Dynamic unloading and real-time scheduling method based on near earth orbit satellite-ground network

Abstract

The invention requests to protect a dynamic unloading and real-time scheduling method based on a near earth orbit satellite and a ground internet of vehicles. The dynamic unloading and real-time scheduling method is decomposed into a real-time bandwidth allocation and task unloading problem and a joint task unloading and cost ratio optimization problem. According to the invention, rewriting conversion is carried out on the sub-problems 1 and 2, and a greedy algorithm based on utility gain is provided to obtain an optimal channel resource configuration strategy of the first sub-problem. And then, designing a joint iterative optimization algorithm based on Lagrange improvement to obtain an optimal task unloading strategy of the second sub-problem, and obtaining an optimal solution. Utility gain task scheduling evaluation indexes are provided, it is proved that the converted problem has the optimal substructure and the optimal solution, and therefore the algorithm is provided to obtain the optimal solution of the subproblem 1. And for the sub-problem 2, solving the sub-problem 2 based on an improved joint optimization algorithm of Lagrange, and further obtaining an optimal solution of a utility maximization optimization problem by using a KKT condition.

Description

Technical Field

[0001] The present invention relates to a method for joint dynamic offloading and real-time scheduling of near-earth orbit satellites and ground vehicle networks, and particularly to an improved joint optimization scheduling method based on Lagrange. Background Art

[0002] As one of the typical application scenarios, the near-earth orbit satellite network can provide global coverage, high-rate data transmission, stable orbits, and scalability for ground network applications. The low-earth orbit satellite network plays a key role in the real-time communication and computing of vehicle network data, especially in remote areas or dense urban areas with communication obstacles. With the continuous increase in the number of users accessing the satellite communication network, several key issues still need to be addressed in combination with the random characteristics of satellite-ground channels and the dynamic nature of ground access. First, in a traffic-dense vehicle network, timely scheduling and transmission of tasks are particularly important. Although low-earth orbit satellites can provide extensive ground network coverage and efficient edge computing services for vehicle networks, the high mobility of vehicles and low-earth orbit satellites may lead to the complication of real-time task scheduling, as well as frequent channel competition and conflicts, posing significant challenges to efficient data transmission and task computing. Second, roadside units always play a role in task scheduling, collecting the demands of vehicles within the communication range and uploading the processed tasks to the satellite. However, a large number of offloaded tasks result in a surge in data transmission between roadside units and satellites, and the coexistence of multiple roadside units leads to concurrent transmission tasks. Therefore, efficient communication scheduling and bandwidth allocation between multiple roadside units and satellites are crucial. Third, vehicles can choose to offload computationally intensive tasks not only to roadside units but also to satellites for processing. Satellites provide high data processing capabilities and communication bandwidth, significantly improving task processing efficiency, but at high cost. Roadside units have low latency, but due to resource limitations, they cannot process a large number of tasks in a timely manner. How to design an optimal joint dynamic offloading task according to real-time task requirements and network conditions is extremely challenging.

[0003] Therefore, it is necessary to further study the algorithms for joint dynamic offloading and real-time scheduling of low-earth orbit satellites and ground vehicle networks.

[0004] After retrieval, the application publication number is CN115835302A, a task offloading method and device for a near-earth orbit satellite and a ground edge computing network, belonging to the field of wireless communication. The method includes: determining the current state of device offloading allocation and calculating the total system cost of the current state; for each device, iteratively traverse the decision results of processing the task to be processed locally, offloading it to the base station or satellite for processing, and determine the decision result that minimizes the total system cost among all decision results, and select candidate decision results from the decision results of all devices; if the total system cost of the candidate decision result is less than the total system cost of the current state, update the current state; repeat the above update process until the total system cost of the candidate decisions of each device is greater than the current state. The game theory decision method proposed by this method has excellent performance compared with common benchmark methods, and can greatly reduce the system energy consumption and improve the task offloading efficiency.

[0005] The main decision basis of this invention is carried out through system cost. However, in actual user decisions, in addition to cost, factors such as task delay, communication quality, and user satisfaction also affect the decision-making, which are ignored by this invention. And how to balance these factors to make a reasonable decision becomes a challenging problem. To overcome this problem, this invention takes system utility as the evaluation goal. In the construction of the utility model, the task utility gain is constructed as the utility sensitivity index, considering the task satisfaction of users including delay factors, and it is used as an important evaluation factor.

