Energy-consumption-optimized satellite-ground collaborative unloading method for low-orbit satellite internet of things

Abstract

The invention relates to a low earth orbit satellite Internet of Things satellite-ground collaborative unloading method with optimized energy consumption. The method comprises the following steps: constructing an LEO satellite MEC network model and initializing system parameters; calculating a system energy consumption sum of user terminal UE calculation, idle terminal AE unloading calculation and LEO satellite MEC unloading calculation, and establishing an optimization problem; the whole problem is divided into five sub-problems, namely user matching, unloading decision, unloading proportion, and computing resource allocation of AE and LEO. Solving a user matching sub-problem by adopting a maximum weight matching algorithm; solving an unloading decision problem by adopting a relaxation and rounding technology; solving an AE and LEO satellite resource allocation scheme by adopting a Lagrange multiplier method and a KKT condition; and the optimal unloading decision, the optimal unloading proportion and the computing resource allocation of the AE and the LEO are obtained through alternate iterative optimization. The method has the remarkable advantages that the energy consumption problem existing in the satellite communication network can be effectively relieved, and the excellent performance is shown in the aspect of energy consumption reduction through the innovative technical means and the unique design thought.

Description

Technical Field

[0001] The present invention belongs to the technical field of satellite communications and relates to a low-orbit satellite Internet of Things satellite-to-ground collaborative unloading method with optimized energy consumption. Background Art

[0002] In recent years, with the rapid development of the Internet of Things (IoT) and big data technology, satellite edge computing, as an emerging computing paradigm, has gradually attracted widespread attention. At present, satellite edge computing also brings huge potential for applications in aerospace, intelligent transportation, agriculture and other fields, and provides new technical support for data collection, analysis and application in these fields. However, although SEC has made significant progress in many aspects, there are still some problems that need to be solved in existing research and applications. First, the existing SEC system does not consider the energy consumption problem enough. In a satellite environment with limited resources, how to effectively reduce energy consumption is a key challenge. High energy consumption will not only increase the operating cost of the system, but also shorten the service life of the equipment and affect the stability and reliability of the system. Secondly, most of the existing research focuses on the coordination of computing resources between satellites and ground equipment or introduces high altitude platforms (HAPs) and unmanned aerial vehicles (UAVs) into the ground-satellite architecture, etc., without fully considering the collaborative computing potential of idle satellite terminals. As valuable computing resources, idle satellite terminals, if fully utilized, will significantly improve the computing power and communication efficiency of low-orbit satellite Internet of Things.

[0003] However, in order to further improve the performance and application potential of ground-satellite networks, the academic community has begun to explore a new paradigm of satellite-ground collaborative computing. Although SEC can transcend geographical limitations and provide reliable communication and computing services to ground terminals compared to MEC, the communication and computing load of LEO satellites is heavy and energy is limited, requiring ground terminals to assist in mission calculations. In remote or extreme environments, such as deserts, polar regions, or oceans, a large number of sensors are deployed for environmental monitoring. These devices collect large amounts of data in real time and transmit and process them through LEO satellites. These terminal devices are equipped with solar panels and are charged by solar energy to extend their working time and ensure continuous operation. By utilizing the computing resources of idle ground terminals, the communication and computing burden of LEO satellites can be reduced, and the overall energy consumption of low-orbit satellite networks can also be reduced. Summary of the invention

[0004] In view of this, the purpose of the present invention is to provide an energy-optimized low-orbit satellite Internet of Things satellite-to-ground collaborative unloading method, which can effectively alleviate the energy consumption problem existing in the satellite communication network, and show excellent performance in reducing energy consumption through innovative technical means and unique design ideas.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A low-orbit satellite Internet of Things satellite-ground collaborative unloading method with energy consumption optimization comprises the following steps:

[0007] Step S1: Construct a satellite-ground collaborative LEO satellite network model, including working terminals, idle terminals, and LEO satellites; initialize satellite communication system parameters, including the terminal task size range, terminal equipment computing power range, LEO satellite computing power, UE and AE communication thresholds, etc.

[0008] Step S2: According to the system model constructed in S1, the system energy consumption of the user terminal UE calculation, the idle terminal AE offloading calculation, and the LEO satellite MEC offloading calculation is obtained, and the optimization problem is constructed with the goal of minimizing the system energy consumption;

[0009] Step S3: According to the optimization problem constructed in S2, the maximum weight matching algorithm is used to perform device matching, solve the terminal matching problem, and then obtain the joint optimization problem of offloading decision, offloading ratio, and computing resource allocation of AE and LEOS;

[0010] Step S4: Based on the unloading decision, unloading ratio, and computing resource allocation of AE and LEO obtained in S3, the unloading decision problem is solved by relaxation and rounding techniques; the unloading ratio problem is solved by the solver; the AE and LEO satellite resource allocation scheme is obtained by Lagrange multiplier method and KKT condition;

[0011] Step S5: Repeat the iteration of step S4 until convergence or reaching the set maximum number of iterations, and find the best offloading decision, offloading ratio, and computing resource allocation of AE and LEOS.

