Online resource management and control method based on time delay perception

Abstract

The invention discloses an online resource management and control method based on time delay perception, and belongs to the field of electric power Internet of Things. The method specifically comprises the following steps of: firstly, initializing access selection and a backhaul transmission strategy of each user for a time slot t in an electric power Internet of Things scene, and forming a decision vector; for the task under the time slot, respectively calculating the transmission rate of accessing the small base station and the macro base station, and the backhaul transmission rate; then, the total execution time of the task is calculated, the average delay cost of the long-term execution time of all the tasks is further calculated, the cost minimization serves as an optimization target, and an online resource management and control model of task scheduling is constructed; and finally, solving by using a user online classification algorithm, an MBS-based online resource management and control algorithm, an SBS-based online resource management and control algorithm and a control parameter updating algorithm to obtain an access strategy of the user and a backhaul transmission strategy of the base station. According to the invention, load balancing between the base stations is realized, and efficient execution of tasks is ensured.

Description

Technical Field

[0001] The present invention belongs to the field of electric power Internet of Things, relates to online management and control of resources, and specifically is an online resource management and control method based on delay perception. Background Art

[0002] In recent years, with the rapid development of the Internet of Electric Power Things (IoT), the number of instant response applications has grown rapidly, which has led to a sharp increase in resource requirements and more stringent processing standards. In the IoT environment of Electric Power Things, which pursues high efficiency, stability and real-time performance, the increase in latency may have a significant impact on the overall performance of the system and user experience.

[0003] Mobile edge computing (MEC) technology, with its natural advantage of being close to the data source, has shown great potential in reducing task processing latency and has become a key technology for supporting latency-sensitive applications. However, in the power Internet of Things scenario, when using edge servers for resource scheduling, traditional resource management strategies have two problems:

[0004] First, in the power Internet of Things, the existing edge server deployment strategy is often relatively simple, mainly limited to macro base stations (MBS) or small base stations (SBS), and fails to fully consider the diversity and dynamics of computing tasks. In particular, in certain specific areas or time periods, computing tasks may increase dramatically, and this single-mode deployment solution often cannot respond to sudden changes in task volume in a timely and flexible manner [1][2].

[0005] In addition, when computing tasks between macro base stations and small base stations are unevenly distributed, it is difficult to achieve effective resource scheduling and optimization by relying solely on a single type of base station for edge server deployment. For example, reference [3] focuses on the task scheduling strategy in small cell mobile edge computing networks and studies the distributed collaboration mechanism of edge servers, which has the problem of high algorithm complexity. Reference [4] only considers the service migration strategy between small base stations, ignoring that when computing tasks surge in the coverage area of ​​small base stations, macro base stations may have relatively sufficient computing resources and processing capabilities, while small base stations are overloaded and cannot process all tasks in time. This imbalance in resource allocation will further increase the system load pressure and reduce the efficiency and quality of task processing.

[0006] Considering that the backhaul link between the macro base station and the small base station can alleviate the computing pressure of the small base station server, a dynamic backhaul transmission strategy is designed based on this, which allows the tasks to be offloaded from the small base station to the edge server of the macro base station for execution, ensuring load balancing between base stations while improving resource utilization efficiency.

[0007] Second, in the electric power Internet of Things, the existing edge server deployment scheme is often based on the predicted task requirements to pre-configure resources. Although this method can improve resource utilization efficiency and service response speed to a certain extent, it also has significant disadvantages:

[0008] First, the accuracy of the forecast is difficult to guarantee, especially when faced with rapidly changing task requirements. The pre-configured resources may not be able to meet the actual changes in demand, resulting in insufficient or excessive resources.

[0009] Secondly, the prediction-based pre-configuration method lacks flexibility and is difficult to quickly adapt to sudden mission requirements or emergencies.

[0010] For example, in order to realize the awareness of the distribution network situation, reference [5] designs a resource block reservation mechanism to maximize the energy efficiency of the network. Reference [6] considers a wirelessly powered MEC system, in which the base station is equipped with renewable energy, which can transmit energy to wireless devices and trade energy with the power grid. By optimizing the amount of user unloaded data and the transmission power, the total energy cost of the system is minimized. Reference [7] establishes a communication model of a MEC system based on relay node assistance in smart grids, which minimizes the energy consumption cost of the system while satisfying the constraints of relay node power and task calculation delay.

[0011] However, the above literature often fails to effectively adapt to the highly dynamic network environment. The limitations of these strategies are mainly reflected in the large deviation between the preset task conditions and the complex and changing network environment. Relying on static or relatively fixed prior knowledge for resource pre-configuration cannot meet the needs of actual applications.

[0012] References are as follows:

[0013] [1]Y.Zheng, T.Zhang and J.Loo, "Dynamic Multi-Time Scale User Admission and Resource Allocation for Semantic Extraction in MEC Systems," in IEEETransactions on Vehicular Technology, vol.72, no.12, pp.16441-16453, Dec.2023.