[0006] Secondly, this invention ignores the link change problem caused by user movement and satellite movement. This problem not only affects the offloading decision but also affects the success of communication between users and the base station or satellite. This invention fully considers the movement problems of vehicle users and satellites, and the decision algorithm is also carried out dynamically and in real time. During the movement of vehicle users, a movement delay is generated. This invention constructs a movement delay model, as Figure 3 shown, which is a geometric relationship diagram of vehicle user movement and network coverage. In addition, to solve the problem of unstable links caused by the inability of satellites to continuously cover ground users due to satellite movement, this invention proposes a coverage delay model and an inter-satellite link switching model. When a user's task is offloaded to a satellite and faces the problem of unable to be continuously covered, the satellite selects the nearest satellite to itself for inter-satellite link transmission to ensure the reliability of communication, and at the same time considers the impact of the resulting delay on user satisfaction. Based on these evaluations, a decision that maximizes the system utility is made, as Figure 2 shown as the inter-satellite link switching schematic diagram.

[0007] Finally, the cost of the present invention mainly refers to the evaluation index of energy consumption. However, in practical applications, there is not only this one cost. To more comprehensively evaluate the cost of the system, the present invention also considers the link communication cost and channel occupancy cost between the roadside unit and the satellite, and proposes the definition of the cost rate to dynamically evaluate the relationship between the cost rate and the offloading rate, and proposes a cost incentive and service pricing strategy for the satellite-ground collaborative network. As Figure 7 shown, it is a diagram showing the influence relationship between the offloading rate and the cost rate with the change of the task size. Summary of the Invention

[0008] The present invention aims to solve the above problems of the prior art. A dynamic offloading and real-time scheduling method based on a low-earth orbit satellite-ground network is proposed. The technical solution of the present invention is as follows:

[0009] A dynamic offloading and real-time scheduling method based on a low-earth orbit satellite-ground network, which includes the following steps:

[0010] Step 1, construct a system model to determine the communication and offloading, delay, and system utility models;

[0011] Step 2, analyze and transform the system utility maximization problem in Step 1;

[0012] Step 3, use the greedy algorithm based on utility gain and the improved joint iterative optimization algorithm based on Lagrangian to solve the sub-problems in Step 2. The improvement lies in that the algorithm first derives the upper and lower bounds of the offloading rate, thus greatly narrowing the search range of the optimal solution; secondly, in the face of the optimization problem and constraint conditions, the KKT condition is used to find the optimal solution of sub-problem two; finally, the global optimal task offloading strategy and the optimal resource allocation strategy are obtained by combining the UGG algorithm again.

[0013] Further, the construction of the system model in Step 1 specifically includes:

[0014] Construct a system model, which includes M low-earth orbit (LEO) satellites, R roadside units (RSUs) on the ground, and V dynamically moving vehicles, equipped with single antennas and driving on a multi-lane two-way road. In addition, the ground station can collect the LEO satellite status information, RSU status information, vehicle service requests, and the channel status information between the LEO satellite and the RSU, and between the RSU and the vehicle, and promotes the LEO satellite-ground collaborative scheduling. The communication between the ground vehicle and the RSU is assumed to use non-orthogonal multiple access technology and successive interference cancellation technology. When vehicle i transmits the generated task to RSU j at time slot t, the channel gain between vehicle i and RSU j is:

[0015]

[0016] Wherein, and represent small time scale channel fading and large time scale channel fading respectively, d i,j (t) represents the distance between vehicle i and RSU j in time slot t, and α is the path loss factor. In addition, the present invention considers the interference of other vehicles in the channel on the current vehicle signal. First, the gains of multiple vehicles are sorted in descending order, that is, m refers to the number of vehicles connected to RSU j in time slot t. According to this sequential decoding, vehicles with higher channel gains will be interfered with by other vehicles with relatively lower channel gains. In time slot t, the data upload rate between vehicle i and RSU j is:

[0017]

[0018] Among them, the variable is the task k assigned by RSU to vehicle i i The bandwidth of (t) is variable is the data transmission power of vehicle i at time slot t, σ 2 is the noise power; represents the data transmission power of vehicle n in the same channel at time slot t, and n∈[i+1,m].

[0019] LEO satellite-ground communication uses orthogonal frequency division multiple access technology. RSU transmits tasks to the satellite through the Ka-band wireless backhaul link. RSU can access at most one satellite in one time slot. Each satellite allows a limited number of links within its coverage area. The LEO satellite orbit is pre-planned, so the ground station can obtain the LEO satellite S m The height, speed and position of RSU j and satellite S in time slot t. m The channel gain between is:

[0020]

[0021] Among them, the variable It is satellite S m The distance to RSU j, is the Rician decay factor, is the path loss factor. At time slot t, RSU j and satellite S m The data upload rate between is:

[0022]

[0023] Among them, the variable represents the allocatable bandwidth of the satellite to RSU j, and the variable Indicates interference, and They represent the transmission from RSU j to LEO satellite S in time slot t.m The transmit power and noise power. In order to effectively allocate the bandwidth resources between LEO satellites and the ground network, the total bandwidth constraint needs to be satisfied: The satellite movement coverage time factor is considered. Specifically, for satellite S m The coverage time is Where And Represent the radius of the earth and the height of LEO satellite S from the ground respectively, and β m Is the coverage angle of LEO satellite S m,j Is m The coverage angle of LEO satellite S.

