IRS-based Task Offloading and Resource Allocation Method in Wireless Heterogeneous Networks

Abstract

The present invention discloses a task offloading and resource allocation method based on IRS in a wireless heterogeneous network, which relates to the field of wireless communication technologies, and includes: obtaining the total energy consumption in the heterogeneous network, where the total energy consumption includes user energy consumption and micro base station energy consumption; in the first stage, optimizing the user energy consumption; wherein, the user energy consumption includes the energy consumption of user local computing and the transmission energy consumption for offloading; in the second stage, optimizing the micro base station energy consumption; wherein, the micro base station energy consumption includes the local energy consumption of the micro base station and the offloading energy consumption of the micro base station; obtaining the optimized total energy consumption. This application can obtain the minimum energy consumption of the system.

Description

Technical Field

[0001] The present invention belongs to the technical field of wireless communication, and particularly relates to a method for task offloading and resource allocation based on IRS in a wireless heterogeneous network. Background Art

[0002] In recent years, people's demand for the quality of communication services has been increasing and is gradually reflected in all aspects of daily life; among them, the fifth-generation wireless communication technology (5th-Generation, 5G) aims to meet people's demand for communication services and has developed rapidly in recent years, bringing many conveniences to people's lives.

[0003] In the prior art, researchers have deeply studied key 5G technologies, including massive MIMO technology and ultra-dense network (Ultra-Dense Network, UDN), as well as mobile edge computing (Mobile Edge Computing, MEC), etc.; among them, massive MIMO technology uses a large number of antennas for data transmission, and its antenna array forms a relatively narrow beam in space using beamforming technology. By changing the direction of the main beam, it can reduce the interference of signals in space, improve the communication distance, and can also greatly improve the spectral efficiency; the ultra-dense network can improve the coverage range and spectral efficiency of existing macro cells by deploying a large number of micro base stations. In addition, with the sharp increase in the number of smart devices and the gradual popularization of various computationally intensive and high-latency applications (such as online games, virtual reality, etc.), most existing ordinary smart devices cannot provide powerful computing capabilities to meet the needs of these applications, and their battery capacity is limited, which brings challenges to the development of the next-generation wireless system; among them, based on MEC technology, smart devices can offload local computing tasks to a server connected to the base station and use the computing resources of the MEC server to process tasks, thereby further reducing the latency of task execution and reducing the energy consumption of edge devices.

[0004] However, in practice, the performance of wireless networks mainly depends on the propagation channel, and the wireless channel is random and usually exhibits uncontrollable fading; in a wireless backhaul heterogeneous network (Heterogeneous Networks, HetNets) based on massive MIMO and UDN, wireless backhaul will introduce additional wireless interference; when MEC is applied to this network, the system performance and computing efficiency will be limited.

[0005] Therefore, it is urgent to improve the problems of low system performance and computing efficiency in the existing system. Summary of the Invention

[0006] To solve the above problems existing in the prior art, the present invention provides a method for task offloading and resource allocation based on IRS in a wireless heterogeneous network. The technical problems to be solved by the present invention are achieved through the following technical solutions:

[0007] In a first aspect, the present application provides a method for task offloading and resource allocation based on IRS in a wireless heterogeneous network, including:

[0008] Obtain the total energy consumption in the heterogeneous network, where the total energy consumption includes user energy consumption and micro-base station energy consumption;

[0009] In the first stage, optimize the user energy consumption; where the user energy consumption includes the energy consumption of user local computing and the transmission energy consumption for offloading.

[0010] In the second stage, optimize the micro-base station energy consumption; where the micro-base station energy consumption includes the local energy consumption of the micro-base station and the offloading energy consumption of the micro-base station.

[0011] Obtain the optimized total energy consumption.

[0012] Advantages of the present invention:

[0013] For the method for task offloading and resource allocation based on IRS in a wireless heterogeneous network provided by the present invention, users of the micro-base station can offload their own tasks to the micro-base station, and users of the macro-base station can offload tasks to the macro-base station, and the micro-base station can further offload its own tasks to the macro-base station with the help of IRS; that is, in the first stage, minimize the user energy consumption; in the second stage, minimize the micro-base station energy consumption, and offload tasks to the macro-base station with the help of IRS; it can achieve low energy consumption in the heterogeneous network.

[0014] The following will further elaborate on the present invention in detail with reference to the accompanying drawings and embodiments. Description of the Drawings

[0015] Figure 1 is a flowchart of a method for task offloading and resource allocation based on IRS in a wireless heterogeneous network provided by an embodiment of the present invention;

[0016] Figure 2 is a schematic structural diagram of the distribution of a system provided by an embodiment of the present invention;

[0017] Figure 3 is another flowchart of a method for task offloading and resource allocation based on IRS provided by an embodiment of the present invention;

[0018] Figure 4 is a schematic diagram of a heterogeneous network schematic model based on IRS assistance provided by an embodiment of the present invention;

[0019] Figure 5 is a schematic diagram of the convergence of an algorithm provided by an embodiment of the present invention;

[0020] Figure 6 It is a schematic diagram of the cumulative distribution function of the total energy consumption of user task offloading provided by an embodiment of the present invention;

[0021] Figure 7 It is a relationship diagram between the total energy consumption of different comparison schemes and the number of micro base stations provided by an embodiment of the present invention;

[0022] Figure 8 It is a schematic diagram verifying the impact of using IRS on the total energy consumption provided by an embodiment of the present invention;

[0023] Figure 9 It is a schematic diagram verifying the impact of using IRS on the total energy consumption provided by an embodiment of the present invention;

[0024] Figure 10 It is a schematic diagram showing the relationship between the number of IRS reflection elements and the total energy consumption provided by an embodiment of the present invention;

[0025] Figure 11 It is a schematic diagram showing the relationship between the total energy consumption of a large-scale MIMO and IRS-assisted heterogeneous network task offloading system and the number of macro base station antennas provided by an embodiment of the present invention. Detailed implementation manners

[0026] The following further describes the present invention in detail with reference to specific embodiments, but the implementation manners of the present invention are not limited thereto.

