RIS-assisted latency MEC HetNet resource allocation method

Abstract

The application discloses a resource allocation method for RIS-assisted delay MEC HetNet and relates to the technical field of wireless communication, which comprises the following steps: obtaining the uplink offloading transmission link and the uplink offloading transmission rate of macro users and micro users; obtaining the total delay of the tasks of the macro users and the micro users according to the uplink offloading transmission rate of the macro users and the micro users, and constructing the total delay problem of the tasks; dividing the total delay problem of the tasks into a communication sub-problem and a calculation sub-problem, and obtaining the optimized calculation resource allocation of the macro users, the calculation resource allocation of the micro users and the calculation resource allocation of the micro base stations; and obtaining the minimum total delay of the tasks of the macro users and the micro users according to the optimized calculation resource allocation of the macro users, the calculation resource allocation of the micro users and the calculation resource allocation of the micro base stations. The application can jointly use the service caching strategy, the task caching strategy and the resource allocation to minimize the total calculation delay.

Description

Technical Field

[0001] This invention belongs to the field of wireless communication technology, specifically relating to a resource allocation method for RIS-assisted delay MEC HetNet. Background Technology

[0002] Mobile edge computing (MEC) running on heterogeneous networks (HetNets) not only facilitates task offloading for mobile devices in IoT networks but also harmonizes deployment costs and compatibility with existing networks. In an MEC HetNet system, each base station (including macrocell base stations (MBS) and small cell base stations (SBS)) can be configured with servers to provide caching services and computing resources to mobile devices. Offloading computing tasks to the edge cloud, closer to the user, can provide low-latency and high-bandwidth services. Backhaul and massive MIMO (Multiple-input Multiple-output) can be utilized in MEC HetNet systems to further enhance capacity and coverage performance, and correspondingly improve task offloading efficiency. Among various advanced backhaul technologies, wireless backhaul stands out due to its greater practicality, flexibility, and ease of deployment, and is widely recognized as a more promising and indispensable backhaul technology compared to wired backhaul (such as fiber optics).

[0003] In practical applications, the base station server can only help compute specific types of tasks associated with its mobile devices when the server caches the database of the corresponding tasks. Therefore, service caching decisions and task offloading decisions influence each other, affecting the performance of MEC HetNet. Existing technologies have conducted some research on MEC HetNets, with most studies focusing on task offloading in HetNets with ideal wired backhaul. These studies aim to minimize energy consumption during task computation by optimizing frequency resources, transmission power, computational resource allocation, and offloading decisions. However, they do not consider computational resource allocation from the MBS to the Smallcell User (SUE), nor do they consider intra-layer and cross-layer interference. Furthermore, they do not investigate the actual service caching problem corresponding to the base station's caching capacity and the types of mobile user tasks offloaded.

[0004] Therefore, it is urgent to improve the aforementioned defects in the existing technology. Summary of the Invention

[0005] To address the aforementioned problems in the existing technology, this invention provides a resource allocation method for RIS-assisted latency MECHetNet. The technical problem to be solved by this invention is achieved through the following technical solution:

[0006] In a first aspect, the present invention provides a resource allocation method for RIS-assisted latency MEC HetNet, comprising:

[0007] Obtain the uplink offloading transmission links and their uplink offloading transmission rates for macro users and micro users, including the uplink offloading transmission rate from the k-th macro user to the macro base station, the uplink offloading transmission rate from the s-th micro user to the micro base station, and the uplink offloading transmission rate from the s-th micro base station to the macro base station; wherein, macro users and macro base stations are located in the same macro cell, and micro users and micro base stations are located in the same micro cell, and micro cells are randomly distributed in macro cells;

[0008] Based on the uplink offload transmission rates of macro users and micro users, the total latency of tasks for macro users and micro users is obtained, and a total latency problem for tasks is constructed. The total latency of tasks includes the latency of the j-th type of task for the k-th macro user and the latency of the j-th type of task for the s-th micro user. The latency of the j-th type of task for the k-th macro user includes the local execution latency of the j-th type of task for the k-th macro user and the offload latency of the j-th type of task for the k-th macro user. The offload latency includes the sum of the offload transmission latency of the k-th macro user offloading the task to the macro base station and the macro base station execution latency. The latency of the j-th type of task for the s-th micro user includes the local execution latency of the j-th type of task for the s-th micro user, the first type of offload latency of the j-th type of task for the s-th micro user, and the second type of offload latency of the j-th type of task for the s-th micro user. The first type of offload latency includes the sum of the offload transmission latency of the s-th micro user offloading the task to the micro base station and the micro base station execution latency. The second type of offload latency includes the sum of the offload transmission latency of the s-th micro base station offloading the task to the macro base station and the macro base station execution latency.

[0009] The total latency problem of the task is divided into a communication subproblem and a computation subproblem. The communication subproblem is optimized using a local search algorithm and a genetic algorithm to obtain an optimized passive beamforming matrix for the RIS. The bandwidth is then optimized based on the optimized passive beamforming matrix of the RIS to obtain a bandwidth allocation factor. The computation subproblem is then optimized based on the optimized passive beamforming matrix of the RIS and the bandwidth allocation factor to obtain optimized computational resource allocations for macro users, micro users, and micro base stations.

[0010] Based on the optimized computing resource allocation for macro users, micro users, and micro base stations, the minimum total latency of tasks for macro users and micro users is obtained.

[0011] The beneficial effects of this invention are:

[0012] This invention provides a resource allocation method for RIS-assisted latency MEC HetNet. It constructs a RIS-based wireless backhaul MEC HetNet system, jointly employing service caching, task caching, and resource allocation strategies to minimize total computational latency. Furthermore, considering service caching, task offloading, passive beamforming, and resource allocation, the optimization problem is described as an NP-hard mixed-integer nonlinear programming problem. Given the difficulty in solving this problem, this invention decomposes it into two sub-problems: a communication sub-problem and a computation sub-problem, and proposes a heuristic algorithm to solve these sub-problems to obtain an approximate optimal solution.