[0024] Furthermore, the construction of the communication model in step 1 specifically includes:

[0025] When the delay of satellite S m Processing the task is less than the coverage time The link between satellite S m And RSU j remains unchanged, and the task result is directly fed back to RSU j; however, when the delay of satellite S m Processing the task exceeds the coverage time The satellite S m Will send the task to the adjacent satellite S m+1 That currently covers RSU j through the inter-satellite link; satellite S m+1 Completes the task calculation and feeds back the result to RSU j; the switching variable is represented as When the satellite link needs to be switched, Otherwise A communication model of the inter-satellite link is constructed. The data transmission rate from satellite S m To satellite S m+1 Is:

[0026]

[0027] Where, Represents the transmit power of LEO satellite S m ; The symbols And Represent the transmit antenna gain of satellite S m And the receive antenna gain of satellite S m+1 Respectively; the symbol ε is the Boltzmann constant, and T S Represents the molar temperature; the variable E S Represents the received energy per bit required relative to the noise density, and M S Represents the link margin; the free space loss is represented as l S (t), and the expression is V Ldenotes the speed of light, and F denotes the communication frequency of the inter-satellite link. In addition, L m,m+1 (t) is the slant range between satellite S m and S m+1 .

[0028] Furthermore, in step 1, the system delay is modeled, which includes two cases:

[0029] Case 1: When RSU j receives a task, it uses parallel processing to complete the task calculation; thus, for task k i (t), the delay includes the transmission delay of task k i (t) transmitted from vehicle i to RSU j and the calculation delay of RSU j processing task k i (t). The transmission and calculation delays are calculated as follows:

[0030]

[0031] where z i (t) is the task size, and when RSU j processes task k i (t), c i represents the CPU cycles required for each bit, represents the CPU frequency that RSU j can allocate to task k i (t);

[0032] Case 2: The delay includes the transmission delay from vehicle i to RSU j, the transmission delay from RSU j to the LEO satellite, and the calculation delay of the LEO satellite; the transmission delay from RSU j to the LEO satellite S m and the calculation delay of the LEO satellite can be expressed as:

[0033]

[0034] where represents the CPU frequency that the LEO satellite can allocate to task k i (t) in the t time slot; the delay includes the transmission delay of the task from the current satellite to the target satellite, considering the satellite link handover model:

[0035]

[0036] is the waiting delay when the current task faces link handover, which is transmitted according to the FIFO order and calculated according to the M / G / 1 queuing model, and μ i,j(m,m+1) represents the task transmission intensity from the LEO satellite S m to S m+1 , represents from satellite Sm to S m+1 The average transmission delay, δ, of the task 2 represents the variance of the transmission delay; The transmission delay from the LEO satellite S m to S m+1 is calculated as follows:

[0037]

[0038] The LEO satellite S m+1 calculates the computation delay of task k i (t) as:

[0039]

[0040] To ensure that the vehicle task is completed within the current RSU coverage area, the mobility delay is defined as the time from the current position of vehicle i within the coverage area of RSU j to leaving this position, and the mobility delay is specifically expressed as:

[0041]

[0042] where the variable represents the position of vehicle i at time slot t, while represents the boundary position where vehicle i leaves the coverage area of RSU j. The variable v i (t) is the current speed of vehicle i. The symbols α and L represent the direction of vehicle i and the coverage radius of RSU j respectively, and the operator ||·|| represents the Euclidean distance between two positions; Therefore, for task k i (t), it should satisfy

[0043] Furthermore, step 1 models the system utility, specifically including:

[0044] Considering the satisfaction factor of the vehicle, a vehicle utility model is established:

[0045]

[0046] where the expression [x] + ensures that the satisfaction factor is non - negative. The variable γ i,j is the satisfaction factor parameter, represents the energy cost generated when the vehicle transmits the task to RSU j, while represents the task computation cost and channel occupancy cost paid by vehicle i to RSU j. The variable represents the average cost expenditure of energy consumption, is the average cost expenditure per CPU cycle, and the variable It represents the cost expenditure required for each bit of the transmission task to occupy the channel; at the same time, considering the utility of the network operator, it is calculated as follows:

[0047]

[0048] Among them, represents the revenue obtained from vehicle i, and the variable represents the cost of processing task k i (t) at the RSU, which is mainly composed of the energy consumption of task calculation. The variables and represent the costs of task calculation and task transmission to satellite transmission respectively;

[0049] The optimization objective is to maximize the system utility, and the optimization problem is formulated as follows:

[0050]

[0051] Constraint (1) indicates that the mobile vehicle is associated with the RSU; Constraint (2) stipulates that the number of communication links provided by the LEO satellite should not exceed the maximum allowable number; Constraints (3) and (4) respectively point out that the bandwidth allocation from the vehicle to the RSU and from the RSU to the satellite should not exceed the total available bandwidth in the current time slot; Constraint (5) is a delay constraint to ensure the timely transmission and processing of tasks; Constraints (6) and (7) define the value range of the binary variable; Constraint (8) represents the value range of the unit cost variable.

