov optimization module. The virtual queue backlog value and the corresponding constraint cost function are extracted as exploratory sampling results to eliminate time coupling.

[0022] Step S222: Define the mean-field state characteristics through the mean-field game approximation module, simplify the resource competition of cooperative agents into a response of group statistical law, and eliminate spatial coupling;

[0023] Step S223: After spatiotemporal decoupling, the candidate access nodes and cooperative agents capture vehicle features and node features respectively, and establish context vectors.

[0024] Preferably, in step S221, the long-term constraints include asynchronous delay fairness constraints and switching frequency stability constraints; the virtual queues include virtual queues that track the degree of delay deviation from the mean and virtual queues that track the degree of switching frequency violations.

[0025] Preferably, step S222 specifically includes:

[0026] For each candidate access node, the average number of associated vehicles, average received signal-to-noise ratio, and congestion index are extracted to define the mean field state. The candidate access node then broadcasts this state to the cooperative agent, simplifying the resource competition among the cooperative agents into a response based on group statistical laws and eliminating spatial coupling.

[0027] Preferably, step S23 includes:

[0028] Step S231: The adaptive context Thompson sampling module uses a context Thompson sampling algorithm based on Bayesian linear regression, presets a reward parameter vector, and combines the context vectors to construct the expected reward model of the candidate access node;

[0029] Step S232: Establish a decision scoring function in the online decision-making process by combining the expected reward model and the exploratory sampling results, and select the optimal access node;

[0030] Step S233: Introduce a dual-timescale update mechanism to adjust the model parameters of the distributed decision-making model based on hierarchical decoupling in real time according to feedback, in order to cope with non-stationary changes in the environment.

[0031] Preferably, step S233 specifically includes:

[0032] An exponential forgetting mechanism is adopted to adapt to the gradual decay of V2RSU links, and a periodic soft reset mechanism is used to adapt to the topology changes of V2Sat links. The model parameters of the distributed decision-making model based on hierarchical decoupling are corrected in real time according to feedback to cope with non-stationary changes in the environment.

[0033] To address the problems of the prior art, the present invention also provides: an electronic device, the device comprising: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus, and the processor calls logical instructions in the memory to execute the spatiotemporal decoupling decision-making method for multi-vehicle access in a space-ground fusion network as described in any of the preceding claims.

[0034] The present invention has the following beneficial effects:

[0035] 1. To address the distributed multi-vehicle cooperative access problem in existing space-ground converged networks, a spatiotemporal decoupling decision-making method for multi-vehicle access in space-ground converged networks is proposed. In other words, a mean-field decoupling architecture based on Lyapunov guidance is proposed, which eliminates temporal coupling and spatial coupling respectively through Lyapunov optimization module and mean-field game approximation module, effectively solving the complex spatiotemporal access decision coupling problem and ensuring the scalability and stability of the system under long-term QoS constraints. For the non-stationary dynamic environment caused by the dual mobility of LEO satellites and vehicles, an adaptive context Thompson sampling module is adopted to effectively deal with the multi-agent access problem in heterogeneous environments.

[0036] 2. The electronic device provided by this invention has the same beneficial effects as the spatiotemporal decoupling decision-making method for multi-vehicle access in a space-ground fusion network provided by this invention, and will not be described in detail here. Attached Figure Description

[0037] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 This is a diagram illustrating a spatiotemporal decoupling decision-making method for multi-vehicle access in a space-ground converged network, as provided in an embodiment of the present invention.

[0039] Figure 2 The average reward value of a spatiotemporal decoupling decision-making method for multi-vehicle access in a space-ground converged network provided in one embodiment of the present invention and a comparison algorithm under scheduling time;

[0040] Figure 3 This is a schematic diagram illustrating the dynamic evolution of a delay violation virtual queue, which is a spatiotemporal decoupling decision-making method and comparison algorithm for multi-vehicle access in a space-ground converged network, provided in one embodiment of the present invention.

[0041] Figure 4 This is a schematic diagram illustrating the dynamic evolution of the virtual queue for switching penalties in a spatiotemporal decoupling decision-making method and comparison algorithm for multi-vehicle access in a space-ground converged network, provided in one embodiment of the present invention.

[0042] Figure 5 This is a schematic diagram illustrating the trade-off between rewards and queue backlog under different Lyapunov control parameters for a spatiotemporal decoupling decision-making method for multi-vehicle access in a space-ground converged network, provided in an embodiment of the present invention.

[0043] Figure 6 This is a schematic diagram comparing the end-to-end cumulative delay distribution function of a spatiotemporal decoupling decision-making method for multi-vehicle access in a space-ground converged network, as provided in an embodiment of the present invention, with a comparative algorithm. Detailed Implementation

[0044] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a spatiotemporal decoupling decision-making method for multi-vehicle access in a space-ground converged network proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0045] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0046] The following description, in conjunction with the accompanying drawings, details a specific scheme for a spatiotemporal decoupling decision-making method for multi-vehicle access in a space-ground converged network provided by the present invention.

[0047] Please see Figure 1 The first embodiment of this invention proposes a spatiotemporal decoupling decision-making method for multi-vehicle access in a space-ground converged network, the method comprising:

[0048] Step S1: In the space-ground fusion network composed of ground roadside units that follow the Poisson point distribution and low-orbit satellites, combine multiple networked autonomous vehicles to perform environmental assessments and establish a multi-vehicle collaborative access problem;

[0049] Step S2: Construct a distributed decision model based on hierarchical decoupling for the multi-vehicle access problem, including a Lyapunov optimization module, a mean-field game approximation module, and an adaptive context Thompson sampling module, and perform spatiotemporal decoupling decision-making.