[0027] Please refer to Figure 1 , Figure 1 It is a flowchart of a method for task offloading and resource allocation based on IRS in a wireless heterogeneous network provided by an embodiment of the present invention. A method for task offloading and resource allocation based on IRS in a wireless heterogeneous network provided in this application includes:

[0028] S101. Obtain the total energy consumption in the heterogeneous network, where the total energy consumption includes user energy consumption and micro base station energy consumption;

[0029] S102. In the first stage, optimize the user energy consumption; where the user energy consumption includes the energy consumption of user local computing and the transmission energy consumption for offloading;

[0030] S103. In the second stage, optimize the micro base station energy consumption; where the micro base station energy consumption includes the local energy consumption of the micro base station and the offloading energy consumption of the micro base station;

[0031] S104. Obtain the optimized total energy consumption.

[0032] Specifically, in a method for task offloading and resource allocation based on IRS in a wireless heterogeneous network provided in this embodiment, users of the micro base station can offload their own tasks to the micro base station, users of the macro base station can offload tasks to the macro base station, and the micro base station can further offload its own tasks to the macro base station with the help of IRS; that is, in the first stage, the energy consumption of users is minimized; in the second stage, the energy consumption of the micro base station is minimized, and tasks are offloaded to the macro base station with the help of IRS; low energy consumption of the heterogeneous network can be achieved.

[0033] In an optional embodiment of the present application, before optimizing the user energy consumption in the first stage, it further includes:

[0034] According to the delay constraint T of the user max , obtain the optimal computing frequency of the user

[0035] According to the optimal computing frequency of the user Use the interior point method to obtain the user offloading decision

[0036] Fix the user offloading decision And obtain the optimal computing resources allocated by the macro base station to the user And the computing resources allocated by the micro base station to the user Based on the computing resources allocated to the user Obtain the user transmission rate

[0037] Use the sequential convex approximation method to obtain the user transmit power P u,m .

[0038] In an optional embodiment of the present application, in the first stage, optimizing the user energy consumption includes:

[0039] Based on the local user transmit frequency P u,m , the reflection surface phase shift Θ 0 And the offloading decision vector (v I ) 0 , obtain the energy consumption of user local computing and the transmission energy consumption for offloading And obtain the updated user transmit power P u,m , optimize the user energy consumption; where the offloading decision vector (v I ) 0 Is formed by the user offloading decision .

[0040] In an optional embodiment of the present application, after optimizing the user energy consumption in the first stage, it further includes:

[0041] By optimizing the phase Θ, optimize the reflection surface phase shift Θ 0; where the phase Θ is only related to the m-th row ||w of the null-forcing beamforming matrix W2 2,m || 2 of the micro base station, and the transmission energy consumption of the micro base station is a monotonically increasing function of ||w 2,m || 2 ;

[0042] Then optimizing the phase Θ is equivalent to:

[0043]

[0044] where w 2,m is the m-th row of the null-forcing beamforming matrix W2, and θ n is the phase of the n-th reflecting element;

[0045] When optimizing the phase Θ, discrete phase shift optimization is adopted, and the available phase range of each reflecting element depends on the number of bits of the intelligent reflecting surface;

[0046] The number of bits of the intelligent reflecting surface is optimized by using a local search algorithm to further optimize the phase Θ.

[0047] In an optional embodiment of the present application, in the second stage, before optimizing the energy consumption of the micro base station, it further includes:

[0048] Obtaining the optimal computing frequency of the micro base station according to the delay constraint of the micro base station

[0049] Adopting the interior point method to obtain the offloading decision of the micro base station according to the optimal computing frequency of the micro base station

[0050] Fixing the user offloading decision and obtaining the optimal computing resources allocated by the macro base station to the micro base station Based on the optimal computing resources allocated by the macro base station to the micro base station Obtaining the transmission rate of the micro base station

[0051] Adopting the continuous convex approximation method to obtain the transmission power P of the micro base station m .

[0052] In an optional embodiment of the present application, in the second stage, optimizing the energy consumption of the micro base station includes:

[0053] Based on the transmission power vector P of the micro base station 0 and the micro base station offloading decision vector (v П ) 0 , obtaining the local energy consumption and offloading energy consumption of the micro base station and obtaining the updated transmission power of the micro base station ​​Optimize the energy consumption of the micro base station; where the transmit power vector P of the micro base station 0 is formed by the transmit power P of the micro base station m and the micro base station offloading decision vector (v П ) 0 is formed by the micro base station offloading decision .

[0054] In an optional embodiment of the present application, after obtaining the optimized total energy consumption, it further includes:

[0055] Judge whether the signal-to-interference-plus-noise ratio γ R,m of the m-th micro base station is greater than or equal to the signal-to-interference-plus-noise ratio γ u,m of the u-th user of the m-th base station; if so, output the optimized total energy consumption; if not, sequentially execute the optimization of the user energy consumption in the first stage and the optimization of the micro base station energy consumption in the second stage.

[0056] In an optional embodiment of the present application, please refer to Figure 2 , Figure 2 which is a schematic diagram of the distribution of the system provided by the embodiment of the present invention. In this embodiment, an intelligent reflecting surface-assisted heterogeneous network task offloading model is provided. The model includes a large-scale MIMO macro base station (Macro Base Station, MBS) with Q antennas, multiple micro base stations (Small Base Station, SBS) with single antennas, and an intelligent reflecting surface (Intelligent Reflective Surface, IRS) with N reflecting elements; each base station is equipped with a server as a computing node to provide computing resources for relevant users. It is set that the edge server providing computing resources and the base station are located at the same location and are connected by high-throughput and low-latency optical fibers. Therefore, the data communication between the base station and the server is without delay; let M s ={1, 2,..., M} be the set of all micro base stations, where M is the number of micro base stations, and M all ={0}∪M s is the set of all base stations (Base Station, BS), and M0 = {0} is the macro base station; the macro base station serves a group of K macro cell users, and U0 = {1,..., K} is used to represent a group of K macro cell users. All macro cell users share the spectrum resources of the macro base station, and users can offload their tasks to the macro base station; it is set that the number of macro base station antennas Q >> K + M, and the macro base station uses a zero-forcing receiving beamforming receiver. Therefore, there is no interference between macro user offloadings. It is set that the m-th microcell base station serves U m microcell users, and U m ={1} is defined as the number of users in this microcell, and Denote the set of all users as (Macro Users, MUE) and Small Cell Users (SUE). In the heterogeneous network task offloading model provided in this embodiment, the links between the macro base station and macro cell users, as well as between the small cell base station and small cell users, are called access links, and the link between the macro base station and the small cell base station is called the wireless backhaul link. Since the small cell base station can communicate with both small cell users and the macro base station simultaneously, there is self-interference, which limits performance. Different frequency bands are used between the communication of the small cell base station with its associated users and the wireless backhaul to avoid severe co-channel interference between the two functions.