[0013] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the MEC HetNet system provided in an embodiment of the present invention;

[0015] Figure 2 This is a flowchart of a resource allocation method for RIS-assisted latency MEC HetNet provided in an embodiment of the present invention;

[0016] Figure 3 This is another flowchart of the resource allocation method for RIS-assisted latency MEC HetNet provided in this embodiment of the invention;

[0017] Figure 4 This is a schematic diagram illustrating the CDF performance of user single task computation latency provided in an embodiment of the present invention;

[0018] Figure 5 This is a schematic diagram illustrating the impact of different distances from SBS to MBS on the sum of computational delays for all tasks of all users, as provided in an embodiment of the present invention. Detailed Implementation

[0019] The present invention will be further described in detail below with reference to specific embodiments, but the implementation of the present invention is not limited thereto.

[0020] In existing technologies, considering that ideal wired backhaul is too costly in practical applications, no offloading and caching cooperation strategies between MEC servers are taken into account, which may lead to congestion of computing tasks and waste of resources.

[0021] In view of this, the present invention provides a resource allocation method for RIS-assisted latency MEC HetNet, which uses a reconfigurable intelligent surface (RIS) to assist the wireless backhaul MEC HetNet system, jointly optimizing service caching, task offloading and resource allocation issues, and minimizing the total system latency.

[0022] Please see Figure 1 As shown, Figure 1 This is a schematic diagram of an MEC HetNet system provided in an embodiment of the present invention. The present invention provides a RIS-assisted latency MEC HetNet system, a RIS-assisted two-layer MEC HetNet system, and the MEC HetNet system includes:

[0023] A macrocell comprises multiple macro users and a macro base station. The macro base station is located at the center of the macrocell, while the macro users are randomly distributed throughout. Each macro base station includes multiple antennas, and each macro user has a single antenna. Only one macro base station (MBS) is configured, and each MBS is equipped with N... T There are K antennas, and the macrocell user (MUE) can be configured.

[0024] A microcell comprises micro-users and micro-base stations, randomly distributed within the microcell. Both micro-users and microcells include a single antenna. Macro base stations communicate with both macrocells and micro-base stations. The communication link between a macro base station and a micro-base station is a backhaul link. Micro-base stations communicate with microcells. There are also communication links between macro users and macro base stations, and between micro users and micro-base stations—micro-access links. There are S microcells, each containing only one micro-base station (SBS) and one micro-user (SUE). Macro base stations serve both macro users and micro-base stations, while micro-base stations serve micro-users within the same microcell. Micro-users do not communicate directly with macro base stations but can transmit data to them through their associated micro-base stations. It should be noted that the communication link between a micro-base station and a macro base station is defined as a backhaul link, while the communication links between macro users and macro base stations, and between micro-users and micro-base stations, are defined as access links. The backhaul link utilizes dedicated frequency resources to avoid self-interference from micro-base stations, while the access link utilizes the remaining frequency resources.

[0025] Reconfigurable smart surfaces, placed near macro base stations, assist wireless backhaul links, thereby mitigating backhaul limitations; these reconfigurable smart surfaces are equipped with N... R Array elements;

[0026] The server is located in both micro and macro base stations and is used to cache task data and execute tasks. It should be noted that in the RIS-assisted wireless backhaul MEC HetNet system, the MEC server is integrated in each base station, that is, the server is integrated in both micro and macro base stations to cache task data and execute tasks.

[0027] In an optional embodiment of the invention, each user (including macro users and micro users) has There are indivisible computational tasks to be processed; specifically, the j-th type of task for the k-th macro user can be handled by... It means that, among them, and These represent the size of the task's input data, the CPU cycles required for task computation, and the size of the task's cache data, respectively; in addition, The value can be obtained by collecting user information during task execution. To distinguish between macro users and micro users' task representations, the j-th type task of the s-th micro user is represented as... It's important to note that servers at base stations can only assist in computing tasks when the base station caches the database for such tasks. Macro base stations can cache all types of tasks, while micro base stations can only cache databases for certain types of tasks due to server cache capacity limitations. Considering HetNet's operating mechanism, macro users can either execute certain types of tasks locally or send task bits to macro base stations for task offloading. Micro users have three possible options for executing different types of tasks: 1. Execute certain types of tasks locally; 2. Offload certain types of tasks to their associated micro base stations when the micro base station can cache tasks; 3. Offload certain types of tasks to macro base stations for computation via the micro base stations. It's crucial to understand that if a task type doesn't match the cache database of the associated micro base station, the micro user can only choose option 1 or 3 to execute that specific type of task.

[0028] In an optional embodiment of the present invention, For user sets, and These are respectively macro user and micro user sets; For the base station's index set, and They are sets of macro base stations and micro base stations, This is a set of task types.

[0029] Please see Figure 2 As shown, Figure 2 This is a flowchart of a resource allocation method for RIS-assisted latency MEC HetNet provided in an embodiment of the present invention. Figure 3This is another flowchart of the resource allocation method for RIS-assisted latency MEC HetNet provided in this embodiment of the invention. The resource allocation method for RIS-assisted latency MEC HetNet provided by this invention includes:

[0030] S101. Obtain the uplink offloading transmission links and their uplink offloading transmission rates for macro users and micro users, including the uplink offloading transmission rate from the k-th macro user to the macro base station, the uplink offloading transmission rate from the s-th micro user to the micro base station, and the uplink offloading transmission rate from the s-th micro base station to the macro base station; wherein, macro users and macro base stations are located in the same macro cell, and micro users and micro base stations are located in the same micro cell, and micro cells are randomly distributed in the macro cell.

[0031] Specifically, in this embodiment, the uplink offloading link from macro user to macro base station, the uplink offloading link from micro user to micro base station, and the uplink offloading link from micro base station to macro base station are obtained respectively.

[0032] Among them, in the uplink offloading link from macro user to macro base station;

[0033] In the HetNet MEC system, the received signal of the macro base station in the access link frequency band is represented as y0 = H0x0 + u0, where H0 and x0 are the channel matrices N from all users (i.e., K macro users and S micro users) to the macro base station. T ×(K+S) and transmission signal matrix (K+S)×1, u0 is the additive white Gaussian noise (AWGN) vector of the macro base station; its expression is:

[0034] H0=[G0,L0];

[0035] in, G0 and L0 represent the channel matrices from macro users and micro users to the macro base station, respectively. k and l s,0 Let represent the Rayleigh distribution channel matrices from the k-th macro user and the s-th micro user to the macro base station, respectively.