[0050] Step S3: Verify the effectiveness of the distributed decision-making model based on hierarchical decoupling using simulation experiments.

[0051] This paper explains that in a heterogeneous environment of space-ground convergence constructed by roadside units and low-orbit satellites, a spatiotemporal decoupling decision-making method for multi-vehicle access in a space-ground converged network is proposed. This method addresses the severe environmental non-stationarity caused by the dual mobility of multiple connected autonomous vehicles and satellites, as well as the congestion and spatiotemporal decision-making challenges resulting from multiple vehicles competing for and sharing limited resources under long-term QoS constraints. The method transforms the problem of the entire environmental feedback into a multi-vehicle collaborative access problem and constructs a distributed decision-making model based on hierarchical decoupling. Specifically, it constructs a distributed online learning and decision-making system with the goal of maximizing the long-term average utility of the system and strictly satisfying the constraints of asynchronous delay and switching stability. The reliability of the system is verified through simulation experiments.

[0052] Further, step S1 includes:

[0053] Step S11: Define the ground roadside units and low-orbit satellites in the space-ground fusion network as candidate access nodes, and the networked autonomous driving vehicles as cooperative intelligent agents. Then, connect the candidate access nodes and cooperative intelligent agents to obtain V2RSU links and V2Sat links respectively.

[0054] Specifically, in the space-ground integrated network, it includes For each connected autonomous vehicle, the ground roadside units are counted as follows: One, each scheduling cycle Low-Earth orbit satellites, that is, the dynamic subset of LEO satellites is The scheduling cycle is the time interval in the space-ground fusion network used for discretizing dynamic satellite coverage and batch matching of roadside units with low-Earth orbit satellite resources; that is, discretizing the time range of analysis into scheduling cycles of equal length. The duration of each scheduling cycle is To balance channel coherence and decision response delay; ground roadside units and low-Earth orbit satellites in the space-ground fusion network are defined as candidate access nodes, thereby constructing a candidate access node set, i.e. In actual operation, following the single-connection paradigm, each networked autonomous vehicle, i.e. each collaborative intelligent agent, associates with at most one candidate access node in each scheduling cycle.

[0055] In high-speed vehicle mobility and complex propagation environments, vehicle-to-everything (V2X) channels often exhibit significant unique characteristics. When satellites are introduced to participate in communication, the channels change rapidly, resulting in high Doppler shift and multipath propagation effects. V2RSU and V2Sat links exhibit distinctly different propagation characteristics. V2RSU links operate in the 2.4 GHz band and are primarily non-line-of-sight (NLoS) urban propagation, where vehicle movement leads to gradual shadow fading. V2Sat links operate in the 20 GHz (Ka band) band and are primarily line-of-sight (LoS) propagation, making them susceptible to sudden statistical mutations caused by low Earth orbit (LEO) orbital dynamics and satellite switching.

[0056] Step S12: Define the shared unified signal-to-noise ratio model for V2RSU and V2Sat links, perform environmental assessment, define the instantaneous utility of connected autonomous vehicles, and establish the multi-vehicle collaborative access problem.

[0057] It is explained that during each scheduling cycle, based on the space-ground fusion network, an environmental assessment is performed, and each connected autonomous vehicle obtains its local buffer status and receives macroscopic statistical information broadcast from candidate access nodes.

[0058] Further, step S12 includes:

[0059] Step S121: Obtain the transmit power and thermal noise density from the candidate access nodes and the cooperative agent respectively, and establish a signal-to-noise ratio model by combining the allocated bandwidth and the time-varying channel coefficients of each scheduling cycle.

[0060] Specifically, during execution, it is assumed that the V2RSU link and the V2Sat link share a unified signal-to-noise ratio model, and the corresponding calculation formula is as follows:

[0061]

[0062] in, Indicates during the scheduling period , No. Connected autonomous vehicles and the first The signal-to-noise ratio of each candidate access node; Indicates the transmission power; Indicates during the scheduling period , No. Connected autonomous vehicles and the first Time-varying channel coefficients of candidate access nodes; This represents the absolute value operation; Indicates thermal noise density; Indicates the first The bandwidth allocated to each candidate access node.

[0063] It can be explained that the transmission power Radio frequency power output by the transmitter of a connected autonomous vehicle; time-varying channel coefficients The thermal noise density is accurately obtained by the connected autonomous vehicle using conventional channel estimation in each scheduling cycle. This represents the noise power spectral density of the candidate access node receiver, determined by thermal noise.

[0064] Step S122: Solve the signal-to-noise ratio model to obtain the uplink capacity and instantaneous queuing delay for the corresponding scheduling period, and determine the end-to-end delay.

[0065] Specifically, the signal-to-noise ratio (SNR) model is calculated, and the differences between heterogeneous satellite-to-ground channels are evaluated using relevant parameters in the SNR model to obtain the uplink capacity. , Indicates during the scheduling period , No. Connected autonomous vehicles and the first Uplink capacity of each candidate access node; instantaneous queuing delay , Indicates during the scheduling period , No. Queuing delay of each candidate access node; Indicates during the scheduling period , No. Instantaneous queue backlog of candidate access nodes; Indicates the first The service rate of each candidate access node.