[0057] Assume that each user has a latency-constrained task, and let L u,m , u ∈ U m , m ∈ M all (in bits) represent the task input data size for each user, and ρ u,m (in cycles / bit) represent the number of CPU cycles required to complete one bit of the task, and T max (in seconds) represent the maximum allowed latency; the total number of CPU cycles required for a user to complete the latency-constrained task is C u,m = L u,m ρ u,m ; Considering that the latency-constrained task is a divisible task and the bits of the task are independent of each other, it can be divided into different task blocks and executed on multiple servers.

[0058] In the offloading mode, each user can either execute its computation task locally or offload its task to the associated base station, where represents the offloading rate of user u at the corresponding base station.

[0059] The user task offloading process in the small cell includes two stages, namely the user task offloading stage and the small cell base station task offloading stage. During the first stage, that is, the user task offloading stage, the users in the small cell transmit their corresponding tasks to the small cell base station. After the small cell base station receives all the tasks offloaded by the users, the small cell base station can further offload them to the macro base station for processing according to the secondary offloading strategy. Among them, the small cell base station uses a reflecting surface to assist the offloading. Assume the matrix This matrix can control the reflection coefficient of the IRS elements, where η n ∈ [0, 1] and θ n ∈ [0, 2π) are the reflection amplitude and phase shift of the nth reflecting element respectively; the IRS phase shift is calculated at the macro base station according to the channel dynamics and then sent to the IRS controller through a dedicated channel. For simplicity, assume the reflection amplitude of the intelligent reflecting surface is

[0060] The above-mentioned backhaul link and access link adopt different bandwidth resources. Specifically, for a macro cell, a portion of the frequency resources accounting for β (0 ≤ β ≤ 1) is used for each backhaul link, and the MBS uses the remaining 1 - β proportion of the frequency resources to communicate with all MUEs; for a micro cell, the frequency resources of β are used for the backhaul link from the current SBS to the MBS, and the SBS uses the same frequency resources as the MUEs to communicate with the SUEs; optionally, let β = 0.5, that is, the backhaul link and the access link each account for half of the bandwidth.

[0061] During the macro user offloading transmission process, let represent the channel coefficient matrix from K macro users to the macro base station, where the k-th element h R1,k represents the channel coefficient vector between the k-th macro user and the macro base station; the macro base station adopts a large-scale antenna array. During the macro user offloading phase, all macro users simultaneously send their symbols to the macro base station, which is given by the following formula:

[0062] x1 = P 1 ·s;

[0063] where, P 1 = [P 1,0 ,..., P K,0 T represents the transmit power of K macro users, s = [s1,..., s K T represents the symbols transmitted from K macro users to the macro base station; for the uplink transmission from the MUE to the MBS, the signal received by the MBS from one MUE will be affected by co-channel interference from other MUEs and cross-layer interference from the SUEs; since orthogonal frequency bands are used for the wireless backhaul link and the access link, there is no co-channel interference from the SBS at the MBS during the signal transmission from the MUEs; due to the frequency orthogonality between the users and the IRS, there is also no interference from the IRS. Therefore, the signal received by the MBS in the frequency band used by the users is represented as:

[0064] y R1 = H R1 x1 + H R0 x0 + n R1 ;

[0065] where, represents the zero-mean additive white Gaussian noise (AWGN) at the MBS, represents the channel coefficient matrix from M micro users to the macro base station, where the m-th element h R0,m represents the channel coefficient vector between the m-th micro user and the macro base station, and x0 represents the symbols sent by all micro users. ​​

[0066] Considering that the MBS adopts a massive number of antennas, the MBS can use zero-forcing beamforming (ZFBF) in the frequency band used by users to completely mitigate the co-tier interference among MUEs and the cross-tier interference of SUEs. The MBS receives its received signal y R1 and performs receive filtering to obtain the filtered signal vector The expression is as follows:

[0067] x R1 = W1y R1 ;[[ID=1३]]

[0068] where, represents the ZF receive beamforming matrix, and the k-th row of W1 represents the receive filtering for macro user k.

[0069] In the scenario of unified bandwidth allocation, the proportion of bandwidth resources allocated to MUEs is 1-β. Assuming that all MUEs equally share these resources, according to Shannon's theorem, the achievable rate of the k-th MUE can be obtained, which is expressed as:

[0070] R k,0 = (1-β)Blog2(1 + γ k,0 );

[0071] where, B represents the bandwidth of the system, represents the signal-to-interference-plus-noise ratio of macro user transmission, γ k,0 where σ 2 is the noise power, P k,0 represents the transmit power of the k-th MUE offloading task, w 1,k represents the k-th row of the receive beamforming matrix W1.

[0072] Similarly, for the small cell, the proportion of bandwidth allocated to the wireless backhaul link is β, and the remaining bandwidth is used for data transmission of SUEs; each small base station communicates with one SUE. If multiple SUEs are considered, the NOMA protocol can be applied to offload tasks to the same small base station, and there will be intra-cell interference in this case; the small base station can apply successive interference cancellation (SIC) technology to decode the signals transmitted from multiple users in the same small cell. Assuming that the channel gain of small cell users is g u,m , m ∈ M s , so the achievable rate of the SUE in the m-th small cell is:

[0073] R u,m = (1-β)Blog2(1 + γ u,m );

[0074] where, represents the signal-to-interference-plus-noise ratio of micro user transmission, γu,m Medium P u,m Indicates the transmission power of microcell users, g u,m Represents microcell users and micro base stations m∈M s The channel gain between represents the cross-layer interference from the macro user, g k Indicates the channel gain of the interference signal transmitted from the macro user to the micro base station; It represents the same-layer interference offloaded from users in other micro cells. It is assumed that all users have been associated with the macro base station and the micro base station before the task is offloaded, and the micro base station knows the best channel state information for user transmission.