[0036] When implementing massive MIMO in a macro base station, zero-forcing beamforming (ZFBF) can be used in the receiver to mitigate intra-layer and inter-layer interference between users. The interference-to-noise ratio (ORR) of the received signal of the k-th macro user after ZFBF processing is γ. k The Signal to Interference plus Noise Ratio (SINR) is obtained using the following formula:

[0037]

[0038] Where, p M For macro users' transmit power, w k ZFBF matrix The k-th column, σ1 is the noise power at the k-th macro user;

[0039] The uplink offloading transmission rate R from the kth macro user to the macro base station k The expression is:

[0040] R k = (1-β)Blog2(1+γ) k );

[0041] Where β is the allocated bandwidth factor, B is the system bandwidth, and γ k The interference-to-noise ratio of the received signal for the k-th macro user.

[0042] Among them, in the uplink offloading link from micro-user to micro-base station;

[0043] For transmission from a micro-user to its associated micro-base station, both intra-layer and cross-layer interference are involved. In the access link frequency band, the SINR of the micro-user served by the micro-base station on the s-th micro-base station is obtained by the following formula:

[0044]

[0045]

[0046]

[0047] Where, p S l s,s , and Let l represent the transmit power of the micro-user, the channel matrix from the s-th micro-user to its serving micro-base station, in-layer interference, cross-layer interference, and noise power at the s-th micro-base station, respectively. n,s and g k,s These are the Rayleigh distribution channel matrices from the nth micro-user and the kth macro-user to the sth micro-base station, respectively.

[0048] The uplink offloading transmission rate R from the s-th micro-user to the micro-base station s,s The expression is:

[0049] R s,s = (1-β)Blog2(1+γ) S,s );

[0050] Where, γ S,s The interference plus noise ratio of the received signal of the micro user at the s-th micro base station;

[0051] Among them, in the uplink offloading link from micro base station to macro base station;

[0052] For transmission from micro base stations to macro base stations, it can be done via y 0,S =(G 0,S +FΘH R )x 0,S +u2 calculates the received signal of the macro base station on the backhaul link frequency band. and Let represent the channel matrices from all micro base stations to the macro base station, from the RIS to the macro base station, and from all micro base stations to the RIS, respectively, and let Θ represent the passive beamforming matrix at the RIS. Let r be the reflection coefficient of the r-th array element. Let the reflection phase shift of the r-th element be expressed as:

[0053]

[0054] Where c is an integer, b is the bit resolution of the discrete phase; x 0,S A matrix representing the information symbols of all micro base stations. At the macro base station, it is AWGN, G 0,S =[g 1,0 ,g 2,0 ,…,g S,0 ], g s,0 Let H be the channel matrix from the s-th micro base station to the macro base station. R =[h1,h2,…,h S ], h s Let be the channel matrix from the s-th micro base station to the RIS; it should be noted that ZFBF can be used in the receiver to reduce interference between micro base stations. The received SINR of a micro base station after ZFBF processing by the macro base station is obtained by the following formula, which is expressed as:

[0055]

[0056] Where, p B For the transmit power of the micro base station, w s Let W1 be a ZFBF vector in a micro base station, which is equal to a column of the ZFBF matrix W1. The calculation of W1 is similar to the calculation of W0.

[0057] The uplink offloading transmission rate R from the s-th micro base station to the macro base station 0,s The expression is:

[0058] R 0,s =βBlog2(1+γ 0,s );

[0059] Where, γ0,s This is the interference plus noise ratio of the received signal of the macro base station at the s-th micro base station.

[0060] It should be noted that there is only one micro base station and one micro user in a micro cell. The number of micro base stations is equal to the number of micro users, and the s-th micro base station corresponds to the s-th micro user.

[0061] In an optional embodiment of the present invention, the task cache capacity of the macro base station and the task cache capacity of the micro base station are obtained; wherein, the server at the macro base station can cache all types of tasks, and the expression for the task cache capacity of the server at the micro base station is:

[0062]

[0063] in, For the caching decision of the j-th type of task data in the s-th micro base station, A set of task types Let Q be the size of the cached data for the j-th type of task in the s-th micro base station. s Let be the task cache capacity of the s-th micro base station. It is a set of micro base stations; among which, The server at the micro base station has the ability to cache related types of tasks. The server at the micro base station does not have the ability to cache related types of tasks.

[0064] Specifically, in this embodiment, the MEC servers at macro base stations and micro base stations can cache data for different types of tasks. Macro base station servers typically have a large cache capacity, capable of caching all types of tasks, while micro base station servers, due to their limited cache capacity, can only cache certain types of tasks. Only when the servers cache databases related to the same type of task can a user offload a task to the corresponding MEC server for execution. When a user's task is offloaded to the MEC server, the server first executes the task data and then returns the execution result to the user.

[0065] S102. Based on the uplink offload transmission rates of macro users and micro users, obtain the total delay of tasks for macro users and micro users, and construct the total delay problem of tasks; wherein, the total delay of tasks includes the delay of the j-th type of task of the k-th macro user and the delay of the j-th type of task of the s-th micro user; the delay of the j-th type of task of the k-th macro user includes the local execution delay of the j-th type of task of the k-th macro user and the offload delay of the j-th type of task of the k-th macro user, the offload delay includes the sum of the offload transmission delay of the k-th macro user offloading the task to the macro base station and the macro base station execution delay; the delay of the j-th type of task of the s-th micro user includes the local execution delay of the j-th type of task of the s-th micro user, the first type of offload delay of the j-th type of task of the s-th micro user and the second type of offload delay of the j-th type of task of the s-th micro user, the first type of offload delay of the j-th type of task of the s-th micro user, the first type of offload delay includes the sum of the offload transmission delay of the s-th micro user offloading the task to the micro base station and the micro base station execution delay; the second type of offload delay includes the sum of the offload transmission delay of the s-th micro base station offloading the task to the macro base station and the macro base station execution delay.