[0066] The end-to-end delay is determined by the following formula:

[0067]

[0068] in, Indicates during the scheduling period , No. The latency when connected autonomous vehicles associate with candidate access nodes; Indicates during the scheduling period , No. The amount of buffered data for connected autonomous vehicles; Indicates during the scheduling period , No. Connected autonomous vehicles and the first The maximum amount of data that can be transmitted on the access link corresponding to each candidate access node. [KB / period].

[0069] It can be explained that during the scheduling cycle The actual data throughput within the buffer is limited by the available buffered data and transmission capacity, i.e. , Indicates during the scheduling period , No. The actual data throughput of connected autonomous vehicles.

[0070] Step S123: Evaluate the switching cost of connected autonomous vehicles between consecutive scheduling cycles, define instantaneous utility by combining uplink capacity and end-to-end latency, and establish a multi-vehicle collaborative access problem.

[0071] Specifically, during the operation of the entire space-ground integrated network, in terms of queue evolution, the data arrival process on the vehicle side is a Poisson point process. Modeling, This represents the data generation requirements for collaborative perception in L3-L4 level autonomous driving. It should be noted that L3 refers to conditionally automated vehicles, and L4 refers to highly automated vehicles. Data from L3-L4 level connected autonomous vehicles is used for testing here, ensuring backward compatibility. Simultaneously, the node queue backlog summarizes the uplink traffic of all associated connected autonomous vehicles; the queue evolution of connected autonomous vehicles and candidate access nodes can be jointly represented as:

[0072]

[0073]

[0074] in, , , , These represent the scheduling period. and , No. Connected autonomous vehicles and the first The instantaneous queue of candidate access nodes is backlogged. Indicates during the scheduling period , No. The data arrival process of connected autonomous vehicles; Indicates during the scheduling period , No. The actual data throughput of connected autonomous vehicles; Indicates during the scheduling period Vehicle data that successfully enters the candidate access node queue, i.e., uplink traffic; Indicates the first Service rate of each candidate access node; Indicates the scheduling period The duration.

[0075] It can be noted that, due to limited service capacity, It may be strictly less than the amount of data transmitted, i.e., the actual data throughput. The remaining traffic will be buffered at the vehicle end; conversely, if the total incoming traffic exceeds the capacity of the candidate access nodes, the resulting backlog will exacerbate queuing delays. This delay actually acts as an implicit congestion indicator, affecting the access decision of subsequent candidate access nodes. Furthermore, instantaneous utility is defined, and the corresponding calculation formula is:

[0076]

[0077] in, Indicates during the scheduling period , No. The instantaneous utility of connected autonomous vehicles; Indicates during the scheduling period , No. Uplink capacity of connected autonomous vehicles; Indicates during the scheduling period , No. The latency when connected autonomous vehicles associate with candidate access nodes; Indicates an indicator function; , These represent the scheduling period. and , No. Candidate access nodes selected for connected autonomous vehicles; , , All represent weighting coefficients, and all .

[0078] It can be explained that the indicator function This represents the handover cost if handover feedback occurs between consecutive scheduling cycles; that is... A value of 1 indicates a switch has occurred; otherwise, a value of 1 indicates a switch has occurred. This indicates a switch, with a value of 0; subsequently, the weighting coefficient is used... , , Balancing uplink capacity, end-to-end latency, and handover costs.

[0079] The calculation formula for establishing multi-vehicle collaborative access is as follows:

[0080]

[0081]

[0082] in, Indicates during the scheduling period The global system state; Indicates during the scheduling period , No. Local observation information of connected autonomous vehicles; , Both indicate long-term constraints.

[0083] It can be noted that the global system state includes all candidate access nodes and connected autonomous vehicles during the scheduling cycle. Complete information; long-term constraints By limiting the relative latency deviation of each connected autonomous vehicle relative to the global system average to a threshold Internally, it ensures the fairness of asynchronous delays; long-term constraints. Limit the long-term average switching frequency to a threshold. The following measures aim to reduce signaling overhead.

[0084] Understandably, spatial coupling leads to a joint action space. Exponential growth and long-term constraints and The introduced cross-time slot feasibility dependency makes the establishment of multi-vehicle cooperative access problem difficult to handle. Therefore, a hierarchical decoupling framework is proposed. In order to overcome the scalability bottleneck and convergence problem in multi-agent stochastic optimization, a distributed decision model based on hierarchical decoupling is constructed to enter the decision-making stage.

[0085] Further, step S2 includes:

[0086] Step S21: Based on the multi-vehicle access problem, with the optimization objective of maximizing long-term average utility and satisfying time delay and stability constraints, construct a distributed decision model based on hierarchical decoupling.

[0087] Specifically, the distributed decision-making model based on hierarchical decoupling includes three core modules: a Lyapunov optimization module, a mean-field game approximation module, and an adaptive context Thompson sampling module. Dimensionality reduction and decoupling are achieved through the collaborative operation of these three modules. During the operation of the model, the Lyapunov optimization module and the mean-field game approximation module perform temporal decoupling and spatial decoupling, respectively. After the two modules complete the spatiotemporal decoupling, the adaptive context Thompson sampling module is used to select the optimal candidate access node to correct the model parameters.

[0088] Step S22: Decouple the Lyapunov optimization module and the mean-field game approximation module in time and space, establish the context vector, and output the exploratory sampling results.