[0052] Furthermore, in step 2: Analyze and transform the system utility maximization problem in step 1, specifically including:

[0053] The optimization problem described in step 1 is an NP-hard problem and thus cannot be directly solved. The utility maximization problem in step 1 is decomposed into two sub-problems: Sub-problem 1 is the real-time bandwidth allocation and task offloading problem, and Sub-problem 2 is the joint task offloading and cost ratio optimization problem.

[0054] Furthermore, in step 3: Use the greedy algorithm based on utility gain and the improved Lagrangian optimization algorithm theory to solve the sub-problems in step 2, specifically including:

[0055] First, rewrite and transform the optimization problem, and propose a UGG algorithm to obtain the optimal channel resource allocation strategy for the first sub-problem; then, design an improved joint optimization algorithm based on Lagrangian to obtain the optimal task offloading strategy for the second sub-problem and obtain the optimal solution.

[0056] ​

[0057] Step 1: Obtain the status of all vehicle users in the research area of the current time slot, including moving speed, direction, and location, obtain the status of the currently covered satellites, including moving speed and location, obtain the task flow of the current moving vehicle and the link status with the roadside unit RSU, as well as the allocable resource status of the roadside unit.

[0058] Step 2: Evaluate the delay requirements for each task to be processed in the current time slot, including the calculation of the moving delay and the coverage delay .

[0059] Step 3: Evaluate the cost model under different decisions and the cost generated by transmitting the task to the RSU and having the RSU complete it, the cost generated by offloading the task to the satellite.

[0060] Step 4: Calculate the utility gain ΔU′ of the current task according to the utility gain evaluation index proposed by the present invention i,j (t) and assign it to the evaluation value V = ΔU′ i,j (t).

[0061] Step 5: Arrange the evaluation values V of all tasks received by the current RSU in the current time slot in descending order and execute them to form a sequence of V.size(), and then sequentially execute the utility gain offloading preference theorem proposed by the present invention for the tasks in the sequence. If and satisfy the constraints C3 and C5, then and φ represents the decision of task k i (t), and when executing this decision, it satisfies the constraints C3 and C5. According to the decision set defined by the present invention Execute the decision: φ u = φ u ∪{φ} and complete the corresponding channel resource allocation. Otherwise, according to the decision set defined by the present invention Execute the decision: φ s = φ s ∪{φ} and complete the corresponding channel resource allocation.

[0062] Step 6: According to Step 5, obtain the optimal channel resource allocation strategy

[0063] Furthermore, an improved joint optimization algorithm based on Lagrange is designed to obtain the optimal task offloading strategy for the second sub-problem and obtain the optimal solution. Specifically, it includes:

[0064] Step 1: Within a single time slot, update the vehicle status of the current time slot, including moving speed, direction, and position; update the status of the current coverage satellite, including moving speed and position; obtain the task flow of the current moving vehicle and the link status with the roadside unit (RSU) and the allocatable resource status of the roadside unit; and update the tolerable delay of the current task.

[0065] Step 2: Combine the UGG algorithm to derive the lower bound of the offload rate of the current task flow and upper bound

[0066] Step 3: Convert the optimization problem, replace the cost variable with the cost rate, and convert the problem into an optimization convex problem with the unloading rate and cost rate as variables.

[0067] Step 4: Use KKT conditions to solve it and get the optimal unloading rate ε * and cost rate δ * .

[0068] Step 5: Combine the fourth and fifth steps in UGG to obtain the global optimal solution of the optimization problem.

[0069] The advantages and beneficial effects of the present invention are as follows:

[0070] This paper constructs a satellite-ground collaborative dynamic scheduling framework for efficient, real-time resource allocation between ground and satellite networks for multiple dynamic vehicle users. To ensure optimal system utility, the present invention first decomposes the system utility maximization problem into two subproblems: real-time bandwidth allocation and task offloading, and joint task offloading and cost ratio optimization. Furthermore, the present invention rewrites and transforms subproblems 1 and 2, proposing a greedy algorithm based on utility gain to obtain the optimal channel resource allocation strategy for the first subproblem. Then, an improved joint optimization algorithm based on Lagrangian finite element theory is designed to obtain the optimal task offloading strategy for the second subproblem, resulting in an optimal solution. The present invention proposes a utility gain task scheduling evaluation metric and proves that the transformed problem has an optimal substructure and optimal solution, thereby proposing an algorithm to obtain the optimal solution to subproblem 1. For subproblem 2, the dual optimization of offloading rate and cost ratio makes the problem more complex. The present invention solves it based on an improved joint optimization algorithm based on Lagrangian finite element theory and applies the KKT condition to obtain the optimal solution to the utility maximization optimization problem. Experimental results demonstrate that the present invention can achieve effective satellite-ground collaborative dynamic real-time scheduling and is highly efficient in terms of system utility. The present invention provides a new method for dynamic unloading and real-time scheduling applied to low-Earth orbit satellites and ground vehicle networks.