[0066] Specifically, in this embodiment, the total latency of tasks for macro users and micro users is obtained, and a total latency problem is constructed; wherein, the total latency includes the latency of the j-th type of task for the k-th macro user and the latency of the j-th type of task for the s-th micro user;

[0067] The latency of the j-th type of task for the k-th macro user includes the local execution latency of the j-th type of task for the k-th macro user and the unloading latency of the j-th type of task for the k-th macro user, specifically:

[0068] Macro users can execute their tasks locally or remotely at the macro base station, where the server at the macro base station has the ability to cache all types of tasks. This represents the unloading decision for the j-th task of the k-th macro user. Local computing means and Offloading the task to a macro base station for computation represents... and This invention mainly studies binary unloading, and its unloading constraints are as follows:

[0069]

[0070] Specifically, obtain the local execution latency of the j-th type of task for the k-th macro user;

[0071] when as well as When the k-th macro user executes its j-th type of task locally, it should be noted that different types of tasks of a macro user share the macro user's computing resources equally for local computation; therefore, the computation latency of the j-th type of task of the k-th macro user is equal to the local execution latency, and its expression is:

[0072]

[0073] in, The computing resources allocated to the k-th macro user for the j-th type of task are determined by the following constraints:

[0074]

[0075] in, For the index of local execution task types for macro users, F M For the maximum computing power of macro users, For macro user set.

[0076] Specifically, obtain the unloading delay of the j-th task of the k-th macro user;

[0077] when as well as At that time, the k-th macro user offloads the j-th type of task to the macro base station for remote computation; the computation delay of the j-th type of task of the k-th macro user is the sum of the offloading transmission delay and the MBS execution delay, and its expression is:

[0078]

[0079] in, The offloading transmission delay for the k-th macro user to offload the task to the macro base station. Due to macro base station execution delay, Let be the size of the input data for the j-th type of task for the k-th macro user. R is the number of CPU cycles required to compute the j-th type of task for the k-th macro user. k Let be the uplink offloading transmission rate from the k-th macro user to the macro base station. The computing resources allocated by the macro base station for the j-th type of task of the k-th macro user. An index set for all macro user tasks executed by the macro base station;

[0080] When a macro base station includes multiple antennas, the delay of the j-th type of task for the k-th macro user is:

[0081]

[0082] in, Let N be the uplink offloading transmission rate from the k-th macro user to the macro base station when the macro base station includes multiple antennas. T This refers to the number of antennas included in a macro base station.

[0083] The latency of the j-th type of task for the s-th micro-user includes the local execution latency of the j-th type of task for the s-th micro-user, the first type of unloading latency of the j-th type of task for the s-th micro-user, and the second type of unloading latency of the j-th type of task for the s-th micro-user, specifically:

[0084] There are three possible options for executing micro-user tasks: 1. Calculate locally by the micro-user; 2. Calculate remotely at the micro-base station associated with the micro-user (first-stage remote calculation); 3. Calculate remotely at the macro base station via the associated micro-base station (second-stage remote calculation). This represents the task unloading decision for the j-th type of task of the s-th micro-user. in and Perform local calculations at any time; and The first phase of remote computation is executed at that time. and Perform the second phase of remote computation; the expression for the binary computation constraint of the j-th task of the micro-user is:

[0085]

[0086] Specifically, obtain the local execution latency of the j-th type of task for the s-th micro-user;

[0087] for and At that time, the s-th micro-user executes its j-th type of task locally; the local execution delay of the j-th type of task of the s-th micro-user is... The expression is:

[0088]

[0089] in, Allocate computing resources to the j-th type of task for the s-th micro-user. The number of CPU cycles required to compute the j-th type of task for the s-th micro-user; wherein the constraints for the micro-user computing resource allocation decision are:

[0090]

[0091] Among them, F S For the maximum computing power of micro users, For the local execution task type index set of the s-th micro-user, For micro-users.

[0092] Specifically, obtain the first type of unloading delay for the j-th type of task of the s-th micro-user;

[0093] when and At that time, the s-th micro-user offloads its j-th type of task to the associated micro-base station; the first type of offloading delay of the j-th type of task of the s-th micro-user. The expression is:

[0094]

[0095] in, The offloading transmission delay for the s-th micro-user to offload the task to the micro base station. For micro base station execution delay, Let be the input data size for the j-th type of task for the s-th micro-user. R is the number of CPU cycles required to compute the j-th type of task for the s-th micro-user. s,s Let be the uplink offloading transmission rate from the s-th micro-user to the micro-base station. The computing resources allocated by the s-th micro base station to the s-th micro user for the j-th type of task; wherein the constraints for the computing resource allocation decision of the micro base station are:

[0096]

[0097] in, This represents the maximum computing power of a micro base station. This is the set of task type indexes for the s-th micro-user to be offloaded to the micro-base station for computation.

[0098] Specifically, obtain the second type of unloading delay for the j-th type of task of the s-th micro-user;

[0099] when and At that time, the s-th micro user offloads its j-th type of task to the macro base station through its associated micro base station forwarding; the second type of offloading delay of the j-th type of task of the s-th micro user. The expression is:

[0100]

[0101] in, The offloading transmission delay for the s-th micro base station to offload the task to the macro base station. Due to macro base station execution delay, Let the macro base station allocate computing resources to the j-th type of task for the s-th micro user; where the constraints of the macro base station's computing resource allocation decision are micro:

[0102]

[0103] in, The maximum computing power of a macro base station, This is the index set of task types for the s-th micro-user to be offloaded to the macro base station for computation.

[0104] When the macro base station includes multiple antennas, the delay of the j-th type of task for the s-th micro user is:

[0105]

[0106] Where, N T This refers to the number of antennas included in a macro base station. Let be the uplink offloading transmission rate from the s-th micro base station to the macro base station when the macro base station includes multiple antennas.

[0107] Based on the various delays obtained above, the total delay T of the tasks for macro users and micro users is obtained. total Its expression is:

[0108]

[0109] in, For the task unloading decision of the j-th type of task in the k-th MUE, in and Indicates local computation. and This indicates that the task will be unloaded to MBS; For the task unloading decision of the j-th type of the s-th SUE, in and Indicates local computation. and This indicates the execution of the first phase of remote computation. and This indicates the second stage of remote computing; Cache decision for the j-th type of task data in the s-th SBS;

[0110] Based on the total task latency obtained, a joint optimization problem was developed to address passive beamforming at the RIS, system bandwidth allocation, micro base station caching decisions, offloading decisions for all users, and computational resource allocation between micro and macro base stations, in order to minimize the task computation time. The expression for the constructed task latency problem is as follows:

[0111]

[0112] in, For the task unloading decision of the j-th type of task in the k-th MUE, For the task unloading decision of the j-th type of the s-th SUE, N R The number of elements in the RIS array. Let θ be the reflection phase shift of the r-th array element, c be an integer, b be the bit resolution of the discrete phase, Θ be the passive beamforming matrix of RIS, (a) to (c) be binary offloading and caching, (d) be the cache constraint of the relevant database at the micro base station as a prerequisite for micro base station execution, (e) be the bandwidth allocation factor constraint, (f) be the backhaul capacity constraint, (g) be the RIS reflection stage constraint, and (h) to (j) be the allocated computing resource constraints.

