Space-air-ground integrated network reasoning efficiency optimization method based on sub-modulo optimization

Abstract

The invention discloses a space-air-ground integrated network reasoning efficiency optimization method based on sub-modulo optimization, and belongs to the technical field of artificial intelligence and wireless communication in a space-air-ground integrated network. The method comprises the following steps: (1) establishing an edge cooperation reasoning system in a sky-ground integrated network, providing a plurality of cooperation modes for reasoning tasks on a vehicle, and obtaining reasoning time delays of the reasoning tasks in different cooperation modes; (2) constructing a resource allocation model and a terminal association model of the edge cooperative reasoning system to obtain total time delays experienced by the vehicles in different allocation modes; (3) constructing a resource allocation model of the edge cooperative reasoning system, and proposing an objective function and constraint conditions to form an optimization problem; and (4) decomposing the optimization problem into two sub-problems, and alternately solving the two sub-problems until an optimal reasoning efficiency optimization strategy is obtained. According to the invention, efficient resource allocation with minimized total time delay of the system is realized; the method has lower reasoning time delay and higher resource utilization efficiency.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence and wireless communication technology in an air-ground integrated network, and in particular to an air-ground-ground integrated network reasoning efficiency optimization method based on submodule optimization. Background Art

[0002] The fifth generation (5G) mobile communication network is designed to support enhanced mobile broadband (eMBB), massive machine type communication (mMTC), and ultra-reliable low latency communication (URLLC). With the commercialization of 5G technology, the research community has turned its attention to the sixth generation (6G) network, which is expected to provide ubiquitous coverage and anytime, anywhere ultra-broadband access, which is essential for future applications such as the Internet of Things (IoT), remote area coverage, emergency communications, and ecological remote sensing. However, the current 5G network architecture is not sufficient to support these advanced services, especially the demand for ubiquitous ultra-broadband access, artificial intelligence services, and numerous IoT applications. To address this challenge, the Sky-Ground Integrated Network (SAGIN) is proposed as a promising architecture that can meet these service and application requirements. At the same time, the rapid development of artificial intelligence (AI) technology has led to a surge in the demand for model reasoning. AI has become the driving force of new technological and industrial revolutions with its ability to simulate and extend human intelligence. In the context of 6G networks, AI applications are about to become popular, and with the development of AI technology, the demand for model reasoning continues to grow. Despite the great potential of SAGIN, research on collaborative reasoning within this network is still in its infancy. Resource allocation in SAGIN is crucial for efficiently completing reasoning tasks, which are characterized by the need for high-quality completion within limited resources and time. Summary of the invention

[0003] Purpose of the invention: To provide a method for optimizing the efficiency of space-ground integrated network reasoning based on sub-module optimization, so as to solve the problem that the collaborative reasoning research in the SAGIN network is still in its infancy, but it is impossible to complete the task with high quality within limited resources and time.

[0004] The present invention provides a method for optimizing the reasoning efficiency of an air-ground integrated network based on sub-module optimization, comprising the following steps:

[0005] S1. Building an edge cooperative reasoning system for reasoning on the air-space-ground integrated network, wherein the air-space-ground integrated network includes vehicles, drones, satellites, and base stations. The edge cooperative reasoning system is built based on the vehicles, drones, satellites, and base stations in the air-space-ground integrated network, and several cooperation modes are proposed for the reasoning tasks on the vehicles, and the reasoning delays of the reasoning tasks under the different cooperation modes are obtained;

[0006] S2. Based on the edge cooperative reasoning system, construct a resource allocation model and a terminal association model of the edge cooperative reasoning system, and obtain the total delay experienced by the vehicle under different allocation modes;

[0007] S3. Construct a resource allocation model for the edge cooperative reasoning system, and propose an objective function and constraints, wherein the objective function is defined as:

[0008]

[0009] Where W represents the task offloading decision variable, I represents the model segmentation decision variable, Z represents the resource allocation decision variable, and T j represents the total delay experienced by car j;

[0010] The constraints are the limitation of computing resources and the limit of the number of vehicles that can be associated with drones, satellites and base stations;

[0011] S4. Decompose the optimization problem into sub-problems P1 and P2, namely, the joint terminal association and resource allocation problem and the model segmentation problem. The joint terminal association and resource allocation problem proposes a low-complexity association strategy based on sub-module optimization. The model segmentation problem is solved by traversing search to determine the best splitting point. The joint terminal association and resource allocation problem and the model segmentation problem are solved alternately until the optimal reasoning efficiency optimization strategy is obtained.