[0089] The Lyapunov optimization module eliminates long-term constraint coupling across time slots, absorbing complex long-term cross-time slot service quality constraints and transforming them into real-time virtual queue stability objectives. The mean-field game approximation module is responsible for decoupling the spatial dimension, using group-level macroscopic statistical information such as average load and average signal-to-noise ratio to replace explicit interactions between high-dimensional cooperative agents, reducing coordination overhead.

[0090] Further, step S22 includes:

[0091] Step S221: The long-term constraints corresponding to the multi-vehicle access problem are mapped to real-time virtual queue pressure through the Lyapunov optimization module. The virtual queue backlog value and the corresponding constraint cost function are extracted as exploratory sampling results to eliminate time coupling.

[0092] Provide an explanation and establish long-term constraints. and By linking the decision-making feasibility of time intervals corresponding to multiple scheduling cycles, strict time coupling is introduced. In order to decouple this cross-time-period dependency, the Lyapunov Drift Plus Penalty (DPP) optimization framework is used based on the Lyapunov optimization module to transform the long-term constraint into an easy-to-handle queue stability problem.

[0093] Furthermore, in step S221, the long-term constraints include asynchronous delay fairness constraints and switching frequency stability constraints; the virtual queues include virtual queues that track the degree of delay deviation from the mean and virtual queues that track the degree of switching frequency violations.

[0094] It can be noted that during the operation of the Lyapunov optimization module, two dedicated virtual queues are maintained for each connected autonomous vehicle to track historical constraint violations of long-term constraints; among them, the asynchronous delay fairness constraint is a long-term constraint. The switching frequency stability constraint is a long-term constraint. .

[0095] Specifically, first, determine the virtual queue whose tracking delay deviates from the mean, denoted as This indicates asynchronous delay fairness constraints. For virtual queues that violate delay fairness rules, their update dynamics are represented as follows:

[0096]

[0097] in, Indicates the instantaneous average system delay. ; Indicates the threshold. (Through...) The calculation ensures that only values ​​exceeding a predefined threshold are processed. Only the relative deviation will affect the backlog of the virtual queue.

[0098] Then, determine the virtual queue for tracking the degree of switching frequency violations, denoted as... This indicates a switching frequency stability constraint. The evolution of the virtual queue that tracks the switching frequency deviation is as follows:

[0099]

[0100] in, The increment occurs as connected autonomous vehicles switch to candidate access nodes, at a permissible long-term rate, i.e., the threshold. Decreasing.

[0101] Next, by aggregating the two virtual queues into the network state vector, i.e. Define Lyapunov ,Right now The objective function for minimizing the DPP (Drift-Plus-Penalty) at each step is:

[0102]

[0103]

[0104] in, This indicates Lyapunov drift; Indicates control parameters, and .

[0105] It can be explained that the control parameters This is used to strictly control system stability, i.e., the trade-off between queue boundaries and utility maximization; this transformation converts the time-coupled optimization problem into a series of deterministic time-slot problems, requiring no prior knowledge of future states. Furthermore, the well-known... Theoretical performance trade-offs This represents the complexity function; that is, the trade-off between the expected virtual queue backlog boundary and the utility optimality gap.

[0106] From a physical perspective, the long-term constraints of the multi-vehicle access problem are mapped to real-time virtual queue pressure. The virtual queue acts as a dynamic pricing mechanism, regulating the behavior between candidate access nodes and collaborating agents. When connected autonomous vehicles violate their asynchronous delay fairness or handover stability constraints, the corresponding virtual queue rapidly accumulates and backs up. In the constructed DPP objective function, this backlog is equivalent to a high penalty weight, naturally forcing the decision-making process to prioritize constraint satisfaction, such as queue stability, rather than the instantaneous utility optimization of subsequent cycles. In the objective function, control parameters... A larger value leads to a greedy pursuit of short-term performance, but may result in larger queue boundaries; while control parameters... Smaller sizes ensure strict enforcement of constraints.

[0107] Step S222: Define the mean-field state characteristics through the mean-field game approximation module, simplify the resource competition of the cooperative agents into a response of group statistical laws, and eliminate spatial coupling.

[0108] It should be noted that the analysis of the multi-vehicle access problem exhibits spatial coupling because the instantaneous utility of each connected autonomous vehicle depends on the behavior of its companions generated through shared bandwidth and queuing resources. This leads to a joint action space, which, within the corresponding scheduling period, is represented by the global information formed by the entire pool of candidate access nodes and cooperating agents. The scale is such that, therefore, the mean-field game approximation module is used, which takes advantage of the fact that in large-scale groups, individual utility depends mainly on the behavior of the group as a whole, rather than the behavior of specific peers.

[0109] Further, in step S222, specifically:

[0110] For each candidate access node, the average number of associated vehicles, average received signal-to-noise ratio, and congestion index are extracted to define the mean field state. The candidate access node then broadcasts this state to the cooperative agent, simplifying the resource competition among the cooperative agents into a response based on group statistical laws and eliminating spatial coupling.

[0111] Specifically, for each candidate access node During the scheduling cycle The average number of associated vehicles, average received signal-to-noise ratio, and congestion index are extracted respectively, and denoted as follows: , and To provide feedback on candidate access nodes The mean field state, i.e. Among them, the three state parameters—average number, average received signal-to-noise ratio, and congestion index—correspond to the average value of the networked autonomous vehicles accessing the candidate access node, and all state parameters are determined by a forgetting factor. The exponential moving average is updated.

