A resource allocation method for wireless power supply edge computing networks

Abstract

The present invention relates to the field of wireless power supply edge computing and task scheduling technology, and specifically to a resource allocation method for a wireless power supply edge computing network. First, heterogeneous access points, mobile terminals, time-varying channels, and heterogeneous tasks are comprehensively considered, and corresponding system models, equipment and service models, and task model calculation models are established; then, a selection algorithm based on hybrid delay tasks is designed by comprehensively considering various factors affecting computing performance; finally, considering the network model characteristics of the wireless power supply edge computing network, a coordinate descent algorithm based on one-dimensional time variable binary search is used to solve the mobile terminal's offloading decision and system time resource allocation problems. The present invention can effectively make heterogeneous access point selection, offloading decision-making, and time resource allocation schemes under time-varying channels, thereby improving the overall computing efficiency of the system.

Description

Technical Field

[0001] The present invention relates to the field of wireless power supply edge computing and task scheduling technology, and in particular to a resource allocation method for a wireless power supply edge computing network. Background Art

[0002] Wirelessly powered mobile edge computing integrates mobile edge computing and wireless power transmission technologies to address the challenges of insufficient computing power and continuous energy supply for mobile terminals. However, when multiple mobile terminals are involved in a wirelessly powered mobile edge computing network, not only must mobile terminal offloading decisions be considered, but also the trade-offs between heterogeneous access points providing power to the mobile terminals and allocating task processing time, posing significant challenges to resource allocation and offloading decisions.

[0003] Traditional resource allocation methods and decision-making approaches only consider scenarios with a single heterogeneous access point. Computing networks with a single heterogeneous access point can only provide limited energy and computing services, limiting improvements in mobile terminal quality of experience and failing to overcome the dual near-far effect between multiple mobile terminals.

[0004] Therefore, designing efficient resource allocation and offloading decision-making methods that match the needs of multiple heterogeneous access points in wireless edge computing networks is crucial. In this scenario, different heterogeneous access points have varying transmit power and processing capabilities, mobile terminals may have heterogeneous latency requirements for tasks in different time slots, and the wireless channels between mobile terminals and heterogeneous access points can vary over time. Therefore, comprehensively considering these factors to design a rational and efficient resource allocation and offloading decision-making scheme to improve the overall system's computing performance is a key challenge currently facing research. Summary of the Invention

[0005] The purpose of the present invention is to provide a resource allocation method for wireless power supply edge computing networks, aiming to solve the problem of unreasonable resource allocation and unloading decision-making schemes in scenarios with multiple heterogeneous access points and multiple mobile terminals, and to improve the computing performance of wireless power supply edge computing networks.

[0006] To achieve the above objectives, the present invention provides a resource allocation method for a wireless power supply edge computing network, comprising the following steps:

[0007] Step 1: In a scenario with multiple heterogeneous access points and multiple mobile terminals, build a system model, device and service model, task model, and computational model to obtain the optimization objective.

[0008] Step 2: Solve the heterogeneous access point selection problem based on the selection algorithm of mixed delay tasks;

[0009] Step 3: A coordinate descent algorithm based on one-dimensional time-variable binary search solves the mobile terminal offloading decision and time resource allocation problems in the subnetwork generated after heterogeneous access point selection;

[0010] Step 4: Use the binary search algorithm of one-dimensional time variable to obtain the time resource allocation strategy;

[0011] Step 5: Top-level problem solving, generating the entire network time resource allocation plan and offloading decision.

[0012] Optionally, the execution process of step 1 includes the following steps:

[0013] Step 1.1: Consider a network with multiple heterogeneous access points and multiple mobile terminals and establish a system model.

[0014] Step 1.2: Establish equipment and service models based on the division of system time frames;

[0015] Step 1.3: Consider the heterogeneous delay requirements that may exist in different time slots and establish a task model;

[0016] Step 1.4: Establish a calculation model based on different offloading decisions of mobile terminals;

[0017] Step 1.5: Based on the established model, obtain the optimization target of the entire network.

[0018] Optionally, the execution process of step 2 includes the following steps:

[0019] Step 2.1: Calculate the correlation coefficient between local computing and offload computing for all mobile terminals;

[0020] Step 2.2: Normalize the correlation coefficient to its maximum and minimum values;

[0021] Step 2.3: Introduce the system delay-sensitive task ratio γ and calculate the relevant value;

[0022] Step 2.4: Obtain the heterogeneous access point selection result, and divide the overall network into multiple sub-networks.

