Intelligent on-demand service method and device for vehicle networking deterministic latency guarantee

Abstract

The present application relates to the technical field of mobile communication, in particular to a smart on-demand service method and device for vehicle networking deterministic delay guarantee. The method comprises: establishing a deterministic delay model under a multi-mode link; modeling a wireless channel, introducing a stochastic network calculus (SNC) theory, and deducing a general formula of delay violation probability under different channel conditions; dynamically establishing a transmission power model according to link characteristics, and giving a minimum power threshold value that meets the required channel conditions; taking the maximization of the number of successful vehicles as the optimization objective, and determining the constraint conditions; using an LA2C reinforcement learning algorithm improved based on an A2C mechanism to perform strategy learning and solving, and completing the smart on-demand service for vehicle networking deterministic delay guarantee. The present application adaptively configures key communication parameters such as transmission mode, access mode, resource allocation, and semantic compression ratio, thereby significantly improving the task completion rate and meeting the development trend of the future communication network towards intelligentization and adaptive evolution.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of mobile communication technology, in particular to an intelligent on-demand service method and device for vehicle networking deterministic latency guarantee. BACKGROUND

[0002] With the rapid development of mobile communication technology, especially under the promotion of the fifth generation (5G) and the future sixth generation (6G) mobile communication network, the communication system is more and more strict in meeting the requirements of large bandwidth and high reliability, and the requirement for ultra-low latency is also increasing. In key application scenarios such as industrial Internet, remote medical treatment, and vehicle networking, short data packet services occupy a dominant position, and millisecond or even sub-millisecond level challenges are put forward for end-to-end latency. Taking vehicle networking as an example, the communication between vehicles and base stations (Vehicle to Vehicle, V2V) requires the whole process of "perception-decision-brake" to be less than 10ms, and the air transmission can only occupy about 1ms. The communication between vehicles and infrastructure (Vehicle to Infrastructure, V2I) needs to issue high-precision maps or collaborative perception results in the scene of high-speed vehicle movement, and also needs to maintain millisecond-level link round trip. Therefore, the existing system shortens the transmission time by simplifying the protocol, increasing the bandwidth, and using high-order modulation. However, these methods are mostly based on fixed network resource allocation, and lack the ability to adapt to the dynamic changes of vehicle speed, road conditions, and traffic flow.

[0003] As a new communication paradigm, semantic communication is committed to transmitting "semantics" rather than "bits" as the core goal, and by extracting semantic information critical to the task from raw data for encoding and transmission, it can significantly reduce the amount of data that needs to be transmitted over the air. Compared with traditional methods, semantic communication can improve information transmission efficiency when network bandwidth is limited or channel quality is poor, and is an important supplement to achieve high-reliability and low-latency communication in vehicle networking and industrial control scenarios. However, in practical applications, semantic communication still faces two major challenges: first, the computational overhead of semantic extraction and reconstruction may offset the latency advantage brought by bandwidth saving; second, existing systems mostly use static or fixed semantic compression strategies, which are difficult to dynamically adjust with real-time changes in vehicle speed, vehicle distance, channel fading, and traffic density, resulting in suboptimal system performance.

[0004] On the other hand, although traditional bit transmission schemes can reduce latency by adjusting frame structure or enhancing scheduling flexibility, they rarely consider the semantic redundancy of information content, so it is difficult to meet the dual requirements of resource efficiency and ultra-low latency for high-frequency short packet communication in future vehicle networking. Existing semantic communication frameworks often only focus on a single indicator (such as compression ratio or reconstruction accuracy), lack global coordination design of system-level parameters (resource allocation, access link mode, etc.), and are particularly limited in complex wireless environments with frequent V2V / V2I handovers.

[0005] Therefore, how to dynamically adjust the transmission mode (bit transmission or semantic transmission), access mode (V2V or V2I), bandwidth resource allocation, and semantic compression ratio according to the real-time needs and channel conditions of services such as vehicle-to-everything (V2X) while ensuring task completion accuracy and service reliability has become a key research direction for meeting user needs in V2X scenarios. This invention constructs a perceptible, controllable, and learnable semantic-bit joint transmission mechanism, integrating semantic communication and intelligent decision-making methods, which can achieve a low-latency, highly robust communication system in extremely latency-sensitive scenarios such as V2X.

[0006] In summary, while existing technologies have made some progress in reducing the latency of vehicle-to-everything (V2X) communication, they generally suffer from the following core problems: (1) high computational latency in semantic communication makes it difficult to meet the extremely low latency requirements in different scenarios; (2) resource adaptation methods lack dynamic response capabilities, making it difficult to adapt to changing network environments; and (3) the lack of an intelligent transmission framework that can uniformly schedule semantic compression, resource allocation, and link modes results in limited room for overall system performance optimization. These problems are particularly prominent in latency-sensitive V2X scenarios, severely restricting the evolution and actual deployment effectiveness of in-vehicle communication systems. Summary of the Invention

[0007] To address the technical challenge of varying deterministic latency and reliability requirements among multiple vehicles in dynamic vehicular networks in existing technologies, embodiments of the present invention provide an intelligent on-demand service method and apparatus for ensuring deterministic latency in vehicular networks. The technical solution is as follows:

[0008] On the one hand, a method for providing intelligent on-demand services with deterministic latency guarantees for vehicle-to-everything (V2X) networks is provided, characterized by the following:

[0009] S1. Establish a deterministic delay model under multi-mode links;

[0010] S2. Model the wireless channel, introduce the Stochastic Network Calculus (SNC) theory and Meijer G function, and derive the general formula for the delay violation probability under different channel conditions.

[0011] S3. Dynamically establish a transmit power model based on link characteristics and provide the minimum power threshold value when the required channel conditions are met.

[0012] S4. The optimization objective is to maximize the number of vehicles that successfully complete the mission, and the constraints are determined.

[0013] S5. The LA2C reinforcement learning algorithm based on the A2C mechanism is used to learn the policy and solve the optimization objective to obtain the power space output result, thus completing the intelligent on-demand service for deterministic latency guarantee of vehicle networking.