[0075] For uplink transmission from SBS to MBS, the signal received by MBS from one SBS includes the same-layer interference from other SBSs and the reflected signal from IRS, because the dedicated spectrum is used for IRS and wireless backhaul link; represents the channel coefficient matrix from M micro base stations to the macro base station, where the mth element h R2,m represents the channel coefficient vector between the mth micro base station and the macro base station; in addition, and They represent the channel coefficient matrices from M micro base stations to IRS and from IRS to macro base stations, respectively, where the mth element h S2,m represents the channel coefficient vector between the IRS and the mth micro base station.

[0076] Assume that the server of the macro base station executes the tasks received from the micro base station after receiving all the tasks. Therefore, in the offloading phase, all micro base stations send their symbols to the macro base station at the same time, which is given by:

[0077] x2=P·s;

[0078] Where P=[P1,...,P M ] T represents the transmission power of M micro base stations, s=[s1,...,s M ] T represents the symbol transmitted from M micro base stations to the macro base station, "·" represents the multiplication of the corresponding terms of the vector, and the signal received by the macro base station is Expressed as:

[0079] y R2 =(H R2 +H I2 ΘH S2 )x2+n R2 ;

[0080] in, represents the zero-mean additive white Gaussian noise (AWGN) at MBS, with variance Θ = diag(exp{jθ1}, exp{jθ2},..., exp{jθ N}) represents the N×N passive beamforming matrix at the IRS, where θ n is the configurable phase of the nth IRS element.

[0081] Similar to the above macro user transmission, ZFBF can be used to completely mitigate the interference between SBSs in the frequency band used for wireless backhaul. The MBS filters its received signal y R2 and obtains the filtered signal vector which is expressed as:

[0082] x R2 = W2y R2 ;

[0083] where, is expressed as the decoding matrix; for simplicity, the effective channel gain of the macro base station is defined as follows:

[0084] G = H R2 + H I2 ΘH S2 ;

[0085] Using a zero-forcing receiver, the MBS performs ZF precoding, and the decoding matrix of the MBS can be expressed as:

[0086]

[0087] The SINR corresponding to the mth SBS can be expressed as:

[0088]

[0089] where, P m represents the transmission power of the micro base station, w 2,m represents the mth row of the ZFBF matrix W2, then the uplink transmission data rate unloaded by the mth SBS can be obtained, which is expressed as:

[0090] R R,m = βBlog2(1 + γ R,m ).

[0091] The process of user local computing is:

[0092] Set f u,m as the computing power of user u ∈ U m , and the task execution delay for local computing is expressed as:

[0093]

[0094] The energy consumption of tasks for local computing is expressed as:

[0095]

[0096] where, represents the task offloading rate of the user, and C u,m represents the total number of CPU cycles required for each user to complete the task. The computing power of hundreds of users, δ depends on the power consumption coefficient of the chip architecture.

[0097] The process of edge computing is as follows:

[0098] For the edge computing mode, the user first uploads its computing task to the associated BS, and then the equipped server executes the task; according to the description of the communication model, for user u∈U m The corresponding transmission time during the task upload process can be given by the following formula:

[0099] <ocke>

[0100] The energy consumption is given by the following formula:

[0101]

[0102] where, R u,m represents the transmission rate of the user to offload the task, and L u,m represents the task data size of each user, and P u,m represents the transmit power of the user.

[0103] Assume that the computing resources of each server on the macro base station and the micro base station are different, and is expressed as the computing power of base station m. Then the execution time of the server at the MBS for the MUE task is expressed as:

[0104]

[0105] Then the execution time of the server at the SBS for the SUE task is expressed as:

[0106]

[0107] where, F u,m represents the computing resources allocated by the macro base station to user u, and the computing resources allocated by the micro base station to the user are represented by F m and satisfy the constraint

[0108] Since the time and energy consumption for transmitting the task calculation results from the server at the base station to the user are small, the time and energy consumption of the downlink transmission part are ignored. Only the energy consumption of the user's own task processing and the transmission energy consumption of offloading are considered. Therefore, the offloading time of edge computing is expressed as:

[0109]

[0110] The offloading energy consumption of edge computing is expressed as:

[0111]

[0112] The total energy consumption of task offloading for each user in the first stage is expressed as:

[0113] Among them,

[0114] Since the macro base station is usually plugged in, the computing energy consumption of the macro base station is not considered. However, the micro base station is deployed temporarily according to the actual scenario during deployment, and some small base stations also use batteries. Therefore, in this method, the computing energy consumption of task processing at the micro base station is considered. Then, the total energy consumption of task processing at the micro base station and the user is For the m-th micro base station, when all the tasks offloaded by users to the micro base station are processed locally by the micro base station, the energy consumption of server computing at each micro base station is In order to further reduce the computing energy consumption of the micro base station, it can be considered to offload the computing tasks at the micro base station to the macro base station for processing. Therefore, at this time, the micro base station can be regarded as a user, and the computing task of each micro base station is the computing task offloaded by the micro users in its cell. The maximum execution time for the server at each micro base station to process the task is At this time, the local processing time of the micro base station is expressed as:

[0115]

[0116] The energy consumption of the micro base station is expressed as:

[0117]

[0118] Among them, f m represents the computing power allocated by the micro base station itself, represents the offloading rate of the total task of the m-th micro base station, represents the computing power consumption of the user, and δ is the power consumption coefficient depending on the chip architecture.