[0023] Optionally, during the execution of step 3, first, all mobile terminals in the subnet randomly generate an initial offloading decision, and randomly select between two offloading decisions: local calculation / offloading to the selected heterogeneous access point for calculation; then, after calculating the calculation rate value after each mobile terminal flips the offloading mode, the calculation rate increment is selected, that is, the mobile terminal flipping strategy with the largest difference between the calculation rate after flipping and the calculation rate obtained by the initial decision is selected as the current strategy; finally, the above process is iterated until the calculation rate increment no longer increases after all mobile terminals flip their decisions, and the final offloading decision of the subnet is obtained.

[0024] Optionally, a binary search convex optimization algorithm is combined to obtain a computing rate increment, and the binary search algorithm is called to respectively calculate computing rate values ​​after flipping the offloading decision for all mobile terminals in each sub-network.

[0025] Optionally, the execution process of step 4 is specifically to first determine the interval to be optimized, that is, the time range to be optimized, then calculate the midpoint value of the interval, and finally update the interval according to the strategy.

[0026] Optionally, the execution process of step 5 is specifically as follows: after obtaining the heterogeneous access point selection strategy through the hybrid delay task selection algorithm, the coordinate descent algorithm of the one-dimensional time variable binary search and the binary search algorithm are used to obtain the offloading decision and time resource allocation scheme of different subnetworks, for each subnetwork, the coordinate descent algorithm of the one-dimensional time variable binary search is called successively to obtain the heterogeneous access point selection, offloading decision and resource allocation scheme of the entire network.

[0027] This invention provides a resource allocation method for wireless power edge computing networks. First, it comprehensively considers heterogeneous access points, mobile terminals, time-varying channels, and heterogeneous tasks, and establishes corresponding system models, device and service models, and task model computation models. Then, it comprehensively considers various factors that affect computing performance and designs a selection algorithm based on mixed-delay tasks. Finally, considering the network model characteristics of the wireless power edge computing network, a coordinate descent algorithm based on a one-dimensional time-variable binary search solves the mobile terminal offloading decision and system time resource allocation problems. Compared with traditional single-heterogeneous access point research, the proposed method can meet the efficient offloading and power supply requirements of mobile terminals in multi-heterogeneous access point scenarios, effectively overcome the dual near-far effect caused by a single heterogeneous access point, and improve the system's task computing efficiency. Compared with conventional multi-heterogeneous access point research, the proposed method considers the delay-sensitive heterogeneity between different tasks, significantly reduces task deadline violations, can handle heterogeneous tasks with multiple delay requirements, and improves the system task computing rate. Furthermore, it further adapts to various device distributions and channel variations. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0029] Figure 1 It is a system principle flow chart of a resource allocation method for a wireless power supply edge computing network of the present invention.

[0030] Figure 2It is a schematic diagram of the BMDT algorithm flow of the present invention.

[0031] Figure 3 It is a flow chart of the CD algorithm of the present invention.

[0032] Figure 4 2 is a schematic diagram showing the effective calculation rate of CD and its comparison algorithm in a specific embodiment of the present invention.

[0033] Figure 5 Schematic diagram of the impact of different delay-sensitive task proportions of the present invention. DETAILED DESCRIPTION

[0034] The following describes embodiments of the present invention in detail, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and are not to be construed as limiting the present invention.

[0035] The following is an explanation of the English abbreviations used in this invention:

[0036] BMDT: a hybrid delayed task-based selection algorithm;

[0037] CD: Coordinate descent algorithm based on one-dimensional time-variant binary search.

[0038] The present invention provides a resource allocation method for a wireless power supply edge computing network, comprising the following steps:

[0039] Step 1: In a scenario with multiple heterogeneous access points and multiple mobile terminals, build a system model, device and service model, task model, and computational model to obtain the optimization objective.