[0112] Based on mean field state Each connected autonomous vehicle autonomously assesses its resource competitiveness and compares it with the macro-group state, rather than exchanging explicit transmission intentions with its peers; by utilizing these group-level statistical data, it predicts its potential bandwidth allocation, a complex... The multi-agent random game is cleverly simplified into a single-agent interaction with a deterministic mean field; this effectively eliminates the high cost of... Signaling overhead and severe scalability bottlenecks reduce the required coordination overhead to a level comparable to irrelevant strict , This represents the total number of candidate access nodes reported by the candidate access node set. Theoretically, the resulting average-field game equilibrium strategy constitutes... Vehicle game -Nash equilibrium, .

[0113] Step S223: After spatiotemporal decoupling, the candidate access nodes and cooperative agents capture vehicle features and node features respectively, and establish context vectors.

[0114] Specifically, after spatiotemporal decoupling, a context vector is constructed between the connected autonomous vehicle and the candidate access node, denoted as... It connects three features: vehicle features, node features, and the aforementioned mean field state. Vehicle features are obtained based on the vehicle's local task requirements and buffer status, including the vehicle's service priority, data arrival rate, and local queue health indicators after non-linear transformation, used to quantify the urgency of local data transmission. Node features are obtained based on physical layer channel estimation and node feedback, including node type identification (RSU, Road Side Unit) or satellite, service rate, instantaneous achievable capacity, node queue availability, normalized signal-to-noise ratio, and discrete indicators reflecting channel trends with mobility, comprehensively characterizing the node's current service capability and link quality. The mean field state... The group statistics information is extracted from the periodic broadcasts of candidate access nodes during the aforementioned spatial decoupling stage. This includes the average number of associated vehicles, average SNR, and normalized congestion index data broadcast by the candidate access nodes. This enables each collaborative agent to perceive the macro-resource competition situation in order to avoid congestion.

[0115] It can be explained that the structurally difficult problem is broken down by the synergistic integration of the Lyapunov optimization module and the mean-field game approximation module; that is, the temporal coupling introduced by long-term QoS constraints is absorbed by the dynamics of the virtual queue, while the spatial coupling caused by resource competition is alleviated by statistical perception at the group level; by implementing hierarchical spatiotemporal decoupling through these two modules, the stochastic optimization of multi-cooperative agents is seamlessly transformed into a series of scalable and independent context optimization problems.

[0116] Step S23: Utilize the adaptive context Thompson sampling module to comprehensively analyze the context vector and exploratory sampling results, select the optimal access node, and adjust the model parameters of the distributed decision-making model based on hierarchical decoupling in real time based on feedback.

[0117] Understandably, in the face of unknown and dynamically changing satellite-ground integrated wireless communication environments, traditional deterministic or rule-based access strategies often fall into suboptimal due to a lack of exploration of the environment. Therefore, an adaptive context Thompson sampling module is adopted to model the access decision as a multi-armed gambling machine problem, that is, each step of the long subproblem requires online node selection under random and non-stationary channel conditions. Specifically, after completing the spatiotemporal decoupling through the aforementioned two modules, the adaptive context Thompson sampling module is used to make uncertainty-aware decisions using Bayesian linear regression, and an exponential forgetting factor and a periodic soft reset mechanism are introduced to simultaneously adapt to the gradual fading of V2RSU links and the topological changes of V2Sat links.

[0118] Further, step S23 includes:

[0119] Step S231: The adaptive context Thompson sampling module uses a context Thompson sampling algorithm based on Bayesian linear regression, presets a reward parameter vector, and combines the context vectors to construct the expected reward model of the candidate access node.

[0120] Specifically, a contextual Thompson sampling algorithm based on Bayesian linear regression is adopted, which achieves an effective balance between utilizing existing high-yield candidate access nodes and exploring potentially better candidate access nodes. That is, it constructs an expected reward model for candidate access nodes, and the corresponding calculation formula is as follows:

[0121]

[0122] in, Indicates during the scheduling period , No. Connected autonomous vehicles in the selection of the first The instantaneous utility calculated after each candidate access node, i.e., expected reward modeling; This represents the preset reward parameter vector. ; This represents the vector transpose operation.

[0123] Step S232: Establish a decision scoring function in the online decision-making process by combining the expected reward model and the exploratory sampling results, and select the optimal access node.

[0124] It can be noted that the exploratory sampling results refer to the network state vectors formed by aggregating from the virtual queues. The obtained virtual queue backlog value and the corresponding constraint cost function are respectively as follows: and , , represents the virtual queue index.

[0125] Specifically, in the actual analysis process, it is assumed that the prior distribution is a Gaussian distribution. , This represents a preset reward parameter vector; Indicates a Gaussian distribution; The precision parameter representing the prior Gaussian distribution; The matrix represents the identity matrix; the observation noise follows a Gaussian distribution. Then the posterior distribution is still a Gaussian distribution. The sufficient statistics are updated recursively with low complexity using the Sherman-Morrison rank-one method, and the corresponding calculation formula is as follows:

[0126]

[0127]

[0128]

[0129] in, Let represent the posterior covariance matrix, reflecting the variance of the first digit. Uncertainty in estimating the rewards for each candidate access node; This represents the cumulative reward weighted context vector, recording the overall situation of historical returns; Indicates during the scheduling period The posterior mean of the reward parameter vector obtained from historical data, i.e., for the th... The optimal reward estimate for each candidate access node.