[0012] Furthermore, the step S1 specifically includes:

[0013] Step S11: In the satellite-ground collaborative LEO satellite MEC communication model, there are several working terminals UE and idle terminals AE in the LEO satellite coverage area, and there are three task offloading decisions: terminal local computing, terminal collaborative computing, and satellite edge computing.

[0014] Step S12: Assume that the computational tasks generated by the UE are divisible. The i-th UE is represented as Its task attributes can be expressed as a triple (D i ,X i ,T i max ), where D i Indicates the data size of the computing task, X i Indicates that the task calculation D is completed i Required computational CPU cycles, Ω i Indicates the maximum tolerable delay of the task.

[0015] Step S13: This method sets the decision variable u i , i∈N to represent different offloading strategies. Set the offloading ratio variable u i =1, there are two cases: when c i =0, indicating that UE i If you choose not to offload, all tasks will be calculated locally; i ≠0,UE i Select to offload some tasks to AE devices. i =0,UE i Choose to offload part of the task to LEO.

[0016] Furthermore, the step S2 specifically includes:

[0017] S21、UE i With AE j The achievable data rate for the link between is:

[0018]

[0019] Among them, B i,j is the bandwidth of the subchannel, For UE i The transmission power, H i,j is the channel gain of the link, and N0 is the power of the additive white Gaussian noise (AWGN) of the link. i,j =|h i,j | 2 , channel coefficient Among them, σ i,j is a Rayleigh random variable, and γ is the path loss coefficient.

[0020] S22, UE i The achievable data rate for the link with a LEO satellite is:

[0021]

[0022] Among them, B i,s is the bandwidth of the channel. In the uplink and downlink, it is assumed that the satellite access bandwidth is divided into Q sub-channels, and the bandwidth of each channel is W. For UE i Transmit power in satellite mode, g i For UE i The antenna gain, G r is the receiving antenna gain of the LEO satellite, H i,sis the channel gain of the link, and N0 is the Gaussian white noise power of the link. r ≈(4D RA f u / c) 2 , where D RA is the diameter of the LEO satellite receiving antenna, f u is the uplink carrier frequency, and c is the speed of light. i,S =|h i,S 2 , channel coefficient Among them, ρ i,S is a Ricean random variable, is the free space path loss. Among them, d i,S For UE i The distance from the LEO satellite.

[0023] S23, when c i =0, indicating that UE i All generated tasks are offloaded locally, UE i The latency of local computation of all tasks i can be expressed as:

[0024]

[0025] where f i L For UE i computing power.

[0026] The energy consumption of the device CPU is related to its operating frequency, so the local energy consumption can be expressed as:

[0027] E i,L =κ(1-c i )X i (f i L ) 2

[0028] In the above formula, κ is the energy factor constant, which depends on the micro integrated circuit architecture, and κ(f i L ) 2 It indicates the energy consumption of each CPU cycle when the local terminal performs computing tasks.

[0029] S24, when the terminal chooses to uninstall on an idle terminal device, c i =1,u i = 1. Transfer another part of the task to AE j The transmission delay can be expressed as

[0030]

[0031] AE j The latency of computing this part of the task can be expressed as

[0032]

[0033] in For AE j computing power.

[0034] The energy consumption of offloading to adjacent nodes can be expressed as:

[0035]

[0036] Idle Equipment AE j The computational energy consumption can be expressed as:

[0037]

[0038] Because local computation and device offload are done in parallel, when u i =1, the task completion delay can be expressed as:

[0039]

[0040] Mission CompleteUE i and AE j The total energy consumed can be expressed as:

[0041]

[0042] S25. When the terminal chooses to unload at the LEO satellite MEC, c i =1,u i =0.UE i The transmission delay of transmitting another part of the mission to the LEO satellite can be expressed as

[0043]

[0044] UE i The propagation delay between the LEO satellite and the

[0045]

[0046] Since the distance between the UE and the LEO satellite is very far, the corresponding propagation delay cannot be ignored.

[0047] The latency of the MEC server of the LEO satellite to calculate this part of the task can be expressed as

[0048]

[0049] where f i,S Computational resources allocated to task i by the MEC server of the LEO satellite.