[0014] [2]T.H.Hoang,C.T.Nguyen,T.N.Do and G.Kaddoum,"Joint Task Offloadingand Radio Resource Management in Stochastic MEC Systems,"in IEEE Transactionson Communications,vol.72,no.5,pp.2670-2686,May 2024.

[0015] [3]H.Li,K.Xiong,Y.Lu,W.Chen,P.Fan and K.B.Letaief,"Collaborative TaskOffloading and Resource Allocation in Small-Cell MEC:A Multi-Agent PPO-BasedScheme,"in IEEE Transactions on Mobile Computing,pp.1-13,Nov.2024.

[0016] [4]Y.Shi,C.Yi,B.Chen,C.Yang and J.Cai,"A Two-Timescale OnlineOptimization for Balancing Service Migration and Task Rerouting in MEC,"GLOBECOM 2023-2023IEEE Global Communications Conference,Kuala Lumpur,Malaysia,2023,pp.5500-5505.

[0017] [5]Q.Li,H.Tang,Z.Liu,J.Li,X.Xu and W.Sun,"Optimal Resource Allocationof 5GMachine-Type Communications for Situation Awareness in ActiveDistribution Networks,"in IEEE Systems Journal,vol.16,no.3,pp.4187-4197,Sept.2022.

[0018] [6] N.Li, W.Hao, F.Zhou, M.Zeng and S.Yang, "Smart Grid EnabledComputation Offloading and Resource Allocation for SWIPT-Based MEC System," in IEEE Transactions on Circuits and Systems II: Express Briefs, vol.69, no.8, pp.3610-3614, Aug.2022.

[0019] [7] P.Liu, J.Wang, K.Ma and Q.Guo, "Joint Cooperative Computation and Communication for Demand-Side NOMA-MEC Systems With Relay Assistance in SmartGrid Communications," in IEEE Internet of Things Journal, vol.11, no.19, pp.30594-30606, 1Oct.1, 2024. Summary of the invention

[0020] In view of the existing technology, when designing resource management in the power Internet of Things, the deployment of edge servers is limited to macro base stations or small base stations, the deployment mode is relatively single, and resources are often pre-configured based on predicted task requirements, which cannot adapt to the diversity and dynamics of tasks. The present invention proposes an online resource management method based on delay perception in the power Internet of Things, which adjusts the task scheduling scheme according to the real-time occupancy of macro base station and small base station resources, thereby achieving load balancing between base stations and ensuring efficient execution of tasks.

[0021] The proposed online resource control method based on latency perception has the following specific steps:

[0022] Step 1: Build a dynamic scenario including a macro base station, M small base stations and I user in the power Internet of Things:

[0023] The macro base station and M small base stations use {0}∪ M ={0}∪{1,2,...,m,...,M} represents; multiple MEC servers are deployed in small base stations and macro base stations, named MEC-SBS and MEC-MBS respectively.

[0024] User collection I={1,2,...i,...,I} means that each user only publishes one task. I That is, a collection of tasks.

[0025] Each task uses a four-tuple Indicates; where L i (t) represents the task size of user i in time slot t, A i represents the task release time of user i, represents the bandwidth resources required by user i, f i Represents the computing resources required by user i.

[0026] The T time slots of this scenario are represented by the set T ={1,2,...t,...,T} represents the subcarriers of the base station. K ={1,2,...k,...,K} represents.

[0027] Step 2: For time slot t, initialize the access selection of each user and the backhaul transmission strategy from the small base station to the macro base station to form the system decision vector X(t) = [a i,k,0 (t),a i,k,m (t),s i,m,0 (t)].

[0028] element a i,k,0 (t) and a i,k,m (t) represents the binary access selection variable of user i; specifically, a i,k,0 (t) = 1 means that user i accesses the macro base station through subcarrier k in time slot t and offloads its tasks to MEC-MBS, i,k,m (t) = 1 means that user i accesses small base station m through subcarrier k in time slot t and offloads its tasks to the mth MEC-SBS, a i,k,0 (t) = 0 or a i,k,m (t)=0 means that user i does not access any base station in time slot t.

[0029] Element i,m,0 (t) indicates that task i on the mth MEC-SBS is offloaded to the MEC-MBS at time slot t.

[0030] Step 3: For task i in time slot t, calculate the transmission rate r of the task accessing the mth small base station through subcarrier k i,k,m (t) and the transmission rate r of the access macro base station i,k,0 (t):

[0031] The calculation formula is as follows:

[0032]

[0033] Where α={α 1 ,...,α k ,...,α K} represents the bandwidth allocation vector of the subcarrier, α k represents the bandwidth allocation vector of subcarrier k; g i,m (t) represents the channel gain between user i and small base station m, I i,k,m (t) represents the interference from other SBSs or MBSs occupying the same subcarrier k, p i represents the transmission power of user i, n 0 represents the one-sided noise spectral density.

