A method for offloading tasks in a wireless network

Abstract

The present invention provides a method for task offloading of a wireless network, which belongs to the field of communication technology. The method includes: constructing a STAR‑RIS-assisted drone network model, a communication model, and an energy consumption model, and determining a first objective function and a first constraint; for the task offloading problem, optimizing the task offloading using the KKT condition to obtain an optimal task offloading decision vector; for the transmission resource problem, optimizing the resource allocation using the continuous convex approximation method SCA; performing alternate solutions until convergence to obtain the optimal solution for task offloading and resource allocation. The present invention jointly optimizes the two sub-problems of task offloading and resource allocation to minimize the total energy consumption of the system, adopts a hovering drone equipped with a server as a base station, and jointly optimizes the amplitude, phase shift, and task offloading ratio of STAR‑RIS. The energy consumption in the energy splitting mode is minimized.

Description

Technical Field

[0001] The present invention belongs to the field of communication technology, and in particular relates to a task offloading method for a wireless network. Background Art

[0002] Reconfigurable Intelligent Surface (RIS) is a key technology in the sixth generation (6G) communication system. It can dynamically reconfigure the wireless propagation environment through passive beamforming to improve channel conditions and coverage. Specifically, RIS is a panel that integrates many low-cost passive reflective elements and implements fine-grained beamforming by jointly optimizing the phase of each reflective element. This not only increases the signal strength of the actual receiver, but also mitigates its signal strength interference. The energy consumption of RIS without active radio frequency (RF) chain is low, which improves the energy efficiency of the system. Applying RISs to mobile edge computing (MEC) networks can optimize the performance of task offloading. Simultaneously Transmitting And Reflecting Reconfigurable Intelligent Surface (STAR-RIS)-assisted task offloading in MEC networks has attracted great attention in improving system performance and user quality of experience (QoE).

[0003] The combination of drones and RIS can obtain additional air-to-ground communication advantages and further improve the communication quality. In order to maximize the average confidentiality rate of drone-assisted drone wireless networks, the flight trajectory, transmission beam and phase shift of drones are jointly optimized in the prior art. The prior art jointly optimizes the position of drones in three-dimensional space and the phase shift of RIS, aiming to improve the data transmission rate of drones and reduce the propulsion energy of drones. The prior art studies RIS-assisted multi-drone NOMA networks, jointly optimizing the spatial position and transmission power of drones, the reflection matrix of RIS and the SIC decoding order between users to maximize the total rate.

[0004] However, the RIS covering the building surface can only provide reflected signals to users on one side, which limits the coverage of the RIS communication system. The existing STAR-RIS can effectively solve this problem. The prior art studies the drone communication system of the Intelligent OmniSurface (IOS) and improves the execution rate by jointly optimizing the flight trajectory of the drone and the phase shift of the IOS. The prior art improves the total rate of the system by optimizing the STAR-RIS beamforming vector, the drone flight trajectory and the power allocation. The prior art actively adjusts the spatial trajectory of the drone and the active beamformer through reinforcement learning (RL) to maximize the total rate of all users.

[0005] The traditional RIS covering the surface of the building can only provide reflected signals to users on one side, which limits the coverage of the RIS communication system. Although the existing STAR-RIS can effectively solve this problem, the STAR-RIS at the edge of the cell only has the functions of transmitting and reflecting signals. At present, with the popularization of the Internet and the advancement of science and technology, the user base is showing an exponential growth trend, and the phenomenon of massive users accessing the network concurrently is becoming more and more common. In order to alleviate the problem of severe resource consumption and tight allocation, the computing center needs to be further tilted to the edge of the network, so that the computing-intensive and delay-sensitive tasks from mobile users can be processed at the edge as much as possible, so as to further alleviate the allocation pressure of key resources of the backbone network such as computing, storage and transmission resources. Summary of the invention

[0006] In view of the above-mentioned deficiencies in the prior art, the present invention provides a method for offloading tasks in a wireless network, which solves the problem of minimizing the total energy consumption of the network under the assisted network of a UAV and STAR-RIS.

[0007] In order to achieve the above object, the technical solution adopted by the present invention is: a task offloading method for a wireless network, comprising the following steps:

[0008] S1. Construct and calculate the STAR-RIS-assisted drone network model, communication model and energy consumption model, and determine the first objective function and the first constraint condition;

[0009] S2. For the task offloading problem, according to the first objective function and the first constraint condition, the Carlo-Kuhn-Tucker condition KKT is used to optimize the task offloading and obtain the optimal task offloading decision vector;

[0010] S3. For the transmission resource problem, the continuous convex approximation method SCA is used to optimize the resource allocation according to the optimal task offloading decision vector;

[0011] S4. Step S2 and step S3 are solved alternately until convergence, and an optimal solution for task offloading and resource allocation is obtained, thereby completing task offloading of the wireless network.

[0012] The beneficial effects of the present invention are as follows: the present invention provides a scenario of applying computational STAR-RIS to wireless communication networks, proposes a new task offloading and resource allocation problem to minimize the energy consumption of computational STAR-RIS-assisted wireless networks, and studies the energy-saving task offloading and resource allocation problems in STAR-RIS with computing capabilities in unmanned aerial vehicle communication networks for the first time. The present invention decomposes the energy minimization problem into a task offloading problem and a transmission resource allocation problem. For the task offloading problem, the (Karush-Kuhn-Tucker, KKT) condition is used, and for the transmission resource allocation sub-problem, the STAR-RIS beamforming is optimized using the Successive Convex Approximation (SCA). The present invention uses a hovering drone equipped with a server as a base station to jointly optimize the amplitude, phase shift and task offloading ratio of STAR-RIS, and the energy consumption in the energy splitting (ES) mode is minimized.

[0013] Furthermore, the step S1 includes the following steps:

[0014] S101, analyzing and calculating resource allocation in a STAR-RIS-assisted drone network model, and constructing a STAR-RIS-assisted drone network model, wherein:

[0015] The computing STAR-RIS-assisted drone network model includes a single-antenna drone with a MEC server, a single-antenna mobile user, and a computing STAR-RIS having M elements and installed on a building surface, wherein the computing STAR-RIS element transfers the task offloaded by the user to the MEC server deployed on the drone; the computing STAR-RIS includes M integrated on a panel c computing elements and M s Reflective and transmissive elements;

[0016] S102, building a communication model;

[0017] S103, splitting the system energy consumption into energy consumption for calculating STAR-RIS, energy consumption for users, energy consumption for MEC servers and energy consumption for drones, and constructing energy consumption models for calculating STAR-RIS, energy consumption for users, energy consumption for MEC servers and energy consumption for drones based on the communication model;

[0018] S104. According to the determined problem of minimizing the total energy consumption of STAR-RIS, mobile users, MEC servers and drones, a first objective function and a first constraint condition are obtained.

[0019] The beneficial effect of the above further scheme is: the energy-saving task offloading and resource allocation problem of computing STAR-RIS in a complex UAV communication network is converted into an easy-to-understand and process mathematical model through abstraction and simplification. Based on the system model, the problem of minimizing the total energy consumption of computing STAR-RIS, mobile users, MEC servers and UAVs is determined, and the first objective function and constraints are obtained, which lays the foundation for the subsequent model parameter adjustment and optimization to improve the system performance, thereby providing an effective strategy for solving the problem of minimizing the total energy consumption.