[0050] The energy consumption of offloading to the satellite edge node can be expressed as:

[0051]

[0052] The computational energy consumption of LEO satellites can be expressed as:

[0053]

[0054] Because local computing and MEC offloading are parallel, when u i = 0, the task completion delay can be expressed as

[0055]

[0056] Because the size of the calculation result is usually much smaller than the size of the input data, this method ignores the sending of the result from the MEC server of the LEO satellite to the UE. i delay.

[0057] Mission CompleteUE i The total energy consumed by LEO can be expressed as:

[0058]

[0059] S26, construct the energy consumption minimization optimization problem as:

[0060]

[0061] Furthermore, in step S3, the maximum weight matching algorithm is used to solve the terminal device matching problem, which specifically includes:

[0062] S31. When constructing a bipartite graph, add distance constraints and energy constraints between devices. If the constraints are not met, remove the connection between the two devices. i,j Represents the maximum calculation frequency of AE, ω i,j =0 means UE i and AE j There is no connection between them. The number of AEs connected to the user UE may be multiple or none.

[0063] S32. Construct a bipartite graph, represent the computing frequencies of working devices and idle devices as a left vertex set and a right vertex set of the bipartite graph, and fully connect the devices.

[0064] S33, the goal of this sub-problem is to find the matching solution with the largest total weight so that the entire system can obtain more computing power. The optimal matching matrix between devices is as follows

[0065]

[0066] Indicates that the Nth working device and the Mth idle device are connected.

[0067] S34, narrowing the problem space to an optimization problem between matched terminal devices, and obtaining an optimization function P1, thereby reducing constraints that need to be considered and the complexity of solving the optimization function.

[0068]

[0069] Furthermore, in step S4, using relaxed rounding and KKT conditions to solve the optimization problem P1 specifically includes:

[0070] S41, use relaxation and rounding techniques to solve the unloading decision problem. Optimize the unloading pattern matrix U, and fix the unloading amount matrix C and the idle device AE calculation resource allocation matrix F AE and the LEO edge computing resource allocation matrix F LEO , we get problem P2:

[0071]

[0072] Since the objective function and constraints are represented by binary variables u i Problem P2 is a "0-1" integer programming problem. The present invention first converts the integer variable u i Relax to a continuous variable and solve for u i , using rounding techniques to recover the binary value. i Relaxation is a continuous variable u i ∈[0,1], we get P2a:

[0073]

[0074] Since the function expression of problem P2 is a convex function and the constraint condition is a convex set, the definition of convex problem shows that problem P2 is a convex problem. The offloading probability matrix can be directly solved. Since there are limits on the number of users that LEO can serve at the same time, the smallest Q u i The value is 0, and the rest are 1.

[0075] S42. Use the solver to solve the unloading ratio optimization problem, optimize the unloading ratio matrix C, and fix the unloading mode matrix U and the AE calculation resource allocation matrix F. AE and the LEO edge computing resource allocation matrix FLEO ,get

[0076]

[0077] Problem P3 is a linear programming (LP) problem that can be solved by using the CVX-Gurobi solver.

[0078] S43, use the Lagrange multiplier method and KKT conditions to solve the AE computing resource allocation optimization problem, and fix the LEOS edge computing resource allocation matrix F LEO As well as the unloading ratio matrix C and the unloading mode matrix U, the AE computing resource allocation matrix F is optimized AE ,get:

[0079]

[0080] It can be proved that problem P4 is a convex optimization problem. The KKT (Karush-Kuhn-Tucker) method is used to solve the problem.

[0081] Specifically, the Lagrangian function of problem P4 is expressed by the following formula:

[0082]

[0083] where variables λ1, λ2, and λ3 are non-negative Lagrange multipliers. The optimal computational resource allocation matrix in AE is Optimal Lagrange multiplier λ1 * ,λ2 * ,λ3 * The KKT condition should be satisfied, and the solution can be obtained directly.

[0084] Then solve to get:

[0085]

[0086] S44, using Lagrange multiplier method and KKT conditions to solve the LEO computing resource allocation optimization problem, given the AE computing resource allocation matrix F AE As well as the offloading ratio matrix C and the offloading mode matrix U, the LEOS edge computing resource allocation matrix F is optimized LEO ,get:

[0087]

[0088] Problem P5 is also a convex optimization problem. In order to find the optimal solution, we can solve the KKT condition and get the result:

[0089]

[0090] where the variables μ1, μ2, and μ3 are non-negative Lagrange multipliers. The optimal computational resource allocation in LEOS is i LEO Under this condition, the optimal Lagrange multiplier μ1 * 、μ2 * 、μ3 * The KKT condition should be satisfied, and f can be directly solved i LEO* .