[0034]

[0035] Among them I i,k,0 (t) represents the interference from other SBSs occupying the same subcarrier k; g i,0 (t) represents the channel gain between user i and the macro base station.

[0036] Step 4: For task i in time slot t, calculate the backhaul transmission rate r of the task from small base station m to macro base station through subcarrier c i,m,0 (t):

[0037]

[0038] Among them, the subcarrier set of the backhaul link is expressed as C ={1,2,...c,...,C},α c represents the bandwidth of subcarrier c. m represents the transmission power of small base station m. m,0 (t) represents the channel gain between the small base station m and the macro base station.

[0039] Step 5: For task i in time slot t, calculate the total execution time T of the task i :

[0040]

[0041] Among them, E {x} If event x occurs, then E {x} =1, otherwise E {x} =0; T represents the number of time slots.

[0042] Step 6: Calculate the average delay cost of the long-term execution time of all tasks

[0043]

[0044] I represents the number of tasks;

[0045] Step 7: Taking the minimization of average delay cost as the optimization goal, jointly optimize the user's access decision and the base station's backhaul transmission strategy, and build an online resource management and control model for task scheduling;

[0046] The online resource control model is expressed as:

[0047]

[0048] Constraint C1 indicates that in one time slot, the same task can only choose to access one base station or perform local calculations.

[0049] Constraint C2 indicates that the bandwidth resources allocated by SBS in a time slot should be subject to the total amount limit, where B SBS Indicates the maximum value of the small base station bandwidth.

[0050] Constraint C3 indicates that the bandwidth resources allocated by MBS in a time slot should be subject to the total amount limit, where B MBS Indicates the maximum value of the macro base station bandwidth.

[0051] Constraint C4 indicates that in a time slot, for any small base station, the sum of all link transmission capabilities cannot exceed its own maximum computing capability. represents the maximum value of the m-th MEC-SBS computing resources, η m represents the computing resources required by the m-th MEC-SBS to process 1 bit of data, θ m (t-1) represents the control parameters of the mth MEC-SBS in the t-1 time slot.

[0052] Constraint C5 indicates that in a time slot, the sum of all link transmission capabilities for a macro base station cannot exceed its own maximum computing capability. represents the maximum value of MEC-MBS computing resources, η 0 represents the computing resources required by MEC-MBS to process 1 bit of data, θ 0 (t-1) represents the control parameters of MEC-MBS in time slot t-1.

[0053] Constraint C6 indicates that the user's access selection and the base station's backhaul transmission strategy are both binary variables.

[0054] Step 8: Use the user online classification algorithm, the MBS-based online resource management algorithm, the SBS-based online resource management algorithm, and the control parameter update algorithm to solve the online resource management model, and finally obtain the user access strategy and the base station backhaul transmission strategy.

[0055] The specific solution process is:

[0056] Step 801: Create a task classification set: a set S of tasks to be executed locally local (t), the set S executed in MEC-SBS SBS (t) and the set S executed in MEC-MBS MBS (t).

[0057] Step 802: Calculate the computational latency of each task executed locally. The computation latency of the m-th MEC-SBS And the computational latency performed in MEC-MBS

[0058] The computation latency for local execution is: C i represents the computational density of task i.

[0059] The computation delay performed at the mth MEC-SBS is:

[0060] The calculation delay in MEC-MBS is:

[0061] Step 803: Using the user online classification algorithm, by comparing the three calculation delays and The size of , divides task i into the corresponding classification set;

[0062] Specifically:

[0063] like The task performs local computation and has a i,k,0 (t) = a i,k,m (t) = s i,m,0 (t) = 0, and divide task i into the set S local (t) in.

[0064] like The task performs MEC-MBS calculations, with a i,k,0 (t) = 1, a i,k,m (t) = 0, s i,m,0 (t) = 0, then divide task i into set S MBS (t) medium;

[0065] like The task performs the mth MEC-SBS calculation, with a i,k,0 (t) = 0, a i,k,m (t) = 1, s i,m,0 (t) = 0, then divide task i into set S SBS (t) in.

[0066] Step 804: Use the MBS-based online resource management algorithm to manage the set S MBS Each task in (t) performs access judgment:

[0067] First, determine whether the remaining bandwidth of the MBS is greater than the bandwidth of the subcarrier k allocated to task i, that is, whether the following conditions are satisfied:

[0068]

[0069] If not, divide this task into set S local (t) and reset a i,k,0 (t) = a i,k,m (t) = s i,m,0 (t) = 0; if yes, further determine whether the sum of all link transmission capabilities of the macro base station is less than its own maximum computing capability, that is, determine whether:

[0070]

[0071] If not, divide this task into set S local (t) and reset a i,k,0 (t) = a i,k,m (t) = s i,m,0 (t) = 0. If yes, task i accesses the macro base station through subcarrier k and performs calculations in MEC-MBS, then there is a i,k,0 (t) = 1, a i,k,m (t) = 0, s i,m,0 (t)=0.