[0071] The main innovation of the present invention lies in that, firstly, comprehensively considering the vehicle task flow, mobility, real-time resource status of roadside units and satellite networks, task utility sensitivity, vehicle user satisfaction, as well as transmission and offloading costs, a system model based on the cooperation of low Earth orbit satellite-ground networks is constructed. Secondly, the present invention fully considers the mobility of vehicle users and satellites, and makes decisions dynamically and in real time. During the movement of vehicle users, mobile delay is generated, and the present invention constructs a mobile delay model. In addition, to solve the problem of unstable links caused by the inability of satellites to continuously cover ground users due to satellite movement, the present invention proposes a coverage delay model and an inter-satellite link handover model. When a user task is offloaded to a satellite and faces non-continuous coverage, the satellite selects the satellite closest to itself for inter-satellite link transmission to ensure communication reliability, and at the same time considers the impact of the resulting delay on user satisfaction, and makes a decision to maximize the system utility based on these evaluations. Thirdly, based on the link communication cost, channel occupancy cost, and task calculation cost between roadside units and satellites, the present invention proposes the definition of a cost rate, dynamically evaluates the relationship between the cost rate and the offloading rate, and proposes a cost incentive and service pricing strategy for the satellite-ground collaborative network. Fourthly, the problem proposed by the present invention belongs to an NP-hard problem. The present invention transforms and derives the problem, decomposes it into two sub-problems, solves the first sub-problem by designing a greedy algorithm based on utility gain to obtain the optimal uplink bandwidth allocation strategy between roadside units and satellites. The second sub-problem is solved by designing an improved joint optimization algorithm based on Lagrange to obtain an efficient offloading strategy with optimal system utility. Finally, using a real dataset in Shanghai (China) to evaluate the offloading strategy and system utility, compared with representative schemes, the experimental results show that the method of the present invention has superiority in multiple indicators. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 FIG. is a system model diagram of a dynamic real-time collaborative network of low-orbit satellites and ground vehicle networks provided by the preferred embodiment of the present invention.

[0073] Figure 2 FIG. is a schematic diagram of inter-satellite link handover for a low-orbit satellite to cover a ground network.

[0074] Figure 3 FIG. is a schematic diagram of the dynamic network communication and position relationship between multiple vehicle users and roadside units.

[0075] Figure 4 FIG. shows the performance of the IJIA algorithm designed by the present invention and five other baseline algorithms in terms of system average utility under different numbers of roadside unit deployments.

[0076] Figure 5 FIG. shows the performance of the IJIA algorithm designed by the present invention and five other baseline algorithms in terms of system average utility under different CPU computing frequencies of roadside units.

[0077] Figure 6 Performance of the IJIA algorithm designed for the present invention and five other baseline algorithms in terms of system average utility at different task sizes.

[0078] Figure 7 The relationship between the offloading rate and cost rate of the IJIA algorithm designed for this invention under different task size changes. DETAILED DESCRIPTION

[0079] The following will describe the technical solutions in the embodiments of the present invention in detail with reference to the accompanying drawings. The described embodiments are only a part of the embodiments of the present invention.

[0080] The technical solution of the present invention to solve the above technical problems is:

[0081] A joint optimization real-time scheduling algorithm based on utility gain has the following steps:

[0082] Step 1: Build a system model to determine the communication and offloading, latency, and system utility models.

[0083] Step 2: Analyze and transform the system utility maximization problem in step 1).

[0084] Specifically, we first rewrite the goal of the original problem as

[0085]

[0086] This paper proposes utility gain as an evaluation metric and decomposes the original problem into subproblems 1 and 2 based on the rewritten objective. Subproblem 1 is about optimizing real-time bandwidth allocation and task offloading, while subproblem 2 is about optimizing joint task offloading and cost ratio. Since the second term in the objective function U does not affect the solution of the optimization problem, this paper formulates the optimization problem of subproblem 1 as follows:

[0087]

[0088] In addition, the optimization problem of subproblem 2 is formulated as:

[0089]

[0090] Step 3: Use the greedy algorithm based on utility gain and the improved Lagrangian optimization algorithm theory to solve the subproblems in step 2).