[0020] Furthermore, the communication model in step S102 includes:

[0021] The channel where the task is executed locally:

[0022]

[0023] Among them, R i,l represents the information transmission rate from user i to relay l for computing STAR-RIS, B i,l represents the link bandwidth from user i to relay l for computing STAR-RIS, P i represents the transmission power of user i, h i,l represents the channel coefficient from user i to relay l for computing STAR-RIS, represents the variance of the noise power in the link from user i to relay l where STAR-RIS is calculated;

[0024] Channels for task offloading:

[0025]

[0026] Among them, R i,u represents the information transmission rate from user i to the UAV, B i,u represents the transmission bandwidth of the link from user i to the UAV through the calculation of the STAR-RIS surface, represents the channel coefficient matrix from STAR-RIS to the UAV, H i,s represents the channel coefficient matrix from user i to calculate STAR-RIS, represents the variance of the noise power in the link from the user to the drone, Θ i Represents the transmission and reflection coefficient matrix of the signal of user i.

[0027] The beneficial effect of the above further scheme is: the communication model established by the present invention is further analyzed and divided into two parts: the channel where the task is executed locally and the channel where the task is offloaded, and the corresponding link information transmission rate is derived.

[0028] Furthermore, the energy consumption model in step S103 includes:

[0029] Calculate the STAR-RIS energy consumption model:

[0030]

[0031]

[0032] in, represents the energy consumption of STAR-RIS for processing local tasks, κ represents the capacitance coefficient, τ l represents the time consumed by STAR-RIS to process local tasks, f s Indicates the CPU frequency for calculating STAR-RIS, α i represents the task offloading rate ratio of user i, L i represents the size of the task transferred by user i, C i represents the computing resources required to process the task of user i, where i, r and t all represent users;

[0033] User energy consumption model:

[0034]

[0035] in, Represents user energy consumption, P i represents the transmission power of user i, R i,l and R i,u denote the information transmission rate from user i to the relay l for computing STAR-RIS and from user i to the UAV, respectively;

[0036] MEC server energy consumption model:

[0037]

[0038]

[0039] in, represents the energy consumption of the MEC server deployed on the drone to perform the offloading task, τ o represents the time consumed by the MEC server to process the user's offload task, f u Indicates the CPU main frequency of the MEC server deployed on the drone;

[0040] UAV energy consumption model:

[0041]

[0042] in, represents the minimum energy required for the drone to hover at the same position, m u represents the mass of the drone, g represents the gravitational acceleration at the drone's position, A represents the area of ​​the rotor, and ρ represents the air density.

[0043] The beneficial effect of the above further scheme is: the energy consumption model established by the present invention is further analyzed and divided into four parts: calculating STAR-RIS energy consumption, user energy consumption, MEC server energy consumption and drone energy consumption, providing a basis for the subsequent objective function construction.

[0044] Furthermore, the expression of the first objective function in step S104 is as follows:

[0045]

[0046]

[0047]

[0048] Among them, α represents the task offloading ratio, Θ represents the transmission and reflection coefficient matrix, and E tot (α, Θ) represents the total energy of the STAR-RIS-assisted UAV communication network, min represents the minimization operation, represents the energy consumption of calculating STAR-RIS to process local tasks, Indicates the user energy consumption, represents the energy consumption of the MEC server deployed on the drone to perform the offloading task, represents the minimum energy required for the drone to hover at the same position, i, r and t all represent users, Indicates the Mth s The reflection coefficient of each element, T' represents the total delay of task offloading;

[0049] The expression of the first constraint condition is as follows:

[0050]

[0051]

[0052]

[0053]

[0054]

[0055]

[0056] Among them, L i represents the size of the task transmitted by user i, α i represents the task offloading ratio of user i, R i,l represents the information transmission rate from user i to relay l for computing STAR-RIS, C i represents the computing resources required to process the task of user i, f s Indicates the CPU frequency for calculating STAR-RIS, R i,u represents the information transmission rate from user i to the drone, f u Indicates the CPU main frequency of the MEC server deployed on the drone. represents the amplitude of the mth STAR-RIS element of the user r signal, represents the amplitude of the mth STAR-RIS element of the user t signal, represents the phase shift of the m-th STAR-RIS element of the user r signal, Represents the phase shift of the mth STAR-RIS element of user t's signal.

[0057] The beneficial effect of the above further scheme is: based on the above system modeling, around the research goal of the present invention, that is, minimizing the total network energy consumption, the expressions of the first objective function and the first constraint condition are constructed.

[0058] Furthermore, step S2 includes the following steps:

[0059] S201. For the task offloading problem, fix the transmission and reflection coefficient matrices Θ, and rewrite the first objective function and the first constraint condition:

[0060] S202, introducing an approximate quadratic regularization term into the rewritten first objective function to obtain a second objective function and a second constraint condition;

[0061] S203: According to the second objective function and the second constraint, the task offloading is optimized using the Carlo-Kuhn-Tucker condition (KKT) to obtain an optimal task offloading decision vector.

[0062] The beneficial effect of the above further scheme is: for the task offloading sub-problem, the first objective function and the first constraint condition are rewritten by mathematical means, the task offloading is optimized using the KKT condition, and the optimal task offloading decision vector is obtained.

[0063] Furthermore, the second objective function expression in step S202 is as follows:

[0064]

[0065] The expression of the second constraint is as follows:

[0066]

[0067]

[0068]

[0069]

[0070] Among them, min represents the minimization operation, E tot (α, Θ 0 ) represents the total energy consumption function, α represents the task offloading ratio, Θ 0 Represents the signal transmission and reflection coefficient matrix, R i,l represents the information transmission rate from user i to relay l for computing STAR-RIS, L i represents the size of the task transmitted by user i, α i represents the task offloading ratio of user i, R i,l represents the information transmission rate from user i to relay l for computing STAR-RIS, f s Indicates the CPU frequency for calculating STAR-RIS, C i represents the computing resources required to specify the task of user i, represents the information transmission rate from user i to the UAV under fixed Θ, Θ represents the transmission and reflection coefficient matrix, f u represents the CPU main frequency of the MEC server deployed on the drone, T' represents the total delay of task offloading, and B i,u represents the link bandwidth from user i to relay l for computing STAR-RIS, represents the channel coefficient matrix from STAR-RIS calculation to the UAV, represents the transmission and reflection coefficient matrix of the signal of a given user i, H i,s represents the channel coefficient matrix from user i to calculate STAR-RIS, represents the variance of the noise power in the link from the mobile user to the drone.

[0071] Furthermore, the expression of the optimal task offloading decision vector in step S203 is as follows:

[0072]

[0073] ξ a =(ξ a,r ,ξ a,t ) T

[0074] φ a =(φ a,r ,φ a,t ) T

[0075] ζ a =(ζ a,r ,ζ a,t ) T

[0076] ω a =(ω a,r ,ω a,t ) T

[0077] Among them, α* represents the optimal task offloading decision vector, min represents the minimization operation, and E tot (α, Θ 0 ) represents the total energy consumption function, α represents the unloading ratio vector, Θ represents the transmission and reflection coefficient matrix, ξ a ,φ a ,ω a And a Both represent the non-negative Lagrange multiplier vector of the second constraint, ξ a,r ,φ a,r ,ζ a,r ,ω a,r and a,t ,φ a,t ,ζ a,t ,ω a,t denote the non-negative Lagrange multipliers of user r and user t respectively, and T denotes the transpose.