[0130] The complexity of each update is O(n). , Represents the context vector The dimension; in each scheduling cycle, a parameter sample is drawn from the posterior distribution for each candidate access node: , Indicates the exploration coefficient. This is used to control the exploration scale; the decision scoring function in the online decision-making process is determined, that is, the access decision combines the sampled reward estimate with the Lyapunov constraint pressure to perform a comprehensive score, in order to select the optimal access node. The corresponding calculation formula is:

[0131]

[0132]

[0133] in, Indicates the first The decision score corresponding to each candidate access node; Indicates during the scheduling period , for the The parameter samples of each candidate access node drawn from the posterior distribution; Indicates during the scheduling period , No. Connected autonomous vehicles select the first After the first candidate access node, the second... One virtual queue backlog value; This represents the corresponding constraint cost function; Indicates during the scheduling period , No. The optimal access node is selected from the network of autonomous vehicles.

[0134] Step S233: Introduce a dual-timescale update mechanism to adjust the model parameters of the distributed decision-making model based on hierarchical decoupling in real time according to feedback, in order to cope with non-stationary changes in the environment.

[0135] The explanation is that a dual-time-scale update mechanism is introduced to handle the dual mobility non-stationarity of HSTINs, and separate processing is applied to V2RSU links and V2Sat links.

[0136] Furthermore, in step S233, specifically:

[0137] An exponential forgetting mechanism is adopted to adapt to the gradual decay of V2RSU links, and a periodic soft reset mechanism is used to adapt to the topology changes of V2Sat links. The model parameters of the distributed decision-making model based on hierarchical decoupling are corrected in real time according to feedback to cope with non-stationary changes in the environment.

[0138] Specifically, to address the asymptotic attenuation of V2RSU links caused by ground vehicle movement, an exponential forgetting mechanism is applied to the Bayesian sufficient statistics, and the corresponding calculation formula is as follows:

[0139]

[0140]

[0141] in, Represents the exponential forgetting factor. This produced a length of approximately The effective memory window of the cycle.

[0142] For abrupt changes in the topology of a V2Sat link, such as the replacement of visible LEO satellites, this is achieved by... This is resolved by performing a soft reset once every cycle, that is, by using a periodic soft reset mechanism to adapt to topology changes in the V2Sat link. The corresponding calculation formula is:

[0143]

[0144]

[0145] in, Indicates the soft reset intensity coefficient. This term describes the extent to which one regresses to a priori states, which can partially restore the uncertainty of exploration while preserving accumulated structural knowledge.

[0146] It can be shown that the dual-timescale update mechanism based on dual links perfectly matches the physical characteristics of the HSTINs environment; among them, the exponential forgetting mechanism can smoothly and continuously track V2RSU channel fading caused by ground vehicle movement; conversely, the periodic soft reset provides a key correction mechanism to prevent the algorithm from using outdated posterior estimates after the sudden appearance or switching of low Earth orbit satellites causes abrupt changes in the V2Sat topology. During operation, under conditions of... In a piecewise stationary environment with a number of change points, the dynamic regret bound of the adaptive context Thompson sampling module is: , This represents the total time step of the contextual Thompson sampling algorithm based on Bayesian linear regression; the computational complexity of each step is O(n log n). For real-time vehicle deployment, it still maintains extremely high efficiency.

[0147] Understandably, the non-stationary environment processing method based on contextual Thompson sampling is complemented by the spatiotemporal decoupling access strategy. This involves the coordinated operation of the Lyapunov optimization module, the mean-field game approximation module, and the adaptive contextual Thompson sampling module. The Lyapunov optimization module first addresses the long-term QoS constraints across time slots, transforming the difficult-to-handle latency and handover limitations into real-time virtual queue pressure, thus achieving decoupling in the temporal dimension. The mean-field game approximation module replaces the high-dimensional explicit communication between vehicles, reducing coordination overhead by utilizing statistical characteristics at the group competition level, achieving decoupling in the spatial dimension. Subsequently, the adaptive contextual Thompson sampling module, based on dual-timescale contextual Thompson sampling, relies on the simplified single-agent framework formed after the above decoupling. It utilizes an uncertainty-aware exploration mechanism to perfectly fill the gap in online adaptation under non-stationary channel mutations. These three modules complement each other, jointly forming a closed-loop decision system that combines theoretical long-term stability, large-scale scalability, and dynamic channel robustness, providing a solid theoretical foundation for solving the multi-vehicle cooperative access problem in heterogeneous space-ground integrated networks.

[0148] As explained, in step S3, simulation experiments are used to verify the effectiveness of the distributed decision-making model based on hierarchical decoupling; and the simulation experiments focus on key performance indicators such as scheduling efficiency, delay distribution, virtual queue stability, and the theoretical trade-off between long-term quality of service (QoS) constraints and system utility. To ensure physical authenticity, the implementation of real LEO satellite ephemeris data is driven by the UltraStar simulator, and the deployment of RSUs follows a Poisson distribution.

[0149] As an optional implementation, in the simulation experiment provided in this embodiment, core communication and network parameters are set, including: time parameters, with a total simulation time of 6000 scheduling cycles and a single scheduling cycle step size of 0.1 seconds; network scale and topology, with the scenario including 40 networked autonomous vehicles, a vehicle speed set to 30 km / h, 10 RSUs deployed on the ground, and a LEO satellite constellation scale of [missing information]. Channel and frequency band: The terrestrial V2RSU link uses a 2.4 GHz carrier frequency, allocated a bandwidth of 15 MHz, and a transmit power of 0.5 W; the satellite V2Sat link uses 20 GHz (Ka band), allocated a bandwidth of 10 MHz, and the transmit power is compensated to 10 W. Algorithm learning parameters: exponential forgetting factor in the adaptive context Thompson sampling module. It provides a smooth memory window of approximately 100 cycles; soft reset cycle Set to 300 cycles, which aligns with the typical LEO satellite visibility window; exploration coefficient This ensures sufficient exploration under non-stationary conditions. Control parameters: Control parameters in the Lyapunov optimization module. Sensitivity scans from 2.0 to 100.0 were performed in the simulation, and the default settings were used in the comparative experiments. As the optimal balance between queue constraint stability and system utility maximization.