[0012] Preferably, the step S1 specifically comprises:

[0013] S11, in the edge cooperative reasoning system, including the vehicle, the drone, the satellite, the base station and the reasoning task on the vehicle, obtaining the data of the vehicle, the drone, the satellite, the base station and the neural network parameters;

[0014] The step S11 is specifically as follows: there are J vehicles in the edge cooperative reasoning system, the index of the vehicle is defined as J∈{1,2,…,J}, the base station includes U base stations, the index of the base station is defined as U∈{1,2,…,U}, there are K drones in the edge cooperative reasoning system, the index of the drone is defined as K∈{1,2,…,K}, and a satellite S; each vehicle in the network in the edge cooperative reasoning system has a reasoning task A jTo perform, the ground vehicle performs cooperative reasoning with the base station, UAV or satellite within the communication range through the wireless channel; the base station allocates N computing units to perform the tasks of the ground vehicle, and the index of its computing unit is N∈{1,2,…,N}; the UAV allocates L computing units to perform the tasks of the ground vehicle, and the index of its computing unit is L∈{1,2,…,L}; the satellite allocates M computing units to perform the tasks of the ground vehicle, and the index of its computing unit is M∈{1,2,…,M}; the index of the neural network layer in the reasoning task is i∈{0,1,2,…,I}, where I is the total number of layers of the neural network;

[0015] S12, constructing a task offloading decision variable, wherein the task offloading decision variable indicates that the reasoning task of the vehicle is completed locally or in cooperative reasoning with other terminals and servers;

[0016] The S12 step is specifically as follows:

[0017] The vehicle j∈J decides to perform its reasoning task locally, defining is the task offloading decision; the vehicle j completes its reasoning task A locally j ,but otherwise Considering the insufficient reasoning performance of the vehicle itself, the vehicle j∈J cooperates with the base station u∈U to complete the reasoning task;

[0018] definition For task offloading decision, the vehicle j and the base station u cooperate to complete the reasoning task, then otherwise like The model segmentation point between the vehicle j and the base station u is defined as

[0019] The vehicle j considers whether the reasoning task is assigned to the UAV, where the UAV refers to the unmanned aerial vehicle, and defines is the task offloading decision, the vehicle j and the UAV k cooperate to complete the reasoning task, then otherwise The split point of the model between the vehicle j and the UAV k is defined as

[0020] After receiving the offloaded cooperative reasoning task from the vehicle associated with the UAV k, the UAV k decides to process the task locally at the UAV k or offload the UAV k to the satellite according to the computing power available at the UAV k; define is the terminal association strategy, the vehicle j completes the reasoning task through the drone k relay and the satellite s. otherwise The split point of the model at this time is defined as

[0021] S13, constructing a resource allocation decision variable, wherein the resource allocation decision variable indicates that the computing units on the drone, the satellite, and the base station are allocated to the reasoning task of the vehicle;

[0022] The S13 step is specifically as follows:

[0023] definition Assign decision variables to the computing units, the computing unit n at the base station u is assigned to the vehicle j, then otherwise

[0024] definition Assign decision variables to computing units, computing unit l on the UAV k is assigned to the vehicle j, then otherwise

[0025] definition Assign decision variables to computing units, computing unit m on the satellite s is assigned to the vehicle j, then otherwise

[0026] S14, the data of the vehicle, the drone, the satellite, the base station and the neural network parameters, based on the task offloading decision variables and the resource allocation decision variables, calculate the unloading waiting delay of the reasoning task on the vehicle j in different allocation modes in the edge cooperative reasoning communication system;

[0027] The communication rate from the vehicle j to the base station u is At the model split point The unloading transmission time for vehicle j to offload part of the reasoning task to base station u is in, For model The output of the layer;

[0028] The communication rate from vehicle j to UAV k is The vehicle j is at the model segmentation point The transmission delay experienced when offloading part of the reasoning task to the UAV k is in, For model The output of the layer.

[0029] Preferably, the step S2 is specifically as follows:

[0030] Calculate the total delay experienced by the vehicle in different allocation modes in the edge cooperative reasoning system based on the resource allocation decision variable, the task offloading decision variable, the vehicle, the drone, the satellite, the base station data, and the neural network parameters;

[0031] The vehicle j decides to perform its reasoning task locally, i.e. Then the total delay experienced by vehicle j to complete the reasoning task is:

[0032]

[0033] where ξ j is the computing power of vehicle j, α j is the number of CPU cycles required to process one bit of data;

[0034] When , the time required for the base station u to complete the inference task is:

[0035]

[0036] in For Layer Post-reasoning task A j The amount of data that needs to be processed, the total delay experienced by the vehicle j and the base station u to complete the cooperative reasoning task is:

[0037]

[0038] definition Assign decision variables to computing units, computing unit l on the UAV k is assigned to the vehicle j, then otherwise

[0039] When , the time required for UAV k to complete the reasoning task is:

[0040]

[0041] in For Layer Post-reasoning task A j The amount of data that needs to be processed, the total delay experienced by the vehicle j and the UAV k to complete the cooperative reasoning task is:

[0042]

[0043] definition Assign decision variables to computing units, computing unit m on the satellite s is assigned to the vehicle j, then otherwise

[0044] When , the time required for the satellite s to complete the reasoning task is:

[0045]

[0046] When the inference task of vehicle j is offloaded to the satellite, the total delay experienced by vehicle j is:

[0047]

[0048] Preferably, the S3 step is specifically as follows:

[0049] The objective function is specifically:

[0050]

[0051] The constraints include:

[0052] Represents constraints on computing resources; It means that each vehicle can only choose one cooperative reasoning mode;

[0053] It means that each computing unit can be assigned to at most one vehicle in one allocation cycle;

[0054] Indicates that the base station, the drone, and the satellite can be associated with a limited number of vehicles;

[0055] Indicates resource allocation strategy and terminal association strategy;

[0056] Indicates the model split point;

[0057] in, represents the upper limit of the number of vehicles that can be accessed by the base station, the drone, and the satellite, N max , L max , M max Indicates the upper limit of the computing resources of the base station, the drone, and the satellite.