[0078] The beneficial effect of the above further solution is: according to the solution of the optimal task offloading decision vector obtained, the expression of the optimal task offloading decision vector is rewritten.

[0079] Furthermore, step S3 includes the following steps:

[0080] S301. For the transmission resource problem, according to the optimal task offloading decision vector, a task offloading ratio at a certain point in the feasible solution set is given, and the first objective function and the first constraint condition are rewritten;

[0081] S302, introducing an approximate variable into the rewritten first objective function to obtain a third objective function and a third constraint condition;

[0082] The expression of the third objective function is as follows:

[0083]

[0084] The expression of the third constraint condition is as follows:

[0085]

[0086]

[0087]

[0088]

[0089]

[0090] Among them, min represents the minimization operation, v i represents the transmission and reflection coefficient matrix of user i, represents the information transmission rate from user i to the UAV including the approximate variable, L i represents the size of the task transferred by user i, represents the optimal uninstall ratio of user i, P i represents the transmission power of user i, B i,u represents the transmission bandwidth of the link from user i to the UAV through the calculation of the STAR-RIS surface, An approximate variable representing the signal-to-noise ratio of the link from user i to the drone, f u Indicates the CPU main frequency of the MEC server deployed on the drone. represents the transpose of the transmission and reflection coefficient matrix of user i after the nth iteration, Y represents the intermediate variable for simplifying the operation, represents the transmission and reflection coefficient moments of user i after the nth iteration, represents the transpose of the transmission and reflection coefficient matrix of user i, represents the variance of the noise power in the link from the mobile user to the drone, m represents the mth computing STAR-RIS element, represents the square of the transmission and reflection coefficient matrix of user i, Indicates user The square of the transmission and reflection coefficient matrix, represents the transpose of the channel gain matrix from STAR-RIS calculation to the UAV, H i,s represents the channel gain matrix from user i to calculate STAR-RIS, represents the transpose of the channel gain matrix from user i to calculate STAR-RIS, φ s,u represents the channel gain matrix from STAR-RIS to the UAV;

[0091] S303, according to the third objective function and the third constraint condition, deriving the lower bound of the non-convex constraint by using the continuous convex approximation method SCA;

[0092] S304. According to the derivation results, the Lagrange dual method is used to solve the fixed transmission and reflection coefficient matrices to obtain the optimal resource allocation.

[0093] The above further scheme has the beneficial effect that when the task offloading decision vector is given, the original problem objective is transformed into determining the optimal resource allocation strategy to minimize energy consumption. The solution for the optimal resource allocation is obtained by a continuous convex approximation method.

[0094] Furthermore, the expression of the optimal resource allocation is as follows:

[0095]

[0096]

[0097]

[0098] in, represents the optimal transmission and reflection coefficient matrix, v i represents the transmission and reflection coefficient matrix of user i, min represents the minimization operation, represents the Lagrangian dual function, L i represents the size of the task transferred by user i, represents the optimal uninstall ratio of user i, P i represents the transmission power of user i, R i,u represents the information transmission rate from user i to the drone, represents the Lagrange multiplier of user i, represents the information transmission rate from user i to the UAV including the approximate variables, C i represents the computing resources required to process the task of user i, f u Indicates the CPU main frequency of the MEC server deployed on the drone. represents the Lagrange multiplier of user i, f s represents the CPU main frequency on STAR-RIS, T' represents the total delay of task offloading, represents the Lagrange multiplier of user i, m represents the mth computational STAR-RIS element, represents the square of the transmission and reflection coefficient matrix of user i, Indicates user The square of the transmission and reflection coefficient matrix, η i represents the Lagrange multiplier of user i, An approximate variable representing the signal-to-noise ratio of the link from user i to the drone, represents the variance of the noise power in the link from the mobile user to the drone, represents the transpose of the transmission and reflection coefficient matrix of user i after the nth iteration, Y represents the intermediate variable for simplifying the operation, represents the transmission and reflection coefficient matrix of user i after the nth iteration, Re represents the real part operation of a complex number, represents the transpose of the transmission and reflection coefficient matrix of user i, represents the transpose of the channel gain matrix from STAR-RIS calculation to the UAV, H i,s represents the channel gain matrix from user i to calculate STAR-RIS, represents the transpose of the channel gain matrix from user i to calculate STAR-RIS, φ s,u represents the channel gain matrix from STAR-RIS to the UAV.

[0099] The beneficial effect of the above further solution is: according to the obtained solution of the optimal resource allocation, the expression for obtaining the optimal resource allocation is rewritten. BRIEF DESCRIPTION OF THE DRAWINGS

[0100] Figure 1 The present invention is a flow chart of the method.

[0101] Figure 2 The STAR-RIS-assisted UAV wireless communication scene graph is calculated for the present invention.

[0102] Figure 3 This is a simulation diagram of the impact of the number of iterations on the performance of the present invention.

[0103] Figure 4 This is a simulation diagram of the impact of network bandwidth on performance in the present invention.

[0104] Figure 5 This is a simulation diagram of the impact of user transmission power on performance in the present invention. DETAILED DESCRIPTION

[0105] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0106] Example

[0107] like Figure 1 As shown, the present invention provides a method for offloading tasks in a wireless network, and the implementation method thereof is as follows:

[0108] S1. Construct and calculate the STAR-RIS-assisted drone network model, communication model and energy consumption model, and determine the first objective function and the first constraint condition, which are specifically:

[0109] S101, analyzing and calculating resource allocation in a STAR-RIS-assisted drone network model, and constructing a STAR-RIS-assisted drone network model, wherein:

[0110] The computing STAR-RIS-assisted drone network model includes a single-antenna drone with a MEC server, a single-antenna mobile user, and a computing STAR-RIS having M elements and installed on a building surface, wherein the computing STAR-RIS element transfers the task offloaded by the user to the MEC server deployed on the drone; the computing STAR-RIS includes M integrated on a panel c computing elements and M s Reflective and transmissive elements;

[0111] S102, building a communication model;

[0112] S103, splitting the system energy consumption into energy consumption for calculating STAR-RIS, energy consumption for users, energy consumption for MEC servers and energy consumption for drones, and constructing energy consumption models for calculating STAR-RIS, energy consumption for users, energy consumption for MEC servers and energy consumption for drones based on the communication model;

[0113] S104. According to the determined problem of minimizing the total energy consumption of STAR-RIS, mobile users, MEC servers and drones, a first objective function and a first constraint condition are obtained.

[0114] In this embodiment, the present invention first analyzes the resource allocation in the computing STAR-RIS-assisted drone network model, and then constructs the computing STAR-RIS-assisted drone network model, communication model and energy consumption model.