[0040] Step 2: Solve the heterogeneous access point selection problem based on the selection algorithm of mixed delay tasks;

[0041] Step 3: A coordinate descent algorithm based on one-dimensional time-variable binary search solves the mobile terminal offloading decision and time resource allocation problems in the subnetwork generated after heterogeneous access point selection;

[0042] Step 4: Use the binary search algorithm of one-dimensional time variable to obtain the time resource allocation strategy;

[0043] Step 5: Top-level problem solving, generating the entire network time resource allocation plan and offloading decision.

[0044] The specific implementation principle block diagram is as follows Figure 1As shown, the resource allocation method for the wireless power supply edge computing network is divided into three stages, which correspond to the above steps 1, 2, 3 to 5 respectively. The following is further explained in conjunction with the implementation steps:

[0045] In the first stage, in the scenario of multiple heterogeneous access points and multiple mobile terminals, the corresponding system model, device and service model, and task model calculation model are established, and finally the optimization target is obtained.

[0046] The specific steps are as follows:

[0047] 1. Consider a network with multiple heterogeneous access points and multiple mobile terminals and establish a system model.

[0048] Specifically, the wireless power supply edge computing network model considered consists of multiple heterogeneous access points and mobile terminals. The heterogeneous access points have a long-term stable wired power supply, which can broadcast radio frequency to supply energy to mobile terminals and receive and process computing tasks offloaded by mobile terminals. Mobile terminals may be within the coverage area of ​​one or more heterogeneous access points. Mobile terminals should select the appropriate heterogeneous access point to provide services (energy supply services and task offloading services) based on their needs.

[0049] To avoid mutual interference, the communication and energy collection circuits of each mobile terminal operate in a time-division multiplexing manner, and the entire system uses TDMA for multiple access to wireless channels. TDMA (Time Division Multiple Access) is a communication technology used for multiple users to share limited spectrum resources. In a TDMA system, time is divided into continuous time slots, and each user communicates in a different time slot to achieve parallel data transmission. The system time is divided into continuous time frames of equal length T, which is set to be less than the channel coherence time. Within each time frame T, energy must be supplied to all mobile terminals served by the heterogeneous access point before task offloading. When offloading tasks, a binary offloading strategy is considered, so that the task is either calculated locally on the mobile terminal or offloaded to a designated heterogeneous access point for execution.

[0050] 2. Establish equipment and service models based on the division of system time frames.

[0051] Specifically, consider a wireless powered edge computing network consisting of multiple heterogeneous access points and multiple mobile terminals. Different heterogeneous access points have heterogeneous transmission power and computing capabilities. The subchannels between heterogeneous access points and mobile terminals are block fading channels, meaning that the wireless channel gain is constant within each time frame but varies between different time frames. In the same time slot, each mobile terminal can only select a specific heterogeneous access point to bind its service relationship (energy supply service and task offloading service), while heterogeneous access points can serve multiple mobile terminals simultaneously.

[0052] 3. Consider the heterogeneous delay requirements that different time slots may have and establish a task model.

[0053] Specifically, the task list that mobile terminal i needs to perform in the time slot is Tasks are represented by triples in represents the number of clock cycles required to calculate task k of mobile terminal i, represents the input data size of task k of mobile terminal i (in bits, i.e., the size of data that needs to be uploaded when unloading), represents the delay tolerance of task k for mobile terminal i, and max represents the maximum number of tasks that mobile terminal i can execute in that time slot. In each time slot T, the system can obtain a proportion of delay-sensitive tasks of the entire wireless power edge computing network, which is γ∈(0,1). The tasks to be executed by each mobile terminal arrive randomly in different time slots, and the proportion of delay-sensitive tasks in the system varies in different time slots.

[0054] 4. Establish a calculation model based on different uninstall decisions of mobile terminals.

[0055] Task local calculation: During local calculation, after obtaining energy supply, the mobile terminal locally executes its own computing tasks. Since the energy collection circuit and computing unit in the mobile terminal are separate, the mobile terminal can collect energy and calculate its tasks at the same time in local calculation mode. It is assumed that the energy consumed by each mobile terminal during calculation cannot be greater than the energy obtained during the energy supply time, the execution of the task must be completed within the delay requirement time, and the mobile terminal randomly reaches tasks with different delay requirements in different time slots. The effective task calculation rate (bit / s) of mobile terminal i during task local calculation is obtained as follows:

[0056]

[0057] Where a represents the power supply time of the heterogeneous access point to the mobile terminal within the time frame, The number of tasks to meet the latency requirement.