[0113] It should be noted that, by fixing other optimization variables and only considering task unloading and service caching decisions, the optimization problem in the above equation can be simplified to the classic binary knapsack problem, which is a nonlinear mixed integer programming problem that is difficult to solve directly.

[0114] In view of this, the present invention proposes a joint service caching decision, task unloading determination, and resource allocation optimization problem to minimize the total latency of computing user tasks. Specifically, the joint optimization problem is first described, and then it is decomposed into two sub-problems, namely a communication sub-problem and a computation sub-problem, and the sub-problems are further solved.

[0115] S103. Divide the total delay problem of the task into a communication subproblem and a computation subproblem; use a local search algorithm and a genetic algorithm to optimize the communication subproblem, and obtain the optimized passive beamforming matrix of the RIS; optimize the bandwidth based on the optimized passive beamforming matrix of the RIS, and obtain the bandwidth allocation factor; optimize the computation subproblem based on the optimized passive beamforming matrix of the RIS and the bandwidth allocation factor, and obtain the optimized computational resource allocation for macro users, micro users, and micro base stations.

[0116] For details, please continue to see Figure 3 As shown, in this embodiment, the total delay problem of the task is divided into a communication subproblem and a computation subproblem.

[0117] In this context, good communication performance, such as a high uplink transmission rate, is beneficial for reducing transmission latency during task computation. Therefore, the communication subproblem is formulated as minimizing transmission latency by optimizing communication resources, as shown in the expression:

[0118]

[0119] Among them, T tra Offloading transmission delay for tasks of macro users and micro users.

[0120] To address the communication subproblem, this embodiment uses a scheme based on the Local-search algorithm (LSA) and the Genetic-algorithm (GA) to optimize the passive beamforming matrix at the RIS, and then provides a closed-form solution to the system bandwidth allocation problem.

[0121] Specifically, the passive beamforming matrix of the RIS is first optimized. For a given bandwidth of the backhaul link, minimizing the uplink transmission time is equivalent to maximizing the uplink data transmission rate of the SBS. In this case, the passive beamforming matrix of the RIS is obtained, and its expression is:

[0122]

[0123] in, Let c be the reflection phase shift of the r-th array element, where c is an integer and b is the bit resolution of the discrete phase.

[0124] It should be noted that the passive beamforming subproblem in RIS is discrete and non-convex; this embodiment obtains a suboptimal solution based on a local search algorithm and a genetic algorithm. In practical applications, either the local search algorithm or the genetic algorithm can be selected according to the system performance requirements.

[0125] Optimizing the passive beamforming matrix of RIS using a local search algorithm includes:

[0126] Initialize the initial phase of all RIS array elements; fix the other N elements. R -1 array element, search all phases of the r-th array element, and obtain the maximum R. s,s The phase of the r-th element is taken as the optimal solution; the phase of the r-th element is fixed to the optimal solution it has obtained; the optimal solutions of all RIS elements are obtained in turn, and the corresponding optimal solutions are fixed.

[0127] Optimizing the passive beamforming matrix of RIS using a genetic algorithm involves: first, initializing the population; then, using selection, crossover, and mutation as genetic operators to evolve the next generation; repeating this process until the required number of iterations is reached or overall convergence is achieved. Specifically:

[0128] In the genetic algorithm, chromosomes are set as the reflection phase of each element in the RIS array, and the gene encoding of the chromosome can be represented by discrete phases. The initial population is randomly generated within the constraints of discrete phase shifts.

[0129] For the selection operation, this embodiment employs a tournament selection method and elitism. Specifically, firstly, through R... s,s =βBlog2(1+γ 0,sCalculate the fitness of all chromosomes, select the chromosome with the best number of elites to be directly inherited to the next generation; then randomly select several chromosomes according to the tournament size; then select the chromosome with the best fitness as part of the parent generation.

[0130] For the crossover operation, two chromosomes are randomly selected from the parent generation, and a portion of the genes on the two chromosomes are exchanged. Using the two-point crossover method, two points are randomly selected from the chromosomes, and then a segment of gene from chromosome 1 is replaced with a segment of gene from chromosome 2, and vice versa. The resulting two new chromosomes are then added to the next generation, and the crossover operation is repeated until the number of the next generation is the same as the initial population size.

[0131] For mutation operations, chromosomes are selected based on mutation probabilities, and their genes are then altered. The original phase is replaced with a new gene randomly generated within the constraints of discrete phase shifts; the mutated chromosomes in the next generation are replaced with the mutated new chromosomes.

[0132] Each iteration yields a new generation, which serves as the input for the next iteration, undergoing a new round of selection, crossover, and mutation operations. The fitness is then recalculated. If the optimal fitness no longer changes or the maximum number of iterations is reached, then the chromosome with the optimal fitness is the suboptimal solution for the passive beamforming matrix of RIS.

[0133] Based on the suboptimal solution of the passive beamforming matrix of RIS, the bandwidth allocation subproblem is obtained, and its expression is:

[0134]

[0135] According to the return constraint R s,s ≤R 0,s The optimal solution for the bandwidth allocation factor is obtained, and its expression is:

[0136]

[0137] Where, r s,s Let r be the uplink offloading transmission rate from the s-th micro-user to the micro-base station under unit bandwidth. s,s =log2(1+γ) s,s ), r 0,s Let r be the uplink offloading transmission rate from the s-th micro base station to the macro base station per unit bandwidth. 0,s =log2(1+γ) 0,s S represents the number of micro base stations.

[0138] After optimizing communication resources, a computational subproblem is executed. Specifically, the computational subproblem jointly optimizes task offloading, SBS service caching, and base station computational resource allocation. The expression for the computational subproblem is:

[0139]

[0140] To address the computational subproblem, this embodiment develops a heuristic algorithm to obtain suboptimal solutions to task unloading decisions and service caching decisions, providing a closed-form solution to the computational resource allocation problem.