[0115] In this embodiment, the STAR-RIS-assisted drone network model is analyzed and constructed as follows:

[0116] The proposed STAR-RIS-assisted drone access network deploys a MEC server, such as Figure 2 The UAV access network consists of a single-antenna UAV with a MEC server, two single-antenna users, and a computing STAR-RIS with M elements. The computing STAR-RIS includes M c computing elements and M s reflective and transmissive elements are integrated on a panel, M = M c +M s, the computing STAR-RIS is installed on the surface of the building to help air-to-ground communication by changing the propagation environment. The relay of STAR-RIS is represented by l, and the user uploads the task to the CPU unit for processing. The STAR-RIS element can transfer part of the task offloaded by the user to the MEC deployed on the drone for processing. It is assumed that the STAR-RIS element works in ES mode. The computing STAR-RIS also has the ability to process some tasks uploaded by the user and transmit and reflect another part of the task to the MEC server. User t and user r are located inside and outside the building, respectively. Since the connection between MEC and the drone uses low-latency, interference-free, and high-communication capacity optical fiber, the data transmission delay of the link between the drone and the MEC server is not considered in this process. In addition, considering that drones often have higher transmission power and the calculation results are usually small, we ignore the delays in the local execution and task offloading stages.

[0117] In this embodiment, the communication model is analyzed and constructed as follows:

[0118] The channel coefficients from user i∈{r,t} to relay l, from user i∈{r,t} to the STAR-RIS calculation, and from the STAR-RIS calculation to the drone are denoted by h i,l , and express. is the transmission and reflection coefficient matrix of STAR-RIS for the incident signal from user i, where are the amplitude and phase shift of the mth STAR-RIS element of the user i signal, M s Indicates the number of components, Indicates that the user i signal is in the Mth s The amplitude coefficient on the element, j represents the imaginary unit, Indicates that the user i signal is in the Mth s Therefore, the following constraints are required when calculating STAR-RIS in ES mode:

[0119]

[0120] in, represents the amplitude coefficient of user i’s signal on m elements, Indicates user The amplitude coefficient of the signal at the mth element, when i = r, on the contrary

[0121] Orthogonal Frequency Division Multiple Access (OFDMA) is considered in both phases of computing the STAR-RIS assisted computing system. Therefore, the interference in the local execution and task offloading phases can be modeled as follows.

[0122] 1) Channels where tasks are executed locally

[0123] The signal-to-noise ratio (SNR) and transmission rate of the wireless link from user i to relay l are expressed as:

[0124]

[0125] Among them, B i,l represents the link bandwidth from user i to relay l, P i are the transmission powers of user r and user t respectively, let is the variance of the noise power in the link from user to relay l, R i,l represents the information transmission rate from user i to relay l for computing STAR-RIS, h i,l represents the channel coefficient from user i to relay l for computing STAR-RIS.

[0126] 2) Channel for task offloading

[0127] Similarly, the link from user i to the drone is expressed by the signal-to-noise ratio and transmission rate of STAR-RIS after calculation:

[0128]

[0129] Among them, R i,u represents the information transmission rate from user i to the UAV, B i,u represents the transmission bandwidth of the link from user i to the UAV through the calculation of the STAR-RIS surface, represents the channel coefficient matrix from STAR-RIS to the UAV, H i,s represents the channel coefficient matrix from user i to calculate STAR-RIS, represents the variance of the noise power in the link from the user to the drone, Θ i Represents the transmission and reflection coefficient matrix of the signal of user i.

[0130] In this embodiment, the energy consumption model is analyzed and constructed as follows:

[0131] First, the system energy consumption is divided into four parts: calculating STAR-RIS energy consumption, user energy consumption, MEC server energy consumption, and drone energy consumption. Then, based on the communication model, energy consumption models are constructed for the four parts respectively.

[0132] 1) Calculate the energy consumption of STAR-RIS

[0133] Since the energy consumption of RIS itself is very low, the energy consumption of computing STAR-RIS is mainly dominated by the local task processing of the CPU unit. r and L t They are respectively represented as the task size (measured in bits) transmitted by user r and user t. Since the partial task offloading scenario is considered in the system of the present invention, the time consumed by STAR-RIS to process the local task is written as:

[0134]

[0135] Among them, τ l represents the time consumed by STAR-RIS to process local tasks, α i ∈{0,1} is the task offloading rate of user i, and α=(α r ,α t ) T , α r ,α t represent the uninstall ratios of user r and user t, respectively, and C i Specifies the computing resources required to process the task of user i, that is, the number of CPU cycles required to complete 1 bit of input data, f s is the CPU frequency (cycles per second) used to calculate STAR-RIS, L i represents the size of the task transmitted by user i (measured in number of bits).

[0136] For the CPU elements of Compute STAR-RIS, a dynamic voltage frequency scaling (DVFS) model can be used to reduce energy consumption. The energy consumption of Compute STAR-RIS processing local tasks can be expressed as:

[0137]

[0138] The capacitance coefficient is affected by the design of the computational STAR-RIS chip and is denoted by κ.

[0139] 2) User energy consumption

[0140] The energy of the user is consumed in transmitting local and offloaded tasks. Considering the different transmission links in the local and offloaded task phases, the energy consumption of user i can be expressed as:

[0141]

[0142] in, Represents user energy consumption, P i represents the transmission power of user i, R i,l and R i,u denote the information transmission rate from user i to relay l for computing STAR-RIS and from user i to the UAV, respectively.

[0143] 3) MEC server energy consumption

[0144] The time consumed by the MEC server to process the user's offload task is recorded as

[0145]

[0146] Among them, f u is the CPU main frequency of the MEC server deployed on the drone. The energy consumption of the MEC server deployed on the drone to perform the offload task can be expressed as:

[0147]

[0148] in, represents the energy consumption of the MEC server deployed on the drone to perform the offloading task, τ o represents the time consumed by the MEC server to process the user's offload task, f u Indicates the CPU main frequency of the MEC server deployed on the drone.

[0149] 4) UAV energy consumption

[0150] The minimum energy required for a drone to hover in the same position is:

[0151]

[0152] in, represents the minimum energy required for the drone to hover at the same position, m u represents the mass of the drone, g represents the gravitational acceleration at the drone's position, A represents the area of ​​the rotor, and ρ represents the air density.

[0153] In this embodiment, the first objective function and the first constraint condition are determined and optimized as follows:

[0154] From equations (5), (6), (8), and (9), we can see that the total energy of the STAR-RIS-assisted UAV communication network proposed by us is E tot It is expressed as follows:

[0155]

[0156] The problem of minimizing the total energy consumption of STAR-RIS, mobile users, MEC servers, and drones can be expressed as (11), where Θ represents the transmission and reflection coefficient matrix, Indicates the Mth s The reflection coefficient of each element, T' represents the total delay of task offloading.

[0157]

[0158]

[0159]

[0160]

[0161]

[0162]

[0163]

[0164] Among them, L i represents the size of the task transmitted by user i, α i represents the task offloading ratio of user i, R i,l represents the information transmission rate from user i to relay l for computing STAR-RIS, C i represents the computing resources required to process the task of user i, f s Indicates the CPU frequency for calculating STAR-RIS, R i,u represents the information transmission rate from user i to the drone, f u Indicates the CPU main frequency of the MEC server deployed on the drone. represents the amplitude of the mth STAR-RIS element of the user r signal, represents the amplitude of the mth STAR-RIS element of the user t signal, represents the phase shift of the m-th STAR-RIS element of the user r signal, Represents the phase shift of the mth STAR-RIS element of user t's signal.