[0014] Optionally, in S1, a deterministic latency model under a multi-mode link is established, including:

[0015] The deterministic latency composition under different communication links and transmission modes is modeled and analyzed;

[0016] The different communication links include V2I mode and V2V mode;

[0017] In the V2I mode and the V2V mode, the deterministic latency of the semantic user includes two parts of communication latency and semantic calculation latency;

[0018] In the V2V mode, the waiting time of relay establishment also needs to be considered, wherein the total latency of the bit user in the V2V mode includes the transmission time and the time required for the relay vehicle to enter the communication range; the semantic user further superimposes the semantic processing time on this basis;

[0019] The larger value of the semantic calculation latency and the communication establishment latency is included in the analysis.

[0020] Optionally, in S2, the wireless channel is modeled, the stochastic network calculus SNC theory and the Meijer G function are introduced, and the general formula of latency violation probability under different channel conditions is derived, including:

[0021] By constructing the cumulative arrival process and service process model of the service queue, the generalized α-κ-μ channel model is adopted to establish the mathematical relationship between the data arrival rate, service capability, queue length and latency.

[0022] Optionally, in S3, a transmission power model is dynamically established according to the link characteristics, and the minimum power threshold value that meets the required channel conditions is given, including:

[0023] Two types of channel characteristics are considered comprehensively: one is large-scale path loss, and the other is small-scale channel fading;

[0024] The large-scale loss is mainly determined by the distance between the transmitting end and the receiving end, and is affected by the path loss exponent and the reference channel gain; the small-scale fading is based on statistical modeling under complex scenarios, and is described uniformly by the generalized α-κ-μ channel model;

[0025] For the V2I link, the model considers the position change between the service vehicle and the base station and the coverage range of the base station, and calculates the average power consumption from when the vehicle enters the coverage area to when it leaves the area;

[0026] For the V2V link, the relative motion trajectory between the service vehicle and the relay vehicle is analyzed to determine the relative speed, approach and separation time of the two vehicles in the communication range, and then the communication power consumption is estimated.

[0027] Optionally, in S4, the optimization objective is to maximize the number of task success vehicles, and the constraint conditions are determined, including:

[0028] By dynamically deciding the transmission mode, access mode, semantic compression ratio, bandwidth, and other parameters of each vehicle, the number of vehicle users successfully completing the communication task in the system is maximized, while meeting multiple constraints of system resources and quality of service;

[0029] The optimization objective is to maximize the number of task success vehicles, and nine constraint conditions are designed.

[0030] Optionally, the nine constraint conditions include:

[0031] C1-C3 limits the delay of vehicles from the perspective of quality of service to be less than a threshold, the delay violation probability is below a set threshold, and can be extended to other task indicators related to semantic restoration accuracy or error tolerance;

[0032] C4-C6 ensures that the total bandwidth of the system, the total transmission power of V2I, and the single-link power of V2V do not exceed the physically available range;

[0033] C7-C9 discretely constrains the control variables of the system, including the selection of transmission mode, access mode, and semantic compression ratio.

[0034] Optionally, in S5, the LA2C reinforcement learning algorithm improved based on the A2C mechanism is used for policy learning and solving the optimization objective, including:

[0035] The LA2C reinforcement learning algorithm improved based on the A2C mechanism is used for policy learning;

[0036] Among them, the long short-term memory (LSTM) neural network is combined as a feature extractor; and a policy-value parallel update mechanism is combined;

[0037] A multi-dimensional heterogeneous modeling mechanism is used, in which actions are decomposed into multiple sub-dimensions;

[0038] Parallel multi-environment simulation is realized through SubprocVecEnv to speed up the algorithm running efficiency;

[0039] The task success probability is used as the reward function, and delay and reliability penalty terms are introduced.

[0040] On the other hand, an intelligent on-demand service device for vehicle networking deterministic delay guarantee is provided, which is used to implement the intelligent on-demand service method for vehicle networking deterministic delay guarantee. The device includes:

[0041] The deterministic delay model establishment module is used to establish a deterministic delay model under a multi-mode link.

[0042] a channel modeling module, configured to model a wireless channel, introduce a stochastic network calculus (SNC) theory and a Meijer G function, and derive a general formula of a delay violation probability under different channel conditions;

[0043] a transmit power model establishing module, configured to dynamically establish a transmit power model according to link characteristics, and give a minimum power threshold value that meets a required channel condition;

[0044] an optimization module, configured to maximize a number of successful vehicles as an optimization target, and determine a constraint condition;

[0045] a solving module, configured to perform policy learning and solve the optimization target by using an LA2C reinforcement learning algorithm improved based on an A2C mechanism, obtain a power space output result, and complete the intelligent on-demand service for the deterministic delay guarantee of the vehicle networking.

[0046] In another aspect, an intelligent on-demand service device for the deterministic delay guarantee of the vehicle networking is provided, and the intelligent on-demand service device for the deterministic delay guarantee of the vehicle networking comprises a processor and a memory.

[0047] In another aspect, a computer-readable storage medium is provided, and the storage medium stores at least one instruction.

[0048] The technical scheme provided by the embodiments of the present application has at least the following beneficial effects:

[0049] In the embodiments of the present application, an intelligent communication method suitable for the vehicle networking scene is provided, which fuses semantic communication and bit communication, introduces a bandwidth dynamic allocation mechanism, and realizes deterministic delay optimization and maximum resource utilization at a system level.

[0050] In this invention, each vehicle user can choose to compress and transmit data using semantic communication or traditional bit communication, based on their own communication needs and channel conditions. Users simultaneously possess the ability to select access methods, adjust semantic compression ratios, and dynamically control bandwidth usage. The network side, based on a unified optimization objective (maximizing task completion rate), jointly decides the optimal parameter combination for all users, thereby improving the overall system performance. Attached Figure Description

[0051] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0052] Figure 1 This is a flowchart illustrating the intelligent on-demand service method for deterministic latency assurance in the Internet of Vehicles provided in an embodiment of the present invention.

[0053] Figure 2 This is a block diagram of an intelligent on-demand service device for deterministic latency assurance in the Internet of Vehicles provided in an embodiment of the present invention;

[0054] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0055] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0056] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0057] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0058] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0059] The embodiment of the present application provides a smart on-demand service method for vehicle networking deterministic delay guarantee, which can be realized by a smart on-demand service device for vehicle networking deterministic delay guarantee, and the smart on-demand service device for vehicle networking deterministic delay guarantee can be a terminal or a server. Figure 1 As shown in the smart on-demand service method flow chart for vehicle networking deterministic delay guarantee, Figure 1 As shown in the smart on-demand service method flow chart for vehicle networking deterministic delay guarantee,

[0060] S1, a deterministic delay model under a multi-mode link is established;

[0061] In a feasible implementation manner, in S1, the deterministic delay model under the multi-mode link is established, and the establishment includes:

[0062] The deterministic delay constitution under different communication links and transmission modes is modeled and analyzed;

[0063] The different communication links include a V2I mode and a V2V mode.