[0058] Preferably, the S4 step is specifically as follows:

[0059] S41. Questions The sub-problem P1 is given as:

[0060]

[0061] The constraints include:

[0062] Represents constraints on computing resources; It means that each vehicle can only choose one cooperative reasoning mode;

[0063] It means that each computing unit can be assigned to at most one vehicle in one allocation cycle;

[0064] Indicates that the base station, the drone, and the satellite can be associated with a limited number of vehicles;

[0065] Indicates resource allocation strategy and terminal association strategy;

[0066] Among them, the objective function D j (w,z)=T0-T j Aims to minimize delay costs, T0 is the task waiting time;

[0067] Given the subproblem P2:

[0068]

[0069] The constraints include:

[0070] Indicates the model split point;

[0071] S42, converting the joint terminal association and resource allocation problem into a submodule problem for solution;

[0072] The subproblem P1 is transformed into a submodular problem, and the computing resource allocation and terminal association strategy are defined as

[0073]

[0074] in represents the action that the computing resource l on the drone k is allocated to the vehicle j, represents the action that the computing resource n on the base station u is allocated to the vehicle j, represents the action of associating the base station u with the vehicle j, represents the action associated with the drone k and the vehicle j; the basic set A can be divided into V disjoint sets (V′=UV), that is, for any v≠v′, in The basis set B can be divided into Disjoint sets The basis set C can be divided into U disjoint sets, where The basis set D can be divided into K disjoint sets;

[0075] Define a pair E = (F, G), where F = {A, B, C, D} is a global set, A set of independent subsets of F is used to replace the original constraints, defined as

[0076]

[0077] Where X is the set of final solutions; prove that the set family E = (F, G) is a pseudo-matrix, delay reduction function is a monotone submodule function on X∈G; on this basis, by replacing the constraints in the problem with set pairs, the subproblem P1 is expressed as a pseudo-matrix constrained monotone submodule maximization problem:

[0078]

[0079] stX∈G

[0080] initialization is an empty set, where v = 1, ..., V′, u=1,...,U,k=1,...,K, and define X={A,B,C,D} and Y X = F; in each iteration, the new element with the highest marginal gain is Add to Collection X X and from the candidate action set Y X Remove the candidate action set Y X is the set of all currently selectable strategies; In the case of Y=YA, the computing units on other base stations and drones are assigned to the action of the vehicle j, and the actions of other base stations and drones associated with the vehicle j are removed from the candidate action set, that is: v′ ,v′=1,...,V′

[0081] v′≠(u-1)V+1,(u-1)V+2,...,uV,

[0082]

[0083] and

[0084]

[0085] Repeat this iteration until the candidate action set Y X Empty or marginal gain Δ D (a|X) is zero; wherein the marginal gain refers to: in a submodular function, when a new element is added to a set, the contribution of the element to the increase in the function value; in the present invention, the marginal gain Δ D(a|X)=D(a∪{X})-D(X);

[0086] When considering satellites, a redistribution method based on resource accumulation is used to check whether each vehicle can be connected to the satellite through a drone through a loop. The specific steps are as follows:

[0087] Initialize X, Y x=F and the global optimal allocation strategy; when as well as When the conditions are met, enter the loop to calculate the inference latency; if a suitable satellite access strategy is found and the inference latency can be reduced, update the global optimal allocation strategy; after the update, allocate the corresponding computing resources and mark the allocation as used; if the satellite's resource allocation is better than other options, choose to connect to the satellite through the drone, and update the vehicle's computing power and latency;

[0088] S43, model segmentation algorithm to place the model,

[0089] The step S43 specifically includes:

[0090] The best split point is determined by traversal search. Considering the cooperative reasoning between vehicles and base stations, the model segmentation algorithm is proposed as follows:

[0091] Input: total number of layers of the DNN model I, bandwidth B, computing power of base station u The computing power of vehicle j ξ j , the amount of data output per layer Optimal split point i best ;

[0092] ⑤Minimum time to initialize vehicle j Calculate time locally

[0093] ⑥ Assume that the initial optimal segmentation layer i best =I;

[0094] ⑦ For each layer i∈{1,2,…,I} of the neural network in the inference task, perform the following operations:

[0095] Calculate the transmission time of the current layer Processing time at base station u and the processing time on vehicle j

[0096] If the total time to perform the inference task The minimum update time is And i best Set to current layer i;

[0097] ⑧After the cycle ends, return to the best segmentation point ibest ;

[0098] S44. Solve the two sub-problems alternately until the optimal strategy is obtained.

[0099] Compared with the prior art, the present invention adopts the above technical solution and has the following technical effects:

[0100] (1) This method minimizes the total system delay by optimizing the inference efficiency in the integrated space-ground network;

[0101] (2) This method is based on low-complexity submodule optimization and has lower inference latency and higher resource utilization efficiency compared to traditional methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0102] Figure 1 This is a diagram of an integrated air-ground-space collaborative reasoning network architecture.