[0165] Constraint (11b) Constraint (11c) indicates that the task processed by the CPU must be transmitted within the allocated time interval. Constraint (11d) requires that the transmission time of the user offloaded task must not exceed the allocated offload time interval τ o Constraint (11e) indicates that the task offloading ratio of each user varies between 0 and 1. Constraints (11f) and (11g) specify the requirements for the amplitude and phase shift of the incident signal to the mth STAR-RIS unit, respectively.

[0166] S2. According to the first objective function and the first constraint condition, for the task offloading problem, the KKT condition is used to optimize the task offloading, and the optimal task offloading decision vector is obtained, which is specifically:

[0167] S201. For the task offloading problem, fix the transmission and reflection coefficient matrices Θ, and rewrite the first objective function and the first constraint condition:

[0168] S202, introducing an approximate quadratic regularization term into the rewritten first objective function to obtain a second objective function and a second constraint condition;

[0169] S203. According to the second objective function and the second constraint condition, the task offloading is optimized using the KKT condition to obtain an optimal task offloading decision vector.

[0170] In this embodiment, KKT condition optimization task offloading is adopted: First, given a fixed Θ, solve problem (11). 0 When , the goal of problem (11) is to make the optimal unloading decision to minimize the energy consumption.

[0171] In this embodiment, the transmission and reflection coefficient matrices Θ are fixed, and the objective function and constraints are rewritten, so:

[0172]

[0173] Among them, E tot (α, Θ 0 ) represents the total energy consumption function, P i represents the transmission power of user i, R i,l represents the information transmission rate from user i to relay l for computing STAR-RIS, represents the information transmission rate from user i to the UAV under a given transmission and reflection coefficient matrix Θ.

[0174] The first objective function and the first constraint can be rewritten to obtain the second objective function and the second constraint as follows:

[0175]

[0176]

[0177]

[0178]

[0179] is defined as:

[0180]

[0181] Among them, min represents the minimization operation, E tot (α, Θ 0 ) represents the total energy consumption function, α represents the task offloading ratio, Θ 0 represents the fixed transmission and reflection coefficient matrix, Ri,l represents the information transmission rate from user i to relay l for computing STAR-RIS, L i represents the size of the task transmitted by user i, α i represents the task offloading ratio of user i, R i,l represents the information transmission rate from user i to relay l for computing STAR-RIS, f s Indicates the CPU frequency for calculating STAR-RIS, C i represents the computing resources required to specify the task of user i, represents the information transmission rate from user i to the UAV under fixed Θ, Θ represents the transmission and reflection coefficient matrix, f u represents the CPU main frequency of the MEC server deployed on the drone, T' represents the total delay of task offloading, and B i,u represents the link bandwidth from user i to relay l for computing STAR-RIS, represents the channel coefficient matrix from STAR-RIS calculation to the UAV, represents the transmission and reflection coefficient matrix of the signal of a given user i, H i,s represents the channel coefficient matrix from user i to calculate STAR-RIS, represents the variance of the noise power in the link from the mobile user to the drone.

[0182] In this embodiment, an approximately quadratic regularization term is introduced: Consider the non-convexity of problem (12) generated by the optimization objective (12a) under a given Θ. Add a nearly quadratic regularization term, that is, This problem can be overcome by using (12a), where ε is a positive scalar parameter. The regularized problem is equivalent to the original problem when ∈→0. Therefore, Problem (12) can be written as:

[0183]

[0184] st(11e),(12b)-(12d) (14b)

[0185] Right now:

[0186]

[0187]

[0188]

[0189]

[0190] In this embodiment, the feasible solution set is obtained through the KKT condition:

[0191] Lemma 1: For a given Θ 0 , the approximate task offloading optimization problem (14) is convex.

[0192] Proof: Considering is a function of α, and its Hessian matrix can be expressed as:

[0193]

[0194]

[0195]

[0196] in, represents the total energy consumption function after the regularization term is introduced, Indicates user The uninstall ratio, Represents the total energy consumption function after the regularization term is introduced.

[0197] Each element in equation (15) is non-negative, so all eigenvalues ​​of the Hessian matrix H are positive. Therefore, H is a symmetric positive definite matrix. Based on the second-order condition, we can conclude that is convex. Since all constraints are linear in α, problem (14) can be classified as a convex optimization problem.

[0198] According to Lemma 1, the following theorem can be established using the KKT condition.

[0199] Theorem 1: Given Θ, according to the KKT condition, we can obtain the optimal task offloading decision vector for problem (14):

[0200]

[0201] Proof: According to Lemma 1, under given Θ, the Lagrangian function L(α,ξ) of problem (14) α ,φ α ,ζ α ,ω α ) can be expressed as (17), where ξ α =(ξ α,r ,ξ α,t ) T ,φ α =(φ α,r ,φ α,t ) T , And ω α =(ω α,r ,ω α,t ) Tare the non-negative Lagrange multipliers corresponding to constraints (11e), (12b), (12c), and (12d) respectively. If is a feasible solution set For any point in that satisfies the following KKT conditions, the optimal task offloading decision can be obtained, as shown in (16).

[0202] Among them, α* represents the optimal task offloading decision vector, min represents the minimization operation, and E tot (α, Θ) represents the total energy of the STAR-RIS-assisted UAV communication network, α represents the task offloading ratio, Θ represents the transmission and reflection coefficient matrix, ξ a ,φ a ,ω a And a Both represent the non-negative Lagrange multiplier vector of the second constraint, ξ a,r ,φ a,r ,ζ a,r ,ω a,r and a,t ,φ a,t ,ζ a,t ,ω a,t denote the non-negative Lagrange multipliers of user r and user t respectively, and T denotes the transpose.

[0203]

[0204] in, represents the optimal transmission and reflection coefficient matrix, v i represents the transmission and reflection coefficient matrix of user i, min represents the minimization operation, represents the Lagrangian dual function, L i represents the size of the task transferred by user i, represents the optimal uninstall ratio of user i, P i represents the transmission power of user i, R i,u represents the information transmission rate from user i to the drone, represents the Lagrange multiplier of user i, represents the information transmission rate from user i to the UAV including the approximate variables, C i represents the computing resources required to process the task of user i, f u Indicates the CPU main frequency of the MEC server deployed on the drone. represents the Lagrange multiplier of user i, f s represents the CPU main frequency on STAR-RIS, T' represents the total delay of task offloading, represents the Lagrange multiplier of user i, m represents the mth computing STAR-RIS element, represents the square of the transmission and reflection coefficient matrix of user i, Indicates user The square of the transmission and reflection coefficient matrix, η i represents the Lagrange multiplier of user i, An approximate variable representing the signal-to-noise ratio of the link from user i to the drone, represents the variance of the noise power in the link from the mobile user to the drone, represents the transpose of the transmission and reflection coefficient matrix of user i after the nth iteration, Y represents the intermediate variable for simplifying the operation, represents the transmission and reflection coefficient matrix of user i after the nth iteration, Re represents the real part operation of a complex number, represents the transpose of the transmission and reflection coefficient matrix of user i, represents the transpose of the channel gain matrix from STAR-RIS calculation to the UAV, H i,s represents the channel gain matrix from user i to calculate STAR-RIS, represents the transpose of the channel gain matrix from user i to calculate STAR-RIS, φ s,u represents the channel gain matrix from STAR-RIS to the UAV.