[0141] Specifically, based on the obtained optimized passive beamforming matrix and bandwidth allocation factor, the optimization computation subproblem is further decomposed into subproblems of task offloading, micro base station caching, and allocation of computational resources between micro base stations and macro base stations.

[0142] Specifically, assuming that macro base stations and micro base stations distribute computing resources equally across all tasks, and obtaining the task decision subproblem, its expression is:

[0143]

[0144] First, optimize the macro user offloading decision; specifically, for binary offloading decisions, calculate the latency of the j-th type of macro user tasks under both local and remote offloading calculation scenarios at the macro base station. The expression for the latency of the j-th type of macro user tasks is:

[0145]

[0146] if Then the j-th type of task for the macro user is executed locally; otherwise, the task is offloaded and executed at the macro base station.

[0147] Secondly, optimize the uninstallation decision for micro-users; the uninstallation decision for micro-users is as follows:

[0148]

[0149] The correspondence between the micro-user's uninstallation decision and the micro-base station's caching decision is obtained, and its expression is:

[0150]

[0151] Based on the correspondence between the micro-user's offloading decision and the micro-base station's caching decision, the latency of the j-th type of task for the micro-user is obtained, and its expression is:

[0152]

[0153] In the above process, if the cache of the s-th micro base station exceeds the maximum cache capacity, then subsequent tasks with type numbers ranked last can only use... or

[0154] Finally, based on the optimized macro-user offloading decision and the optimized micro-user offloading decision, the computational resource allocation subproblem for the servers at the macro base station and the servers at the micro base station is obtained, and its expression is:

[0155]

[0156] Based on the Lagrange multiplier method and the KKT conditions, the optimized computational resource allocation for macro users is obtained. Computing resource allocation for micro users Computing resource allocation for micro base stations The expressions are as follows:

[0157]

[0158]

[0159]

[0160] S104. Based on the optimized computing resource allocation for macro users, micro users, and micro base stations, obtain the minimum total latency of tasks for macro users and micro users.

[0161] In an optional embodiment of the present invention, this example simulates a macro cell with a radius of 350m. A macro base station with a server is located at the center of the macro cell, with coordinates (0m, 0m). Four macro users are randomly distributed within the macro cell. Eight micro cells with a radius of 10m are also evenly distributed within the macro cell, equidistant from the macro base station. Each micro cell has a micro base station with a server and a randomly distributed micro user at its center. A 64-element RIS is placed near the macro base station, with coordinates (0m, 10m). The maximum transmit power of the macro and micro users is 27dBm (0.5W), the maximum transmit power of the micro base station is 33dBm (2W), and the noise power is -80dBm. For small-scale fading, a Rayleigh fading channel model is considered. For large-scale fading, the expression for spatial propagation path loss is:

[0162]

[0163] Where C0 = -80dB is the path loss at the reference distance D0 = 1m, d is the link distance, and κ is the path loss exponent; the RIS-MBS channel is set to κ = 2.2, the SBS-RIS channel is set to κ = 2.8, and the SUE-SBS, SUE-MBS, MUE-MBS, MUE-SBS, and SBS-MBS channels are set to κ = 3.5; the system bandwidth B is set to 5MHz, each user has J = 4 tasks, and the input data size of each task is... For 800-1200 kbits, the required CPU cycles The task cache is randomly set within the range of 8000-12000 cycles / s. The maximum cache size is randomly set between 2-7GB, and the maximum cache size for each SBS is randomly set between 5-15GB; the maximum computing resources for MBS SBS's maximum computing resources are 200 Gcycles / s. The maximum computational resources for MUE and SUE are 60 Gcycles / s, and 2 Gcycles / s. For the genetic algorithm to optimize the passive beamforming matrix of RIS, the population size popSize is set to 100, the tournament size tourSize is set to 5, the elite size eliteSize is set to 10, the mutation probability p is set to 0.1, and the maximum number of iterations maxIteration is set to 100.

[0164] Please see Figure 4 As shown, Figure 4 This is a schematic diagram illustrating the CDF performance of a single user task computation latency provided in an embodiment of the present invention. "LSA+HA" and "GA+HA" represent the passive beamforming optimization methods based on LSA and GA proposed in this invention, respectively. "local" indicates that all computations are performed locally. "w / o RIS+HA" indicates that a heuristic algorithm is used to optimize offloading and caching decisions, but does not involve RIS. Figure 4 As can be seen, the curves for the "LSA+HA" and "GA+HA" schemes proposed in this embodiment are on the far left, proving that the RIS-assisted joint task offloading and service cache resource allocation scheme proposed in this invention has the best effect in reducing computational latency. Furthermore, the "LSA+HA" and "GA+HA" curves also show that for the RIS passive beamforming subproblem, the local search algorithm is better at reducing computational latency because it has a more comprehensive search space. However, while the computational complexity of the local search algorithm is low when the number of RIS elements and discrete phase bits is small, it increases significantly as the number of RIS elements and discrete phase bits increases. In this case, the genetic algorithm becomes less complex than the local search algorithm. Therefore, the local search algorithm is recommended when the number of RIS elements and discrete phase bits is small, as it is more effective. Conversely, when the number of RIS elements and discrete phase bits is large, the genetic algorithm is recommended due to its lower computational complexity. Furthermore, the "w / o RIS+HA" curve is to the right of the "LSA+HA" and "GA+HA" curves, and is only better than the "local" scheme. This proves that the computational latency is large without the assistance of RIS. Adding RIS can help with task offloading, reduce the transmission latency of task offloading, and thus reduce the total computational latency.

[0165] Please see Figure 5 As shown, Figure 5 This is a schematic diagram illustrating the impact of different distances from SBS to MBS on the sum of computational delays for all tasks of all users, provided by an embodiment of the present invention. Figure 5 It is evident that computational latency increases with increasing distance because the transmission distance of the secondary offloading task from SBS to MBS increases, leading to a corresponding increase in transmission time. Simultaneously, the performance difference between adding RIS and not adding RIS gradually widens with increasing distance. This demonstrates that the auxiliary role of RIS is more significant and provides greater assistance to the system at greater distances. Furthermore, considering the "LSA+HA" and "GA+HA" schemes, the local search algorithm shows a more pronounced effect at greater distances. Therefore, as the SBS distance increases, the SUE task, constrained by the SBS cache, needs to be secondary offloaded to MBS. A more refined optimization algorithm can better reduce the transmission latency of the secondary offloading, meaning a more refined optimization of the RIS passive beamforming method is needed.