[0150] It can be noted that, for performance comparison, the proposed distributed decision-making model based on hierarchical decoupling is compared with two representative baseline schemes: first, a greedy algorithm that maximizes the immediate utility at each step without exploration; second, the standard Q-learning algorithm. Two ablation variants are introduced, one without the Lyapunov optimization module and the other without the mean-field game approximation module, to explicitly separate and verify the respective contributions of the time-constrained control module and the spatial mean-field coordination module.

[0151] Please see Figure 2It demonstrates the steady-state average performance of the model reward; the distributed decision model based on hierarchical decoupling achieves the highest average reward and shows excellent scheduling efficiency; in contrast, among the comparative algorithms, the standard Q-Learning algorithm has difficulty handling high-dimensional continuous state spaces, resulting in poor rewards; the ablation variant without mean-field coordination exhibits catastrophic reward fading, and due to the lack of mean-field coordination, vehicles greedily tend to the same high-gain channels, causing severe congestion.

[0152] Please combine Figure 3 and Figure 4 To verify the theoretical stability guarantee of the Lyapunov optimization module, two virtual queues are used in a distributed decision-making model based on hierarchical decoupling. and All are strictly bounded and converge quickly to a stable operating region defined by the control parameter boundaries; conversely, the greedy algorithm and Q-learning strategy cannot take into account long-term cross-slot coupling, resulting in continuous queue growth and excessive fluctuations, indicating that time decoupling is crucial for enforcing long-term delays and switching constraints without deviating from expectations.

[0153] Please see Figure 5 It demonstrates the fundamental theoretical trade-off between reward optimality and queue stability, used to illustrate the control parameters in Lyapunov optimization modules. The sensitivity of the control parameters With the increase of , the algorithm prioritizes the optimization of instantaneous utility, resulting in a higher average reward, but also a significant increase in queue backlog, indicating a weakening of the enforcement of long-term constraints; conversely, overly conservative settings will severely penalize queue growth, ensuring strict satisfaction of constraints, but will significantly sacrifice the overall utility of the system.

[0154] Please see Figure 6 Delay reliability was assessed by plotting the cumulative distribution function of transmission delay. The hierarchical decoupling-based distributed decision model exhibited the fastest convergence speed, indicating that the transmission delay was highly concentrated in the low-latency region and had a steep tail decay. In contrast, the distribution tail in the comparison algorithm was significantly longer, indicating that passive channel selection without predictive spatial coordination leads to highly unstable service time. The hierarchical decoupling-based distributed decision model can consistently limit delay variations and meet the stringent reliability requirements of dynamic cooperative vehicle applications.

[0155] It can be seen that, through the deployed experimental environment, the performance characteristics of the hierarchical decoupling-based distributed decision-making model in a heterogeneous space-ground integrated vehicle network environment were evaluated. According to relevant data, the hierarchical decoupling-based distributed decision-making model significantly improves system utility while reducing average latency and QoS default rate. In other words, the simulation results verify the theoretical trade-off between stability and optimality, and confirm that the proposed scheme can guarantee long-term stable QoS in non-stationary environments, providing a feasible technical path for the deployment of space-ground integrated vehicle networks.

[0156] Understandably, this paper proposes a spatiotemporal decoupling decision-making method for multi-vehicle cooperative access in existing space-ground converged networks, addressing the distributed multi-vehicle cooperative access problem. In other words, it proposes a mean-field decoupling architecture based on Lyapunov guidance, which eliminates temporal and spatial coupling through Lyapunov optimization and mean-field game approximation modules respectively, effectively solving the complex spatiotemporal access decision coupling problem and ensuring the system's scalability and stability under long-term QoS constraints. For the non-stationary dynamic environment caused by the dual mobility of LEO satellites and vehicles, an adaptive context Thompson sampling module is adopted to effectively address the multi-agent access problem in heterogeneous environments.

[0157] A second embodiment of the present invention provides an electronic device, which includes a processor, a communication interface, a memory, and a communication bus. The processor, the communication interface, and the memory communicate with each other through the communication bus. The processor calls logical instructions in the memory to execute a spatiotemporal decoupling decision-making method for multi-vehicle access in a space-ground fusion network as described in any embodiment of the present invention.

[0158] When it is in operation, it needs to use a spatiotemporal decoupling decision-making method for multi-vehicle access in a space-ground converged network. Therefore, whether the device and program data are integrated or different hardware is configured to produce functions with similar effects to those achieved by this invention, they all fall within the protection scope of this invention. This device has the same beneficial effects as the aforementioned spatiotemporal decoupling decision-making method for multi-vehicle access in a space-ground converged network, and will not be elaborated here.