[0058] Task offloading to the selected heterogeneous access point for computation: Due to time-division multiplexing circuit limitations, mobile terminals in offloading mode can only offload their tasks to heterogeneous access points after collecting energy. Since the computing power of heterogeneous access points is generally several orders of magnitude higher than that of mobile terminals, the processing delay caused by the heterogeneous access points integrated with edge computing servers is negligible compared to the processing delay caused by offloading and transmission. In addition, the size of the calculation result is much smaller than the size of the original task, so the duration of the download phase is also negligible, and each time slot is only occupied by wireless power transmission (i.e., energy supply) and task offloading. Although from the overall perspective of the system, the execution time of the task is ignored, for delay-sensitive tasks, slight time differences may also be fatal. Therefore, from the perspective of the task, by considering the execution delay of the task on the heterogeneous access point, when determining whether the task execution completion time is offloaded within the delay requirement time, the effective computation rate (ECR) (bit / s) of mobile terminal i is:

[0059] During unloading, the ECR (bit / s) of mobile terminal i is:

[0060]

[0061] Where a represents the power supply time of the heterogeneous access point to the mobile terminal within the time frame, τ i represents the uninstallation time of the i-th mobile terminal, The number of tasks to meet the latency requirement.

[0062] 5. Based on the established model, the optimization target is obtained.

[0063] Specifically, according to the established model, the sum effective computation rate (SECR) of a single sub-network is expressed as:

[0064]

[0065] where x={x i |i∈N j}, τ={τ i |i∈N j}, uninstall indicator variable x i represents the task offloading decision of mobile terminal i in time slot T, x i =0 means local calculation, x i =1 indicates task offloading.

[0066] For the entire wireless power supply edge computing network with multiple heterogeneous access points and multiple mobile terminals, the optimization goal is to maximize the total effective computing rate of the system:

[0067]

[0068] That is, when the channel condition h is known, the heterogeneous access point selection, offloading decision x and time resource allocation scheme (τ, a) are made to maximize the effective task computing rate of the system.

[0069] In the second stage, the hybrid delay task-based selection algorithm (BMDT) solves the heterogeneous access point selection problem.

[0070] 1. Calculate the correlation coefficient between local computing and offload computing for all mobile terminals.

[0071] Specifically, first according to the formula and Calculate the correlation coefficients of local computing and offload computing for all mobile terminals respectively.

[0072] 2. The correlation coefficient is normalized to its maximum and minimum values.

[0073] Specifically, in order to make all correlation values ​​comparable, the sum of the locally calculated correlation values ​​is offloaded to the heterogeneous access point to perform maximum-minimum normalization processing on the offloaded correlation values.

[0074] 3. Introduce the system delay-sensitive task ratio γ and calculate the related value.

[0075] Specifically, after the data is normalized, the system delay-sensitive task ratio γ and the common task ratio 1-γ are introduced, and then according to the formula and Calculate the correlation values ​​between all mobile terminals and heterogeneous access points.

[0076] 4. Obtain the heterogeneous access point selection results and divide the overall network into multiple sub-networks.

[0077] Specifically, for each mobile terminal, select the relevant values ​​of all heterogeneous access points The largest heterogeneous access point is sufficient, and this operation is repeated iteratively to obtain the mobile terminal selection result for the entire system. Correspondingly, after the heterogeneous access point selection is performed, the network is divided into multiple sub-networks consisting of a single heterogeneous access point and the mobile terminals it serves.

[0078] The corresponding BMDT algorithm flowchart is shown in the figure.

[0079] In the third stage, a coordinate descent (CD) algorithm based on one-dimensional time variable binary search is designed to solve the sub-network mobile terminal offloading decision and time resource allocation problems. The CD algorithm flow chart is as follows: Figure 3 shown.

[0080] The specific process is as follows:

[0081] 1. Initialize uninstall decision.

[0082] Specifically, the CD algorithm needs to calculate the system's computing rate in advance based on the initial decision. Each sub-network has N mobile terminals, and introduces N auxiliary binary variables x = [x1,...,x N ]', where x i =0(or x i =1) indicates that mobile terminal i selects local computing tasks (or offloads tasks to its selected heterogeneous access point for execution) and randomly generates the initial offloading decision x 0 , where all x i Randomly choose between the two uninstall decisions of 0 / 1.