[0064] In the V2I mode and the V2V mode, the deterministic delay of a semantic user includes two parts of a communication delay and a semantic calculation delay.

[0065] In the V2V mode, the waiting time of relay establishment also needs to be considered, wherein the total delay of a bit user in the V2V mode includes a transmission time and a time required for a relay vehicle to enter a communication range; and the semantic user further superimposes a semantic processing time on the basis.

[0066] The larger value of the semantic calculation delay and the communication establishment delay is included in the analysis.

[0067] In a feasible implementation manner, before the communication system scheduling optimization mechanism is constructed, the deterministic delay constitution under different communication links and transmission modes is modeled and analyzed, so as to provide a delay reference basis for subsequent scheduling decision and resource allocation.

[0068] In V2I, the delay of a common bit transmission user is mainly determined by the amount of data to be transmitted and the current link transmission rate. For a user using semantic communication, in addition to the basic transmission delay, the processing delay introduced in the semantic encoding and decoding process also needs to be considered, and the delay is closely related to the semantic information complexity, terminal computing capacity and other factors. Therefore, in the V2I mode, the deterministic delay of the semantic user includes two parts of a communication delay and a semantic calculation delay.

[0069] In V2V, in addition to the transmission and calculation delay mentioned above, the waiting time of relay establishment also needs to be considered. The total delay of bit users in V2V mode includes the transmission time and the time required for the relay vehicle to enter the communication range; semantic users further superimpose semantic processing time on this basis. In order to ensure that the delay model is more in line with the actual situation, the system will take the larger value of semantic calculation delay and communication establishment delay into account to more accurately reflect the delay overhead in the worst case.

[0070] In addition, the estimation of semantic decoding delay is based on the hardware characteristics of the receiving end, considering factors such as its unit data processing capacity and semantic reconstruction complexity. Through this mechanism, the system can estimate the expected communication delay under each combination strategy in advance, and make intelligent transmission parameter decisions accordingly, ensuring that the final scheduling result meets the delay requirement while taking into account the calculation feasibility and transmission efficiency.

[0071] This step lays the foundation for joint optimization in the construction of the overall intelligent on-demand service framework, enabling the subsequent decision-making process to fully consider factors such as link differences, semantic processing capabilities, and vehicle dynamics, achieving precise control of deterministic delay and overall improvement of system performance.

[0072] In a feasible implementation, in S2, the wireless channel is modeled, the SNC theory and Meijer G function are introduced, and the general formula of delay violation probability under different channel conditions is derived.

[0073] In a feasible implementation, in S2, the wireless channel is modeled, the SNC theory and Meijer G function are introduced, and the general formula of delay violation probability under different channel conditions is derived.

[0074] By constructing the cumulative arrival process and service process model of the service queue, a generalized alpha-kappa-mu channel model is adopted to establish the mathematical relationship between data arrival rate, service capability, queue length, and delay.

[0075] In a feasible implementation, the method establishes the mathematical relationship between data arrival rate, service capability, queue length, and delay by constructing the cumulative arrival process and service process model of the service queue, providing a theoretical basis for system design.

[0076] In order to enhance the applicability of the general formula under diversified channel conditions, the application adopts a generalized alpha-kappa-mu channel model, which can cover various typical channel types (such as Rayleigh, Rician, Nakagami-m, etc.) as special cases, and has strong generalization ability. The probability statistical characteristics of the channel are characterized by the parameters in the model. When these parameters take certain specific values, the alpha-kappa-mu channel can be simplified as some common channels, and the specific parameters are shown in Table 1:

[0077] Table 1 Simplification table of alpha-kappa-mu channel

[0078]

[0079] The application proposes a delay analysis method for alpha-kappa-mu channel, and the specific steps are as follows:

[0080] (1) Define the cumulative process: establish the queuing system of the service to the base station, represent the data arrival process and the service process as the cumulative arrival function and the cumulative service function respectively, and use them to characterize the running state of the system over time.

[0081] (2) Derive the queue evolution formula: based on the conservation relationship among arrival, service and departure, give the explicit expression of the queue length change over time, which lays the foundation for delay calculation.

[0082] (3) Introduce the delay violation probability: take "queue length exceeding the maximum tolerance value" as an event, define the delay violation probability, and derive its upper bound form as a reliability evaluation index.

[0083] (4) Inequality transformation: use Markov inequality and inequality to transform the upper bound of the delay violation probability into the expected form, which is convenient for subsequent solution.

[0084] (5) Model the arrival process: assume that the service arrival conforms to the Poisson distribution, combine the average arrival rate and the packet length to obtain the probability characteristics of the cumulative arrival process, and evaluate its influence on the delay.

[0085] (6) Model the service process: under the framework of finite block length coding, characterize the instantaneous channel gain by the probability density function of the alpha-kappa-mu channel, and derive the single packet service amount distribution and its moment generating function.

[0086] (7) Introduce Meijer G function to solve complex integrals and simplify the expression: in order to solve the expected item, use Meijer G function to transform the complex integral into an analytical form, and simplify the result by parameter substitution, obtain a unified expression applicable to Rayleigh, Rician, Nakagami-m, etc. Special cases.

[0087] (8) Forming the delay violation probability formula: Based on the above steps, the final formula of delay violation probability containing α-κ-μ parameters is given, which can be directly used for delay evaluation and resource scheduling constraints under different channel conditions.

[0088] Finally, the probability of delay violation can be expressed as:

[0089]

[0090] Where: ε k denotes the delay violation probability of the kth user. A(τ,t) denotes the cumulative arrival process in the time interval (τ,t);

[0091] denotes the maximum delay that the kth user can tolerate, ρ a (·), σ a (·) denotes the related parameters of the arrival process; ρ a (·), σ a (·) denotes the related parameters of the service process; when the arrival process and the service process are independent, p=q=1, θ is a constant greater than 0.