[0119] According to the description of the communication model, the corresponding transmission time of the m-th micro base station during the task upload process is expressed as:

[0120]

[0121] The energy consumption of the micro base station m during the task upload process is expressed as:

[0122]

[0123] The task execution time of the micro base station unloaded to the server of the macro base station is expressed as:

[0124]

[0125] where, f R,m represents the computing resources allocated by the macro base station to the micro base station m; considering the task processing energy consumption of the micro base station, the total execution time of the micro base station in edge computing is expressed as:

[0126]

[0127] The total energy consumption of the micro base station in edge computing is respectively expressed as:

[0128]

[0129] Then in the second stage, the total task processing energy consumption of each micro base station is expressed as:

[0130]

[0131] In the above embodiments of the present application, the goal of task offloading optimization is to minimize the total energy consumption of user task processing; since the macro base station is plugged in, the computing energy consumption of the macro base station is not considered, so for macro users, the total energy consumption includes two parts: the energy consumption of local processing and the offloading transmission energy consumption; for microcell users, the total energy consumption includes four parts: the user's local energy consumption, the transmission energy consumption unloaded to the micro base station, the local processing energy consumption of the micro base station, and the transmission energy consumption of the micro base station unloaded to the macro base station; then the overall optimization problem can be expressed as:

[0132]

[0133] Among them, constraints (a) and (b) respectively indicate that the computing resources allocated to offloading users do not exceed the computing resource amount of each server, and since the SBS can be offloaded to the MBS, the MBS can handle the tasks of all MUEs and SUEs; (c) represents the local computing capacity constraint of users; (d) and (e) respectively indicate that when the SBS is offloaded, the computing resources allocated by the MBS to the SBS cannot exceed the total resources of the MBS minus the resources allocated to the MUEs; (f) represents the actual local computing capacity constraint of the SBS; (g) represents the phase shift of the nth reflection element of the IRS during the offloading transmission of the SBS; (h) represents that the total time delay of user task offloading is less than the maximum allowable delay of one time slot; (i) represents that the total time delay of SBS offloading is less than the task processing time when the SBS energy consumption is not optimized; (j) represents the power limit of all offloading users; (k) represents the power limit of all SBS offloadings; (l) and (m) respectively represent the task offloading rates of users and SBSs; (n) represents the backhaul limit, and the transmission rate of the SBS offloaded to the MBS is greater than or equal to the rate of the SUE offloaded to the SBS.

[0134] The above problem involves multiple variables and there are couplings between the variables, which is a non-convex problem and difficult to solve directly; therefore, it is necessary to decompose the above problem into two sub-problems that are easier to handle. This embodiment proposes a method of optimizing in two stages, that is, in the first stage, optimizing the local processing and offloading energy consumption of users, and in the second stage, further offloading the tasks of the micro base station to the macro base station for energy consumption optimization of the micro base station; finally, when jointly optimizing the total energy consumption, the two optimized sub-problems are iteratively solved.

[0135] Optionally, the total energy consumption of each user is composed of Given, for the optimization of the first stage, the problem of minimizing the energy consumption in the first stage can be expressed as:

[0136]

[0137] In the first stage, among them, the optimization parameters include the computing resources allocated by the base station to the offloading users, the local computing capacity of the users, the task offloading rates of all users, and the transmission power of the users.

[0138] Optionally, the total energy consumption of each micro base station is composed of Given, for the optimization of the second stage, the problem of minimizing the energy consumption in the second stage can be expressed as:

[0139]

[0140] In the second stage, among them, the optimization parameters include the computing resources allocated by the macro base station to the offloading of the micro base station, the local computing resources of the micro base station, the task offloading rates of all micro base stations, the transmission power of the micro base station, and the reflection surface phase.

[0141] In an optional embodiment of the present application, see Figure 3 As shown, Figure 3 This is another flow chart of the IRS-based task offloading and resource allocation method provided by an embodiment of the present invention. The parameters to be optimized in the first and second stages are obtained through the above process. Next, optimization is performed through the following embodiments, including optimization in the first and second stages, specifically:

[0142] In the first stage;

[0143] The expression of user energy consumption is:

[0144]

[0145] S201, solving the task scheduling sub-problem to obtain task and computing resource allocation under fixed power allocation;

[0146] Specifically, the non-convex optimization problem of energy consumption minimization in the first stage is transformed into an easy-to-handle convex optimization problem;

[0147] First, verify that the objective function of the energy consumption minimization problem in the first stage is Monotonically increasing;

[0148] Secondly, based on the task execution delay expression and energy consumption expression of local calculation, the user's optimal calculation frequency is obtained Its expression is:

[0149]

[0150] Calculate frequency based on user's best Get the best energy consumption for local computing Its expression is:

[0151]

[0152] S202, the optimal energy consumption of local calculation Substitute this into the above expression of user energy consumption and we get:

[0153]

[0154] The expression of the energy consumption minimization problem in the first stage is equivalently transformed into:

[0155]

[0156] The transformed problem is still a non-convex problem and needs to be further decomposed into two sub-problems: offloading decision and power allocation.

[0157] First, fix the transmission power of each user. The task allocation sub - problem related to the task offloading rate can be expressed as:

[0158]

[0159] Through the above transformation, a convex optimization problem is obtained. It is calculated using the classical interior - point method with polynomial computational complexity. By taking the derivative of the objective function with respect to obtain:

[0160]

[0161] Combining constraints (a) and (b) of the above formula, the feasible region of can be determined as By taking the derivative of the objective function with respect to When , then is 's monotonically increasing function; when , then is 's monotonically decreasing function; then 's optimal value is expressed as:

[0162]

[0163] where

[0164] S203. Fix the offloading decision For the macro - base station, it can be found that the more computing resources the base station allocates to users, the shorter the time consumed in the task execution process, which makes the transmission - time constraint under the total - delay requirement become looser; obviously, this will be beneficial to reducing the transmission power. Therefore, the computing - resource allocation problem is to minimize the total task - execution delay of offloading users served by the same BS, that is:

[0165]

[0166] It can be solved by using the partial Lagrangian function; the optimal computing resources allocated by the macro - base station to each user are The minimum task - upload rate requirement of macro - cell users is

[0167] The micro - base station needs to consider the offloading transmission process. So the task - computing time for the micro - base station to offload tasks of micro - cell users needs to ensure that the micro - base station can transmit the tasks to the macro - base station and the macro - base station needs to process and complete them. Under the premise of ensuring the transmission - delay constraint, the computing resources allocated by the micro - base station to users can be obtained from where Denoted as the sum of the time for micro base station offloading transmission and the time for macro base station task processing, the micro base station calculates the delay constraint From this, the minimum task upload rate requirement for microcell users can be correspondingly obtained, and its calculation formula is Then, by optimizing the transmit power of the user, the transmission rate that satisfies the delay constraint can be obtained.

[0168] S204. When optimizing the user power, the optimization problem of offloading users on the same bandwidth is expressed as:

[0169]

[0170] where, P = {P u,m , m ∈ M all , u ∈ U m} represents the power set of offloading users. Also, due to the non-convexity of the objective function and constraint (b), the above formula is not a convex optimization problem; the sequential convex approximation (SCA) method can be used to solve the non-convex problem.