[0083] 2. Combine the binary search convex optimization algorithm to obtain the calculation rate increment.

[0084] Specifically, after obtaining the initial offloading decision, for all mobile terminals in each subnetwork, a binary search algorithm is called to calculate the calculation rate value after flipping the offloading decision (local calculation flips to offloading calculation, and offloading calculation flips to local calculation). After calculating the calculation rate value after each mobile terminal flips the offloading mode, the calculation rate increment is selected. That is, the mobile terminal flipping strategy with the largest difference between the calculation rate after flipping and the calculation rate obtained by the initial decision is selected as the current strategy. The above process is iterated until the calculation rate increment no longer increases after the flipping decision of all mobile terminals, and the final offloading decision of the subnetwork is obtained.

[0085] The main process of the binary search algorithm is as follows: first, determine the interval to be optimized, that is, the time range to be optimized. Then, calculate the midpoint value of the interval, and finally update the interval according to the strategy.

[0086] 3. Time resource allocation plan and offloading decision generation.

[0087] Specifically, after obtaining the heterogeneous access point selection strategy through the BMDT algorithm and the CD algorithm obtaining the offloading decision and time resource allocation scheme of different sub-networks, the CD algorithm is called one by one for each sub-network to obtain the heterogeneous access point selection, offloading decision and resource allocation scheme of the entire network.

[0088] The present invention differs from existing methods in that it studies the problem of maximizing the effective computing rate of tasks in a wireless power edge computing network with multiple mobile terminals and multiple heterogeneous access points, and proposes an efficient joint scheduling algorithm. The present invention can effectively make heterogeneous access point selection, offloading decisions and time resource allocation plans under time-varying channels, thereby improving the overall computing efficiency of the system. By jointly optimizing the time-varying channel gain, the heterogeneous delay requirements of the system tasks, and the different transmission powers and computing capabilities of the heterogeneous access points, the BMDT selection strategy is used to make heterogeneous access point selection, and the CD algorithm combined with the time binary search algorithm suitable for wireless power edge computing networks is used to solve the offloading decision and time resource allocation problems. The three-stage iterative optimization maximizes the effective computing rate of the system.

[0089] Furthermore, the present invention further demonstrates the beneficial effects of the present invention through comparison with specific embodiments:

[0090] As shown in Table 1 below, at different system scales, both the system's weighted compute rate (WCR) and the weighted effective compute rate (WECR) (taking into account task latency requirements) are significantly improved. The improvement becomes more pronounced as the system scale increases. HAP represents a heterogeneous access point, and MD represents a mobile terminal.

[0091] Table 1 Comparison of the computational rates of BMDT and its comparison algorithms at different device scales

[0092]

[0093]

[0094] Figure 4 and Figure 5 The figures respectively show the computational performance comparison between the CD algorithm combined with BMDT and other comparison algorithms at different system scales, and the effective completion of tasks of the present invention at different ratios of delay-sensitive tasks.

[0095] Specifically, Figure 4 The performance results of various offloading algorithms under various system scales and distributions are shown. Each scale is tested under multiple random distributions, and the calculation results are the average value of each distribution within 2000 time slots. Figure 4 As can be seen in Figure 2, compared with various baseline solutions, the BMDT_CD method proposed in this invention can achieve the best computational efficiency.

[0096] Figure 5 The effect of the BMDT_CD method on the effective completion ratio of system tasks under different delay-sensitive task ratios is shown. The horizontal axis in the figure represents the different delay-sensitive task ratios from 20% to 80%, and the vertical axis represents the effective completion ratio of tasks under this ratio, that is, how many tasks are completed within the delay time. Figure 5 It can be seen that the BMDT_CD algorithm proposed in the present invention can perform well in scenarios with different system scales and delay-sensitive task ratios, ensuring the effective completion of system tasks. Moreover, this guarantee becomes more stable as the system scale increases. For example, at 8HAP and 30MD, the system's effective task completion ratio remains above 90%.

