Environment and resource state sensing resource allocation method and system

Abstract

The invention provides an environment and resource state sensing resource distribution method and system, and the method comprises the steps: obtaining user environment states and network resource states in a coverage range of a plurality of cells, the user environment states comprise task information, user importance and maximum tolerable delay of tasks, and the network resource states comprise network resource states and network resource states; the network resource state comprises availability of communication resources in the network, availability of computing resources and channel state information from a user to a base station; and iteratively allocating communication resources and computing resources to the selected user according to the user environment state and the network resource state by taking an average user satisfaction degree, which maximizes the satisfaction degree of the selected user and the satisfaction degree of the rejected user at the same time, as an optimization target, the rejected user is a user which is estimated according to the task information and cannot meet the maximum tolerable delay.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of communication, in particular to the field of wireless communication technology, and more particularly to an environment and resource state aware resource allocation method and system. BACKGROUND

[0002] With the advent of the Internet of Everything (IoE) era, the sixth generation (6G) mobile cellular network is expected to connect everything and support various services, such as autonomous driving, industrial Internet, and augmented / virtual reality (AR / VR). This may impose higher requirements on high-density computing capability and low end-to-end delay. Due to limited volume, mobile devices such as virtual reality (VR) headsets lack sufficient computing power and battery capacity. In order to meet the requirements of high-density computing and low delay, it is necessary to offload services to mobile edge computing (MEC) servers. However, due to the limited coverage of single-cell mobile edge computing networks, multi-cell mobile edge computing networks (MC-MEC networks) are widely used.

[0003] In MC-MEC networks, communication and computing resources are always limited. In order to provide wireless offloading to users performing computing-intensive tasks, the same communication frequency may be repeatedly used by multiple users, so communication interference is inevitable. At the same time, on the MEC server, computing resource competition occurs when performing multiple computing-intensive tasks. Both communication interference and computing resource competition can cause an increase in end-to-end delay and reduce user satisfaction (USD). Therefore, it is crucial to study the mutual influence of limited communication and computing resources on end-to-end delay and develop intelligent resource allocation schemes to improve user satisfaction. A lot of research has been done in this field, especially considering MEC networks based on non-orthogonal multiple access (NOMA).

[0004] As is known, through superposition coding and successive interference cancellation (SIC), NOMA can allow multiple users to share the same frequency resource at the same time, thereby increasing transmission capacity. This is very attractive to systems with limited spectrum. Since the total service delay of users may be limited, multi-cell MEC networks integrating NOMA technology can increase transmission capacity, reduce communication delay, and thus gain more computing time for computing tasks. Conversely, if the competition of multiple computing-intensive tasks is reduced through a reasonable computing resource allocation algorithm, the computing delay is also reduced, which will also gain more time for communication. In summary, in MEC networks integrating NOMA, how to allocate limited resources to improve user satisfaction is crucial.

[0005] Existing research [1]-[4]Assume that the communication and computing resources (including MEC, local, etc.) are sufficient to complete all user tasks. However, in practical applications, there are many scenarios where resources are very limited and cannot support all tasks in the case of personnel gathering, such as large sports venues, large public gatherings and festival activities. In these cases, the existing resource solution cannot work. Therefore, when the total demand of users exceeds the available resources, it is necessary to selectively provide services for part of the users, while delaying the provision of services for other users. Some people have studied user selection in MEC networks [5]-[7] Without considering the size of the task and the channel state, the study [5] A first-come-first-served user selection mechanism is designed, that is, resources are allocated until there are not enough resources to serve new users. Unlike research [5], the system bandwidth is assumed to be divided into N sub-channels, each of which can support up to 2 users through NOMA, and research [6] proposes a user selection mechanism to select 2N users, which is based on information such as channel state information (CSI), where users with better channel conditions are more likely to be selected. The user selection mechanisms in studies [5] and [6] do not consider the quality of experience related to important performance standards such as maximum tolerable delay. Therefore, late users who are sensitive to delay or users with poor channel conditions may not be selected, resulting in poor experience for them. Unlike researches [5][6], the user selection mechanism in research [7] is based on the deadline of the user's task. The smaller the maximum delay that can be tolerated, the higher the priority of the selected user. However, it ignores the impact of channel conditions and available resources (such as MEC computing capacity), which can also reduce user satisfaction. For example, a user with poor channel conditions or high computing requirements who is selected first may fail to complete the task due to limited available resources.

[0006] In summary, in MEC networks, the existing user selection mechanisms are based on information such as CSI or task deadlines, ignoring the availability of resources, which can result in the current resources not being able to meet the needs of the selected users; It can also make part of the users unable to be served for a long time and exacerbate the unfairness phenomenon. Therefore, the existing technology may not be able to provide good service satisfaction for the selected users, and it is also difficult to guarantee the overall service quality of all users within the network service range.

[0007] The information of the above documents is as follows:

[0008] [1] L. Zhang et al., “Digital Twin-Assisted Edge Computation Offloading in Industrial Internet of Things With NOMA,” IEEE Transactions on Vehicular Technology, vol. 72, no. 9, pp. 11935–11950, Sep. 2023, doi:10.1109 / TVT.2023.3270859.