[0171] First, tighten the constraint (b) in the above formula through the inequality where and When , the bound is tight; further, obtain the lower bound of the data rate of user u, and its expression is:

[0172]

[0173] The constraint (b) in the above formula can be rewritten as

[0174] By setting P = e S , the optimization problem of offloading users on the same bandwidth is equivalently reformulated as the following approximate problem:

[0175]

[0176] The above formula belongs to a standard convex optimization problem. By using the SCA method to update S to iteratively solve the convex optimization problem of the above formula, further solve the optimization problem of offloading users on the same bandwidth, and obtain the transmit power P u,m .

[0177] In the specific implementation process, in the first stage, optimize the energy consumption of users through the following Algorithm 1.

[0178] [[ID=�6]]

[0179]

[0180] After receiving the user's offloaded task, each micro base station further integrates all the tasks of the micro base station. At this time, m micro base stations can be approximated as m users. The micro base station can choose to process tasks on the edge server connected to it, which is equivalent to local processing of the micro base station, or it can offload all remaining tasks to the macro base station. Therefore, the total energy consumption of each micro base station in the first stage can also be expressed as:

[0181]

[0182] It should be noted that from the expression of the total energy consumption of each micro base station in the first stage, it can be seen that the passive beamforming matrix only affects the energy consumption during the SBS offloading process; therefore, the fixed power allocation {P m} and task offloading rate The energy consumption minimization problem in the second stage can be expressed as:

[0183]

[0184] Then, based on the SINRγ received at the MBS R,m From the expression of 2,m || 2 is related, and the transmission energy consumption of SBS is ||w 2,m || 2 is a monotonically increasing function; therefore, the energy consumption minimization problem in the second stage is equivalent to:

[0185]

[0186] When optimizing the phase Θ, discrete phase shift optimization is used. The available phase range of each reflector element depends on the bit of the IRS. The local search method shown in the following Algorithm 2 is used to solve this problem; keep the other N-1 phase shift values fixed, and for each element θ n , traverse all possible values and select the best value, then, use this optimal solution θ n * As θ n The new value of θ is used for the optimization of another phase shift, until all phase shifts in the set Θ are optimized.

[0187] The process of Algorithm 2 can be expressed as:

[0188]

[0189] In the second phase;

[0190] The total energy consumption expression of the micro base station is:

[0191]

[0192] S301. Transform the non-convex optimization problem of minimizing the energy consumption in the second stage into an easily tractable convex optimization problem;

[0193] First, the objective function of the problem of minimizing the energy consumption in the second stage increases monotonically with f m and f R,m Based on the delay constraint (e) in the expression of the problem of minimizing the energy consumption in the second stage, we get:

[0194]

[0195] Furthermore, the expression for obtaining the optimal CPU cycle frequency at the micro base station is:

[0196]

[0197] S302. Based on the expression of the total energy consumption of the micro base station and the expression of the optimal CPU cycle frequency at the micro base station, transform the problem of minimizing the energy consumption in the second stage into:

[0198]

[0199] It should be noted that the transformed problem is still non-convex. Therefore, it needs to be further divided into two easily tractable sub-problems of task allocation and micro base station power control, and they are solved alternately.

[0200] By fixing the transmit power {P m} of the micro base station, the task allocation sub-problem formula related to the task offloading rate is as follows:

[0201]

[0202] Let where

[0203] So the offloading allocation ratio at the micro base station can be determined by combining the constraints (a) and (b) in the above formula. The feasible region of When When is is a monotonically increasing function of When is is a monotonically decreasing function of The optimal value of

[0204]

[0205] where

[0206] S302. Fixed Micro - base Station Offloading Decision To optimize the computing resources of the macro - base station, it can be analyzed that the more computing resources the macro - base station allocates to the micro - base station, the shorter the time consumed in the task execution process, which makes the transmission time constraint under the micro - base station offloading delay requirement become looser, and this will be conducive to reducing the transmit power of the micro - base station. Therefore, the computing resource allocation problem is transformed into minimizing the total task computing delay of the micro - base station offloading served by the macro - base station, that is:

[0207]

[0208] The optimal computing resource allocation for each micro - base station can be obtained by ; correspondingly, the minimum task upload rate requirement of the micro - base station can be obtained, and its calculation formula is

[0209] S303. Similar to the power optimization of users, the optimization problem of all micro - base stations offloading on the same bandwidth is expressed as:

[0210]

[0211] Among them, \(P = \{P m , m\in M s \}\) is the set of transmit powers of micro - base station offloading. Due to the non - convexity of the objective function and constraint (b), the above - mentioned problem is not a convex optimization problem; therefore, the successive convex approximation method is used to solve this problem, and the lower bound of the data rate of micro - base station \(m\) can be given as:

[0212]

[0213] Then, the constraint condition (b) of the above - mentioned problem can be rewritten as By setting \(P = e S \), the above - mentioned problem can be equivalently reformulated as the following approximate problem:

[0214]

[0215] The above is a standard convex optimization problem, and the solution is iteratively obtained by using the SCA method to update \(S\).

[0216] Among them, the optimization scheme of the second stage is shown in Algorithm 3.

[0217]

[0218] Finally, according to Algorithm 1, it is found that the total energy consumption of users does not increase after each iteration, and the system energy consumption will not increase after updating the offloading decision; Algorithm 3 further offloads the tasks of the micro base stations to the macro base station, and the energy consumption of local processing and transmission of the micro base stations can also converge after a limited number of iterations. By iterating the two algorithms, the optimized total system energy consumption can be obtained.

[0219] Among them, the joint task offloading and resource allocation method is described in Algorithm 4.

[0220]

[0221]

[0222] In an optional embodiment of the present application, the effectiveness of the present application is further verified through simulation experiments.

[0223] In the simulation experiment, consider a macro cell located in the center, M micro base stations distributed around the macro base station at equal distances, K macro users randomly distributed in the macro cell, the radii of the macro cell and the micro cells are 300m and 10m respectively, the micro users are evenly distributed in their corresponding cells, each micro base station serves one micro user, and the macro users are not located in the micro cells. Please refer to Figure 4 shown in Figure 4 is a schematic model diagram of an IRS-assisted heterogeneous network provided by an embodiment of the present invention. The hexagram is the macro base station, the circle is the user, the pentagram is the micro base station, and the square is the IRS.