[0072] Step 805: Use the SBS-based online resource management algorithm to manage the set S SBS Each task in (t) performs access judgment:

[0073] First, determine whether the remaining bandwidth of the SBS is greater than the bandwidth of the subcarrier k allocated to task i, that is, determine whether:

[0074]

[0075] If not, divide this task into set S local (t) and reset a i,k,0 (t) = a i,k,m (t) = s i,m,0 (t) = 0. If yes, further determine whether the sum of the link transmission capabilities of all small base stations is less than its own maximum computing capability, that is, determine whether:

[0076]

[0077] If yes, task i accesses the mth small base station through subcarrier k and performs calculation on the corresponding MEC-SBS, then:

[0078] a i,k,0 (t) = 0, a i,k,m (t) = 1, s i,m,0 (t) = 0

[0079] Otherwise, it is further determined whether the sum of the link transmission capabilities of the macro base station is less than its own maximum computing capability, that is, whether the following conditions are satisfied:

[0080] If not, divide this task i into the set S local (t) and reset a i,k,0 (t) = a i,k,m (t) = s i,m,0 (t) = 0. If yes, task i is offloaded to the macro base station via the backhaul link and calculated on the MEC-MBS, then s i,m,0 (t) = 1, a i,k,0 (t) = 1, a i,k,m (t)=0.

[0081] Step 806: Update the control parameter θ using the control parameter update algorithm 0 (t) and θ m (t), realize load balancing between base stations:

[0082] In time slot t, θ 0 (t) is calculated as:

[0083]

[0084] θ m (t) is calculated as:

[0085] Step 807: In each time slot, the unloading decision of each task must be re-judged based on the classification set to which it was initially located, so as to continuously update the classification set of tasks until the optimal base station access strategy for all tasks and the return transmission strategy between base stations are obtained, which are used to guide the access of users in this time slot and the return transmission method of the base station.

[0086] The advantages of the present invention are:

[0087] 1. The present invention provides an online resource management and control method based on delay perception. After preliminarily classifying tasks according to the computing delays of tasks at different nodes, the access decision of the tasks can be re-judged according to the real-time occupancy of resources of each base station, thereby achieving load balancing between base stations while minimizing the average delay cost of all users.

[0088] 2. The present invention provides an online resource management and control method based on delay perception, which uses simulation to evaluate performance and compares it with existing algorithms, namely, tasks are only executed locally online, tasks are only executed online in small base stations, and tasks are only executed online in macro base stations. This shows the superiority of the present method in reducing task processing delay. BRIEF DESCRIPTION OF THE DRAWINGS

[0089] Figure 1 This is a flow chart of an online resource management and control method based on delay perception of the present invention.

[0090] Figure 2 The figure is a schematic diagram for comparing the performance of the present invention with that of the existing resource management and control solutions.

[0091] Figure 3 These are the system optimization results of the two control parameters of the present invention under different numbers of users. DETAILED DESCRIPTION

[0092] In order to facilitate those skilled in the art to understand and implement the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. Obviously, the described embodiments are only partial embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work should fall within the scope of protection of the present invention.

[0093] In view of the limitations of edge server deployment in the power Internet of Things and the inflexibility of existing resource pre-allocation strategies, the present invention proposes an online resource management method based on delay perception, specifically: first, establish a long-term average delay cost minimization model for all users; according to the size relationship of the calculation delay of the computing task on different nodes, build a matching model between the task and the local computing node or base station; then, design online resource management strategies on small base stations and macro base stations respectively; in order to achieve load balancing between base stations, formulate a control parameter update method for the backhaul transmission rate between base stations; thereby adjusting the task scheduling plan according to the real-time occupancy of base station resources to reduce the long-term average delay cost of all users. Finally, the above algorithms are used in combination to solve the problem of minimizing the long-term average delay cost of all users. By formulating an online resource management strategy, the present invention ensures load balancing between base stations while minimizing the long-term average delay cost of all users.

[0094] The proposed online resource control method based on latency awareness, such as Figure 1 As shown, the specific steps are as follows:

[0095] Step 1: Build a dynamic scenario including a macro base station, M small base stations and I user in the power Internet of Things:

[0096] The macro base station and M small base stations use {0}∪ M ={0}∪{1,2,...,m,...,M} represents; Consider that there is a backhaul link from the small base station to the macro base station that can realize relay transmission in this scenario. Multiple MEC servers with powerful computing capabilities are deployed in the small base station and the macro base station, named MEC-SBS and MEC-MBS respectively.

[0097] User collection I ={1,2,...i,...,I} means that each user only publishes one task. I That is, a collection of tasks.

[0098] Each task uses a quad Indicates; where L i (t) represents the task size of user i in time slot t, A i represents the time when user i’s task was published. represents the bandwidth resources required by user i, f i Represents the computing resources required by user i.

[0099] The T time slots of this scenario are represented by the set T ={1,2,...t,...,T} represents the subcarriers of the base station. K ={1,2,...k,...,K} represents.