[0091] From the KKT conditions, we can know that the solution is:

[0092]

[0093] Furthermore, in step S5, the solutions of the four sub-problems in step S4 are repeatedly iterated, and the optimization is alternately performed until convergence or reaching a set maximum number of iterations, so as to find the best offloading decision, offloading ratio, and computing resource allocation of AE and LEOS.

[0094] The beneficial effects of the present invention are as follows: the present invention designs an energy-optimized low-orbit satellite Internet of Things satellite-ground collaborative unloading method, aiming to minimize the energy consumption of the satellite edge computing system. By effectively utilizing idle terminal equipment (AE), the problem of underutilization of ground terminal equipment in the satellite-ground network is solved, the potential computing resources in the satellite-ground network are fully tapped, the resource utilization rate of the entire network is improved, and the network can handle more user tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0095] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below in conjunction with the accompanying drawings, wherein:

[0096] Figure 1 This is a two-layer satellite edge computing system architecture diagram described in the present invention.

[0097] Figure 2 A schematic diagram of binary matching of the user matching method of the present invention.

[0098] Figure 3 Schematic diagram of the method of the present invention. DETAILED DESCRIPTION

[0099] like Figure 1 , Figure 2 and Figure 3 As shown, an energy consumption optimized low-orbit satellite Internet of Things satellite-to-ground collaborative unloading method provided by an embodiment of the present invention includes the following steps:

[0100] Step S1: Construct a satellite-ground collaborative LEO satellite network model, including working terminals, idle terminals, and LEO satellites; initialize satellite communication system parameters, including the terminal task size range, terminal equipment computing power range, LEO satellite computing power, UE and AE communication thresholds, etc.

[0101] Step S2: According to the system model constructed in S1, the system energy consumption of the user terminal UE calculation, the idle terminal AE offloading calculation, and the LEO satellite MEC offloading calculation is obtained, and the optimization problem is constructed with the goal of minimizing the system energy consumption;

[0102] Step S3: According to the optimization problem constructed in S2, the maximum weight matching algorithm is used to perform device matching, solve the terminal matching problem, and then obtain the joint optimization problem of offloading decision, offloading ratio, and computing resource allocation of AE and LEOS;

[0103] Step S4: Based on the unloading decision, unloading ratio, and computing resource allocation of AE and LEO obtained in S3, the unloading decision problem is solved by relaxation and rounding techniques; the unloading ratio problem is solved by the solver; the AE and LEO satellite resource allocation scheme is obtained by Lagrange multiplier method and KKT condition;

[0104] Step S5: Repeat the iteration of step S4 until convergence or reaching the set maximum number of iterations, and find the best offloading decision, offloading ratio, and computing resource allocation of AE and LEOS.

[0105] The step S1 specifically includes:

[0106] Step S11: In the satellite-ground collaborative LEO satellite MEC communication model, there are several working terminals UE and idle terminals AE in the LEO satellite coverage area, and there are three task offloading decisions: terminal local computing, terminal collaborative computing, and satellite edge computing. The UE calculates part of the task by itself and offloads the other part of the task to a neighboring AE for computing, that is, local computing and idle terminal offloading, abbreviated as terminal collaborative offloading. The UE calculates part of the task by itself and offloads the other part of the task to the LEO satellite, which is calculated by the MEC server of the satellite, that is, local computing and MEC offloading, abbreviated as MEC offloading. The UE calculates all tasks by itself, that is, local computing.

[0107] Step S12: Assume that the computational tasks generated by the UE are divisible. The i-th UE is represented as Its task attributes can be expressed as a triple (D i ,X i ,T i max ), where D i Indicates the data size of the computing task, Xi Indicates that the task calculation D is completed i Required computational CPU cycles, Ω i Indicates the maximum tolerable delay of the task.

[0108] Step S13: This method sets the decision variable u i , i∈N to represent different offloading strategies. Set the offloading ratio variable u i =1, there are two cases: when c i =0, indicating that UE i If you choose not to offload, all tasks will be calculated locally; i ≠0,UE i Select to offload some tasks to AE devices. i =0,UE i Choose to offload part of the task to LEO.

[0109] The step S2 specifically includes:

[0110] S21. Allocate an equal-bandwidth and orthogonal subchannel to each communication link between the UE and the AE. Since non-line-of-sight transmission dominates the communication link, the corresponding channel is modeled as a Rayleigh channel. i With AE j The achievable data rate for the link between is:

[0111]

[0112] Among them, B i,j is the bandwidth of the subchannel, For UE i The transmission power, H i,j is the channel gain of the link, and N0 is the power of the additive white Gaussian noise (AWGN) of the link. i,j =|h i,j | 2 , channel coefficient Among them, σ i,j is a Rayleigh random variable, and γ is the path loss coefficient.