[0224] For the communication channel, both small-scale fading and large-scale path loss are considered. Without loss of generality, the small-scale fading is an independent and identically distributed term and follows a complex Gaussian distribution with zero mean and unit variance. The path loss is given by the following formula:

[0225]

[0226] where, C0 represents the path loss at the reference distance d = 1m, and κ represents the path loss factor; it is set that C0 = -30dB, and the path loss factors of the SBS-to-MBS link, the SBS-to-IRS link, and the IRS-to-MBS link are set to κ SBS→MBS = 3.5, κ SBS→IRS = 2.8 and κ IRS→MBS = 2.2. The carrier frequency of the system is f c = 2GHz, and the transmission bandwidth is B = 20MHz; the computing capabilities of all micro base stations are set to be the same, and the computing capability of the macro base station is stronger; for the computing tasks, it is considered that the task sizes of all users for computing offloading are the same. The number of antennas of the macro base station is set to Q = 40, and the number of reflecting surface elements is set to N = 40; in Table 1, the set simulation parameters are summarized.

[0227] Table 1 Simulation Parameter Settings

[0228]

[0229]

[0230] Please refer to Figure 5 as shown Figure 5 Figure 15 is a schematic diagram of the convergence of the algorithm provided by an embodiment of the present invention, where the number of micro base stations M = 4 and the number of macro users K = 4; from Figure 5 it can be seen that both the energy consumption of optimizing users by Algorithm 1 and the energy consumption of optimizing micro base stations by Algorithm 3 decrease as the number of iterations increases. This proves the convergence of the algorithm, and the convergence speed is relatively stable.

[0231] Please refer to Figure 6 as shown Figure 6 Figure 24 is a schematic diagram of the cumulative distribution function of the total energy consumption of user task offloading provided by an embodiment of the present invention, where the cumulative distribution function (Cumulative Distribution Function, CDF). In the existing work on joint task offloading and resource allocation in wireless heterogeneous networks, the use of IRS-assisted communication has not been considered, so the further improvement of system performance after using IRS in this method is compared. From Figure 6 it can be seen that the performance of optimizing the phase using IRS is significantly better than that without using IRS. Although the performance of using random phase with IRS is relatively close to that without using IRS, it is still better than not using IRS.

[0232] Please refer to Figure 7 as shown Figure 7 Figure 33 is a relationship diagram between the total energy consumption of different comparison schemes and the number of micro base stations provided by an embodiment of the present invention is for all local processing of users, optimizing means only considering Algorithm 1 to optimize the offloading energy consumption of users, and the tasks offloaded by microcell users to micro base stations are all processed by micro base stations. At this time, it is equivalent to sacrificing the energy consumption of micro base stations to minimize the transmission energy consumption of local processing and offloading of users. However, the total energy consumption of this scheme is much higher than that of all local processing of users, mainly because the task processing delay of micro base stations is very small, and more computing resources are allocated to tasks, resulting in high processing energy consumption. Therefore, further offloading of micro base stations is considered. By jointly optimizing the offloading energy consumption of users and the offloading energy consumption of micro base stations through Algorithm 4. It can be seen from the figure that the performance of this method is better than other schemes.

[0233] Please refer to Figure 8 as shown Figure 8It is a schematic diagram provided by an embodiment of the present invention to verify the impact of using IRS on the total energy consumption, comparing the performance at different distances from the micro base station to the macro base station. In the random phase scheme, Algorithm 2 is used to optimize the task offloading rate and power control of the micro base station, without optimizing the IRS phase shift. The random allocation of the IRS phase shift follows [0, 2π). As can be seen from the figure, even when using the IRS with a random phase system for the optimized task offloading of the micro base station, the total energy consumption is lower than that without using the IRS. The performance with the optimized phase is improved by about 10% compared to not using the IRS. When the micro base station is 50 m away from the macro base station, the performance with the optimized phase is improved by 2.7% compared to the random phase, and when the distance is 350 m, it is improved by 9.5%. Therefore, when the distance between the micro base station and the macro base station is farther, the use of IRS brings a greater improvement in performance. This is mainly because the farther the distance, the greater the fading of the wireless transmission of the micro base station, and the IRS-assisted transmission can improve the signal-to-interference-plus-noise ratio of the micro base station signal transmission, thereby reducing the total energy consumption.

[0234] Please refer to Figure 9 as shown in Figure 9 It is a schematic diagram provided by an embodiment of the present invention to verify the impact of using IRS on the total energy consumption, comparing the performance of three schemes: optimizing offloading without using IRS, using IRS but with random phase allocation, and using IRS with optimized phase. As can be seen from the figure, the total energy consumption of the system with optimized phase for the optimized task offloading of the micro base station using IRS is about 10% lower than that without using IRS, and about 8% lower than that with random phase of IRS.

[0235] Please refer to Figure 10 as shown in Figure 10 It is a schematic diagram provided by an embodiment of the present invention showing the relationship between the number of IRS reflection elements and the total energy consumption. When the number of micro base stations and macro users is 4, the performance of optimized phase, random phase, and without using IRS is compared. It can be observed that in the case of micro base station offloading. As the number of IRS elements increases, the total energy consumption of IRS random phase and without using IRS basically remains unchanged; the total energy consumption of the configuration with optimized IRS phase decreases. When the number of reflection elements is 40, the performance with optimized phase is improved by 9.1% compared to random phase, and when the number of reflection elements increases to 130, the performance with optimized phase is improved by 21% compared to random phase. Since it is mainly considered to add IRS-assisted communication during the micro base station offloading transmission and there is no IRS assistance in the user offloading stage, increasing the number of reflection elements will not reduce the energy consumption of user offloading transmission. Therefore, the method of using IRS with optimized phase saves energy compared to random phase, but continuing to increase the number of reflection elements, the total energy consumption will still converge.