[0100] Step 2: For time slot t, initialize the access selection of each user and the backhaul transmission strategy from the small base station to the macro base station to form the system decision vector X(t) = [a i,k,0 (t),a i,k,m (t),s i,m,0 (t)].

[0101] element a i,k,0 (t) and a i,k,m (t) represents the binary access selection variable of user i; specifically, a i,k,0 (t) = 1 means that user i accesses the macro base station through subcarrier k in time slot t and offloads its tasks to MEC-MBS, i,k,m (t) = 1 means that user i accesses small base station m through subcarrier k in time slot t and offloads its tasks to the mth MEC-SBS, a i,k,0(t) = 0 or a i,k,m (t)=0 means that user i does not access any base station in time slot t.

[0102] Element i,m,0 (t) indicates that task i on the mth MEC-SBS is offloaded to the MEC-MBS at time slot t.

[0103] Step 3: For task i in time slot t, calculate the transmission rate r of the task accessing the mth small base station through subcarrier k i,k,m (t) and the transmission rate r of the access macro base station i,k,0 (t):

[0104] The calculation formula is as follows:

[0105]

[0106] Where α={α 1 ,...,α k ,...,α K} represents the bandwidth allocation vector of the subcarrier, α k represents the bandwidth allocation vector of subcarrier k; g i,m (t) represents the channel gain between user i and small base station m, which is calculated as g i,m (t) = Ξ·[d(i,m)] -γ , Ξ represents the path loss of 1m, d(i,m) represents the distance between user i and small base station m, and γ represents the path loss exponent. i,k,m (t) represents the interference from other SBSs or MBSs occupying the same subcarrier k, expressed as:

[0107] I i,k,m (t)=∑ j∈ I ,j≠i ∑ u∈ M ∪{0},u≠m a j,k,u (t)p j g j,u (t)

[0108] p i represents the transmission power of user i, n 0 represents the one-sided noise spectral density.

[0109]

[0110] Among them I i,k,0 (t) represents the interference from other SBSs occupying the same subcarrier k; it is expressed as

[0111] I i,k,0(t)=∑ j∈ I ,j≠i ∑ u∈ M ,u≠m a j,k,u (t)p j g j,u (t)

[0112] g i,0 (t) represents the channel gain between user i and the macro base station.

[0113] Step 4: For task i in time slot t, calculate the backhaul transmission rate r of the task from small base station m to macro base station through subcarrier c i,m,0 (t):

[0114]

[0115] Among them, the subcarrier set of the backhaul link is expressed as C ={1,2,...c,...,C},α c represents the bandwidth of subcarrier c. m represents the transmission power of small base station m. m,0 (t) represents the channel gain between the small base station m and the macro base station.

[0116] Step 5: For task i in time slot t, calculate the total execution time T of the task i :

[0117]

[0118] Among them, E {x} If event x occurs, then E {x} =1, otherwise E {x} =0; T represents the number of time slots.

[0119] Step 6: Calculate the average delay cost of the long-term execution time of all tasks

[0120]

[0121] I represents the number of tasks;

[0122] Step 7: Minimize the average delay cost of all users as the optimization goal, jointly optimize the user's access decision and the base station's backhaul transmission strategy, and build an online resource management and control model for task scheduling;

[0123] The online resource control model is expressed as:

[0124]

[0125] Constraint C1 indicates that in one time slot, the same task can only choose to access one base station or perform local calculations.

[0126] Constraint C2 indicates that the bandwidth resources allocated by SBS in a time slot should be subject to the total amount limit, where B SBS Indicates the maximum value of the small base station bandwidth.

[0127] Constraint C3 indicates that the bandwidth resources allocated by MBS in a time slot should be subject to the total amount limit, where B MBS Indicates the maximum value of the macro base station bandwidth.

[0128] Constraint C4 indicates that in a time slot, for any small base station, the sum of all link transmission capabilities cannot exceed its own maximum computing capability. represents the maximum value of the m-th MEC-SBS computing resources, η m represents the computing resources required by the m-th MEC-SBS to process 1 bit of data, θ m (t-1) represents the control parameters of the mth MEC-SBS in the t-1 time slot.

[0129] Constraint C5 indicates that in a time slot, the sum of all link transmission capabilities for a macro base station cannot exceed its own maximum computing capability. represents the maximum value of MEC-MBS computing resources, η 0 represents the computing resources required by MEC-MBS to process 1 bit of data, θ 0 (t-1) represents the control parameters of MEC-MBS in time slot t-1.

[0130] Constraint C6 indicates that the user's access selection and the base station's backhaul transmission strategy are both binary variables.

[0131] Step 8: Use the user online classification algorithm, the MBS-based online resource management algorithm, the SBS-based online resource management algorithm, and the control parameter update algorithm to solve the online resource management model, and finally obtain the user access strategy and the base station backhaul transmission strategy.