[0091] Subproblem 1 aims to maximize system utility by allocating transmission bandwidth and optimizing computational offloading of vehicle tasks. However, problem P1 is still a mixed integer nonlinear programming problem and is difficult to solve. To address the above issues, we first propose a task scheduling evaluation metric, specifically:

[0092]

[0093] Among them, represents the latency of task offloading to the roadside unit, while the latency of task offloading to the satellite is represented as For the tasks of vehicle users, their utility gains can be derived through the above evaluation metrics, and the present invention proves that for any two vehicle users in the t time slot, preferentially offloading the tasks with higher utility gains to the side with higher gains can obtain a higher utility for the system than offloading the tasks with relatively lower utility gains once, that is: In addition, to facilitate solving the optimal offloading strategy, the present invention defines the task offloading set as φ = {φ u , φ s}, where

[0094]

[0095] represent the task set processed locally by the roadside unit and the task set offloaded to the low-earth orbit satellite for processing, respectively. Therefore, to solve Sub-problem 1, the present invention transforms the optimization problem into:

[0096]

[0097] In addition, the present invention also proves that Sub-problem 1 has an optimal substructure and an optimal solution. Therefore, the UGG algorithm is proposed to solve it. The specific algorithm pseudo-code is shown in Table 1.

[0098]

[0099]

[0100] The present invention proposes an offloading rate variable ε j (t) ∈ [0, 1] and a cost rate variable where and the cost rate are used as variables, and based on Algorithm 1, the upper and lower bounds of the cost rate are derived, that is:

[0101]

[0102] Among them, the variable The symbol The specific expression of is:

[0103]

[0104] Therefore, the present invention further transforms Sub-problem 2 into the following optimal problem form:

[0105]

[0106] We use To express the optimization objective, that is:

[0107]

[0108] Apply KKT conditions to find the optimal solution Then it satisfies:

[0109]

[0110] in represents a non-negative Lagrange multiplier. The symbol K = 4, and as well as

[0111] The designed IJIA algorithm is shown in Table 2.

[0112]

[0113]

[0114] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0115] The above embodiments should be understood as merely illustrating the present invention and not as limiting the scope of protection of the present invention. After reading the contents of the present invention, technicians may make various changes or modifications to the present invention, and these equivalent changes and modifications also fall within the scope defined by the claims of the present invention.

Claims

1. A dynamic offloading and real-time scheduling method based on a low Earth orbit satellite-ground network, characterized in that It includes the following steps: Step 1: Construct a system model to determine the communication and offloading, latency, and system utility models; Step 2: Analyze and transform the system utility maximization problem in Step 1; Step 3: Use the utility gain-based greedy algorithm and the improved Lagrangian-based joint iterative optimization algorithm to solve the sub-problems in Step 2. The improvement lies in that the algorithm first derives the upper and lower bounds of the offloading rate, thus greatly reducing the search range of the optimal solution; secondly, in the face of the optimization problem and constraints, the KKT conditions are used to find the optimal solution of sub-problem two; finally, the global optimal task offloading strategy and the optimal resource allocation strategy are obtained by combining with the UGG algorithm again.

2. The dynamic offloading and real-time scheduling method based on a low Earth orbit satellite-ground network according to claim 1, wherein The specific construction of the system model in Step 1 includes: Construct a system model that includes M low Earth orbit (LEO) satellites, R roadside units (RSUs) on the ground, and V dynamically moving vehicles, equipped with single antennas and driving on multi-lane two-way roads. In addition, the ground station can collect the state information of LEO satellites, the state information of RSUs, vehicle service requests, and the channel state information between LEO satellites and RSUs, and between RSUs and vehicles, and promotes LEO satellite-ground cooperative scheduling. The communication between ground vehicles and RSUs is assumed to use non-orthogonal multiple access technology and successive interference cancellation technology. When vehicle i transmits the generated task to RSU j at time slot t, the channel gain between vehicle i and RSU j is: Among them, and respectively represent the small - time - scale channel fading and the large - time - scale channel fading, d i,j (t) represents the distance between vehicle i and RSU j at time slot t, and α is the path - loss factor; in addition, the present invention considers the interference of other vehicles in the channel on the signal of the current vehicle. First, the gains of multiple vehicles are sorted in descending order, that is m refers to the number of vehicles connected to RSU j at time slot t. Decoding according to this order, vehicles with higher channel gains will be interfered by other vehicles with relatively lower channel gains. At time slot t, the data upload rate between vehicle i and RSU j is: Among them, the variable is the task k assigned by the RSU to vehicle i i (t) bandwidth, and the variable is the data transmission power of vehicle i in time slot t, and σ 2 is the noise power; represents the data transmission power of vehicle n in time slot t in the same channel, and n ∈ [i + 1, m]. LEO satellite-ground communication uses orthogonal frequency division multiple access technology. The RSU transmits tasks to the satellite through the Ka-band wireless backhaul link. The RSU can access at most one satellite within a time slot, and each satellite allows a limited number of links within its coverage area. The LEO satellite orbit is pre-planned, so the ground station can obtain the altitude, speed, and position of LEO satellite S m ; at time slot t, the channel gain between RSU j and satellite S m is as follows: Among them, the variable is the distance from satellite S m to RSU j, is the Rician fading factor, is the path loss factor. At time slot t, the data upload rate between RSU j and satellite S m is as follows: Among them, the variable represents the allocable bandwidth of the satellite for RSU j, and the variable represents interference. and represent the transmit power and noise power from RSU j to LEO satellite S m in time slot t respectively. In order to effectively allocate the bandwidth resources between LEO satellites and the ground network, the total bandwidth constraint needs to be satisfied: The satellite moving coverage time factor is considered. Specifically, the coverage time of satellite S m is where and represent the radius of the earth and the altitude of LEO satellite S m from the ground respectively, and β m,j is the coverage angle of LEO satellite S m .