[0009] [2] H. Hu, D. Wu, F. Zhou, X. Zhu, R. Q. Hu, and H. Zhu, “Intelligent Resource Allocation for Edge-Cloud Collaborative Networks: A Hybrid DDPG-D3QN Approach,” IEEE Trans. Veh. Technol., pp. 1–14, 2023, doi: 10.1109 / TVT.2023.3253905.

[0010] [3] L. Qin, H. Lu, Y. Chen, B. Chong, and F. Guo, “Joint Transmission and Resource Optimization in NOMA-Assisted IoVT With Mobile Edge Computing,” IEEE Transactions on Vehicular Technology, pp. 1–16, 2024, doi: 10.1109 / TVT.2024.3364358.

[0011] [4] B. Liu, C. Liu, and M. Peng, “Resource Allocation for Energy-Efficient MEC in NOMA-Enabled Massive IoT Networks,” IEEE Journal on Selected Areas in Communications, vol. 39, no. 4, pp. 1015–1027, Apr. 2021, doi:10.1109 / JSAC.2020.3018809.

[0012] [5] Y. Qi, L. Tian, Y. Zhou, and J. Yuan, “Mobile Edge Computing-Assisted Admission Control in Vehicular Networks: The Convergence of Communication and Computation,” IEEE Vehicular Technology Magazine, vol. 14, no. 1, pp. 37–44, Mar. 2019, doi: 10.1109 / MVT.2018.2883336.

[0013] [6] J. Du et al., “When Mobile-Edge Computing (MEC) Meets Nonorthogonal Multiple Access (NOMA) for the Internet of Things (IoT): System Design and Optimization,” IEEE Internet of Things Journal, vol. 8, no. 10, pp. 7849–7862, May 2021, doi: 10.1109 / JIOT.2020.3041598.

[0014] [7] Z. Ding, D. Xu, R. Schober, and H. V. Poor, “Hybrid NOMA Offloading in Multi-User MEC Networks,” IEEE Transactions on Wireless Communications, vol. 21, no. 7, pp. 5377-5391, Jul. 2022, doi: 10.1109 / TWC.2021.3139932.

[0015] It should be noted that the background art is only intended to introduce relevant information for understanding the technical solutions of the present application, and does not mean that the relevant information must be prior art. The relevant information is submitted and disclosed together with the present application scheme, and should not be regarded as prior art without evidence that the relevant information has been disclosed before the filing date of the present application. SUMMARY

[0016] Therefore, the purpose of the present application is to overcome the defects of the prior art, and to provide an environment and resource state aware resource allocation method and system.

[0017] The purpose of the present application is achieved by the following technical solutions:

[0018] According to a first aspect of the present application, a resource allocation method is provided, the method comprising: obtaining user environment states and network resource states in the coverage of a plurality of cells, wherein the user environment state comprises task information, user importance and maximum tolerable delay of the task, and the network resource state comprises availability of communication resources, availability of computing resources and channel state information of users to base stations within the network; and iteratively allocating communication resources and computing resources to selected users according to the user environment states and the network resource states, with the optimization target of maximizing the average user satisfaction degree considering the satisfaction degree of the selected users and the satisfaction degree of the rejected users, wherein the rejected users are users whose maximum tolerable delay cannot be satisfied according to the task information. The scheme can at least achieve the following beneficial technical effects: the scheme can obtain user environment states and resource states, and make user selection and resource allocation based on both, reject users whose maximum tolerable delay cannot be satisfied during the allocation process, and consider the average user satisfaction degree of the selected users and the rejected users. Thus, as much as possible, sufficient resources are allocated to the selected users, the delay performance of the selected users is improved, the overall average satisfaction degree is improved, and the unfairness phenomenon that some users cannot be served for a long time is reduced.

[0019] Optionally, the average user satisfaction is calculated using a preset average user satisfaction performance evaluation function, wherein the function is configured to: set the satisfaction of the selected user to be positively correlated with the difference obtained by subtracting the estimated end-to-end delay from the user's maximum tolerable delay; and set the satisfaction of the rejected user to a negative value, which is negatively correlated with the product of the user's importance and the duration of each frame. This solution sets the satisfaction of the selected user to be positively correlated with the difference obtained by subtracting the estimated end-to-end delay from the user's maximum tolerable delay, so as to more accurately evaluate and satisfy the satisfaction of the selected user; while setting the satisfaction of the rejected user to a negative value, which is negatively correlated with the product of the user's importance and the duration of each frame, can achieve a penalty effect, so that users who are rejected for a longer period of time have more opportunities to be selected, better avoid their long-term service rejection, and thus improve the average user satisfaction as a whole.