[0113] S22, UE i Communication with LEO satellites relies on satellite-to-ground links using the Ka band. Since Line of Sight (LOS) transmissions dominate these links, the corresponding channels are modeled as Ricean channels. i The achievable data rate for the link with a LEO satellite is:

[0114]

[0115] Among them, B i,s is the bandwidth of the channel. In the uplink and downlink, it is assumed that the satellite access bandwidth is divided into Q sub-channels, and the bandwidth of each channel is W. For UE i Transmit power in satellite mode, g i For UE i The antenna gain, G r is the receiving antenna gain of the LEO satellite, H i,s is the channel gain of the link, and N0 is the Gaussian white noise power of the link. r ≈(4D RA f u / c) 2 , where D RA is the diameter of the LEO satellite receiving antenna, f u is the uplink carrier frequency, and c is the speed of light. i,S =|h i,S 2 , channel coefficient Among them, ρ i,S is a Ricean random variable, is the free space path loss. Among them, d i,S For UE i The distance from the LEO satellite.

[0116] S23, when c i =0, indicating that UE i All generated tasks are offloaded locally, UE i The latency of local computation of all tasks i can be expressed as:

[0117]

[0118] where f i L For UE i computing power.

[0119] The energy consumption of the device CPU is related to its operating frequency, so the local energy consumption can be expressed as:

[0120] E i,L =κ(1-c i )X i (f i L ) 2

[0121] In the above formula, κ is the energy factor constant, which depends on the micro integrated circuit architecture, and κ(f iL ) 2 It indicates the energy consumption of each CPU cycle when the local terminal performs computing tasks.

[0122] S24. When the terminal chooses to uninstall on an idle terminal device, c i =1,u i = 1. Transfer another part of the task to AE j The transmission delay can be expressed as

[0123]

[0124] AE j The latency of computing this part of the task can be expressed as

[0125]

[0126] in For AE j computing power.

[0127] The energy consumption of offloading to adjacent nodes can be expressed as:

[0128]

[0129] Idle Equipment AE j The computational energy consumption can be expressed as:

[0130]

[0131] Because local computation and device offload are done in parallel, when u i =1, the task completion delay can be expressed as:

[0132]

[0133] Mission CompleteUE i and AE j The total energy consumed can be expressed as:

[0134]

[0135] S25. When the terminal chooses to unload at the LEO satellite MEC, c i =1,u i =0.UE i The transmission delay of transmitting another part of the mission to the LEO satellite can be expressed as

[0136]

[0137] UE i The propagation delay between the LEO satellite and the

[0138]

[0139] Since the distance between the UE and the LEO satellite is very far, the corresponding propagation delay cannot be ignored.

[0140] The latency of the MEC server of the LEO satellite to calculate this part of the task can be expressed as

[0141]

[0142] where f i,S Computational resources allocated to task i by the MEC server of the LEO satellite.

[0143] The energy consumption of offloading to the satellite edge node can be expressed as:

[0144]

[0145] The computational energy consumption of LEO satellites can be expressed as:

[0146]

[0147] Because local computing and MEC offloading are parallel, when u i = 0, the task completion delay can be expressed as

[0148]

[0149] Because the size of the calculation result is usually much smaller than the size of the input data, this method ignores the sending of the result from the MEC server of the LEO satellite to the UE. i delay.

[0150] Mission CompleteUE i The total energy consumed by LEO can be expressed as:

[0151]

[0152] S26, construct the energy consumption minimization optimization problem as:

[0153]

[0154] The constraints in the above formula can be described as follows. Constraint C1 is the delay constraint of the task, which means that the completion delay of task i cannot exceed the maximum value; constraint C2 means that the energy consumption of AE is less than its maximum energy; constraints C3 and C4 mean that a UE can only be paired with another AE; constraint C5 means that a communication link can only be established between UE and AE that meet the communication distance limit; constraint C6 means that UE iOnly one offloading decision can be made to complete task i; constraint C7 indicates that the remaining energy of AE must be greater than a threshold to match the UE; constraint C8 indicates that the computing power allocated by AE to UE must be less than its maximum computing power; constraint C9 indicates that the computing power allocated to all UEs at the LEO edge must be less than its maximum computing power; constraint C10 indicates that the offloading ratio is between 0 and 1; constraint C11 indicates that the energy loss of the LEO satellite must be within a threshold; constraint C12 indicates that the LEO satellite can serve at most Q users at the same time.

[0155] In step S3, the maximum weight matching algorithm is used to solve the terminal device matching problem, specifically including:

[0156] S31. When constructing a bipartite graph, add distance constraints and energy constraints between devices. If the constraints are not met, remove the connection between the two devices. i,j Represents the maximum calculation frequency of AE, ω i,j =0 indicates UE i and AE j There is no connection between them. The number of AEs connected to the user UE may be multiple or none.