[0092] S3, dynamically establish a transmission power model according to the link characteristics, and give the minimum power threshold value that meets the required channel conditions;

[0093] In a feasible implementation, in S3, a transmission power model is dynamically established according to the link characteristics, and the minimum power threshold value that meets the required channel conditions is given, which includes:

[0094] Two types of channel characteristics are considered: one is large-scale path loss, and the other is small-scale channel fading;

[0095] Among them, the large-scale loss is mainly determined by the distance between the transmitting end and the receiving end, and is affected by the path loss exponent and the reference channel gain; the small-scale fading is based on statistical modeling under complex scenarios, and is described uniformly by the generalized α-κ-μ channel model;

[0096] For the V2I link, the model considers the position change between the service vehicle and the base station and the coverage range of the base station, and calculates the average power consumption from when the vehicle enters the coverage area to when it leaves the area;

[0097] For the V2V link, the relative motion trajectory between the service vehicle and the relay vehicle is analyzed to determine the relative speed, approach and separation time of the two vehicles within the communication range, and then the communication power consumption is estimated.

[0098] In an implementable embodiment, in the vehicle communication system constructed by the application, power distribution is a key factor affecting link reliability and energy efficiency. In order to ensure that vehicle users can meet specific quality of service (QoS) requirements in different scenarios, the system needs to dynamically establish a transmission power model according to link characteristics.

[0099] The application comprehensively considers two types of channel characteristics: one is large-scale path loss, and the other is small-scale channel fading. The large-scale loss is mainly determined by the distance between the transmission end and the receiving end, and is affected by the path loss exponent and the reference channel gain; the small-scale fading is statistically modeled according to a complex scenario, and is uniformly described by a generalized alpha-kappa-mu channel model, which can cover special cases of common fading models such as Rayleigh, Rician, and Nakagami. By analyzing the movement trajectory of the vehicle on the road within a given period, the system can estimate the actual communication distance change in the communication process, and combine the channel statistical characteristics to obtain the average transmission power threshold required to meet the service reliability requirements within the period.

[0100] On this basis, the power model further distinguishes between V2I and V2V communication links:

[0101] (1) For the V2I link, the model considers the position change between the service vehicle and the base station and the coverage range of the base station, and calculates the average power consumption from when the vehicle enters the coverage area to when it leaves the area;

[0102] (2) For the V2V link, the relative motion trajectory between the service vehicle and the relay vehicle is analyzed to determine the relative speed of the two vehicles within the communication range, the approach and separation times, and then the communication power consumption is estimated.

[0103] Through the above power model, the system can accurately predict the link power consumption in a dynamic environment, and accordingly flexibly control the transmission power distribution in the scheduling process, to realize the dynamic balance of communication reliability guarantee and energy efficiency. This model also provides necessary physical layer constraint basis for subsequent resource allocation strategies.

[0104] The final formula is as follows:

[0105] The solution of V2I power is as follows:

[0106]

[0107] The solution of V2V power is as follows:

[0108]

[0109] Where, P V2I (t b ,t e) represents the average transmit power of the service vehicle in the time period [t b ,t e ] when communicating with the base station. V2V (t b ,t e ) represents the average transmit power of the service vehicle in the time period [t b ,t e ] when communicating with the base station. a represents the joint antenna gain of the transmitter and the receiver, γ t represents the signal-to-noise ratio (SNR) decoding threshold of the link, σ 2 represents the channel noise power, a p is the path loss exponent, b r is the reference path loss constant, Γ(·) represents the gamma function used to process the integral term of the channel capacity, C, α, Y are channel parameters in α-κ-μ, v s represents the speed of the service vehicle, v r represents the speed of the relay vehicle, D b represents the coverage radius of the base station, D v represents the V2V communication distance limit, d0 represents the initial position of the service vehicle relative to the base station, t b , t e represent the start and end times in the relative time interval. The relative speed between the service vehicle and the relay vehicle is v m = v s + v r , t0 represents the time point when the two vehicles meet.

[0110] S4, taking maximizing the number of task success vehicles as the optimization objective, and determining the constraint conditions.

[0111]

[0112] wherein Φ represents the task success rate, ξ k , φ k , o k , W k respectively represent the access mode, transmission mode, semantic compression ratio and bandwidth of the user. represents the maximum delay requirement of the user, represents the reliability requirement of the user. W max represents the maximum available bandwidth of the base station, P max represents the total power of the base station, P max,v represents the power limit of the V2V link.

[0113] In one possible implementation, in S4, taking maximizing the number of task success vehicles as the optimization objective, and determining the constraint conditions, include:

[0114] By dynamically deciding the transmission mode, access mode, semantic compression ratio, bandwidth and other parameters of each vehicle, the number of vehicle users successfully completing the communication task in the system is maximized, while meeting multiple constraints of system resources and service quality.

[0115] With the optimization goal of maximizing the number of task success vehicles, nine constraint conditions are designed.

[0116] In a feasible implementation, the application aims to maximize the number of vehicle users successfully completing the communication task in the system by dynamically deciding the transmission mode, access mode, semantic compression ratio, bandwidth and other parameters of each vehicle in a multi-user vehicle networking system, while meeting multiple constraints of system resources and service quality.

[0117] In a feasible implementation, the application designs nine constraint conditions (C1-C9) with the optimization goal of maximizing the number of task success vehicles, covering the feasibility of service quality, resource allocation and control variables. The nine constraint conditions include:

[0118] C1-C3 limits the delay of vehicles from the perspective of service quality to be less than the threshold, the delay violation probability to be lower than the set threshold, and can be extended to other task indicators related to semantic restoration accuracy or error tolerance;

[0119] C4-C6 ensures that the total bandwidth of the system, the total transmission power of V2I and the single-link power of V2V do not exceed the physically available range;

[0120] C7-C9 discretely constrains the control variables of the system, including the selection of transmission mode (bit or semantic), access mode (V2I or V2V) and semantic compression ratio, to ensure that the optimization strategy can be realized in actual engineering.

[0121] The optimization model improves the overall task completion rate of the system under multiple constraints, achieving efficient resource scheduling and communication decision for intelligent vehicle networking scenarios.