[0205] S3. According to the optimal task offloading decision vector, for the transmission resource problem, the continuous convex approximation method SCA is used to optimize the resource allocation, which is specifically as follows:

[0206] S301. For the transmission resource problem, according to the optimal task offloading decision vector, a task offloading ratio at a certain point in the feasible solution set is given, and the first objective function and the first constraint condition are rewritten;

[0207] S302, introducing an approximate variable into the rewritten first objective function to obtain a third objective function and a third constraint condition;

[0208] S303, according to the third objective function and the third constraint condition, deriving the lower bound of the non-convex constraint by using the continuous convex approximation method SCA;

[0209] S304. According to the derivation results, the Lagrange dual method is used to solve the fixed transmission and reflection coefficient matrices to obtain the optimal resource allocation.

[0210] In this embodiment, a continuous convex approximation method is used to optimize resource allocation: when α = α * When , the goal of problem (11) is to determine the optimal resource allocation strategy with the goal of minimizing energy consumption.

[0211] In this embodiment, given a certain point task offloading ratio α in the feasible solution set of step 2, rewrite the first objective function and the first constraint condition: The non-convexity of objective (11a), constraint (11c) and constraint (11d) comes from log2(1+γ' i,u ), where The approximate variable represents the signal-to-noise ratio of the link from user i to the drone. Similarly, the non-convexity is overcome by replacing it with an approximate variable. For a given α * , problem (11) can be written as:

[0212]

[0213]

[0214]

[0215]

[0216] Among them, γ' i,u and represents the approximate variable of the signal-to-noise ratio of the link from user i to the UAV, and the approximate variable of the signal-to-noise ratio of the link from user i to the UAV when the offloading ratio α is given, B i,u represents the bandwidth of the link from user i to the UAV.

[0217] In this embodiment, an approximate variable is introduced. It can be written as (19), where represents the square of the overall channel gain modulus from user i to the UAV.

[0218]

[0219] in, and represents the channel coefficient matrix from STAR-RIS to the UAV, Θ i H represents the transmission and reflection coefficient matrix of the signal of user i. i,s represents the channel coefficient matrix from user i to calculate STAR-RIS, represents the simplified transmission and reflection coefficient matrix containing user i. represents the transpose of the channel gain matrix from computing STAR-RIS to the drone, represents the transpose of the channel gain matrix from user i to the STAR-RIS calculation, Y represents the intermediate variable for simplifying the operation, and v i represents the transmission and reflection coefficient matrix of user i, Indicates that the user i signal is in the Mth sThe amplitude coefficient on the element, j represents the imaginary unit, θ i represents the phase shift coefficient of the signal of user i in the calculation of STAR-RIS, represents the transpose of the channel gain matrix from STAR-RIS to the UAV, represents the transpose of the channel gain matrix from computing STAR-RIS to the UAV link, φ s,u represents the channel gain matrix from STAR-RIS to the UAV.

[0220] Therefore, constraints (11f) and (11g) are replaced by the following constraints:

[0221]

[0222] Where m represents the mth computational STAR-RIS element, represents the square of the transmission and reflection coefficient matrix of user i, Indicates user The square of the transmission and reflection coefficient matrix, yes The mth element of . Formula (20) is convex.

[0223] In this embodiment, the scaling method is used to derive the lower bound of the non-convex constraint: In order to solve the non-convex constraint (18d), the SCA method is used to solve problem (18). Specifically, due to Relative to v i is convex, and its lower bound can be derived as:

[0224]

[0225] In the formula, is obtained from the previous iteration. Then, replace constraint (18d) with the following constraint

[0226]

[0227] It is linear. Therefore, problem (18) can be rewritten as the third objective function and the third constraint:

[0228]

[0229] st(18b)(18c)(20)(22), that is:

[0230]

[0231]

[0232]

[0233] In this embodiment, the Lagrangian dual decomposition method is used to solve the fixed transmission and reflection coefficient matrix Θ: Although the convex problem (23) can be solved in CVX, the present invention uses the Lagrangian dual method to improve efficiency. Lagrangian function Given by formula (24), η=(η r ,η t ) T is the Lagrange multiplier associated with the constraints of problem (23). The Lagrange dual function is written as:

[0234]

[0235]

[0236] in, represents the optimal transmission and reflection coefficient matrix, v i represents the transmission and reflection coefficient matrix of user i, min represents the minimization operation, represents the Lagrangian dual function, L i represents the size of the task transferred by user i, represents the optimal uninstall ratio of user i, P i represents the transmission power of user i, R i,u represents the information transmission rate from user i to the drone, represents the Lagrange multiplier of user i, represents the information transmission rate from user i to the UAV including the approximate variables, C i represents the computing resources required to process the task of user i, f u Indicates the CPU main frequency of the MEC server deployed on the drone. represents the Lagrange multiplier of user i, f s represents the CPU main frequency on STAR-RIS, T' represents the total delay of task offloading, represents the Lagrange multiplier of user i, m represents the mth computing STAR-RIS element, represents the square of the transmission and reflection coefficient matrix of user i, Indicates user The square of the transmission and reflection coefficient matrix, η i represents the Lagrange multiplier of user i, An approximate variable representing the signal-to-noise ratio of the link from user i to the drone, represents the variance of the noise power in the link from the mobile user to the drone, represents the transpose of the transmission and reflection coefficient matrix of user i after the nth iteration, Y represents the intermediate variable for simplifying the operation, represents the transmission and reflection coefficient matrix of user i after the nth iteration, Re represents the real part operation of a complex number, represents the transpose of the transmission and reflection coefficient matrix of user i, represents the transpose of the channel gain matrix from STAR-RIS calculation to the UAV, H i,s represents the channel gain matrix from user i to calculate STAR-RIS, represents the transpose of the channel gain matrix from user i to calculate STAR-RIS, φ s,u represents the channel gain matrix from STAR-RIS to the UAV.

[0237] Due to the convexity of problem (23), the optimal solutions of the original problem and problem (25) both satisfy the KKT condition. Find the optimal solution Optimal transmission and reflection coefficient matrix for:

[0238]

[0239] Lagrange multipliers Update via:

[0240]

[0241]

[0242]

[0243]

[0244] In the formula is to update the Lagrange multiplier at the kth iteration The step vector of Δ is updated in each iteration. Without loss of generality, it is assumed that each element in Δ has the same step size in the iteration. represents the Lagrange multiplier of user i, k represents the number of iterations, represents the step vector for updating the Lagrange multiplier, represents the information transmission rate from user i to the UAV including the approximate variables, represents the Lagrange multiplier of user i, represents the step vector for updating the Lagrange multiplier, represents the square of the transmission and reflection coefficient matrix of user i, m represents the mth computing STAR-RIS element, Indicates user The square of the transmission and reflection coefficient matrix, η irepresents the Lagrange multiplier of user i, Δ η represents the step vector for updating the Lagrange multiplier, An approximate variable representing the signal-to-noise ratio of the link from user i to the drone, represents the transpose of the transmission and reflection coefficient matrix of user i after the nth iteration, represents the transmission and reflection coefficient matrix of user i after the nth iteration, Represents the step size vector for updating the Lagrange multipliers.