[0097] In summary, the present invention uses this method in a wireless edge computing environment with multiple mobile terminals and multiple heterogeneous access points to quickly and efficiently select heterogeneous access points, make offloading decisions, generate time resource allocation plans, and improve the system's task computing rate. Compared with traditional single heterogeneous access point research, the method proposed in the present invention can meet the efficient offloading and power supply requirements of mobile terminals in a multi-heterogeneous access point scenario, effectively overcome the dual near-far effect caused by a single heterogeneous access point, and improve the system's task computing efficiency. Compared with conventional multi-heterogeneous access point research, the method proposed in the present invention takes into account the delay-sensitive heterogeneous characteristics between different tasks, significantly reduces task deadline violations, can handle heterogeneous tasks with various delay requirements, improves the system's task computing rate, and further adapts to various device distribution situations and channel changes.

[0098] The above disclosure is only a preferred embodiment of the present invention, and certainly cannot be used to limit the scope of the rights of the present invention. Ordinary technicians in this field can understand that all or part of the processes of the above embodiment and equivalent changes made in accordance with the claims of the present invention are still within the scope of the invention.

Claims

1. A resource allocation method for a wireless power supply edge computing network, characterized in that: The following steps are involved: Step 1: In a scenario with multiple heterogeneous access points and multiple mobile terminals, build a system model, device and service model, task model, and computational model to obtain the optimization objective. Step 2: Solve the heterogeneous access point selection problem based on the selection algorithm of mixed delay tasks; The execution process of step 2 includes the following steps: Step 2.1: Calculate the correlation coefficient between local computing and offload computing for all mobile terminals; Step 2.2: Normalize the correlation coefficient to its maximum and minimum values; Step 2.3: Introduce the system delay-sensitive task ratio γ and calculate the relevant value; Step 2.4: Obtain the heterogeneous access point selection results and divide the overall network into multiple sub-networks; Step 3: A coordinate descent algorithm based on one-dimensional time-variable binary search solves the mobile terminal offloading decision and time resource allocation problems in the subnetwork generated after heterogeneous access point selection; During step 3, all mobile terminals in the subnet first randomly generate an initial offloading decision and randomly choose between two offloading decisions: local calculation or offloading to the selected heterogeneous access point. Then, after calculating the computation rate value after each mobile terminal flips the offloading mode, the computation rate increment is selected. That is, the mobile terminal flipping strategy with the largest difference between the computation rate after flipping and the computation rate calculated from the initial decision becomes the current strategy. Finally, the above process is iterated until the computation rate increment no longer increases after all mobile terminals flip their decisions, thus obtaining the final offloading decision for the subnet. Step 4: Use the binary search algorithm of one-dimensional time variable to obtain the time resource allocation strategy; Step 5: Top-level problem solving, generating the entire network time resource allocation plan and offloading decision.

2. The resource allocation method for a wireless power supply edge computing network according to claim 1, characterized in that: The execution process of step 1 includes the following steps: Step 1.1: Consider a network with multiple heterogeneous access points and multiple mobile terminals and establish a system model. Step 1.2: Establish equipment and service models based on the division of system time frames; Step 1.3: Consider the heterogeneous delay requirements that may exist in different time slots and establish a task model; Step 1.4: Establish a calculation model based on different offloading decisions of mobile terminals; Step 1.5: Based on the established model, obtain the optimization target of the entire network.

3. The resource allocation method for a wireless power supply edge computing network according to claim 2, characterized in that: The calculation rate increment is obtained by combining the binary search convex optimization algorithm, and the binary search algorithm is called to calculate the calculation rate values ​​of all mobile terminals in each subnetwork after the flip offloading decision.

4. The resource allocation method for a wireless power supply edge computing network according to claim 3, characterized in that: The execution process of step 4 is to first determine the interval to be optimized, that is, the time range to be optimized, then calculate the midpoint value of the interval, and finally update the interval according to the strategy.

5. The resource allocation method for a wireless power supply edge computing network according to claim 4, characterized in that: The execution process of step 5 is as follows: after obtaining the heterogeneous access point selection strategy through the hybrid delay task selection algorithm, the coordinate descent algorithm and the binary search algorithm of the one-dimensional time variable binary search are used to obtain the offloading decisions and time resource allocation plans for different subnetworks. Then, the coordinate descent algorithm of the one-dimensional time variable binary search is called for each subnetwork one by one to obtain the heterogeneous access point selection, offloading decision and resource allocation plan for the entire network.

Images