[0103] Figure 2 This is a flow chart of the method for optimizing the reasoning efficiency of the space-ground integrated network based on sub-module optimization.

[0104] Figure 3 An embodiment of the present invention provides a system minimum delay after optimizing the reasoning efficiency of an air-ground integrated network based on sub-module optimization. DETAILED DESCRIPTION

[0105] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. The experimental methods described in the following embodiments are all conventional methods unless otherwise specified.

[0106] According to a method for optimizing the reasoning efficiency of an air-ground integrated network based on sub-module optimization provided by the present invention, the edge cooperative reasoning system in the air-ground integrated network of the present invention is as follows: Figure 1 As shown in Figure 2, it is assumed that the mobile edge computing communication system in the sky-ground integrated network consists of ground vehicles, drones, base stations, and a LEO satellite. The goal is to minimize the total system delay while meeting the number of associated vehicles and drones. Figure 3 This embodiment provides the minimum system delay after optimizing the reasoning efficiency of the integrated air-space-ground network based on sub-module optimization.

[0107] like Figure 2 As shown, a flowchart of a method for optimizing the efficiency of air-ground integrated network reasoning based on sub-module optimization provided by an embodiment of the present invention includes the following steps:

[0108] Step S1 specifically includes:

[0109] S11. Build an edge cooperative reasoning system based on the integrated air-ground-space network, including vehicles, drones, satellites, base stations and reasoning tasks on vehicles, and obtain data of vehicles, drones, satellites, base stations and neural network parameters;

[0110] S12, constructing a task offloading decision variable, which is used to indicate that the vehicle's reasoning task is completed locally or in cooperative reasoning with other terminals and servers;

[0111] S13, construct resource allocation decision variables to represent the reasoning tasks that the computing units on the drone, satellite, and base station are assigned to the vehicle;

[0112] S14, based on the task offloading decision variables and resource allocation decision variables, the data of vehicles, drones, satellites, base stations and neural network parameters, calculate the offloading waiting delay of the reasoning task under different allocation modes in the edge cooperative reasoning communication system;

[0113] In a further embodiment, the cooperative reasoning problem in a sky-ground integrated network is considered, in which there are J vehicles, whose index is defined as J∈{1,2,…,J}, U base stations, whose index is defined as U∈{1,2,…,U}, K drones, whose index is defined as K∈{1,2,…,K}, and a satellite S. Each vehicle in the considered network has a reasoning task A j To perform, the ground vehicle performs cooperative reasoning with a base station, drone or satellite within the communication range through a wireless channel. Assume that the base station allocates N computing units to perform the tasks of the ground vehicle, and the index of its computing units is N∈{1,2,…,N}. Assume that the drone allocates L computing units to perform the tasks of the ground vehicle, and the index of its computing units is L∈{1,2,…,L}. The satellite allocates M computing units to perform the tasks of the ground vehicle, and the index of its computing units is M∈{1,2,…,M}. The index of the neural network layer in the reasoning task is i∈{0,1,2,…,I}, where I is the total number of layers of the neural network. Assume that all computing units have the same computing power, that is, ξ CPU cycles are executed per second, and the neural network is split into at most two parts for processing.

[0114] If vehicle j∈J decides to perform its reasoning task locally, define is the terminal association strategy. If vehicle j completes its reasoning task A locally j ,but otherwise Considering the insufficient reasoning performance of the vehicle itself, vehicle j∈J cooperates with base station u∈U to complete the reasoning task. Definition is the terminal association strategy. If vehicle j and base station u cooperate to complete the reasoning task, then otherwise like Define the model segmentation point between vehicle j and base station u at this time as

[0115] Assume that the distance between vehicle j and base station u is The path loss model can be expressed as:

[0116]

[0117] Among them, P j is the transmission power at vehicle j; For distance The received power of base station u at location; λ is the path loss index; G j and G u are the antenna gains at vehicle j and base station u respectively.

[0118] Then the communication rate from vehicle j to base station u can be calculated as

[0119]

[0120] Among them, σ 2 is the noise power, is the average interference power when vehicle j communicates with base station u.

[0121] Based on the communication rate, at the model split point The unloading transmission time when vehicle j offloads part of the reasoning task to base station u is in, For model The output of the layer.

[0122] Secondly, vehicle j considers whether the reasoning task is assigned to the UAV and defines is the terminal association strategy. If vehicle j and UAV k cooperate to complete the reasoning task, then otherwise if The split point of the model between vehicle j and UAVk is defined as

[0123] Define [x j ,y j ] T and are the horizontal coordinates of ground vehicle j and UAV k, and the distance between ground vehicle j and UAV k is

[0124]

[0125] Then, by adopting the free space path loss model, the channel gain between device j and UAV k is:

[0126]

[0127] in represents the channel gain at the reference distance d0=1m, and θ is the path loss exponent. The communication rate at vehicle j is calculated as:

[0128]

[0129] Among them, P j is the uplink transmission power of j, is the achievable channel gain between vehicle j and UAV k, σ 2 is the noise power, is the average interference power when vehicle j and drone k communicate. According to the communication rate, vehicle j is at the model segmentation point The transmission delay experienced when offloading part of the reasoning task to UAV k is in, For model The output of the layer.