[0245] S4, alternately solving step S2 and step S3 until convergence, obtaining the optimal solution for task offloading and resource allocation, and completing the task offloading of the wireless network.

[0246] In this embodiment, Matlab or other mathematical computing tools are used to alternately solve the above two sub-problems, and a suitable convergence threshold is set until convergence, thereby obtaining an optimal solution for task offloading and resource allocation.

[0247] In this embodiment, the impact of various factors on the performance is analyzed: This section will provide numerical results to demonstrate the performance of the proposed computational STAR-RIS strategy in a dual-user wireless communication system.

[0248] In different M s Quantitative analysis of algorithm convergence performance: Figure 3 In this paper, the convergence performance of the proposed algorithm is studied. The results show that as the number of outer loop iterations increases, the energy consumption of computational STAR-RIS and traditional STAR-RIS schemes decreases faster and converges faster. Although all strategies have the same convergence speed, computational STAR-RIS always has better energy efficiency than traditional STAR-RIS. This is because the computation of STAR-RIS handles some or all user tasks by jointly optimizing the transmission and reflection coefficient matrices and offloading decisions, further reducing the total energy consumption. Therefore, solutions with more elements consume less energy than solutions with fewer elements. At the same time, the enhanced channel state causes more tasks to be sent to the MEC server for processing, reducing the performance gap between computational and traditional STAR-RIS.

[0249] In different M s Analyze the impact of network bandwidth on performance under quantity: Figure 4In this paper, the impact of different network bandwidths on the energy consumption of the proposed computing STAR-RIS strategy is studied. As the network bandwidth increases, the energy consumption of each scheme decreases. The reason is that the increase in network bandwidth increases the transmission rate in the network, reduces the transmission delay and energy consumption. At the same time, the performance gap between the calculation results and the traditional STAR-RIS is reduced. This is because due to the increase in transmission rate, more tasks are offloaded to the MEC server for processing. As the network bandwidth increases, the performance gap between STAR-RISs with and without computing capabilities decreases because the high transmission rate has less impact on different destination channels.

[0250] In different M s Analyze the impact of user transmission power on performance under the number of Figure 5 In the present invention, the energy consumption of the proposed computational STAR-RIS strategy is shown as the user's transmit power changes. Although the increase in transmit power can improve the transmission rate between the user and the computational STAR-RIS, as well as the transmission rate between the user and the drone, according to (6), more user energy will be consumed, reducing the energy efficiency of all strategies. For the computational STAR-RIS, more tasks from mobile users can be processed in the computational STAR-RIS, thereby reducing the total energy consumption and widening the gap between the computational STAR-RIS and the traditional STAR-RIS.

[0251] The simulation results verify that this method effectively reduces energy consumption and converges quickly in heterogeneous simulation environments compared to existing STAR-RIS-assisted wireless networks. The computational power of STAR-RIS has an important impact on strategies with different numbers of elements.

[0252] In this embodiment, Figure 3 As shown in Figure 3, the convergence performance of the proposed algorithm. The results show that as the number of iterations increases, the energy consumption of the computational STAR-RIS and traditional STAR-RIS solutions decreases rapidly and converges quickly. Although all strategies have the same convergence speed, computational STAR-RIS always has better energy efficiency than traditional STAR-RIS. The reason is that computational STAR-RIS further reduces the total energy consumption by jointly optimizing the reflection and transmission coefficient matrices and offloading decisions to handle some or all of the users' tasks. Therefore, the scheme with a larger number of elements consumes less energy than the scheme with a smaller number of elements. At the same time, the enhanced channel state causes more tasks to be sent to the MEC server for processing, reducing the performance gap between computational STAR-RIS and traditional STAR-RIS.

[0253] In this embodiment, Figure 4As shown in Figure 2, the impact of different network bandwidths on the energy consumption of the proposed computational STAR-RIS strategy is studied. As the network bandwidth increases, the energy consumption of each scheme decreases. The reason is that the increase in network bandwidth increases the transmission rate in the network, reducing the transmission delay and energy consumption. At the same time, the performance gap between computational STAR-RIS and traditional STAR-RIS decreases. This is because more tasks are offloaded to the MEC server for processing due to the increase in transmission rate. As the network bandwidth increases, the performance gap between STAR-RIS with and without computing power decreases because the high transmission rate has less impact on the channels of different destinations.

[0254] In this embodiment, Figure 5 As shown, in Figure 5 In , the energy consumption of the proposed Compute STAR-RIS strategy is illustrated as the user transmission power changes. Although the increase in transmission power can improve the transmission rate between the user and the Compute STAR-RIS, as well as the transmission rate between the user and the UAV, according to (6), more user energy is consumed, reducing the energy efficiency of all strategies. For Compute STAR-RIS, it can handle more tasks from mobile users to reduce the total energy consumption, widening the gap between Compute STAR-RIS and traditional STAR-RIS. .

[0255] In summary, the present invention designs a task offloading and resource allocation scheme to reduce the energy consumption of computing STAR-RIS-assisted wireless networks, fills the gap in the relevant research field of integrating computing units in the STAR-RIS plane, has a relatively broad research and development prospect, and has reference value for researchers in the field of wireless communications. The present invention has certain practical value and solves the problem of minimizing the total network energy consumption under the unmanned aerial vehicle and STAR-RIS-assisted network. The analysis results verify the various key factors that affect the delay, laying a relevant foundation for future researchers.