3. The dynamic offloading and real-time scheduling method based on a low Earth orbit satellite-ground network according to claim 2, wherein The specific construction of the communication model in Step 1 includes: When satellite S m The time delay for processing the task is less than the coverage time At this time, satellite S m The link between satellite S and RSU j remains unchanged, and the task result is directly fed back to RSU j; however, when satellite S m The time delay for processing the task exceeds the coverage time At this time, satellite S m Will send the task to the adjacent satellite S that currently covers RSU j via the inter-satellite link m+1 Satellite S m+1 Completes the task calculation and feeds back the result to RSU j; represents the handover variable as When satellite link handover is required Otherwise A communication model of the inter-satellite link is constructed. The data transmission rate from satellite S m To satellite S m+1 Is as follows: Among them, represents the transmission power of the LEO satellite S m ; the symbol and respectively represent the transmitting antenna gain of the satellite S m and the receiving antenna gain of the satellite S m+1 ; the symbol ε is the Boltzmann constant, T S represents the molar temperature; the variable E S represents the received energy per bit relative to the noise density, M S represents the link margin; the free space loss is expressed as l S (t), and the expression is V L represents the speed of light, and F represents the communication frequency of the inter-satellite link. In addition, L m,m+1 (t) is the slant range between the satellite S m and S m+1 .

4. The dynamic offloading and real-time scheduling method based on a low Earth orbit satellite-ground network according to claim 3, characterized in that The system latency in Step 1 is modeled, which includes two cases: Case 1: When RSU j receives a task, it uses parallel processing to complete the task calculation; therefore, for task k i (t), the delay includes the transmission of task k from vehicle i to RSU j i (t) transmission delay and RSU j processing task k i The transmission and computation delays are calculated as follows: where z i (t) is the task size, when RSU j processes task k i (t), c i represents the CPU cycles required for each bit, represents the CPU frequency that RSU j can allocate to task k i (t); Case 2, the time delay includes the transmission delay from vehicle i to RSU j, the transmission delay from RSU j to the LEO satellite, and the computing delay of the LEO satellite; the transmission delay from RSU j to LEO satellite S m The transmission delay and the computing delay of the LEO satellite can be expressed as: Among them, indicates that the LEO satellite can be allocated to task k at time slot t i (t) is the CPU frequency; the time delay includes the transmission delay of the task from the current satellite to the target satellite, and the satellite link switching model is considered: is the waiting delay when the current task faces link switching. This is calculated according to the FIFO order of transmission and the M / G / 1 queuing model, μ i,j(m,m+1) represents the task transmission intensity from the LEO satellite S m to S m+1 ; represents the average transmission delay of the task from satellite S m to S m+1 ; δ 2 represents the variance of the transmission delay; the transmission delay from the LEO satellite S m to S m+1 is calculated as follows: LEO satellite S m+1 Computing task k i (t) The computing delay is as follows: To ensure that vehicle tasks are completed within the current RSU coverage area, the mobile delay is defined That is, the time from the current position of vehicle i within the coverage area of RSU j to leaving that position. The mobile delay is specifically expressed as: Among them, the variable represents the position of vehicle i at time slot t, while represents the boundary position where vehicle i leaves the coverage area of RSU j. The variable v i (t) is the current speed of vehicle i. The symbols α and L represent the direction of vehicle i and the coverage radius of RSU j respectively. The operator ||·|| represents the Euclidean distance between two positions. Therefore, for task k i (t), it should satisfy 5. The dynamic offloading and real-time scheduling method based on the low-earth orbit satellite-ground network according to claim 4, wherein The system utility in Step 1 is modeled, specifically including: Considering the satisfaction factor of the vehicle, a vehicle utility model is established: Among them, the expression [x] + ensures that the satisfaction factor is non - negative, and the variable γ i,j is the satisfaction factor parameter. represents the energy cost generated when the vehicle transmits the task to RSU j, while represents the task calculation cost and channel occupancy cost paid by vehicle i to RSU j. The variable represents the average cost expenditure of energy consumption. is the average cost expenditure per CPU cycle, and the variable represents the cost expenditure required for each bit of the transmitted task to occupy the channel; At the same time, considering the utility of the network operator, it is calculated as follows: Among them, represents the revenue obtained from vehicle i, and the variable represents the cost of processing task k at the RSU i (t), which is mainly composed of the energy consumption for task calculation. The variables and represent the costs of task calculation and task transmission to satellite transmission, respectively; The optimization objective is to maximize the system utility, and the optimization problem is formulated as follows: Constraint (1) indicates that the moving vehicle is associated with the RSU; Constraint (2) stipulates that the number of communication links provided by the LEO satellite should not exceed the maximum allowed number; Constraints (3) and (4) respectively point out that the bandwidth allocation from the vehicle to the RSU and from the RSU to the satellite should not exceed the total available bandwidth in the current time slot; Constraint (5) is a latency constraint to ensure the timely transmission and processing of tasks; Constraints (6), (7), and (8) define the value range of the binary variables; Constraint (9) represents the value range of the unit cost variable.