[0122] S5, an LSTM-Advantage Actor-Critic (LA2C) reinforcement learning algorithm improved based on the Advantage Actor-Critic (A2C) mechanism is used for policy learning and solving the optimization objective to obtain the output results of the power space and complete the intelligent on-demand service for vehicle networking with deterministic delay guarantee.

[0123] In a feasible implementation, in S5, the LA2C reinforcement learning algorithm improved based on the A2C mechanism is used for policy learning and solving the optimization objective, including:

[0124] The LA2C reinforcement learning algorithm improved based on the A2C mechanism is used for policy learning.

[0125] The algorithm combines a long short-term memory (LSTM) neural network as a feature extractor, and a strategy-value parallel update mechanism.

[0126] A multi-dimensional heterogeneous modeling mechanism is used, in which actions are decomposed into multiple sub-dimensions.

[0127] Parallel multi-environment simulation is realized through SubprocVecEnv to speed up the algorithm running efficiency.

[0128] The task success probability is used as the reward function, and a delay and reliability penalty term is introduced.

[0129] In a feasible implementation, the present application uses the LA2C reinforcement learning algorithm improved based on the A2C mechanism to realize policy learning for the joint optimization of transmission mode, access mode, bandwidth resource, and semantic compression ratio. The algorithm combines a long short-term memory (LSTM) neural network as a feature extractor, effectively mining the historical dependencies and implicit patterns in the system state, and combines a strategy-value parallel update mechanism, having the ability to make efficient decisions in a dynamic and continuous state space. To further improve the policy convergence speed and system explainability, the present application designs a structured state space, action space, and reward function, as follows:

[0130] (1) Action space design:

[0131] The action space is defined as a four-dimensional discrete space: The specific meanings are as follows:

[0132] Transmission mode selection: discrete variable, represents semantic transmission, represents traditional bit transmission. The agent flexibly selects the communication mode between compression efficiency and calculation delay according to the current bandwidth, signal-to-noise ratio, and business delay requirement.

[0133] Access mode selection: represents V2V mode, represents V2I mode. The agent selects the access link based on the distance to neighboring vehicles, relative speed, and base station accessibility.

[0134] Semantic compression ratio control: discrete variable, value range [0, 1]. The smaller the value, the higher the compression degree, the smaller the semantic data packet, and the faster the transmission, but the reconstruction calculation amount increases.

[0135] Bandwidth allocation scheme: a discrete variable, dynamically allocates bandwidth resources to each user according to system resource remaining, data volume and service priority, ensures system scheduling efficiency and stability.

[0136] (2) State space design:

[0137] After each step of algorithm execution, the network state is updated and stored in the experience pool, and the state space is represented as Including:

[0138] Channel state information: including current channel quality (such as SNR, shadow fading factor), noise power, etc., used to evaluate link availability and transmission reliability.

[0139] User service quality requirements: including the maximum allowed delay of users, the upper limit of delay violation probability, etc., guiding the strategy to shift to low delay or high reliability.

[0140] Inter-vehicle relative speed and distance information: used to determine whether the vehicle access mode is suitable for V2V, affecting communication duration and coverage.

[0141] The system state is time-encoded by the LSTM network, enabling the agent to make more reasonable decisions using past historical state information, thereby improving generalization ability in dynamic environments.

[0142] (3) Reward function design:

[0143] The invention takes the task success probability as the reward function, that is, the ratio of the number of users meeting the quality of service requirements to the total number of users. At the same time, by introducing delay and reliability penalty terms, the influence of delay and reliability on system performance is effectively captured, providing more detailed feedback for reinforcement learning algorithms. Through this design, the reward function becomes smooth instead of sudden change, enabling the agent to learn the strategy more quickly.

[0144] (4) Advantages of LA2C algorithm:

[0145] The LA2C algorithm has multiple advantages in intelligent scheduling of vehicle-mounted communication. First, by using a multi-dimensional heterogeneous action modeling mechanism, factors such as access mode, transmission mode, bandwidth allocation and semantic compression rate are included in a unified multi-dimensional heterogeneous action space, making the strategy learning more close to the real scheduling problem and enhancing the expression ability and decision flexibility of the model. Second, advanced mathematical tools such as Meijer G function are introduced to analyze and model key performance indicators such as channel capacity, power allocation and delay, significantly improving the accuracy and efficiency of environment simulation. Third, a parallel multi-environment training architecture based on SubprocVecEnv is constructed to realize efficient parallel and index visualization in the training process, accelerate the convergence speed of the strategy, and support the efficient combination of offline training and online deployment. Finally, by introducing an LSTM feature extractor, the time-dependent information in the state is fully exploited, enabling the agent to make more forward-looking decisions using historical trajectories, significantly improving the robustness and generalization ability of the model in dynamic complex scenarios. The above designs together build an efficient, stable and realistic adaptive reinforcement learning scheduling framework.

[0146] In a feasible implementation, the optimization target is solved, and the solution is the output of the action space, that is, the transmission mode, access mode, semantic compression ratio and bandwidth. After obtaining this solution, resources are allocated according to the four parameters, that is, on-demand services are realized.

[0147] In the embodiment of the application, an intelligent on-demand service framework for vehicle networking deterministic delay guarantee is provided, which integrates semantic compression, adaptive bandwidth allocation and reinforcement learning scheduling mechanism, and realizes efficient communication and flexible scheduling in complex environments. Compared with the traditional static transmission scheme, this method significantly improves the communication efficiency, better guarantees the deterministic delay, and improves the task completion rate. With the help of the LA2C algorithm, the system has the ability to optimize the transmission strategy in real time according to the channel state and business demand, enhancing the environmental adaptability and intelligent level; at the same time, by accurately modeling the delay and reliability requirement, the quality of service is guaranteed, and the task completion rate and system robustness are improved; in terms of resources and energy consumption, by jointly optimizing the transmission mode, access mode, bandwidth allocation and semantic compression ratio, the communication performance and energy consumption are effectively balanced, and good actual deployment value is achieved. The technical scheme is applicable to intelligent transportation, vehicle-road cooperation and industrial internet and other low-delay and high-reliability scenarios, and provides efficient and intelligent network service support for the next generation of mobile communication systems.