[0130] After receiving the offloaded cooperative reasoning task from the vehicle associated with UAV k, UAV k decides to process the task locally at UAV k or offload UAV k to the satellite according to the computing power available at UAV k; define As the terminal association strategy, vehicle j completes the reasoning task through the UAV k relay and satellite s cooperation. otherwise The split point of the model at this time is defined as

[0131] The available mmWave backhaul link capacity between UAV k and satellite s can be expressed as:

[0132]

[0133] in is the mmWave bandwidth between UAV k and the satellite, is the transmission power of UAV k, and is the antenna gain of the transmitter and receiver, L r is the attenuation factor, E s is the noise temperature, a is the Boltzmann constant, f c mm is the millimeter wave carrier frequency, reflects the distance between UAV k and satellite s. The transmission delay between UAV k and satellite is expressed as

[0134]

[0135] in, For model The output of the layer.

[0136] To solve the problem of computing resource allocation, we define Assign decision variables to computing units. If computing unit n at base station u is assigned to vehicle j, then otherwise

[0137] definition Assign decision variables to computational units. If computational unit l on UAV k is assigned to vehicle j, then otherwise

[0138] definition Assign decision variables to computational units. If computational unit m on satellite s is assigned to vehicle j, then otherwise

[0139] Step S2 specifically includes:

[0140] Based on the task offloading decision variables and resource allocation decision variables, data from vehicles, drones, satellites, and base stations, calculate the total delay experienced by vehicles in different allocation modes in the edge cooperative reasoning system;

[0141] In a further embodiment, first, if vehicle j decides to perform its reasoning task locally (i.e. ), then the total delay experienced by vehicle j to complete the reasoning task is:

[0142]

[0143] where ξ j is the computing power of vehicle j, α j It is the number of CPU cycles required to process one bit of data.

[0144] When , the time required for base station u to complete the inference task is:

[0145]

[0146] in For Layer Post-reasoning task A j The amount of data that needs to be processed. The total delay experienced by vehicle j and base station u to complete the cooperative reasoning task is:

[0147]

[0148] definition Assign decision variables to computational units. If computational unit l on UAV k is assigned to vehicle j, then otherwise

[0149] When , the time required for UAV k to complete the reasoning task is:

[0150]

[0151] in For Layer Post-reasoning task A j The amount of data that needs to be processed. The total delay experienced by vehicle j and UAV k to complete the cooperative reasoning task is:

[0152]

[0153] definition Assign decision variables to computational units. If computational unit m on satellite s is assigned to vehicle j, then otherwise

[0154] When , the time required for satellite s to complete the reasoning task is:

[0155]

[0156] Then, when the inference task of vehicle j is offloaded to the satellite, the total delay experienced by vehicle j is:

[0157]

[0158] Step S3 specifically includes:

[0159] Based on the goal of minimizing the total delay experienced by the vehicle, a resource allocation model for the edge cooperative reasoning system is constructed, and the proposed objective function is to minimize the total delay of the system while ensuring that the constraint of the number of associations is met. The objective function is defined as:

[0160]

[0161] Where W represents the task offloading decision variable, I represents the model segmentation decision variable, Z represents the resource allocation decision variable, and T j Indicates the total system delay, specifically:

[0162]

[0163] The constraints include:

[0164] Represents the constraints of computing resources. Indicates that each vehicle can only choose one cooperative reasoning mode.

[0165] It means that each computing unit can be assigned to at most one vehicle in one allocation cycle.

[0166] This means that the number of vehicles that base stations, drones, and satellites can associate with is limited.

[0167] Indicates resource allocation policy and terminal association policy.

[0168] at last, Indicates the model split point.

[0169] Step S4 specifically includes:

[0170] S41. Based on the goal of minimizing the total system delay, the original problem is decomposed into two sub-problems. Sub-problem 1 is the joint terminal association and resource allocation problem, and sub-problem 2 is the model segmentation problem.

[0171] S42, convert the sub-problem 1 into a sub-module problem for solution;

[0172] S43. For sub-problem 2, a model segmentation algorithm is proposed to place the model;

[0173] S44, solve these two sub-problems alternately until the optimal strategy is obtained;

[0174] According to one aspect of the present application, step S41 is specifically:

[0175] To solve the problem of associating strategies and resource allocation, Given subproblem P1 as:

[0176]

[0177] The constraints include:

[0178] Represents the constraints of computing resources. Indicates that each vehicle can only choose one cooperative reasoning mode.

[0179] It means that each computing unit can be assigned to at most one vehicle in one allocation cycle.

[0180] This means that the number of vehicles that base stations, drones, and satellites can associate with is limited.

[0181] Indicates resource allocation policy and terminal association policy.

[0182] Among them, the objective function D j (w,z)=T0-T j Aims to minimize delay cost, T0 is the task waiting time.

[0183] To solve the model placement problem, subproblem P2 is given:

[0184]

[0185] The constraints include:

[0186] Indicates the model split point.