[0020] Optionally, the preset average user satisfaction performance evaluation function is:

[0021] ,or

[0022] ,or

[0023] ,

[0024] in, Indicates the number of cells, Indicates the number of users served by a single cell, Indicates the index of the current frame, Indicates a cell Medium users The maximum tolerable delay, Indicates a cell Medium users The end-to-end delay, Indicates the total number of sub-channels, Indicates a cell Medium users In the subchannel User selection on Indicates a cell Medium users User selection on all sub-channels, when When When the user is denied service, Indicates a cell Medium users The importance of users, Indicates the duration of each frame, represents the base of natural logarithms; is a preset constant and The scheme can achieve the following beneficial technical effects: through multiple parallel schemes, the average user satisfaction is effectively calculated to better evaluate the average user satisfaction of all users, and the overall service quality of all users in the network service range is ensured.

[0025] Optionally, the user importance is calculated by using a preset user importance function, wherein the preset user importance function is provided with a specified parameter for adjusting whether the user importance is related to the historical number of denied services, the specified parameter is set to a first value to control the degree of exponential increase of the user importance with the increase of the historical number of denied services, and the specified parameter is set to a second value to control the user importance to be a preset value. The scheme can achieve the following beneficial technical effects: the user importance function is provided with a specified parameter to adjust the calculation method of the user importance, which is used to adjust the degree of exponential increase of the user importance with the increase of the historical number of denied services or set the user importance to a constant value in different scenarios to meet diversified control requirements.

[0026] Optionally, the preset user importance function includes:

[0027] ,

[0028] wherein, represents the user importance of a user in a cell, represents the frame number of a current frame, represents the base number of a natural logarithm, represents a specified parameter, wherein, when the user importance is a preset value, the historical number of denied services is not considered; and when the user importance exponentially increases with the increase of the historical number of denied services. is the historical number of denied services of a user in a cell before the frame. The scheme can achieve the following beneficial technical effects: when the user importance exponentially increases with the increase of the historical number of denied services, the more the user is denied services, the higher the priority weight. Moreover, with the increase of the value of , the sensitivity of the user importance to denied services also increases, thereby reducing the number of denied services of some users and further improving fairness.

[0029] ​​​​​Optionally, the following allocation mechanism is iteratively performed with the optimization objective of maximizing the average user satisfaction: performing the first-stage allocation, including adjusting the communication resource allocation solution according to the current user selection and the computation resource allocation solution, the user environment state, the availability of the communication resource, and the channel state information from the user to the base station, wherein the first execution selects all users and uses the initial computation resource allocation solution; performing the second-stage user selection and computation resource allocation, including selecting users with the estimated end-to-end delay less than the maximum tolerable delay from the estimated end-to-end delay for each user obtained according to the task information and the communication resource allocation solution, and rejecting other users, and adjusting the computation resource allocation solution for the selected users according to the communication resource allocation solution, the user environment state, and the availability of the computation resource; and determining the latest communication resource allocation solution and the computation resource allocation solution according to the average user satisfaction corresponding to the communication resource allocation solution and the computation resource allocation solution obtained in the last adjustment, and the historical average user satisfaction corresponding to the communication resource allocation solution and the computation resource allocation solution. This scheme can at least achieve the following beneficial technical effects: the scheme iteratively performs the communication resource allocation, the user selection, and the computation resource allocation, and uses the optimization objective of maximizing the average user satisfaction, so that the effect of better utilizing the existing resources to improve the satisfaction of the selected users can be achieved by rejecting to provide services to some users who are difficult to meet the requirements.

[0030] According to a second aspect of the present application, a resource allocation system is provided for implementing the method of the first aspect, the system comprising: an internal and external environment perception module for performing the internal and external environment perception of the network, and obtaining the user environment state in the coverage of a plurality of cells and the network resource state, wherein the user environment state comprises the task information, the user importance, and the maximum tolerable delay of the task, and the network resource state comprises the availability of the communication resource, the availability of the computation resource, and the channel state information from the user to the base station in the network; a user selection and resource joint allocation module for iteratively allocating the communication resource and the computation resource to the selected users according to the user environment state and the network resource state, with the optimization objective of maximizing the average user satisfaction considering the satisfaction of the selected users and the satisfaction of the rejected users; a delay-perception-based user average satisfaction evaluation and user importance updating module for evaluating the average user satisfaction corresponding to the user selection, the communication resource allocation solution, and the computation resource allocation solution obtained by the allocation module, and updating the user importance according to the final allocation result.

[0031] According to a third aspect of the present application, an electronic device is provided, comprising: one or more processors; and a memory, wherein the memory is configured to store executable instructions; and the one or more processors are configured to implement the steps of the method of the first aspect by executing the executable instructions. BRIEF DESCRIPTION OF DRAWINGS

[0032] The embodiments of the present application will be further described with reference to the drawings, wherein:

[0033] Figure 1 A flowchart of a resource allocation method according to an embodiment of the present application;

[0034] Figure 2 A framework diagram of a resource allocation method according to an embodiment of the present application;

[0035] Figure 3 A system diagram for implementing a resource allocation method according to an embodiment of the present application;

[0036] Figure 4 A performance comparison diagram of convergence and optimal solution of a resource allocation method according to an embodiment of the present application;

[0037] Figure 5 A variation trend diagram of DA-AveUSD of a comparative experiment under different user numbers according to an embodiment of the present application;

[0038] Figure 6 A variation trend diagram of service success rate of a comparative experiment under different user numbers according to an embodiment of the present application;

[0039] Figure 7 A relationship trend diagram of average CAQA and maximum total energy of a comparative experiment under different maximum service times according to an embodiment of the present application. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application will be further described in detail with reference to the drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.