[0132] The specific solution process is:

[0133] Step 801: Create a task classification set: a set S of tasks to be executed locally local (t), the set S executed in MEC-SBS SBS (t) and the set S executed in MEC-MBS MBS (t).

[0134] Step 802: Calculate the computational latency of each task executed locally. The computation latency of the m-th MEC-SBS And the computational latency performed in MEC-MBS

[0135] The computation latency for local execution is: C i represents the computational density of task i.

[0136] The computation delay performed at the mth MEC-SBS is:

[0137] The calculation delay in MEC-MBS is:

[0138] Step 803: Using the user online classification algorithm, by comparing the three calculation delays and The size of , divides task i into the corresponding classification set;

[0139] Specifically:

[0140] like The task performs local computation and has a i,k,0 (t) = a i,k,m (t) = s i,m,0 (t) = 0, and divide task i into the set S local (t) in.

[0141] like The task performs MEC-MBS calculations, with a i,k,0 (t) = 1, a i,k,m (t) = 0, s i,m,0 (t) = 0, then divide task i into set S MBS (t) medium;

[0142] like The task performs the mth MEC-SBS calculation, with a i,k,0 (t) = 0, a i,k,m (t) = 1, s i,m,0 (t) = 0, then divide task i into set S SBS (t) in.

[0143] Step 804: Use the MBS-based online resource management algorithm to manage the set S MBS Each task in (t) performs access judgment:

[0144] First, determine whether the remaining bandwidth of the MBS is greater than the bandwidth of the subcarrier k allocated to task i, that is, whether the following conditions are satisfied:

[0145]

[0146] If not, divide this task into set S local (t) and reset a i,k,0 (t) = a i,k,m (t) = s i,m,0 (t) = 0;

[0147] If yes, it is further determined whether the sum of the link transmission capabilities of the macro base station is less than its own maximum computing capability, that is, whether the following conditions are met:

[0148]

[0149] If not, divide this task into set S local (t) and reset a i,k,0 (t) = a i,k,m (t) = s i,m,0 (t) = 0;

[0150] If yes, task i accesses the macro base station through subcarrier k and performs calculation in MEC-MBS, then there is a i,k,0 (t) = 1, a i,k,m (t) = 0, s i,m,0 (t)=0.

[0151] Step 805: Use the SBS-based online resource management algorithm to manage the set S SBS Each task in (t) performs access judgment:

[0152] First, determine whether the remaining bandwidth of the SBS is greater than the bandwidth of the subcarrier k allocated to task i, that is, determine whether:

[0153]

[0154] If not, divide this task into set S local (t) and reset a i,k,0 (t) = a i,k,m (t) = s i,m,0 (t) = 0;

[0155] If yes, it is further determined whether the sum of the transmission capabilities of all links of the small base station is less than its own maximum computing capability, that is, whether the following conditions are met:

[0156]

[0157] If yes, task i accesses the mth small base station through subcarrier k and performs calculation on the corresponding MEC-SBS, then:

[0158] ai,k,0 (t) = 0, a i,k,m (t) = 1, s i,m,0 (t) = 0

[0159] Otherwise, it is further determined whether the sum of the link transmission capabilities of the macro base station is less than its own maximum computing capability, that is, whether the following conditions are satisfied:

[0160] If not, divide this task i into the set S local (t) and reset a i,k,0 (t) = a i,k,m (t) = s i,m,0 (t) = 0. If yes, task i is offloaded to the macro base station via the backhaul link and calculated on the MEC-MBS, then s i,m,0 (t) = 1, a i,k,0 (t) = 1, a i,k,m (t)=0.

[0161] Step 806: Update the control parameter θ using the control parameter update algorithm 0 (t) and θ m (t), realize load balancing between base stations:

[0162] In time slot t, θ 0 (t) is calculated as:

[0163]

[0164] θ m (t) is calculated as:

[0165] Step 807: In each time slot, the unloading decision of each task must be re-judged based on the classification set to which it was initially located, so as to continuously update the classification set of tasks until the optimal base station access strategy for all tasks and the return transmission strategy between base stations are obtained, which are used to guide the access of users in this time slot and the return transmission method of the base station.

[0166] Example:

[0167] In the power Internet of Things scenario, macro base stations and multiple small base stations equipped with edge computing servers provide transmission and computing services to multiple users within their coverage area. Considering the limited bandwidth and computing resources of base stations, users need to decide which subcarrier to occupy in which time slot and which base station to access according to the strategy designed by the algorithm.

[0168] (1) Simulation settings

[0169] In order to verify the performance of the proposed solution, this example uses Matlab for scene construction and numerical simulation. Consider a cellular network with one macro base station and three small base stations, with 50 users randomly distributed, and all users evenly publish tasks from time slot 1 to time slot T.