6. The dynamic offloading and real-time scheduling method based on a low Earth orbit satellite-ground network according to claim 5, wherein Step 2: Analyze and transform the system utility maximization problem in Step 1, specifically including: The optimization problem described in Step 1 is an NP-hard problem and cannot be directly solved. The utility maximization problem in Step 1 is decomposed into two sub-problems: Sub-problem 1 is the real-time bandwidth allocation and task offloading problem, and Sub-problem 2 is the joint task offloading and cost ratio optimization problem.

7. The dynamic offloading and real-time scheduling method based on the low Earth orbit satellite-ground network according to claim 6, characterized in that, Step 3: Use the utility gain-based greedy algorithm and the improved Lagrangian optimization algorithm theory to solve the sub-problems in Step 2, specifically including: First, the optimization problem is rewritten and transformed, and a UGG algorithm is proposed to obtain the optimal channel resource allocation strategy for the first sub-problem. Then, an improved joint optimization algorithm based on Lagrange is designed to obtain the optimal task offloading strategy for the second sub-problem and obtain the optimal solution.

8. The dynamic offloading and real-time scheduling method based on a low Earth orbit satellite-ground network according to claim 7, characterized in that The proposed UGG algorithm to obtain the optimal channel resource allocation strategy for the first sub-problem specifically includes: Step 1: Obtain the states of all vehicle users in the research area of the current time slot, including moving speed, direction, and position, obtain the state of the currently covered satellite, including moving speed and position, obtain the task flow of the current moving vehicle and the link state with the roadside unit RSU, and the allocable resource state of the roadside unit. Step 2: Evaluate the delay requirements for each task to be processed in the current time slot, including the calculation of mobile delay and coverage delay calculation. Step 3: Evaluate the cost model under different decisions and is the cost generated when the task is transmitted to the RSU and completed by the RSU, is the cost generated when the task is offloaded to the satellite. Step 4: Calculate the utility gain ΔU′ of the current task according to the utility gain evaluation index proposed by the present invention i,j (t) and assign it to the evaluation value V = ΔU′ i,j (t). Step 5: Evaluate and sort all the tasks received by the current RSU in the current time slot in descending order of the evaluation value V to form a sequence of V.size(), and then sequentially execute the utility gain offloading preference theorem proposed by the present invention on the sequence tasks. If and the constraints C3 and C5 are satisfied, then and φ represents the decision of task k i (t), and when this decision is executed, the constraints C3 and C5 are satisfied. According to the decision set defined by the present invention Execute the decision: φ u = φ u ∪{φ} and complete the corresponding channel resource allocation. Otherwise, according to the decision set defined by the present invention Execute the decision: φ s = φ s ∪{φ} and complete the corresponding channel resource allocation. Step 6: Obtain the optimal channel resource allocation strategy according to Step 5 9. The dynamic offloading and real-time scheduling method based on a low Earth orbit satellite-ground network according to claim 7, wherein An improved joint optimization algorithm based on Lagrange is designed to obtain the optimal task offloading strategy for the second sub-problem and obtain the optimal solution. Step 1: Within a single time slot, update the vehicle state of the current time slot, including moving speed, direction, and position, update the state of the currently covered satellite, including moving speed and position, obtain the task flow of the current moving vehicle and the link state with the roadside unit RSU, and the allocable resource state of the roadside unit, and update the current task's tolerable delay. Step 2: Combine the UGG algorithm to derive the lower bound of the offloading rate of the current task flow and the upper bound Step 3: Transform the optimization problem, equivalently replace the cost variable with the cost rate, and transform the problem into an optimization convex problem with the offloading rate and cost rate as variables. Step 4: Use the KKT conditions to solve it and obtain the optimal offloading rate ε * and the cost rate δ * . Step 5: Combine the fourth and fifth steps in UGG to obtain the global optimal solution of the optimization problem.

Images