[0041] As mentioned in the background section, the prior art can not provide good service satisfaction for selected users, and it is difficult to guarantee the overall service quality of all users in the network service range. Considering the measurement of user service satisfaction, it is necessary to design an average user satisfaction index (AveUSD) which takes into account both selected users and rejected users, wherein the selected users and the rejected users make positive and negative contributions to the average user satisfaction (AveUSD). Unlike the prior art, the method of the present application combines user selection and resource allocation to maximize AveUSD in a multi-cell multi-user MEC network. First, a suitable model should be built to measure the AveUSD of selected users and rejected users. Then, with the goal of maximizing AveUSD, user selection and resource allocation are jointly optimized. For this purpose, the present application designs an environment state and resource aware joint user selection and resource allocation scheme (DA-JUSRA). The main idea is that in the process of resource allocation, users whose maximum tolerable delay cannot be met can be rejected. In this way, enough resources can be allocated to selected users to improve their delay performance and satisfaction.

[0042] According to one embodiment of the present application, referring to Figure 1 , a resource allocation method is provided, comprising: S1, obtaining the user environment state and network resource state in the coverage range of a plurality of cells, wherein the user environment state includes task information, user importance and maximum tolerable delay of the task, and the network resource state includes availability of communication resources, availability of computing resources and channel state information of users to base stations within the network; S2, according to the user environment state and the network resource state, iteratively allocating communication resources and computing resources to selected users with the optimization goal of maximizing the average user satisfaction which takes into account the satisfaction of selected users and the satisfaction of rejected users, wherein the rejected users are users whose maximum tolerable delay cannot be met according to the task information. Preferably, a preset average user satisfaction performance evaluation function is used to calculate the average user satisfaction, wherein the function is configured to: set the satisfaction of selected users positively related to the difference between their maximum tolerable delay and the estimated end-to-end delay; set the satisfaction of rejected users to a negative value, and negatively related to the product of user importance and the length of each frame.

[0043] In order to better understand the scheme of the method of the present application, the following describes the basic scenario, the average user satisfaction performance evaluation and the update scheme of user importance, and the resource allocation mechanism.

[0044] I. Basic scenario description

[0045] The purpose of resource allocation is mainly to offload tasks to MC-MEC network for auxiliary computing. As mentioned above, the present application combines user selection and resource allocation to maximize AveUSD in a multi-cell multi-user MEC network.Figure 2 An illustrative MC-MEC network is shown based on the joint user selection and resource allocation framework based on environmental state and resource perception, including a central control unit (CCU for short), M cells and U users. The CCU is responsible for managing the selection of users and the allocation of network resources. Each cell has a base station (BS for short) at its center, and a MEC server is deployed on the base station. For clarity, the index set of the cell, base station and MEC server is collectively referred to as M = {1,..., M}. Within the coverage of each cell, it is assumed that there are K = U / M users randomly distributed, and the user number set is represented by K = {1,..., K}. Let the system bandwidth be B, which is evenly divided into N orthogonal subchannels, and the orthogonal subchannel number set is represented by N = {1,..., N}. Assuming that the frequency reuse factor is 1, each cell can completely reuse all subchannels, and the users in the cell can share all subchannels through NOMA. Of course, this is only illustrative, and those skilled in the art can also apply the method of the present application based on orthogonal multiple access (OMA) technology.

[0046] The illustrative process of offloading tasks to the MC-MEC network to perform auxiliary computing in this scenario includes:

[0047] ① Waiting time;

[0048] ② Collection of information such as task information and user importance reporting;

[0049] ③ Resource perception, scheduling decision and notification of scheduling results

[0050] ④ Task upload time, i.e. the time required to offload tasks to the edge computing network;

[0051] ⑤ Task computation;

[0052] ⑥ Return of computation results, i.e. returning the computation results to the user.

[0053] Consider the user's compute-intensive and delay-sensitive tasks. Since the user's computing resources and energy are limited, it is assumed that full offloading is used, i.e. the task needs to be completely offloaded to the MEC network for processing. It is assumed that in the time domain, the minimum resource scheduling unit is a frame, and the duration of each frame (i.e. the length of each frame) is represented by , and the frame index is represented by , ∈I = {0, 1,...}. Tasks arriving in a frame will respond in the next frame +1. It is assumed that the base station and the user are both equipped with a single antenna, and each user can only select one subchannel for transmission in each frame. Define a four-tuple to represent the importance of tasks and users in each cell in each frame. Illustratively, is the cell Medium users The maximum tolerable delay varies from user to user. . It's a community Medium users The computational intensity of a task is the number of CPU cycles required to compute the predetermined amount of data in the task. For example, when cycles / bit is used as the unit, it is the number of CPU cycles required to compute one bit. It's a community Medium users The data size of the task. The unit of is, for example, bit. It's a community Medium users The importance of users (or the importance weight of the penalty when users are rejected).