[0148] Figure 2 is a block diagram of an intelligent on-demand service device 300 for vehicle networking deterministic delay guarantee according to an exemplary embodiment, and the device 300 is used for an intelligent on-demand service method for vehicle networking deterministic delay guarantee. Referring to Figure 2The device comprises a deterministic delay model establishing module 310, a channel modeling module 320, a transmission power model establishing module 330, an optimization module 340, and a solving module 350. Among them:

[0149] The deterministic delay model establishing module 310 is configured to establish a deterministic delay model under a multi-mode link.

[0150] The channel modeling module 320 is configured to model a wireless channel, introduce a stochastic network calculus (SNC) theory, and derive a general formula of a delay violation probability under different channel conditions.

[0151] The transmission power model establishing module 330 is configured to dynamically establish a transmission power model according to link characteristics.

[0152] The optimization module 340 is configured to maximize the number of successful vehicles as an optimization objective and determine a constraint condition.

[0153] The solving module 350 is configured to perform policy learning and solving by using an LA2C reinforcement learning algorithm improved based on an A2C mechanism, and complete intelligent on-demand services for deterministic delay guarantee of vehicle networking.

[0154] Optionally, the deterministic delay model under the multi-mode link comprises:

[0155] Modeling and analyzing the deterministic delay composition under different communication links and transmission modes.

[0156] Among them, the different communication links include a V2I mode and a V2V mode.

[0157] Under the V2I mode and the V2V mode, the deterministic delay of a semantic user includes two parts of a communication delay and a semantic calculation delay.

[0158] Under the V2V mode, the waiting time for relay establishment also needs to be considered, wherein the total delay of a bit user under the V2V mode includes a transmission time and a time required for a relay vehicle to enter a communication range; and a semantic user further superimposes a semantic processing time on this basis.

[0159] The larger value between the semantic calculation delay and the communication establishment delay is included in the analysis.

[0160] Optionally, the wireless channel is modeled, the stochastic network calculus (SNC) theory and Meijer G function are introduced, and a general formula of a delay violation probability under different channel conditions is derived, comprising:

[0161] By constructing a cumulative arrival process and a service process model of a service queue, a generalized α-κ-μ channel model is adopted to establish a mathematical relationship between a data arrival rate, a service capability, a queue length, and a delay.

[0162] Optionally, a transmission power model is dynamically established according to link characteristics, giving the minimum power threshold value when the required channel conditions are met, including:

[0163] Two types of channel characteristics are considered comprehensively: one is large-scale path loss, and the other is small-scale channel fading;

[0164] Among them, the large-scale loss is mainly determined by the distance between the transmitting end and the receiving end, and is affected by the path loss exponent and the reference channel gain; the small-scale fading is based on statistical modeling under complex scenarios, and is described uniformly by the generalized α-κ-μ channel model;

[0165] For the V2I link, the model considers the position change between the service vehicle and the base station and the coverage range of the base station, and calculates the average power consumption from when the vehicle enters the coverage area to when it leaves the area;

[0166] For the V2V link, the relative motion trajectory between the service vehicle and the relay vehicle is analyzed to determine the relative speed, approach and separation time of the two vehicles within the communication range, and then the communication power consumption is estimated.

[0167] Optionally, the optimization objective is to maximize the number of task-completed vehicles, and the constraint conditions are determined, including:

[0168] By dynamically deciding the transmission mode, access mode, semantic compression ratio, bandwidth and other parameters of each vehicle, the number of vehicle users successfully completing the communication task in the system is maximized, while meeting multiple constraints of system resources and quality of service;

[0169] The optimization objective is to maximize the number of task-completed vehicles, and nine constraint conditions are designed.

[0170] Optionally, the nine constraint conditions include:

[0171] C1-C3 limit the delay of vehicles from the quality of service perspective to be less than the threshold, the delay violation probability to be lower than the set threshold, and can be extended to other task indicators related to semantic restoration accuracy or error tolerance;

[0172] C4-C6 ensure that the total bandwidth of the system, the total transmission power of V2I and the power of V2V single link do not exceed the physical available range from the resource dimension;

[0173] C7-C9 discretely constrain the control variables of the system, including the selection of transmission mode, access mode and semantic compression ratio.

[0174] Optionally, the LA2C reinforcement learning algorithm improved based on the A2C mechanism is used for policy learning and solution, completing the intelligent on-demand service for vehicle networking with deterministic delay guarantee, including:

[0175] Adopt the LA2C reinforcement learning algorithm improved based on A2C mechanism to carry out policy learning;

[0176] Among them, the algorithm combines long short-term memory (LSTM) neural network as a feature extractor; and combines a strategy-value parallel update mechanism.

[0177] A multi-dimensional heterogeneous modeling mechanism is adopted, wherein the action is decomposed into multiple sub-dimensions.

[0178] Parallel multi-environment simulation is realized through SubprocVecEnv to speed up the algorithm running efficiency.

[0179] The task success probability is taken as a reward function, and a delay and reliability penalty term is introduced.

[0180] In the embodiment of the application, a smart on-demand service architecture (VIDS) for vehicle networking is constructed, which can adaptively configure key communication parameters such as transmission mode, access mode, resource allocation, and semantic compression ratio according to channel dynamic changes and diversified quality of service (QoS) requirements of vehicles, thereby significantly improving the task completion rate and meeting the development trend of future communication networks towards intelligentization and adaptive evolution.

[0181] (1) Design an adaptive communication strategy that integrates semantic / bit transmission and V2V / V2I flexible access:

[0182] The application supports flexible transmission of semantic information and traditional bit information, allows the system to dynamically select V2I or V2V access mode on the basis of adjustable semantic compression ratio, and adaptively matches the requirements of different tasks for delay and reliability, realizes precise matching between communication resources and user services and on-demand service, and improves the overall communication flexibility and system efficiency.

[0183] (2) Introduce SNC and Meijer G function to construct a general expression of delay violation probability:

[0184] Based on the random network calculus theory and Meijer G function, a general formula of delay violation probability suitable for different channel conditions is derived, and V2I and V2V links are modeled and analyzed differently. This method can be extended to other typical channels, providing a new idea for delay and service capability analysis in communication networks.

[0185] (3) Give a general form of the minimum power threshold required to achieve the expected service capability:

[0186] On the basis of delay and reliability constraints, the minimum transmission power expression required to meet the lower limit of service capability is further derived, providing a theoretical basis for power control and energy efficiency optimization, and improving the feasibility and efficiency of communication scheduling under resource limited conditions.