[0166] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations are intended to cover non-exclusive inclusion, such that an article or device comprising a list of elements includes not only those elements but also other elements not expressly listed. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the article or device comprising said element. Terms such as "connected" or "linked" are not limited to physical or mechanical connections but can include electrical connections, whether direct or indirect. The orientations or positional relationships indicated by terms such as "upper," "lower," "left," and "right" are based on the orientations or positional relationships shown in the accompanying drawings and are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as limiting the invention.

[0167] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features or characteristics described may be combined in any suitable manner in one or more embodiments or examples. In addition, those skilled in the art can combine and integrate the different embodiments or examples described in this specification.

[0168] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A resource allocation method for RIS-assisted latency MEC HetNet, characterized in that, include: Obtain the uplink offload transmission links and their uplink offload transmission rates for macro users and micro users, including the first... The uplink offloading transmission rate from the macro user to the macro base station, the first The uplink offloading transmission rate from the micro user to the micro base station and the first The uplink offloading transmission rate from a micro base station to a macro base station; wherein the macro user and the macro base station are located in the same macro cell, the micro user and the micro base station are located in the same micro cell, and the micro cells are randomly distributed in the macro cell; Based on the uplink offload transmission rates of macro users and micro users, the total latency of the tasks for the macro users and micro users is obtained, and a total latency problem for the tasks is constructed; wherein, the total latency of the tasks includes the first... The first macro user's Delay of class tasks and the first The first micro-user's Delay of the class of tasks; the first The first macro user's The delay of this type of task includes the first The first macro user's Local execution latency of the class of tasks and the first The first macro user's The unloading delay of the class of tasks, the unloading delay including the first The sum of the offloading transmission delay and the macro base station execution delay for each macro user to offload tasks to the macro base station, wherein the first... The first micro-user's The delay of this type of task includes the first The first micro-user's Local execution delay of the first type of task, The first micro-user's The first type of unloading delay and the second type of task The first micro-user's The second type of offloading delay for class tasks, the first type of offloading delay including the first The sum of the offloading transmission delay and the micro base station execution delay for each micro user to offload tasks to the micro base station, the second type of offloading delay includes the first The sum of the offloading transmission delay of the micro base station to the macro base station and the execution delay of the macro base station; The total latency problem of the task is divided into a communication subproblem and a computation subproblem. A local search algorithm and a genetic algorithm are used to optimize the communication subproblem to obtain an optimized passive beamforming matrix for the Resonant Array (RIS). The bandwidth is then optimized based on the optimized RIS passive beamforming matrix to obtain a bandwidth allocation factor. The computation subproblem is further optimized based on the optimized RIS passive beamforming matrix and the bandwidth allocation factor to obtain optimized computational resource allocations for macro users, micro users, and micro base stations. The expression for the communication subproblem is: ； ； in, For offloading transmission delays of tasks for macro users and micro users, For the passive beamforming matrix of RIS, For the first Uplink offloading transmission rate from micro-users to micro-base stations For the first Uplink offloading transmission rate from micro base station to macro base station For micro base stations, For the first The reflection phase shift of each array element It is an integer. For discrete phase, the bit resolution is... The number of elements in the RIS array. For macro user set, For micro-user collections, For the first An index set of task types for micro-users to offload computation to micro-base stations. For the first An index set of task types that micro-users offload to macro base station computing. For the first The first macro user's The size of the input data for each type of task. For the first The first micro-user's The size of the input data for the task. For the first Uplink offloading transmission rate from macro users to macro base stations The bandwidth factor to be allocated; The expression for the computational subproblem is: ； in, No. The first macro user's Task unloading decisions for different task types For the first The first micro-user's Task unloading decisions for different task types For the first The first of the micro base stations Caching decisions for different types of task data For macro base stations as the first The first macro user's Computing resources allocated to task classes Assign macro base stations to the first The first micro-user's Computational resources for task-like tasks For the first The micro base station is assigned to the first The first micro-user's Computational resources for task-like tasks A set of task types For the first The first of the micro base stations Size of cached data for each type of task For the first Task cache capacity of each micro base station For the first The first macro user's Task unloading decisions for different task types ,in, and Indicates local computation. and This indicates that the task will be offloaded to the macro base station. For the first The macro user is assigned to the first Computational resources for task-like tasks For the maximum computing power of macro users, For the first The first micro-user's Task unloading decisions for each type of task, among which , and Indicates local computation. , and This indicates the execution of the first phase of remote computation. , and This indicates the execution of the second phase of remote computation. For the first The micro-user is assigned to the first Computational resources for task-like tasks For the maximum computing power of micro users, For the first A set of local execution task types for each micro-user. An index for local execution task types for macro users. This represents the maximum computing power of a micro base station. This represents the maximum computing power of a macro base station. Based on the optimized computing resource allocation for macro users, micro users, and micro base stations, the minimum total latency of tasks for the macro users and the micro users is obtained.

2. The resource allocation method for RIS-assisted latency MEC HetNet according to claim 1, characterized in that, The first Uplink offloading transmission rate from macro users to macro base stations The expression is: ； in, For system bandwidth, For the first The interference plus noise ratio of the received signal for each macro user; The first Uplink offloading transmission rate from micro-users to micro-base stations The expression is: ； in, For micro users in the first The interference-to-noise ratio of the received signal of a micro base station; The first Uplink offloading transmission rate from micro base station to macro base station The expression is: ； in, For macro base stations in the first The interference plus noise ratio of the received signal of a micro base station.

3. The resource allocation method for RIS-assisted latency MEC HetNet according to claim 1, characterized in that, Also includes: Obtain the task cache capacity of the macro base station and the task cache capacity of the micro base station; wherein, the server at the macro base station can cache all types of tasks, and the expression for the task cache capacity of the server at the micro base station is: ； in, ,in, The server at the micro base station has the ability to cache related types of tasks. The server at the micro base station does not have the ability to cache related types of tasks.