[0054] 2. Average User Satisfaction Performance Evaluation and User Importance Update Scheme

[0055] In order to measure AveUSD for users with delay-sensitive tasks, a delay-aware USD model is proposed, which takes into account both selected and rejected users. The USD model for selected users is the end-to-end delay The non-increasing function of must satisfy the maximum tolerable delay requirement, i.e. Delay Gain Should be able to make a positive contribution to USD. The greater the delay gain, the higher the USD. When a user is denied service in the current frame, a corresponding penalty should be introduced. If in the cell Importance weight Users In the current frame If it is rejected, the weighted penalty is , because it has to wait at least one frame to be served. Note that the importance weight Can be adjusted to balance the model's preferences and avoid unfairness, such as some users being selected every time.

[0056] According to one embodiment of the present invention, user importance is calculated using a preset user importance function. The preset user importance function includes a specified parameter for adjusting whether user importance is correlated with the number of historical service denials. The specified parameter is set to a first value (e.g., 1, 2, or 4) to control the exponential increase in user importance as the number of historical service denials increases. The specified parameter is set to a second value (e.g., 0) to control user importance to a preset value. Preferably, user importance is calculated using the following preset user importance function:

[0057]

[0058] wherein, denotes the user importance of a user in a cell denotes the user importance of a user in a cell denotes the user importance of a user in a cell denotes the frame number of the current frame, denotes the base of the natural logarithm, denotes a specified parameter, wherein, when the user importance is a pre-set value, regardless of the number of history of denial of service; when the user importance increases exponentially with the increase of the number of history of denial of service; is the number of history of denial of service of a user in a cell is the number of history of denial of service of a user in a cell is the number of history of denial of service of a user in a cell is the number of history of denial of service of a user in a cell is a hyper parameter to control the sensitivity of the priority weight to the history of denial of service, wherein, when the importance of each user is the same, regardless of the number of history of denial of service; when the more the number of history of denial of service of a user, the higher the priority weight. With the increase of , the sensitivity of the user importance to the denial of service also increases, thereby improving the fairness.

[0059] Therefore, the performance evaluation formula of the DA-AveUSD model, i.e., the average user satisfaction performance evaluation function, is:

[0060]

[0061] wherein, denotes the user selection of a user in a cell denotes the user selection of a user in a cell denotes the user selection of a user in a cell denotes that the user is selected and gets service, denotes that the user is denied. And the delay gain of the selected user is related to the communication and computing resources it gets. In short, the more the communication and computing resources, the smaller the upload delay of the user, thereby the more the delay gain of the user, and the higher the USD. Obviously, DA-AveUSD is affected by both user selection and resource allocation.

[0062] It should be understood that those skilled in the art can also adjust the above formula to obtain a DA-AveUSD model with similar effects, for example:

[0063] or

[0064]

[0065] wherein, denotes the base of the natural logarithm; is a preset constant and .

[0066] III. Resource allocation mechanism

[0067] According to an embodiment of the present application, a resource allocation method is provided, which comprises: obtaining task information of users, user importance, user environment state and network resource state in the coverage of a plurality of cells, wherein the network resource state comprises network internal; according to the task information of the users, the user importance, the user environment state and the network resource state, iteratively allocating communication resources and computing resources to selected users with the optimization target of maximizing the average user satisfaction degree which simultaneously considers the satisfaction degree of the selected users and the satisfaction degree of rejected users, wherein the rejected users are users whose maximum tolerable delay cannot be satisfied according to the task information.

[0068] According to an embodiment of the present application, the following allocation mechanism is iteratively executed multiple times with the optimization target of maximizing the average user satisfaction degree: performing a first-stage allocation, which comprises: adjusting the communication resource allocation solution according to the current user selection and computing resource allocation solution, the user environment state, the availability of the communication resources and the channel state information of the users to the base station, wherein the first execution selects all users and adopts an initial computing resource allocation solution; performing a second-stage user selection and computing resource allocation, which comprises: obtaining the estimated end-to-end delay for each user according to the task information and the communication resource allocation solution, selecting the users whose estimated end-to-end delay is less than the maximum tolerable delay and rejecting other users, and adjusting the computing resource allocation solution for the selected users according to the communication resource allocation solution, the user environment state and the availability of the computing resources; determining the latest communication resource allocation solution and computing resource allocation solution according to the average user satisfaction degree corresponding to the communication resource allocation solution and the computing resource allocation solution obtained by the previous adjustment and the average user satisfaction degree corresponding to the historical communication resource allocation solution and the computing resource allocation solution.