[0170] (2) Simulation results

[0171] like Figure 2 As shown in the figure, the impact of the number of users on the average processing delay of different algorithm tasks is shown. It can be seen that the average processing delay of tasks in this method increases with the increase in the number of users and gradually reaches a stable state. This is because as the number of users increases, network resources are insufficient to support the execution of tasks. Therefore, tasks requested later will choose local calculations, resulting in higher delays and reaching a stable state. In addition, compared with the other three online schemes, this scheme always obtains the smallest delay, showing the superior performance of the proposed algorithm.

[0172] like Figure 3 As shown in Figure 1, the change of control parameters when the number of users ranges from 100 to 900. It can be seen that θ m (t) is higher than θ 0 (t), this is because macro base stations face greater computing pressure than small base stations. In addition, when the number of IoT users increases, θ m (t) and θ 0 (t) are all reduced because the increase in the number of users will lead to an increase in the number of request processing tasks in each time slot. Therefore, in order to alleviate the computing pressure of the MEC server, θ m (t) and θ 0 (t) are all reduced to control the transmission rate between users and MBS / SBSs.

Claims

1. A method for online resource management and control based on latency perception, characterized in that: The following steps are involved: Step 1: Build a dynamic scenario including a macro base station, M small base stations and I user in the power Internet of Things: Step 2: For time slot t, initialize the access selection of each user and the backhaul transmission strategy from the small base station to the macro base station to form the decision vector X(t) of the system; X(t)=[a i,k,0 (t),a i,k,m (t),s i,m,0 (t)] elementa i,k,0 (t) and a i,k,m (t) represents the binary access selection variable of user i; a i,k,0 (t) = 1 means that user i accesses the macro base station through subcarrier k in time slot t and offloads its tasks to MEC-MBS, i,k,m (t) = 1 means that user i accesses small base station m through subcarrier k in time slot t and offloads its tasks to the mth MEC-SBS, a i,k,0 (t) = 0 or a i,k,m (t) = 0 means that user i does not access any base station in time slot t; element s i,m,0 (t) indicates that task i on the mth MEC-SBS is offloaded to the MEC-MBS at time slot t; Step 3: For task i in time slot t, calculate the transmission rate r of the task accessing the mth small base station through subcarrier k i,k,m (t) and the transmission rate r of the access macro base station i,k,0 (t): Transmission rate i,k,m (t) is calculated as follows: Where α={α1,...,α k ,...,α K } represents the bandwidth allocation vector of the subcarrier, g i,m (t) represents the channel gain between user i and small base station m, I i,k,m (t) represents the interference from other SBSs or MBSs occupying the same subcarrier k, p i represents the transmission power of user i, n0 represents the unilateral noise spectral density; Task i in time slot t accesses the macro base station via subcarrier k with a transmission rate r i,k,0 (t), the calculation formula is as follows: Among them I i,k,0 (t) represents the interference from other SBSs occupying the same subcarrier k; g i,0 (t) represents the channel gain between user i and the macro base station; Step 4: For task i in time slot t, calculate the backhaul transmission rate r of the task from small base station m to macro base station through subcarrier c i,m,0 (t); Backhaul transmission rate r i,m,0 (t): The subcarrier set of the backhaul link is represented by C = {1, 2, ... c, ..., C}, α c represents the bandwidth of subcarrier c; p m represents the transmission power of small base station m; g m,0 (t) represents the channel gain between the small base station m and the macro base station; Step 5: For task i in time slot t, calculate the total execution time T of the task i : Among them, E {x} If event x occurs, then E {x} =1, otherwise E {x} =0; T represents the number of time slots; A i represents the task release time of user i; L i (t) represents the task size of user i in time slot t; Step 6: Calculate the average delay cost of the long-term execution time of all tasks I represents the number of tasks; Step 7: Taking the minimization of average delay cost as the optimization goal, jointly optimize the user's access decision and the base station's backhaul transmission strategy, and build an online resource management and control model for task scheduling; Step 8: Use the user online classification algorithm, the MBS-based online resource management algorithm, the SBS-based online resource management algorithm, and the control parameter update algorithm to solve the online resource management model, and finally obtain the user access strategy and the base station backhaul transmission strategy.

2. The online resource management and control method based on latency perception according to claim 1, characterized in that: In step 1, the macro base station and M small base stations use {0}∪ M ={0}∪{1,2,...,m,...,M} represents; multiple MEC servers are deployed in small base stations and macro base stations, named MEC-SBS and MEC-MBS respectively; User collection I ={1,2,...i,...,I} means that each user only publishes one task. I That is, a set of tasks; Each task uses a quad Indicates; among them, represents the bandwidth resources required by user i; f i represents the computing resources required by user i; The T time slots of this scenario are represented by the set T ={1,2,...t,...,T} represents the subcarriers of the base station. K ={1,2,...k,...,K} represents.