[0187] According to one aspect of the present application, step S42 is specifically:

[0188] In order to transform subproblem P1 into a submodular problem, the computing resource allocation and terminal association strategy are defined without considering the satellite as

[0189]

[0190] in represents the action of allocating computing resource l on drone k to vehicle j, represents the action that the computing resource n on base station u is allocated to car j, represents the action of associating base station u with vehicle j, represents the action associated with UAV k and vehicle j. The basic set A can be divided into V disjoint sets (V′=UV), that is, for any v≠v′, in Similarly, the basis set B can be divided into Disjoint sets The basis set C can be divided into U disjoint sets, where Similarly, the basis set D can be divided into K disjoint sets.

[0191] Define a pair E = (F, G), where F = {A, B, C, D} is a global set, A set of independent subsets of F is used to replace the original constraints, defined as

[0192]

[0193] where X is the set of final solutions. It is not difficult to prove that the family of sets E = (F, G) is a pseudo-matrix, and the delayed reduction function is a monotone submodule function on X∈G. On this basis, by replacing the constraints in the problem with set pairs, subproblem P1 can be expressed as a pseudo-matrix constrained monotone submodule maximization problem:

[0194]

[0195] stX∈G

[0196] The embodiment of the present invention proposes a submodule optimization algorithm to solve this problem. First, initialize is an empty set, where v = 1, ..., V′, u=1,...,U,k=1,...,K, and define X={A,B,C,D} and Y X = F. In each iteration, the new element with the highest marginal gain is Add to Collection X X and from the candidate action set Y X Remove the candidate action set Y X is the set of all currently selectable strategies. In the case of , the action of allocating computing units on other base stations and drones to vehicle j, and the action of associating other base stations and drones with vehicle j should be removed from the candidate action set, that is:

[0197]

[0198] v'≠(u-1)V+1,(u-1)V+2,...,uV,

[0199]

[0200] and

[0201]

[0202] Repeat this iteration until the candidate action set Y X Empty or marginal gain Δ D (a|X) is zero. Wherein the marginal gain refers to: in a submodular function, when a new element is added to a set, the contribution of the element to the increase in the function value. In the present invention, the marginal gain Δ D (a|X)=D(a∪{X})-D(X).

[0203] When considering satellites, the embodiment of the present invention further proposes a reallocation method based on resource accumulation.

[0204] This method goes through a loop to check for each vehicle whether it can connect to the satellite via the drone.

[0205] First initialize X, Y X =F and the global optimal allocation strategy. as well as When the conditions are met, enter the loop to calculate the inference latency. If a suitable satellite access strategy is found and the inference latency can be reduced, update the global optimal allocation strategy. After the update, allocate the corresponding computing resources and mark the allocation as used. If the satellite's resource allocation is better than other options, choose to connect to the satellite through the drone and update the vehicle's computing power and latency.

[0206] According to one aspect of the present application, step S43 is specifically:

[0207] Step S43 specifically includes:

[0208] For sub-problem 2, the best split point is determined by traversal search. In the embodiment of the present invention, a model segmentation algorithm is proposed in consideration of the cooperative reasoning between the vehicle and the base station as follows:

[0209] Input: total number of layers of the DNN model I, bandwidth B, computing power of base station u The computing power of vehicle j ξ j , the amount of data output per layer Optimal split point i best .

[0210] 1. Minimum time to initialize vehicle j Calculate time locally

[0211] 2. Assume that the initial optimal segmentation layer i best =I.

[0212] 3. For each layer i∈{1,2,…,I} of the neural network in the inference task, perform the following operations:

[0213] Calculate the transmission time of the current layer Processing time at base station u and the processing time on vehicle j

[0214] If the total time to perform the inference task The minimum update time is And i best Set as current layer i.

[0215] 4. After the loop ends, return to the best segmentation point i best .

[0216] S44, solve these two sub-problems alternately until the optimal strategy is obtained;

[0217] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for optimizing the efficiency of space-ground integrated network reasoning based on sub-module optimization, characterized in that: The following steps are involved: S1. Building an edge cooperative reasoning system for reasoning on the air-space-ground integrated network, wherein the air-space-ground integrated network includes vehicles, drones, satellites, and base stations. The edge cooperative reasoning system is built based on the vehicles, drones, satellites, and base stations in the air-space-ground integrated network, and several cooperation modes are proposed for the reasoning tasks on the vehicles, and the reasoning delays of the reasoning tasks under the different cooperation modes are obtained; S2. Based on the edge cooperative reasoning system, construct a resource allocation model and a terminal association model of the edge cooperative reasoning system to obtain the total delay experienced by the vehicle under different allocation modes; S3. Construct a resource allocation model for the edge cooperative reasoning system, and propose an objective function and constraints, wherein the objective function is defined as: Where W represents the task offloading decision variable, I represents the model segmentation decision variable, Z represents the resource allocation decision variable, and T j represents the total delay experienced by car j; The constraints are the limitation of computing resources and the limit of the number of vehicles that can be associated with drones, satellites and base stations; S4. Decompose the optimization problem into sub-problems P1 and P2, namely, the joint terminal association and resource allocation problem and the model segmentation problem. The joint terminal association and resource allocation problem proposes a low-complexity association strategy based on sub-module optimization. The model segmentation problem is solved by traversing search to determine the best splitting point. The joint terminal association and resource allocation problem and the model segmentation problem are solved alternately until the optimal reasoning efficiency optimization strategy is obtained.