[0187] (4) The LA2C algorithm is proposed, which combines the LSTM-based feature extractor:

[0188] In the reinforcement learning framework, a deep neural network based on LSTM is designed as a feature extractor, which can capture the time dependence and nonlinear association between vehicle state and channel environment, provide a more expressive state representation for policy learning, and improve the learning quality.

[0189] (5) The proposed LA2C algorithm uses SubprocVecEnv to realize multi-environment parallel training:

[0190] Combined with the SubprocVecEnv parallel environment encapsulation mechanism, the sample collection efficiency and concurrent computing capability in the policy training process are significantly improved, the model convergence speed is accelerated, and the training efficiency of the system in complex task scenarios is improved.

[0191] (6) The proposed LA2C algorithm is based on a multi-dimensional heterogeneous action modeling scheme:

[0192] Innovatively, a multi-dimensional heterogeneous action space containing transmission mode, access mode, semantic compression ratio, bandwidth and other control variables is constructed, so that the learning strategy can directly output control decisions to realize unified optimization of joint communication scheduling parameters.

[0193] Figure 3 The structural diagram of the intelligent on-demand service equipment for vehicle networking deterministic delay guarantee provided by the embodiment of the application is shown in Figure 3 The intelligent on-demand service equipment for vehicle networking deterministic delay guarantee can include the intelligent on-demand service device for vehicle networking deterministic delay guarantee shown in Figure 2 Optionally, the intelligent on-demand service equipment for vehicle networking deterministic delay guarantee 410 can include a first processor 2001.

[0194] Optionally, the intelligent on-demand service equipment for vehicle networking deterministic delay guarantee 410 can further include a memory 2002 and a transceiver 2003.

[0195] The first processor 2001, the memory 2002 and the transceiver 2003 can be connected through a communication bus.

[0196] The various constituent components of the intelligent on-demand service equipment for vehicle networking deterministic delay guarantee 410 will be specifically introduced below: Figure 3

[0197] ​The first processor 2001 is a control center of the intelligent on-demand service device 410 for the vehicle networking deterministic latency guarantee, which can be one processor or a collective term of multiple processing elements. For example, the first processor 2001 is one or more central processing units (CPUs), which can also be application specific integrated circuits (ASICs), or one or more integrated circuits configured to implement embodiments of the present application, such as one or more digital signal processors (DSPs), or one or more field programmable gate arrays (FPGAs).

[0198] Optionally, the first processor 2001 can perform various functions of the intelligent on-demand service device 410 for the vehicle networking deterministic latency guarantee by running or executing software programs stored in the memory 2002 and calling data stored in the memory 2002.

[0199] In a specific implementation, as an embodiment, the first processor 2001 can include one or more CPUs, such as the CPU0 and the CPU1 shown in FIG. 2. Figure 3

[0200] In a specific implementation, as an embodiment, the intelligent on-demand service device 410 for the vehicle networking deterministic latency guarantee can also include multiple processors, such as the first processor 2001 and the second processor 2004 shown in FIG. 2. Each of these processors can be a single-CPU or a multi-CPU. The processor here can refer to one or more devices, circuits, and / or processing cores for processing data (such as computer program instructions). Figure 3

[0201] The memory 2002 is used to store software programs for implementing the schemes of the present application, and is controlled by the first processor 2001 to perform, and the specific implementation manner can refer to the above-mentioned method embodiments, which will not be described here again.

[0202] ​​Optionally, the memory 2002 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory 2002 may be integrated with the first processor 2001 or exist independently, and may be accessed via the interface circuit of the intelligent on-demand service device 410 for deterministic latency assurance in the vehicle network. Figure 3 (Not shown in the figure) is coupled to the first processor 2001, and the embodiments of the present invention do not specifically limit this.

[0203] The transceiver 2003 is used to communicate with network devices or with terminal devices.

[0204] Alternatively, transceiver 2003 may include a receiver and a transmitter. Figure 3 (Not shown separately). The receiver is used to implement the receiving function, and the transmitter is used to implement the sending function.

[0205] Optionally, the transceiver 2003 can be integrated with the first processor 2001 or exist independently, and can be connected via the interface circuit of the intelligent on-demand service device 410 for deterministic latency assurance in the Internet of Vehicles. Figure 3 (Not shown in the figure) is coupled to the first processor 2001, and the embodiments of the present invention do not specifically limit this.

[0206] It should be noted that, Figure 3 The structure of the intelligent on-demand service device 410 for deterministic latency assurance in the Internet of Vehicles shown does not constitute a limitation on the router. The actual knowledge structure identification device may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0207] Furthermore, the technical effects of the intelligent on-demand service device 410 for deterministic latency assurance in vehicle-to-everything (V2X) can be referred to the technical effects of the intelligent on-demand service method for deterministic latency assurance in V2X as described in the above method embodiments, and will not be repeated here.

[0208] It should be understood that the first processor 2001 in the embodiments of the present application can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or can also be any conventional processor.

[0209] It should also be understood that the memory in the embodiments of the present application can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically EPROM (EEPROM) or a flash memory. The volatile memory can be a random access memory (RAM) used as an external cache. By way of example but not limitation, many forms of random access memory (RAM) are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced SDRAM (ESDRAM), synchlink DRAM (SLDRAM) and direct RAMBUS (DRAM).

[0210] The above-described embodiments can be implemented in part or in whole through software, hardware (e.g., circuitry), firmware, or any combination thereof. When implemented in software, the above-described embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When loaded and executed by a computer, the computer instructions or computer programs can produce the processes or functions described above in accordance with the embodiments of the present application. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable apparatus. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium, such as from a website, a computer, a server, or a data center to another website, computer, server, or data center through a wired (e.g., infrared, wireless, microwave, etc.) or wireless medium. The computer-readable storage medium can be any available medium or a collection of medium accessible by a computer or a data storage device such as a server, data center, etc. that includes one or more of the available medium. The available medium can be a magnetic medium (e.g., a floppy diskette, a hard disk, a magnetic tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard drive.

[0211] It should be understood that the term "and / or" in this document is merely used to describe associated objects, and can represent three conditions: A and / or B, which can mean that A exists alone, A and B exist together, and B exists alone, where A and B can be singular or plural. In addition, the character " / " in this document generally represents an "or" relationship between the associated objects, but can also represent an "and / or" relationship. The specific meaning can be understood in the context.