[0069] In order to maximize the performance of DA-AveUSD, user selection and resource allocation are required. Importantly, the selected users should not be selected based on the traditional principle of first-come-first-served or good channel state priority, but need to be based on the joint optimization mechanism of multi-dimensional environment and resource state awareness, dynamically reject users who cannot be satisfied and release resources, and cooperatively allocate communication (such as NOMA spectrum) and computing resources, break the limitations of traditional single-factor decision (such as only relying on channel state or maximum tolerable delay), and reduce the performance loss caused by resource fragmentation. Therefore, the present application proposes an environment state and resource state awareness joint user selection and resource allocation method, which can be performed by, for example Figure 3The system shown is implemented, which comprises: internal and external environment perception module, user selection and resource joint allocation module, time delay perception user average satisfaction evaluation and user importance update module. For Figure 2 The DA-AveUSD maximization problem in the basic scenario, the definition of these modules is as follows:

[0070] (1) Internal and external environment perception module: used to perform network internal and external environment perception, wherein, the network internal environment perception includes network internal communication resource availability and computing resource availability and user to base station channel state information, wherein, the user to base station channel state information will affect the upload communication state of the task, and then affect the user upload time delay, the communication resource availability includes the transmitting power and the spectrum, the user to base station channel state information includes large scale fading and small scale fading. And the network internal communication and computing resource availability will affect the communication and computing time delay of the task. These will affect the user's satisfaction.

[0071] The purpose of external environment perception is to obtain user environment state, mainly by the user to actively report, the user environment state includes task information and user importance, wherein, the task information includes the size of the task, the computing intensity and the maximum tolerable delay of the task. These external information concerns whether the user will be rejected, for example, the greater the user importance, the greater the satisfaction penalty caused by rejecting it, so it will tend to receive the task of the user with high user importance, and the size and computing intensity of the task will affect the upload and computing delay of the task, if these are too large, and the maximum tolerable delay of the task is small, even if the user is selected, it will also lead to the failure of the task, it is not better than not selecting the user, and releasing these resources to the remaining selected users to improve their satisfaction.

[0072] (2) User selection and resource joint allocation module: Based on the dynamic input of the internal and external environment perception module, a multi-dimensional decision space is constructed, and intelligent collaboration of user access and resource allocation is realized through joint optimization algorithm. The allocation module is configured to: adopt a two-stage collaborative optimization architecture to realize the dynamic adaptation of communication-computing resources and the closed-loop iteration of user access decision. First stage: communication resource pre-allocation and digital conversion, the communication resource is allocated by using the perceived channel state information and communication resource availability information, and the communication resource solution under the current user selection and computing resource allocation is obtained by using the fractional programming mathematical method to iteratively solve the convex problem after problem conversion. Based on the solution, the communication delay can be estimated, and then based on the communication resource solution and the estimated communication delay, the second stage is entered; second stage: first pass through the filter engine layer, which will eliminate users whose estimated total delay (i.e. end-to-end delay) exceeds the maximum tolerable delay of the task, then select users and allocate computing resources among the remaining users. User selection and computing resource allocation in this part can be allocated and solved by using evolutionary algorithms or deep reinforcement learning algorithms, including but not limited to genetic algorithms, differential algorithms, multi-agent reinforcement learning algorithms (such as Multi-Agent Deep Q-Network (MADQN), Multi-Agent Proximal Policy Optimization (MAPPO), etc.), but the detailed details of how to solve are not within the scope of discussion of the present application. Then through the loop iteration until the DA-AveUSD converges to the optimal value, at this time the corresponding communication resource allocation solution, user selection and computing resource allocation solution are the optimal solution, which constitutes the final scheduling decision.

[0073] (3) Delay-aware user average satisfaction evaluation and user importance updating module: This module constructs a closed-loop feedback system to quantify the impact of user selection and resource allocation on user satisfaction, and dynamically adjusts the user importance weight accordingly. The output of the user selection and resource joint allocation module is used as the input of the satisfaction evaluation, i.e. the delay-aware user average satisfaction index is calculated using the final scheduling decision. And according to the user scheduling result and the historical user scheduling result, the user importance is updated, and the updated result is transmitted to the user through the downlink communication, and the user updates its own user importance information before reporting the task information and user importance in the next frame, forming a long-term dynamic optimization closed loop, avoiding the existence of long-term unfair factors. This architecture is significantly different from traditional static resource allocation schemes, by introducing a delay-weight joint sensitivity mechanism.

[0074] In order to verify the effect of the present application, the inventors also simulated and verified the resource allocation method proposed, the simulation parameters used are shown in Table 1, and the simulation results are shown in Figures 4-7 .

[0075] Table 1 Simulation parameter description

[0076] Cell radius 400m Total bandwidth B 20 MHz Time of one frame 10 ms Total number of frames 100 Noise power -174 dBm / Hz Number of cells M 3，7 Number of users U 9-70 Maximum transmit power of users 23 dBm

[0077] To evaluate the performance of the proposed joint user selection and resource allocation method with consideration of environment state and resource awareness (JUSRA-CEAO), it is compared with three algorithms:

[0078] (1) FURA: first-come-first-serve (FCFS) user selection mechanism, resources are allocated one by one in order until there is no resource for new users.