2. According to the method for optimizing the efficiency of air-ground integrated network reasoning based on sub-module optimization in claim 1, it is characterized in that: The S1 step is specifically as follows: S11, in the edge cooperative reasoning system, including the vehicle, the drone, the satellite, the base station and the reasoning task on the vehicle, obtaining the data of the vehicle, the drone, the satellite, the base station and the neural network parameters; The step S11 is specifically as follows: there are J vehicles in the edge cooperative reasoning system, the index of the vehicle is defined as J∈{1,2,…,J}, the base station includes U base stations, the index of the base station is defined as U∈{1,2,…,U}, there are K drones in the edge cooperative reasoning system, the index of the drone is defined as K∈{1,2,…,K}, and a satellite S; each vehicle in the network in the edge cooperative reasoning system has a reasoning task A j To perform, the ground vehicle performs cooperative reasoning with the base station, UAV or satellite within the communication range through the wireless channel; the base station allocates N computing units to perform the tasks of the ground vehicle, and the index of its computing unit is N∈{1,2,…,N}; the UAV allocates L computing units to perform the tasks of the ground vehicle, and the index of its computing unit is L∈{1,2,…,L}; the satellite allocates M computing units to perform the tasks of the ground vehicle, and the index of its computing unit is M∈{1,2,…,M}; the index of the neural network layer in the reasoning task is i∈{0,1,2,…,I}, where I is the total number of layers of the neural network; S12, constructing a task offloading decision variable, wherein the task offloading decision variable indicates that the reasoning task of the vehicle is completed locally or in cooperative reasoning with other terminals and servers; The S12 step is specifically as follows: The vehicle j∈J decides to perform its reasoning task locally, defining j∈J is the task offloading decision; the vehicle j completes its reasoning task A locally j ,but otherwise Considering the insufficient reasoning performance of the vehicle itself, the vehicle j∈J cooperates with the base station u∈U to complete the reasoning task; definition j∈J, u∈U is the task offloading decision. The vehicle j and the base station u cooperate to complete the reasoning task. otherwise like The model segmentation point between the vehicle j and the base station u is defined as The vehicle j considers whether the reasoning task is assigned to the UAV, where the UAV refers to the unmanned aerial vehicle, and defines j∈J, k∈K is the task offloading decision. The vehicle j and the UAVk cooperate to complete the reasoning task. otherwise The split point of the model between the vehicle j and the UAV k is defined as After receiving the offloaded cooperative reasoning task from the vehicle associated with the UAVk, the UAVk decides to process the task locally at the UAVk or offload the UAVk to the satellite according to the computing power available at the UAVk; define j∈J, k∈K is the terminal association strategy. The vehicle j completes the reasoning task through the UAV k relay and the satellite s. otherwise The split point of the model at this time is defined as S13, constructing a resource allocation decision variable, wherein the resource allocation decision variable indicates that the computing units on the drone, the satellite, and the base station are allocated to the reasoning task of the vehicle; The S13 step is specifically as follows: definition j∈J,u∈U,n∈N, assign decision variables to the computing units, and the computing unit n at the base station u is assigned to the vehicle j, then otherwise definition j∈J, k∈K, l∈L, assign decision variables to computing units, and computing unit l on the UAVk is assigned to the vehicle j, then otherwise definition j∈J,m∈M, assign decision variables to computing units, and computing unit m on the satellite s is assigned to the vehicle j, then otherwise S14, the data of the vehicle, the drone, the satellite, the base station and the neural network parameters, based on the task offloading decision variables and the resource allocation decision variables, calculate the unloading waiting delay of the reasoning task on the vehicle j in different allocation modes in the edge cooperative reasoning communication system; The communication rate from the vehicle j to the base station u is At the model split point The unloading transmission time for vehicle j to offload part of the reasoning task to base station u is u∈U, in, For model The output of the layer; The communication rate from vehicle j to UAVk is The vehicle j is at the model segmentation point The transmission delay experienced when offloading part of the reasoning task to the UAVk is in, For model The output of the layer.

3. The method for optimizing the efficiency of air-ground integrated network reasoning based on sub-module optimization according to claim 1 is characterized in that: The S2 step is specifically as follows: Calculate the total delay experienced by the vehicle in different allocation modes in the edge cooperative reasoning system based on the resource allocation decision variable, the task offloading decision variable, the vehicle, the drone, the satellite, the base station data, and the neural network parameters; The vehicle j decides to perform its reasoning task locally, i.e. Then the total delay experienced by vehicle j to complete the reasoning task is: where ξ j is the computing power of vehicle j, α j is the number of CPU cycles required to process one bit of data; When , the time required for the base station u to complete the inference task is: in For Layer Post-reasoning task A j The amount of data that needs to be processed, the total delay experienced by the vehicle j and the base station u to complete the cooperative reasoning task is: definition j∈J, k∈K, l∈L are the decision variables for computing unit assignment. The computing unit l on the UAVk is assigned to the vehicle j. otherwise When , the time required for UAVk to complete the reasoning task is: in For Layer Post-reasoning task A j The amount of data that needs to be processed, the total delay experienced by the vehicle j and the UAV k to complete the cooperative reasoning task is: definition j∈J,m∈M, assign decision variables to computing units, and computing unit m on the satellite s is assigned to the vehicle j, then otherwise When , the time required for the satellite s to complete the reasoning task is: When the inference task of vehicle j is offloaded to the satellite, the total delay experienced by vehicle j is:

4. The method for optimizing the efficiency of air-ground integrated network reasoning based on sub-module optimization according to claim 1 is characterized in that: The S3 step is specifically as follows: The objective function is specifically: The constraints include: Represents constraints on computing resources; It means that each vehicle can only choose one cooperative reasoning mode; It means that each computing unit can be assigned to at most one vehicle in one allocation cycle; Indicates that the base station, the drone, and the satellite can be associated with a limited number of vehicles; Indicates resource allocation strategy and terminal association strategy; Indicates the model split point; in, represents the upper limit of the number of vehicles that can be accessed by the base station, the drone, and the satellite, N max , L max , M max Indicates the upper limit of the computing resources of the base station, the drone, and the satellite.

5. The method for optimizing the efficiency of air-ground integrated network reasoning based on sub-module optimization according to claim 1 is characterized in that: The S4 step is specifically as follows: S41. Questions The sub-problem P1 is given as: The constraints include: Represents constraints on computing resources; It means that each vehicle can only choose one cooperative reasoning mode; It means that each computing unit can be assigned to at most one vehicle in one allocation cycle; Indicates that the base station, the drone, and the satellite can be associated with a limited number of vehicles; Indicates resource allocation strategy and terminal association strategy; Among them, the objective function D j (w,z)=T0-T j Aims to minimize delay costs, T0 is the task waiting time; Given the subproblem P2: The constraints include: Indicates the model split point; S42, converting the joint terminal association and resource allocation problem into a submodule problem for solution; The subproblem P1 is transformed into a submodular problem, and the computing resource allocation and terminal association strategy are defined as in represents the action that the computing resource l on the drone k is allocated to the vehicle j, represents the action that the computing resource n on the base station u is allocated to the vehicle j, represents the action of associating the base station u with the vehicle j, represents the action associated with the drone k and the vehicle j; the basic set A can be divided into V disjoint sets (V′=UV), that is, for any v≠v′, in v = (u-1) V + n; the basis set B can be divided into Disjoint sets The basis set C can be divided into U disjoint sets, where The basis set D can be divided into K disjoint sets; Define a pair E = (F, G), where F = {A, B, C, D} is a global set, A set of independent subsets of F is used to replace the original constraints, defined as Where X is the set of final solutions; prove that the set family E = (F, G) is a pseudo-matrix, delay reduction function is a monotone submodule function on X∈G; on this basis, by replacing the constraints in the problem with set pairs, the subproblem P1 is expressed as a pseudo-matrix constrained monotone submodule maximization problem: stX∈G Initialize A v , C j , D j is an empty set, where v = 1, ..., V′, u=1,...,U,k=1,...,K, and define X = {A, B, C, D} and Yx = F; in each iteration, the new element with the highest marginal gain is Add to the set X x and remove from the candidate action set Y x, the candidate action set Y x is the set of all currently selectable strategies; In the case of , the computing units on other base stations and drones are assigned to the action of the vehicle j, and the actions of other base stations and drones associated with the vehicle j are removed from the candidate action set, that is: Y=YA v′ ,v′=1,...,V′ v′≠(u-1)V+1,(u-1)V+2,...,uV, and Repeat this iteration until the candidate action set Y x is empty or the marginal gain Δ D (a|X) is zero; wherein the marginal gain refers to: in a submodular function, when a new element is added to a set, the contribution of the element to the increase in the function value; in the present invention, the marginal gain Δ D (a|X)=D(a∪{X})-D(X); When considering satellites, a redistribution method based on resource accumulation is used to check whether each vehicle can be connected to the satellite through a drone through a loop. The specific steps are as follows: Initialize X, Y x=F and the global optimal allocation strategy; when as well as When the conditions are met, enter the loop to calculate the inference latency; if a suitable satellite access strategy is found and the inference latency can be reduced, update the global optimal allocation strategy; after the update, allocate the corresponding computing resources and mark the allocation as used; if the satellite's resource allocation is better than other options, choose to connect to the satellite through the drone, and update the vehicle's computing power and latency; S43, model segmentation algorithm to place the model, The step S43 specifically includes: The best split point is determined by traversal search. Considering the cooperative reasoning between vehicles and base stations, the model segmentation algorithm is proposed as follows: Input: total number of layers of the DNN model I, bandwidth B, computing power of base station u The computing power of vehicle j ξ j , the amount of data output per layer Optimal split point i best ; ① Minimum time to initialize vehicle j Calculate time locally ② Assume that the initial optimal segmentation layer i best =I; ③For each layer i∈{1,2,…,I} of the neural network in the inference task, perform the following operations: Calculate the transmission time of the current layer Processing time at base station u and the processing time on vehicle j If the total time to perform the inference task The minimum update time is And i best Set to current layer i; ④After the cycle ends, return to the best segmentation point i best ; S44. Solve the two sub-problems alternately until the optimal strategy is obtained.

Images