[0157] S32. Construct a bipartite graph, represent the computing frequencies of working devices and idle devices as a left vertex set and a right vertex set of the bipartite graph, and fully connect the devices.

[0158] S33, the goal of this sub-problem is to find the matching solution with the largest total weight so that the entire system can obtain more computing power. The optimal matching matrix between devices is as follows

[0159]

[0160] Indicates that the Nth working device and the Mth idle device are connected.

[0161] S34, narrowing the problem space to an optimization problem between matched terminal devices, and obtaining an optimization function P1, thereby reducing constraints that need to be considered and the complexity of solving the optimization function.

[0162]

[0163] In step S4, solving the optimization problem P1 using relaxed rounding and KKT conditions specifically includes:

[0164] S41, use relaxation and rounding techniques to solve the unloading decision problem. Optimize the unloading pattern matrix U, and fix the unloading amount matrix C and the idle device AE calculation resource allocation matrix F AE and the LEO edge computing resource allocation matrix FLEO , we get problem P2:

[0165]

[0166] Since the objective function and constraints are represented by binary variables u i Problem P2 is a "0-1" integer programming problem. The present invention first converts the integer variable u i Relax to a continuous variable and solve for u i , using rounding techniques to recover the binary value. i Relaxation is a continuous variable u i ∈[0,1], we get P2a:

[0167]

[0168] Since the function expression of problem P2 is a convex function and the constraint condition is a convex set, the definition of convex problem shows that problem P2 is a convex problem. The offloading probability matrix can be directly solved. Since there are limits on the number of users that LEO can serve at the same time, the smallest Q u i The value is 0, and the rest are 1.

[0169] S42. Use the solver to solve the unloading ratio optimization problem, optimize the unloading ratio matrix C, and fix the unloading mode matrix U and the AE calculation resource allocation matrix F. AE and the LEO edge computing resource allocation matrix F LEO ,get

[0170]

[0171] Problem P3 is a linear programming (LP) problem that can be solved by using the CVX-Gurobi solver.

[0172] S43, use the Lagrange multiplier method and KKT conditions to solve the AE computing resource allocation optimization problem, and fix the LEOS edge computing resource allocation matrix F LEO As well as the unloading ratio matrix C and the unloading mode matrix U, the AE computing resource allocation matrix F is optimized AE ,get:

[0173]

[0174] It can be proved that problem P4 is a convex optimization problem. The KKT (Karush-Kuhn-Tucker) method is used to solve the problem.

[0175] Specifically, the Lagrangian function of problem P4 is expressed by the following formula:

[0176]

[0177] where variables λ1, λ2, and λ3 are non-negative Lagrange multipliers. The optimal computational resource allocation matrix in AE is Optimal Lagrange multiplier λ1 * ,λ2 * ,λ3 * The KKT condition should be satisfied, and the solution can be obtained directly.

[0178] Then solve to get:

[0179]

[0180] S44, using Lagrange multiplier method and KKT conditions to solve the LEO computing resource allocation optimization problem, given the AE computing resource allocation matrix F AE As well as the offloading ratio matrix C and the offloading mode matrix U, the LEOS edge computing resource allocation matrix F is optimized LEO ,get:

[0181]

[0182] Problem P5 is also a convex optimization problem. In order to find the optimal solution, we can solve the KKT condition and get the result:

[0183]

[0184] where the variables μ1, μ2, and μ3 are non-negative Lagrange multipliers. The optimal computational resource allocation in LEOS is i LEO Under this condition, the optimal Lagrange multiplier μ1 * 、μ2 * 、μ3 * The KKT condition should be satisfied, and f can be directly solved i LEO* .

[0185] From the KKT conditions, we can know that the solution is:

[0186]

[0187] In step S5, the solutions of the four sub-problems in step S4 are repeatedly and iteratively executed, and the optimization is alternately performed until convergence or the set maximum number of iterations is reached, so as to find the best offloading decision, offloading ratio, and computing resource allocation of AE and LEOS.

[0188] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution, which should be included in the scope of the claims of the present invention.