[0079] (2) LURA: first-serve deadline-earlier (EDSF), resources are allocated one by one in order until there is no resource for new users.

[0080] (3) JUSRA-ESA: to obtain the best DA-JUSRA performance, an exhaustive search algorithm (ESA) is proposed. All possible user selection and resource schemes are exhaustively searched.

[0081] The results of the convergence of the JUSRA-CEAO mechanism and the performance comparison with the optimal solution are shown in Figure 4 Figure 4 The convergence of the proposed JUSRA-CEAO algorithm is verified when the number of cells is 3. It can be seen that, compared with the optimal JUSRA-ESA based on exhaustive search, JUSRA-CEAO can always converge within a few iterations, and the difference between DA-AveUSD and the optimal performance JUSRA-ESA is less than 1% under different total number of users U.

[0082] Two performance indicators, DA-AveUSD and service success rate, are considered when studying the effectiveness of JUSRA-CEAO. Service success rate is an important performance indicator because it directly affects DA-AveUSD, and analyzing it can provide a deeper understanding of the changes in DA-AveUSD.

[0083] When the number of cells is 7, Figure 5 and Figure 6 show the performance of DA-AveUSD and service success rate as a function of different numbers of users, respectively. Figure 5 The trend of DA-AveUSD under different numbers of users is shown, and from Figure 5 it can be seen that the DA-AveUSD performance of all schemes decreases with the increase in the number of users. This is because when the number of users increases, resource competition occurs, leading to a decrease in service success rate (as shown in Figure 6 ​), which leads to a decrease in DA-AveUSD. When the number of users is low (e.g., U=14), the DA-AveUSD performance of all schemes is similar. This is because when the number of users is low, the available resources are usually sufficient to meet the needs of all users, so the service success rate of all schemes is close to 100% (e.g., Figure 6 (as shown), resulting in similar DA-AveUSD. It can also be seen that JUSRA-CEAO consistently achieves the highest DA-AveUSD among all schemes. First, to illustrate the necessity of a joint optimization mechanism for user selection and resource allocation, it is observed that FURA and LURA perform poorly. They prioritize user selection based on arrival time and deadline, respectively. However, the earliest arriving or earliest due task does not guarantee a high USD.

[0084] In addition, the proposed DA-AveUSD model is a penalty mechanism based on user importance weighting, where user importance is based on the number of times a user has been denied service in the past. This mechanism prevents some users from receiving services too frequently while other users are ignored for a long time, which is crucial for user fairness because no user wants to be ignored all the time. To evaluate the effectiveness of this fairness mechanism, a commonly used fairness metric (Jain index) is used to evaluate the fairness of resource allocation among users. It is calculated as follows:

[0085] ,

[0086] in, It's a community Medium users The number of times the service was performed historically, and U is the total number of users.

[0087] Figure 7 Shows different maximum service times The relationship between the average CAQA and the maximum total energy is shown in Figure 2. When the number of cells is 3, Figure 7 The cumulative distribution function (CDF) of the number of times different users are served is shown. is the hyperparameter of the penalty weight in the DA-AveUSD model. As increases, the CDF curve becomes steeper, indicating that the number of services is more concentrated and the differences between users are reduced. The fairness of the user selection mechanism will also improve with the increase of , which can also be proved by the increase of Jain fairness index. This is because increasing the penalty weight of users who are frequently denied service will make the system more inclined to provide services to them. Therefore, increasing It is effective to improve fairness, and the proposed user importance weighted penalty mechanism is also effective.

[0088] In general, the present application proposes a user satisfaction performance evaluation scheme in the case of insufficient resources to serve all users, i.e., delay-aware average user satisfaction (AveUSD), which includes the satisfaction of selected and rejected users, wherein the satisfaction of selected users decreases with the increase of actual delay, and the satisfaction of rejected users is related to the importance of users and the length of each frame, wherein the importance of users increases exponentially with the increase of the number of historical rejections of users, thereby avoiding the unfairness phenomenon that some users cannot be served all the time. Secondly, the present application proposes an environment and resource state-aware user selection and resource allocation system, which jointly makes a user selection and resource allocation scheme based on the perceived environment state information of users and the integrated resource availability information in the multi-MEC network to maximize the average satisfaction of all users within the network service range.

[0089] It should be noted that although the above describes the steps in a certain order, it does not mean that the steps must be performed in the above-mentioned specific order, in fact, some of the steps can be performed concurrently, or even in reverse order, as long as the desired function can be achieved.

[0090] The present application can be a system, a method, and / or a computer program product. The computer program product can include a computer readable storage medium (or media) having computer readable program instructions thereon for causing a processor to carry out aspects of the present application.

[0091] The computer readable storage medium can be a tangible device that can retain and store instructions for use by an instruction execution device. The computer readable storage medium can be, for example, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium include the following: a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically encoded device such as punch cards or punched tape, and any suitable combination of the foregoing. A non-transitory, computer-readable storage medium does not include a signal.