[0212] It should be understood that in various embodiments of the present application, the size of the sequence number of each process does not mean the order of execution, and the execution order of each process should be determined by its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0213] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be realized in electronic hardware, or a combination of computer software and electronic hardware. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0214] The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0215] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present alone, or two or more units can be integrated in one unit.

[0216] The functions, if realized in the form of software functional units and sold or used as independent products, can be stored in a computer readable storage medium. Based on such understanding, the technical scheme of the present application or the part of the present application which essentially contributes to the prior art or the part of the technical scheme can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application.

[0217] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for intelligent on-demand service for deterministic latency guarantee in vehicle-to-everything, characterized in that, The method comprises: S1, establishing a deterministic delay model under a multi-mode link; S2, modeling a wireless channel, introducing a stochastic network calculus SNC theory and a Meijer G function, and deducing a general formula of a delay violation probability under different channel conditions; S3, dynamically establishing a transmission power model according to link characteristics, and giving a minimum power threshold value that meets a required channel condition; S4, taking a maximum number of task success vehicles as an optimization objective, and determining constraint conditions; S5, performing policy learning and solving the optimization objective by using an LA2C reinforcement learning algorithm improved based on an A2C mechanism, obtaining a power space output result, and completing intelligent on-demand services for deterministic delay guarantee of a vehicle Internet; In S4, taking a maximum number of task success vehicles as an optimization objective, and determining constraint conditions, comprising: By dynamically deciding a transmission mode, an access mode, a semantic compression ratio and a bandwidth parameter of each vehicle, a number of vehicle users that successfully complete a communication task in the system is maximized, while multiple constraints of system resources and quality of service are met; Taking a maximum number of task success vehicles as an optimization objective, nine constraint conditions are designed; The nine constraint conditions comprise: C1-C3 limit a delay of a vehicle to be less than a threshold and a delay violation probability to be lower than a set threshold from a quality of service perspective, and can be extended to other task indicators related to semantic restoration accuracy or error tolerance; C4-C6 ensure that a total bandwidth of the system, a total transmission power of V2I and a single-link power of V2V are all within a physically available range; C7-C9 discretely constrain control variables of the system, including selection of a transmission mode, an access mode and a semantic compression ratio; In S5, the LA2C reinforcement learning algorithm improved based on the A2C mechanism is used to perform policy learning and solve the optimization objective, comprising: The LA2C reinforcement learning algorithm improved based on the A2C mechanism is used to perform policy learning; The algorithm combines a long short-term memory network LSTM as a feature extractor, and combines a policy-value parallel update mechanism; A multi-dimensional heterogeneous modeling mechanism is used, in which an action is divided into multiple sub-dimensions; Parallel multi-environment simulation is realized by using SubprocVecEnv, so as to speed up algorithm running efficiency; A task success probability is taken as a reward function, and a delay and reliability penalty term are introduced.

2. The method of claim 1, wherein, In S1, the deterministic delay model under the multi-mode link comprises: Deterministic delay components under different communication links and transmission modes are modeled and analyzed; The different communication links comprise a V2I mode and a V2V mode; In the V2I mode and the V2V mode, deterministic delays of semantic users include two parts of communication delays and semantic calculation delays; In the V2V mode, a waiting time for relay establishment also needs to be considered, in which a total delay of a bit user in the V2V mode includes a transmission time and a time required for a relay vehicle to enter a communication range; a semantic user further superimposes a semantic processing time on this basis; The larger value of the semantic calculation delay and the communication establishment delay is included in the analysis.

3. The method of claim 2, wherein, In S2, the wireless channel is modeled, the random network calculus SNC theory and the Meijer G function are introduced, and a general formula of the delay violation probability under different channel conditions is derived, including: By constructing the cumulative arrival process and service process model of the service queue, a general channel model is adopted to establish the mathematical relationship between the data arrival rate, service capability, queue length and delay.

4. The method of claim 3, wherein, In S3, a transmission power model is dynamically established according to the link characteristics, and the minimum power threshold value that meets the required channel conditions is given, including: Two types of channel characteristics are considered comprehensively: one is large-scale path loss, and the other is small-scale channel fading; Among them, the large-scale loss is mainly determined by the distance between the transmission end and the receiving end, and is affected by the path loss exponent and the reference channel gain; the small-scale fading is modeled according to the statistics under the complex scene, and is described uniformly by the generalized α-κ-μ channel model; For the V2I link, the model considers the position change between the service vehicle and the base station and the coverage range of the base station, and calculates the average power consumption from when the vehicle enters the coverage area to when it leaves the area; For the V2V link, the relative motion trajectory between the service vehicle and the relay vehicle is analyzed to determine the relative speed, approach and separation time of the two vehicles within the communication range, and then the communication power consumption is estimated.

5. A smart on-demand service device for V2X deterministic latency guarantee, the smart on-demand service device is used to implement the smart on-demand service method for V2X deterministic latency guarantee according to any one of claims 1-4, characterized in that, The device comprises: A deterministic delay model establishment module for establishing a deterministic delay model under a multi-mode link; A channel modeling module for modeling the wireless channel, introducing the random network calculus SNC theory and the Meijer G function, and deriving a general formula of the delay violation probability under different channel conditions; A transmission power model establishment module for dynamically establishing a transmission power model according to the link characteristics, and giving the minimum power threshold value that meets the required channel conditions; An optimization module for maximizing the number of successful vehicles as the optimization objective and determining the constraint condition; A solving module for performing policy learning and solving the optimization objective by using the LA2C reinforcement learning algorithm improved based on the A2C mechanism to obtain the power space output result and complete the intelligent on-demand service for the vehicle networking deterministic delay guarantee.

6. An intelligent on-demand service device for vehicle networking deterministic delay guarantee, comprising: A processor; A memory having computer readable instructions stored thereon, the computer readable instructions being executed by the processor to implement any one of the intelligent on-demand service methods for vehicle networking deterministic delay guarantee according to any one of claims 1-4.

7. A computer readable storage medium having at least one instruction stored therein, the at least one instruction being loaded and executed by a processor to implement any one of the intelligent on-demand service methods for vehicle networking deterministic delay guarantee according to any one of claims 1-4.

Images