4. The resource allocation method for RIS-assisted latency MEC HetNet according to claim 1, characterized in that, The first The first macro user's Local execution delay of task-like tasks The expression is: ； in, For the first The first macro user's The CPU cycles required for computation of each type of task are given, and the constraints for the macro user's computational resource allocation decision are as follows: ； The first The first macro user's Task unloading delay The expression is: ； in, For the first The offloading transmission delay for each macro user to offload tasks to the macro base station Delay for macro base station execution; The first The first micro-user's Local execution delay of task-like tasks The expression is: ； in, For the first The first micro-user's The CPU cycles required for task-like computation; where the constraints for micro-user computing resource allocation decisions are: ； in, For the first A local execution task type index set for each micro-user; The first The first micro-user's First type of unloading delay for class tasks The expression is: ； in, For the first The offloading transmission delay for each micro-user to offload tasks to the micro base station The delay is set for micro base stations; the constraints for the computational resource allocation decisions of micro base stations are as follows: ； The first The first micro-user's Second type of unloading delay for class tasks The expression is: ； in, For the first The offloading transmission delay of the micro base station to the macro base station reduces the offloading of tasks. The macro base station execution delay is specified; wherein, the constraints on the macro base station's computing resource allocation decision are micro: 。 5. The resource allocation method for RIS-assisted latency MEC HetNet according to claim 4, characterized in that, When the macro base station includes multiple antennas, the first The first macro user's The delay for this type of task is: ； in, When a macro base station includes multiple antennas, the first... Uplink offloading transmission rate from macro users to macro base stations This refers to the number of antennas included in a macro base station; When the macro base station includes multiple antennas, the first The first micro-user's The delay for this type of task is: ； in, When a macro base station includes multiple antennas, the first... Uplink offloading transmission rate from micro base station to macro base station.

6. The resource allocation method for RIS-assisted latency MEC HetNet according to claim 1, characterized in that, The total latency of the tasks of the macro user and the micro user The expression is: ； in, Indicates the first The first macro user's Local execution delay of similar tasks Indicates the first The first macro user's Task unloading delay, Indicates the first The first micro-user's Local execution delay of similar tasks Indicates the first The first micro-user's The first type of unloading delay for class tasks, No. The first micro-user's The second type of unloading delay for similar tasks; The total latency issue for the task is: ； Among them, (a) to (c) are binary offloading and caching, (d) is the prerequisite for micro base station execution, which is the caching constraint of the relevant database at the micro base station, (e) is the bandwidth allocation factor constraint, (f) is the backhaul capacity constraint, (g) is the RIS reflection stage constraint, and (h) to (j) are the allocated computing resource constraints.

7. The resource allocation method for RIS-assisted latency MEC HetNet according to claim 6, characterized in that, The process of optimizing the communication sub-problem using a local search algorithm and a genetic algorithm to obtain an optimized passive beamforming matrix for the RIS, and then optimizing the bandwidth based on the optimized passive beamforming matrix for the RIS to obtain the bandwidth allocation factor includes: The passive beamforming matrix of the RIS is obtained, and its expression is: ； in, For system bandwidth; The passive beamforming matrix of the RIS is optimized using a local search algorithm, including: initializing the initial phase of all RIS array elements; fixing other... Array element, search for the first All phases of each array element, and obtain the maximum The phase as the first The optimal solution for the nth array element; The phase of each array element is fixed to its optimal solution; the optimal solutions of all RIS array elements are obtained sequentially, and the corresponding optimal solutions are fixed. The passive beamforming matrix of the RIS is optimized using a genetic algorithm, including: setting the chromosomes in the genetic algorithm as the reflection phases of the RIS array elements; obtaining the fitness of the chromosomes, randomly selecting multiple chromosomes according to the tournament scale, and selecting the chromosome with the best fitness as part of the parent generation; randomly selecting two chromosomes from the parent generation, performing crossover operation first, and then mutation operation, until the chromosome with the best fitness is obtained, that is, obtaining the suboptimal solution of the passive beamforming matrix of the RIS. Based on the suboptimal solution of the passive beamforming matrix of the RIS, the bandwidth allocation subproblem is obtained, and its expression is: ； According to return constraints The optimal solution for the bandwidth allocation factor is obtained, and its expression is: ； in, For unit bandwidth Uplink offloading transmission rate from micro-users to micro-base stations , For unit bandwidth Uplink offloading transmission rate from micro base station to macro base station , The number of micro base stations. For macro base stations in the first The interference-to-noise ratio of the received signal of a micro base station For micro users in the first The interference plus noise ratio of the received signal of a micro base station.

8. The resource allocation method for RIS-assisted latency MEC HetNet according to claim 1, characterized in that, The process of optimizing the computational subproblem based on the optimized passive beamforming matrix of the RIS and the bandwidth allocation factor to obtain the optimized computational resource allocation for macro users, micro users, and micro base stations includes: The task decision subproblem is obtained, and its expression is: ； First, optimize the uninstallation decision for macro users; among which, the macro user's first... The expression for the delay of a task is: ； if Then the macro user's first Tasks of this type are executed locally; otherwise, they are executed at the macro base station by unloading the task. in, Indicates the first The first macro user's Local execution delay of similar tasks Indicates the first The first macro user's Task unloading delay; Secondly, optimize the uninstallation decision for micro-users; the uninstallation decision for micro-users is as follows: ； The correspondence between the micro-user's uninstallation decision and the micro-base station's caching decision is obtained, and its expression is: ； Based on the correspondence between the micro-user's offloading decision and the micro-base station's caching decision, the micro-user's first... The delay of a task is expressed as: ； in, Indicates the first The first micro-user's Local execution delay of similar tasks Indicates the first The first micro-user's The first type of unloading delay for class tasks, No. The first micro-user's The second type of unloading delay for similar tasks; Finally, based on the optimized macro-user offloading decision and the optimized micro-user offloading decision, the computational resource allocation subproblem for the servers at the macro base station and the servers at the micro base station is obtained, and its expression is: ； Based on the Lagrange multiplier method and the KKT conditions, the optimized computational resource allocation for macro users is obtained. Micro-user computing resource allocation Computing resource allocation for micro base stations The expressions are as follows: ； ； ； in, For the first The first macro user's The number of CPU cycles required to compute each type of task For the first The first micro-user's The CPU cycles required for the computation of a task.

Images