[0159] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0160] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A spatiotemporal decoupling decision-making method for multi-vehicle access in a space-ground converged network, characterized in that, The method includes: Step S1: In the space-ground fusion network composed of ground roadside units that follow the Poisson point distribution and low-orbit satellites, combine multiple networked autonomous vehicles to perform environmental assessments and establish a multi-vehicle collaborative access problem; Step S2: Construct a distributed decision model based on hierarchical decoupling for the multi-vehicle access problem, including a Lyapunov optimization module, a mean-field game approximation module, and an adaptive context Thompson sampling module, and perform spatiotemporal decoupling decision-making. Step S3: Verify the effectiveness of the distributed decision-making model based on hierarchical decoupling using simulation experiments.

2. The spatiotemporal decoupling decision-making method for multi-vehicle access in a space-ground converged network according to claim 1, characterized in that, Step S1 includes: Step S11: Define the ground roadside units and low-orbit satellites in the space-ground fusion network as candidate access nodes, and the networked autonomous driving vehicles as cooperative intelligent agents. Combine the candidate access nodes and cooperative intelligent agents to obtain V2RSU links and V2Sat links respectively. Step S12: Define the shared unified signal-to-noise ratio model for V2RSU and V2Sat links, perform environmental assessment, define the instantaneous utility of connected autonomous vehicles, and establish the multi-vehicle collaborative access problem.

3. The spatiotemporal decoupling decision-making method for multi-vehicle access in a space-ground converged network according to claim 2, characterized in that, Step S12 includes: Step S121: Obtain the transmit power and thermal noise density from the candidate access nodes and the cooperative agent respectively, and establish a signal-to-noise ratio model by combining the allocated bandwidth and the time-varying channel coefficients of each scheduling cycle; Step S122: Solve the signal-to-noise ratio model to obtain the uplink capacity and instantaneous queuing delay for the corresponding scheduling period, and determine the end-to-end delay; Step S123: Evaluate the switching cost of connected autonomous vehicles between consecutive scheduling cycles, define instantaneous utility by combining uplink capacity and end-to-end latency, and establish a multi-vehicle collaborative access problem.

4. The spatiotemporal decoupling decision-making method for multi-vehicle access in a space-ground converged network according to claim 2, characterized in that, Step S2 includes: Step S21: Based on the multi-vehicle access problem, with the optimization objective of maximizing long-term average utility and satisfying time delay and stability constraints, construct a distributed decision model based on hierarchical decoupling; Step S22: Decouple the Lyapunov optimization module and the mean-field game approximation module in time and space, establish the context vector, and output the exploratory sampling results; Step S23: Utilize the adaptive context Thompson sampling module to comprehensively analyze the context vector and exploratory sampling results, select the optimal access node, and adjust the model parameters of the distributed decision-making model based on hierarchical decoupling in real time based on feedback.

5. The spatiotemporal decoupling decision-making method for multi-vehicle access in a space-ground converged network according to claim 4, characterized in that, Step S22 includes: Step S221: The long-term constraints corresponding to the multi-vehicle access problem are mapped to real-time virtual queue pressure through the Lyapunov optimization module. The virtual queue backlog value and the corresponding constraint cost function are extracted as exploratory sampling results to eliminate time coupling. Step S222: Define the mean-field state characteristics through the mean-field game approximation module, simplify the resource competition of cooperative agents into a response of group statistical law, and eliminate spatial coupling; Step S223: After spatiotemporal decoupling, the candidate access nodes and cooperative agents capture vehicle features and node features respectively, and establish context vectors.

6. The spatiotemporal decoupling decision-making method for multi-vehicle access in a space-ground converged network according to claim 5, characterized in that, In step S221, the long-term constraints include asynchronous delay fairness constraints and switching frequency stability constraints; the virtual queues include virtual queues that track the degree of delay deviation from the mean and virtual queues that track the degree of switching frequency violations.

7. A spatiotemporal decoupling decision-making method for multi-vehicle access in a space-ground converged network according to claim 5, characterized in that, In step S222, specifically: For each candidate access node, the average number of associated vehicles, average received signal-to-noise ratio, and congestion index are extracted to define the mean field state. The candidate access node then broadcasts this state to the cooperative agent, simplifying the resource competition among the cooperative agents into a response based on group statistical laws and eliminating spatial coupling.

8. The spatiotemporal decoupling decision-making method for multi-vehicle access in a space-ground converged network according to claim 4, characterized in that, Step S23 includes: Step S231: The adaptive context Thompson sampling module uses a context Thompson sampling algorithm based on Bayesian linear regression, presets a reward parameter vector, and combines the context vectors to construct the expected reward model of the candidate access node; Step S232: Establish a decision scoring function in the online decision-making process by combining the expected reward model and the exploratory sampling results, and select the optimal access node; Step S233: Introduce a dual-timescale update mechanism to adjust the model parameters of the distributed decision-making model based on hierarchical decoupling in real time according to feedback, in order to cope with non-stationary changes in the environment.

9. A spatiotemporal decoupling decision-making method for multi-vehicle access in a space-ground converged network according to claim 8, characterized in that, In step S233, specifically: An exponential forgetting mechanism is adopted to adapt to the gradual decay of V2RSU links, and a periodic soft reset mechanism is used to adapt to the topology changes of V2Sat links. The model parameters of the distributed decision-making model based on hierarchical decoupling are corrected in real time according to feedback to cope with non-stationary changes in the environment.

10. An electronic device, characterized in that, The device includes a processor, a communication interface, a memory, and a communication bus. The processor, the communication interface, and the memory communicate with each other through the communication bus. The processor calls logical instructions in the memory to execute the spatiotemporal decoupling decision method for multi-vehicle access in a space-ground fusion network as described in any one of claims 1 to 9.

Images