Claims

1. A low-orbit satellite Internet of Things satellite-ground collaborative unloading method with energy consumption optimization, characterized in that: The method comprises the following steps: Step S1: constructing a satellite-ground collaborative LEO satellite network model, including working terminals, idle terminals, and LEO satellites; initializing satellite communication system parameters, including but not limited to the terminal task size range, the terminal equipment computing capability range, the LEO satellite computing capability, and the UE and AE communication thresholds; Step S2: According to the system model constructed in step S1, the system energy consumption of the user terminal UE calculation, the idle terminal AE offloading calculation, and the LEO satellite MEC offloading calculation is obtained, and the optimization problem is constructed with the goal of minimizing the system energy consumption; Step S3: According to the optimization problem constructed in step S2, the maximum weight matching algorithm is used to perform device matching, solve the terminal matching problem, and then obtain the joint optimization problem of offloading decision, offloading ratio, and computing resource allocation of AE and LEOS; Step S4: Based on the unloading decision, unloading ratio, and computing resource allocation of AE and LEO obtained in step S3, the unloading decision problem is solved by using relaxation and rounding techniques; the unloading ratio problem is solved by using a solver; and the AE and LEO satellite resource allocation scheme is solved by using the Lagrange multiplier method and KKT conditions. Step S5: Repeat the iteration of step S4 until convergence or reaching the set maximum number of iterations, and find the best offloading decision, offloading ratio, and computing resource allocation of AE and LEOS.

2. According to the energy consumption optimized low-orbit satellite Internet of Things satellite-ground collaborative unloading method method of claim 1, it is characterized in that: The step S1 specifically includes: Step S11: In the satellite-ground collaborative LEO satellite MEC communication model, there are several working terminals UE and idle terminals AE in the LEO satellite coverage area, and there are three task offloading decisions: terminal local computing, terminal collaborative computing, and satellite edge computing; Step S12: Assuming that the computing task generated by the UE is divisible, the i-th UE is represented as Its task attributes can be expressed as a triple (D i ,X i ,T i max ), where D i Indicates the data size of the computing task, X i Indicates completion of task calculation D i Required computational CPU cycles, Ω i Indicates the maximum tolerable delay of the task; Step S13: Set decision variable u i , i∈N to represent different offloading strategies; set the offloading ratio variable u i =1, there are two cases: when c i =0, indicating that UE i If you choose not to offload, all tasks will be calculated locally; i ≠0,UE i Choose to offload some tasks to AE devices; when u i =0,UE i Choose to offload part of the task to LEO.

3. According to the energy consumption optimized low-orbit satellite Internet of Things satellite-ground collaborative unloading method method of claim 2, it is characterized in that: The step S2 specifically includes: S21、UE i With AE j The achievable data rate for the link between is: Among them, B i,j is the bandwidth of the subchannel, For UE i The transmission power, H i,j is the channel gain of the link, N0 is the power of the additive white Gaussian noise (AWGN) of the link; H i,j =|h i,j | 2 , channel coefficient Among them, σ i,j is the Rayleigh random variable, γ is the path loss coefficient; S22, UE i The achievable data rate for the link with a LEO satellite is: Among them, B i,s is the bandwidth of the channel. In the uplink and downlink, it is assumed that the satellite access bandwidth is divided into Q sub-channels, and the bandwidth of each channel is W; For UE i Transmit power in satellite mode, g i For UE i The antenna gain, G r is the receiving antenna gain of the LEO satellite, H i,s is the channel gain of the link, N0 is the Gaussian white noise power of the link; G r ≈(4D RA f u / c) 2 , where D RA is the diameter of the LEO satellite receiving antenna, f u is the uplink carrier frequency, and c is the speed of light. i,S =|h i,S | 2 , channel coefficient Among them, ρ i,S is a Ricean random variable, is the free space 2 path loss; Among them, d i,S For UE i The distance to the LEO satellite; S23, when c i =0, indicating that UE i All generated tasks are offloaded locally, UE i The latency of local computation of all tasks i is expressed as: where f i L For UE i computing power; The energy consumption of the device CPU is related to its operating frequency, so the local energy consumption is expressed as: E i,L =κ(1-c i )X i (f i L ) 2 In the above formula, κ is the energy factor constant, which depends on the micro integrated circuit architecture, and κ(f i L ) 2 It represents the energy consumption of each CPU cycle when the local terminal performs computing tasks; S24. When the terminal chooses to uninstall on an idle terminal device, c i =1,u i =1; transfer another part of the task to AE j The transmission delay of AE j The latency of computing this part of the task can be expressed as in For AE j computing power; The energy consumption of offloading to adjacent nodes can be expressed as: Idle Equipment AE j The computational energy consumption can be expressed as: Because local computation and device offload are done in parallel, when u i =1, the task completion delay can be expressed as: Mission CompleteUE i and AE j The total energy consumed can be expressed as: S25. When the terminal chooses to unload at the LEO satellite MEC, c i =1,u i =0;UE i The transmission delay of transmitting another part of the mission to the LEO satellite can be expressed as UE i The propagation delay between the LEO satellite and the The latency of the MEC server of the LEO satellite to calculate this part of the task can be expressed as where f i,S The computing resources allocated to task i by the MEC server of the LEO satellite; The energy consumption of offloading to the satellite edge node can be expressed as: The computational energy consumption of LEO satellites can be expressed as: Because local computing and MEC offloading are parallel, when u i = 0, the task completion delay can be expressed as Because the size of the calculation result is usually much smaller than the size of the input data, the sending of the result from the MEC server of the LEO satellite to the UE is ignored. i Delay Mission CompleteUE i The total energy consumed by LEO can be expressed as: S26, construct the energy consumption minimization optimization problem as:

4. According to the energy consumption optimized low-orbit satellite Internet of Things satellite-ground collaborative unloading method method of claim 3, it is characterized in that: This method uses a maximum weight matching algorithm to solve the terminal device matching problem, and the step S3 specifically includes: S31. When constructing a bipartite graph, add distance constraints and energy constraints between devices. If the constraints are not met, remove the connection between the two devices; the weight between the device lines ω i,j Represents the maximum calculation frequency of AE, ω i,j =0 indicates UE i and AE j There is no connection between them. The number of AEs connected to the user UE may be multiple or none; S32, construct a bipartite graph, represent the computing frequencies of working devices and idle devices as a left vertex set and a right vertex set of the bipartite graph, and fully connect the devices; S33. The goal of this sub-problem is to find the matching solution with the largest total weight so that the entire system can obtain more computing power; the optimal matching matrix between devices is as follows Indicates that the Nth working device and the Mth idle device are connected; S34, narrowing the problem space to an optimization problem between matched terminal devices, and obtaining an optimization function P1, thereby reducing constraints that need to be considered and the complexity of solving the optimization function; 5. According to the energy consumption optimized low-orbit satellite Internet of Things satellite-ground collaborative unloading method method of claim 1, it is characterized in that: The optimization problem P1 is solved using relaxed rounding and KKT conditions, and the step S4 specifically includes: S41, use relaxation and rounding techniques to solve the unloading decision problem; optimize the unloading pattern matrix U, and fix the unloading amount matrix C and the idle device AE calculation resource allocation matrix F AE and the LEO edge computing resource allocation matrix F LEO , we get problem P2: Since the objective function and constraints are represented by binary variables u i Problem P2 is a "0-1" integer programming problem. This method first transforms the integer variable u i Relax to a continuous variable and solve for u i , using rounding techniques to recover the binary value; convert the binary variable u i Relaxation is a continuous variable u i ∈[0,1], we get P2a: Since the function expression of problem P2 is a convex function and the constraint condition is a convex set, the definition of convex problem shows that problem P2 is a convex problem. The offloading probability matrix can be directly solved. Since the number of users that LEO can serve at the same time is limited, the smallest Q u i The value is 0, and the rest are 1; S42. Use the solver to solve the unloading ratio optimization problem, optimize the unloading ratio matrix C, and fix the unloading mode matrix U and the AE calculation resource allocation matrix F. AE and the LEO edge computing resource allocation matrix F LEO ,get Problem P3 is a linear programming (LP) problem, which is solved by using the CVX-Gurobi solver; S43, use the Lagrange multiplier method and KKT conditions to solve the AE computing resource allocation optimization problem, and fix the LEOS edge computing resource allocation matrix F LEO As well as the unloading ratio matrix C and the unloading mode matrix U, the AE computing resource allocation matrix F is optimized AE ,get: It can be proved that problem P4 is a convex optimization problem; the KKT method is used to solve the problem; Specifically, the Lagrangian function of problem P4 is expressed by the following formula: where variables λ1, λ2, and λ3 are non-negative Lagrange multipliers; the optimal computational resource allocation matrix in AE is Optimal Lagrange multiplier λ1 * ,λ2 * ,λ3 * The KKT condition should be satisfied, and the solution can be obtained directly. Solving the KKT conditions yields: S44, using Lagrange multiplier method and KKT conditions to solve the LEO computing resource allocation optimization problem, given the AE computing resource allocation matrix F AE As well as the offloading ratio matrix C and the offloading mode matrix U, the LEOS edge computing resource allocation matrix F is optimized LEO ,get: Problem P5 is also a convex optimization problem; in order to find the optimal solution, the result is obtained by solving the KKT condition: where the variables μ1, μ2, and μ3 are non-negative Lagrange multipliers; the optimal computational resource allocation in LEOS is f i LEO Under this condition, the optimal Lagrange multiplier μ1 * 、μ2 * 、μ3 * The KKT condition should be satisfied, and f can be directly solved i LEO* ; Solving the KKT conditions yields:

Images