[0092] Having described various embodiments of the application, it is to be understood that the above description is meant not to limit and not to encompass all of the possible embodiments covered by the claims. Many modifications and variations of this application can be apparent to those of ordinary skill in the art without departing from the spirit and scope of the described embodiments. It is intended that the scope of the application should only be limited by the appended claims.

Claims

1. A resource allocation method characterized by, The method comprises: obtaining user environment states and network resource states in a coverage of a plurality of cells, wherein the user environment states comprise task information, user importance and maximum tolerable delay of a task, and the network resource states comprise availability of communication resources, availability of computing resources and channel state information from users to base stations within the network; iteratively allocating communication resources and computing resources to selected users according to the user environment states and the network resource states, with an optimization objective of maximizing average user satisfaction that simultaneously considers satisfaction of the selected users and satisfaction of rejected users, wherein the rejected users are users whose maximum tolerable delay cannot be met according to the task information.

2. The method of claim 1, wherein, calculating the average user satisfaction by using a preset average user satisfaction performance evaluation function, wherein the function is configured to: positively correlate the satisfaction of the selected users with a difference between their maximum tolerable delay and estimated end-to-end delay; and positively correlate the satisfaction of the rejected users with a product of user importance and time length of each frame.

3. The method of claim 2, wherein, The preset average user satisfaction performance evaluation function is: or or ， wherein, denotes the number of cells, denotes the number of users served by a single cell, denotes the index of the current frame, denotes the cell in which a user is located, denotes the end-to-end delay of a user in a cell , denotes the total number of sub-channels, denotes the user selection of a user in a cell over a sub-channel , denotes the user selection of a user in a cell over all sub-channels, when denotes that the user is selected and served, when denotes that the user is rejected from service, denotes the user importance of a user in a cell , denotes the duration of each frame, denotes the base of the natural logarithm; is a pre-defined constant and .

4. The method according to one of claims 1 to 3, characterized in that, calculating the user importance by using a preset user importance function, wherein the preset user importance function is provided with a specified parameter for adjusting whether the user importance is related to a historical number of times of being rejected, and the specified parameter is set to a first value to control an extent to which the user importance exponentially increases with the historical number of times of being rejected.

5. The method of claim 4, wherein, The specified parameter is set to a second value to control the user importance to be a preset value.

6. The method of claim 4, wherein, The preset user importance function comprises: ， wherein, denotes a cell in which a user has a user importance, denotes a frame number of a current frame, denotes a base number of a natural logarithm, denotes a designated parameter, wherein, when the user importance is a preset value, regardless of a history number of denial of service; and when the user importance is exponentially increased as the history number of denial of service increases; is a cell in which a user has a history number of denial of service before a frame.

7. The method according to one of claims 1 to 3, characterized in that iteratively performing the following allocation mechanism multiple times with the optimization objective of maximizing the average user satisfaction: performing a first-stage allocation, comprising: adjusting a communication resource allocation solution according to a current user selection and computing resource allocation solution, user environment states, availability of communication resources and channel state information from users to base stations, wherein the first time of performing the first-stage allocation is performed with all users being selected and an initial computing resource allocation solution being used; performing a second-stage user selection and computing resource allocation, comprising: selecting users whose estimated end-to-end delay is less than the maximum tolerable delay and rejecting other users according to the estimated end-to-end delay obtained from the task information and the communication resource allocation solution, and adjusting a computing resource allocation solution for the selected users according to the communication resource allocation solution, the user environment states and availability of computing resources; determining a latest communication resource allocation solution and computing resource allocation solution according to the average user satisfaction corresponding to the communication resource allocation solution and the computing resource allocation solution obtained by the last adjustment and the average user satisfaction corresponding to historical communication resource allocation solutions and computing resource allocation solutions.

8. A resource allocation system for implementing the method of any one of claims 1-7, the system comprising: An internal and external environment perception module is configured to perform internal and external environment perception of the network, and obtain user environment states and network resource states in a coverage of a plurality of cells, wherein the user environment states include task information, user importance and maximum tolerable delay of the task, and the network resource states include availability of communication resources, availability of computing resources and channel state information of users to base stations in the network; A user selection and resource joint allocation module is configured to allocate communication resources and computing resources to selected users iteratively according to the user environment states and the network resource states, with an optimization target of maximizing average user satisfaction considering satisfaction of the selected users and satisfaction of rejected users simultaneously, wherein the rejected users are users whose maximum tolerable delay cannot be met according to the task information; A delay-aware user average satisfaction evaluation and user importance updating module is configured to evaluate average user satisfaction corresponding to user selection, communication resource allocation solution and computing resource allocation solution obtained by the allocation module, and update user importance according to a final allocation result.

9. A computer-readable storage medium, characterized in that, A computer program is stored thereon, and the computer program is executable by a processor to implement steps of the method of any one of claims 1-7.

10. An electronic device, comprising: Comprise: One or more processors; And A memory, wherein the memory is configured to store executable instructions; The one or more processors are configured to implement steps of the method of any one of claims 1-7 via execution of the executable instructions.

Images