Claims

1. A method for offloading tasks in a wireless network, characterized in that: The following steps are involved: S1. Construct and calculate the STAR-RIS-assisted drone network model, communication model and energy consumption model, and determine the first objective function and the first constraint condition; The communication model includes: The channel where the task is executed locally: in, Indicates user Relay to Computing STAR-RIS The information transmission rate, Indicates user Relay to Computing STAR-RIS The link bandwidth is Indicates user The transmission power, Indicates that from the user Relay to Computing STAR-RIS The channel coefficients, Indicates that from the user Relay to Computing STAR-RIS The variance of the noise power in the link; Channels for task offloading: in, Indicates that from the user The information transmission rate to the drone, Indicates that from the user The link to the drone is calculated by the transmission bandwidth of the STAR-RIS surface. represents the channel coefficient matrix from STAR-RIS calculation to the UAV, Indicates that from the user To calculate the channel coefficient matrix of STAR-RIS, represents the variance of the noise power in the link from the user to the drone, Indicates user The transmission and reflection coefficient matrix of the signal; The expression of the first objective function is as follows: in, Indicates the task offloading ratio, represents the transmission and reflection coefficient matrix, represents the total energy of the STAR-RIS-assisted UAV communication network. represents the minimization operation, represents the energy consumption of calculating STAR-RIS to process local tasks, Indicates the user's energy consumption, represents the energy consumption of the MEC server deployed on the drone to perform the offloading task, It represents the minimum energy required for the drone to hover in the same position. , and Both represent users, Indicates The reflection coefficient of each element is T represents transpose; The expression of the first constraint is as follows: in, Indicates that the user The size of the task being transferred, Indicates user The task offloading ratio, Indicates that from the user Relay to Computing STAR-RIS The information transmission rate, Indicates the designated processing user The computing resources required for the task, Indicates the CPU frequency for calculating STAR-RIS. Indicates that from the user The information transmission rate to the drone, Indicates the CPU main frequency of the MEC server deployed on the drone. Indicates user The signal The amplitude of each STAR-RIS element, Indicates user The signal The amplitude of each STAR-RIS element, Indicates user The signal The phase shift of each STAR-RIS element, Indicates user The signal The phase shift of each STAR-RIS element, represents the total latency of task offloading; S2. For the task offloading problem, according to the first objective function and the first constraint condition, the Carlo-Kuhn-Tucker condition KKT is used to optimize the task offloading and obtain the optimal task offloading decision vector; The step S2 comprises the following steps: S201. For the task offloading problem, fix the transmission and reflection coefficient matrices , and rewrite the first objective function and the first constraint: S202, introducing an approximate quadratic regularization term into the rewritten first objective function to obtain a second objective function and a second constraint condition; The second objective function expression is as follows: The expression of the second constraint is as follows: in, represents the minimization operation, represents the total energy consumption function, Indicates the task offloading ratio, represents the transmission and reflection coefficient matrix of the signal, Indicates user Relay to Computing STAR-RIS The information transmission rate, Indicated by user The size of the task being transferred, Indicates user The task offloading ratio, Indicates that from the user Relay to Computing STAR-RIS The information transmission rate, Indicates the CPU frequency for calculating STAR-RIS. Indicates the designated processing user The computing resources required for the task, Indicates that the fixed From user The information transmission rate to the drone, represents the transmission and reflection coefficient matrix, Indicates the CPU main frequency of the MEC server deployed on the drone. represents the total latency of task offloading, Indicates user Relay to Computing STAR-RIS The link bandwidth is represents the channel coefficient matrix from STAR-RIS calculation to the UAV, Represents a given user The transmission and reflection coefficient matrix of the signal, Indicates that from the user To calculate the channel coefficient matrix of STAR-RIS, represents the variance of the noise power in the link from the mobile user to the drone, Indicates user The transmission power; S203, according to the second objective function and the second constraint condition, using the Carlo-Kuhn-Tucker condition KKT to optimize the task offloading, and obtain an optimal task offloading decision vector; The expression of the optimal task offloading decision vector is as follows: in, represents the optimal task offloading decision vector, represents the minimization operation, represents the total energy consumption function, represents the unloading ratio vector, , , as well as Both represent the non-negative Lagrange multiplier vectors of the second constraint condition, and Respectively represent users and users The non-negative Lagrange multiplier of , represents transpose, represents the transmission and reflection coefficient matrix of the signal, Represents the feasible solution set Any point in that satisfies the KKT condition; S3. For the transmission resource problem, the continuous convex approximation method SCA is used to optimize the resource allocation according to the optimal task offloading decision vector; The step S3 comprises the following steps: S301. For the transmission resource problem, according to the optimal task offloading decision vector, a task offloading ratio at a certain point in the feasible solution set is given, and the first objective function and the first constraint condition are rewritten; S302, introducing an approximate variable into the rewritten first objective function to obtain a third objective function and a third constraint condition; The expression of the third objective function is as follows: The expression of the third constraint condition is as follows: in, represents the minimization operation, Indicates user The transmission and reflection coefficient matrices are: Indicates the user containing the approximate variable The information transmission rate to the drone, Indicated by user The size of the task being transferred, Indicates user The optimal unloading ratio, Indicates user The transmission power, Indicates that from the user The link to the drone is calculated by the transmission bandwidth of the STAR-RIS surface. Indicates that from the user Approximate variation of the signal-to-noise ratio of the link to the drone, Indicates the CPU main frequency of the MEC server deployed on the drone. Indicates Users after iterations The transpose of the transmission and reflection coefficient matrices, represents the intermediate variable for simplifying the operation, Indicates Users after iterations The transmission and reflection coefficient moments, Indicates user The transpose of the transmission and reflection coefficient matrices, represents the variance of the noise power in the link from the mobile user to the drone, Indicates computing STAR-RIS elements, Indicates user The square of the transmission and reflection coefficient matrix, Indicates user The square of the transmission and reflection coefficient matrix, represents the transpose of the channel gain matrix from computing STAR-RIS to the drone, Indicates that from the user To calculate the channel gain matrix of STAR-RIS, Indicates that from the user To calculate the transpose of the channel gain matrix of STAR-RIS, represents the channel gain matrix from STAR-RIS to the UAV, Indicates the designated processing user The computing resources required for the task, Indicates a given uninstall ratio From user i Approximate variation of the signal-to-noise ratio of the link to the drone, represents the total latency of task offloading; S303, according to the third objective function and the third constraint condition, deriving the lower bound of the non-convex constraint by using the continuous convex approximation method SCA; S304, according to the derivation result, using the Lagrange dual method to solve the fixed transmission and reflection coefficient matrices to obtain the optimal resource allocation; S4, alternately solving step S2 and step S3 until convergence, obtaining the optimal solution for task offloading and resource allocation, and completing the task offloading of the wireless network.

2. The method for offloading tasks in a wireless network according to claim 1, characterized in that: The step S1 comprises the following steps: S101, analyzing and calculating resource allocation in a STAR-RIS-assisted drone network model, and constructing a STAR-RIS-assisted drone network model, wherein: The STAR-RIS-assisted UAV network model includes a single-antenna UAV with a MEC server, a single-antenna mobile user, and M The computing STAR-RIS is a computing STAR-RIS that is integrated on a panel and installed on the surface of a building, wherein the computing STAR-RIS element transfers the tasks offloaded by the user to the MEC server deployed on the drone; the computing STAR-RIS includes a computing elements and Reflective and transmissive elements; S102, building a communication model; S103, splitting the system energy consumption into energy consumption for calculating STAR-RIS, energy consumption for users, energy consumption for MEC servers and energy consumption for drones, and constructing energy consumption models for calculating STAR-RIS, energy consumption for users, energy consumption for MEC servers and energy consumption for drones based on the communication model; S104. According to the determined problem of minimizing the total energy consumption of STAR-RIS, mobile users, MEC servers and drones, a first objective function and a first constraint condition are obtained.

3. The method for offloading tasks in a wireless network according to claim 2, characterized in that: The energy consumption model in step S103 includes: Calculate the STAR-RIS energy consumption model: in, represents the energy consumption of calculating STAR-RIS to process local tasks, represents the capacitance coefficient, It represents the time consumed by STAR-RIS to process local tasks. Indicates the CPU frequency for calculating STAR-RIS. Indicates user The task offloading ratio, Indicates that the user The size of the task being transferred, Indicates the designated processing user The computing resources required for the task, , and All represent users; User energy consumption model: in, Indicates the user's energy consumption, Indicates user The transmission power, and Respectively represent from user Relay to Computing STAR-RIS And from the user The rate at which information is transmitted to the drone; MEC server energy consumption model: in, represents the energy consumption of the MEC server deployed on the drone to perform the offloading task, Indicates the time consumed by the MEC server to process the user's offload task. Indicates the CPU main frequency of the MEC server deployed on the drone; UAV energy consumption model: in, It represents the minimum energy required for the drone to hover in the same position. Indicates the quality of the drone, represents the gravitational acceleration at the drone's position, represents the area of ​​the rotor, Indicates the air density.

Images