[0236] Please refer to Figure 11 as shown in Figure 11It is a schematic diagram showing the relationship between the total energy consumption of the large-scale MIMO and IRS-assisted heterogeneous network task offloading system provided by the embodiments of the present invention and the number of macro base station antennas. Four micro base stations and four macro users are set. It can be observed from the figure that when the number of macro base station antennas increases, the total energy consumption of the joint optimization algorithm decreases, mainly due to the reduction in the energy consumption of the macro user and micro base station offloading transmissions. When Q = 20, the performance of the optimized phase is improved by 13.1% and 12.5% respectively compared with the random phase and without using IRS. When Q = 120, the performance of the optimized phase is improved by 5.1% and 4.7% respectively compared with the random phase and without using IRS. Therefore, as the number of macro base station antennas increases, the amplitude of the energy consumption reduction also decreases. When the number of macro base station antennas Q → ∞, the SINR of the micro base station offloading transmission will converge to a value independent of the antenna number Q, so continuing to increase the antenna number will not bring performance improvement.

[0237] The above content is a further detailed description of the present invention in combination with specific preferred embodiments. It cannot be determined that the specific implementation of the present invention is only limited to these descriptions. For those of ordinary skill in the technical field to which the present invention belongs, without departing from the concept of the present invention, several simple deductions or substitutions can be made, which should all be regarded as belonging to the protection scope of the present invention.

Claims

1. A method for task offloading and resource allocation based on IRS in wireless heterogeneous networks, characterized in that: include: Obtaining total energy consumption in a heterogeneous network, where the total energy consumption includes user energy consumption and micro base station energy consumption; In the first stage, the user energy consumption is optimized; wherein the user energy consumption includes the energy consumption of the user's local calculation and the transmission energy consumption used for offloading. The user of the micro base station offloads his own tasks to the micro base station, and the user of the macro base station offloads his own tasks to the macro base station. In the first stage, before optimizing the user energy consumption, the following steps are further included: According to the user's delay constraints , get the user's optimal calculation frequency ; Optimal calculation frequency based on the user , , represents the set of all users, Indicates the users served by the base station, , represents the set of all base stations, represents the set of macro base stations, represents the set of micro base stations, Represents the number of micro base stations, and uses the interior point method to obtain the user's unloading decision ; Fixed the user uninstall decision , and obtain the optimal computing resources allocated by the macro base station to the user and the optimal computing resources allocated to users by micro base stations ; Based on the computing resources allocated by the user , get the user's minimum transmission rate ; Use continuous convex approximation method to obtain user transmission power ; in, Represents a user At the corresponding base station unloading rate, The optimal value of is expressed as: ； in, , Indicates the transfer rate of user offloaded tasks, Indicates the size of each user's task data, Represents parameters related to the power consumption coefficient of the chip architecture, Indicates the total number of CPU cycles required for each user to complete the task, Indicates the proportion of bandwidth resources allocated to the backhaul link, Indicates base station Medium users The corresponding signal-to-interference-noise ratio is Indicates the bandwidth of the system; The optimal computing resources allocated by the macro base station to each user is , Indicates the computing power of the macro base station; the minimum task upload rate requirement for macro cell users is ; In the first stage, optimizing the user energy consumption includes: Based on local user transmission frequency , phase shift of the reflecting surface and the offloading decision vector , obtain the energy consumption of user local computing and transmission energy consumption for offloading , and obtain the updated user transmission power , optimize the user energy consumption; wherein the uninstall decision vector Uninstall decision by the user form; In the first stage, after optimizing the user energy consumption, the method further includes: By optimizing the phase , Used to control the reflection coefficient of the reflective element, and Respectively represent The reflection amplitude and phase shift of each reflective element are optimized to optimize the phase shift of the reflective surface. ; wherein the phase Only with zero-breaking beamforming matrix No. OK Related, the transmission energy consumption of micro base stations is A monotonically increasing function of Then optimize the phase is equivalent to: ； in, is the zero-breaking beamforming matrix No. OK, For the The phase of each reflective element; In the optimized phase When discrete phase shift optimization is used, the available phase range of each reflective element depends on the bit position of the smart reflective surface; Use local search algorithm to optimize the bit position of smart reflector and further optimize the phase ; In the second stage, the energy consumption of the micro base station is optimized; wherein the energy consumption of the micro base station includes the local energy consumption of the micro base station and the energy consumption of the micro base station offloaded; the micro base station offloads its own tasks to the macro base station through the IRS; in the second stage, before optimizing the energy consumption of the micro base station, the following steps are also included: According to the delay constraints of micro base stations , get the optimal calculation frequency of the micro base station ; According to the optimal calculation frequency of the micro base station , using the interior point method to obtain the offloading decision of the micro base station ; Fixed the user uninstall decision , and obtain the optimal computing resources allocated by the macro base station to the micro base station ; Based on the optimal computing resources allocated by the macro base station to the micro base station , obtain the minimum transmission rate of the micro base station ; Use continuous convex approximation method to obtain the micro base station transmission power ; Indicates micro base station The task offloading rate, The optimal value of is: ； in, , represents the delay constraint of micro base station calculation, Indicates the The uplink transmission data rate unloaded by each micro base station is Indicates the The signal-to-interference-and-noise ratio corresponding to each micro base station; The optimal computing resource allocation for each micro base station can be obtained by get, Indicates that the macro base station is allocated to the user The computing resources of the micro base station can be obtained by the minimum task upload rate requirement, which is calculated as follows: ; In the second stage, optimizing the energy consumption of the micro base station includes: Based on the transmit power vector of the micro base station and the micro base station offloading decision vector , obtain the local energy consumption of the micro base station and the energy consumption of the micro base station unloading , and obtain the updated micro base station transmission power , optimize the energy consumption of the micro base station; wherein the transmission power vector of the micro base station The transmission power of the micro base station Formation, the micro base station offloading decision vector The micro base station offloading decision form; Get the total energy consumption after optimization.

2. The method for task offloading and resource allocation based on IRS in wireless heterogeneous networks according to claim 1, characterized in that: After obtaining the optimized total energy consumption, the method further includes: Judge the Signal-to-interference-and-noise ratio of a micro base station Is it greater than or equal to The first micro base station Signal-to-interference-and-noise ratio (SINR) ; If yes, output the optimized total energy consumption; if not, execute the first stage of optimizing the user energy consumption and the second stage of optimizing the micro base station energy consumption in sequence.

Images