3. The online resource management and control method based on delay perception according to claim 1, characterized in that: In step 7, the online resource management and control model is expressed as: Constraint C1 indicates that in a time slot, the same task can only choose to access one base station or be calculated locally; M indicates the number of small base stations; K indicates the subcarrier set of the base station; I indicates the user set; Constraint C2 indicates that the bandwidth resources allocated by SBS in a time slot should be subject to the total amount limit, where α k represents the bandwidth allocation vector of subcarrier k; B SBS Indicates the maximum value of the small base station bandwidth; Constraint C3 indicates that the bandwidth resources allocated by MBS in a time slot should be subject to the total amount limit, where B MBS Indicates the maximum value of the macro base station bandwidth; Constraint C4 means that in a time slot, for any small base station, the sum of all link transmission capabilities cannot exceed its own maximum computing capability; represents the maximum value of the m-th MEC-SBS computing resources, η m represents the computing resources required by the m-th MEC-SBS to process 1 bit of data, θ m (t-1) represents the control parameters of the mth MEC-SBS in time slot t-1; Constraint C5 indicates that in a time slot, the sum of all link transmission capabilities for the macro base station cannot exceed its own maximum computing capability; represents the maximum value of MEC-MBS computing resources, η 0 represents the computing resources required by MEC-MBS to process 1 bit of data, θ0(t-1) represents the control parameters of MEC-MBS in time slot t-1; Constraint C6 indicates that the user's access selection and the base station's backhaul transmission strategy are both binary variables; T indicates a time slot set.

4. The online resource management and control method based on latency perception according to claim 1, characterized in that: The specific solution process of step eight is: Step 801: Create a task classification set: a set S of tasks to be executed locally local (t), the set S executed in MEC-SBS SBS (t) and the set S executed in MEC-MBS MBS (t); Step 802: Calculate the computational latency of each task executed locally. The computation latency of the m-th MEC-SBS And the computational latency performed in MEC-MBS The computation latency for local execution is: C i represents the computational density of task i; The computation delay performed at the mth MEC-SBS is: The calculation delay in MEC-MBS is: Step 803: Using the user online classification algorithm, by comparing the three calculation delays and The size of , divides task i into the corresponding classification set; Specifically: like The task performs local computation and has a i,k,0 (t) = a i,k,m (t) = s i,m,0 (t) = 0, and divide task i into the set S local (t) medium; like The task performs MEC-MBS calculations, with a i,k,0 (t) = 1, a i,k,m (t) = 0, s i,m,0 (t) = 0, then divide task i into set S MBS (t) medium; like The task performs the mth MEC-SBS calculation, with a i,k,0 (t) = 0, a i,k,m (t) = 1, s i,m,0 (t) = 0, then divide task i into set S SBS (t) medium; Step 804: Use the MBS-based online resource management algorithm to manage the set S MBS Each task in (t) performs access judgment: First, determine whether the remaining bandwidth of the MBS is greater than the bandwidth of the subcarrier k allocated to task i, that is, whether the following conditions are satisfied: If not, divide this task into set S local (t) and reset a i,k,0 (t) = a i,k,m (t) = s i,m,0 (t) = 0; if yes, further determine whether the sum of all link transmission capabilities of the macro base station is less than its own maximum computing capability, that is, determine whether: If not, divide this task into set S local (t) and reset a i,k,0 (t) = a i,k,m (t) = s i,m,0 (t) = 0. If yes, task i accesses the macro base station through subcarrier k and performs calculations in MEC-MBS, then there is a i,k,0 (t) = 1, a i,k,m (t) = 0, s i,m,0 (t) = 0; Step 805: Use the SBS-based online resource management algorithm to manage the set S SBS Each task in (t) performs access judgment: First, determine whether the remaining bandwidth of the SBS is greater than the bandwidth of the subcarrier k allocated to task i, that is, determine whether: If not, divide this task into set S local (t) and reset a i,k,0 (t) = a i,k,m (t) = s i,m,0 (t) = 0. If yes, further determine whether the sum of the link transmission capabilities of all small base stations is less than its own maximum computing capability, that is, determine whether: If yes, task i accesses the mth small base station through subcarrier k and performs calculation on the corresponding MEC-SBS, then: a i,k,0 (t)=0,a i,k,m (t)=1,s i,m,0 (t)=0 Otherwise, it is further determined whether the sum of the link transmission capabilities of the macro base station is less than its own maximum computing capability, that is, whether the following conditions are satisfied: If not, divide this task i into the set S local (t) and reset a i,k,0 (t) = a i,k,m (t) = s i,m,0 (t) = 0. If yes, task i is offloaded to the macro base station via the backhaul link and calculated on the MEC-MBS, then s i,m,0 (t) = 1, a i,k,0 (t) = 1, a i,k,m (t) = 0; Step 806: Use the control parameter update algorithm to update the control parameters θ0(t) and θ m (t), realize load balancing between base stations: At time slot t, θ0(t) is calculated as: θ m (t) is calculated as: Step 807: In each time slot, the unloading decision of each task must be re-judged based on the classification set to which it was initially located, so as to continuously update the classification set of tasks until the optimal base station access strategy for all tasks and the return transmission strategy between base stations are obtained, which are used to guide the access of users in this time slot and the return transmission method of the base station.

Images