Vehicle fleet base station heterogeneous edge computing resource allocation method and system based on vehicle networking

Abstract

The application discloses a kind of based on vehicle networking's fleet base station heterogeneous edge computing resource allocation method and system, to improve the collaborative utilization efficiency of heterogeneous edge resource.First, the definition of maximum benefit resource allocation problem and maximum benefit resource allocation problem is introduced, and based on this, the optimal semi-closed solution of resource allocation under heterogeneous edge for multiple users is derived;In order to further realize the collaborative task offloading and resource allocation of multiple users, the maximum benefit resource allocation problem for single user is constructed, the optimal solution is derived, which is defined as benefit resource, and it is proved that the problem and the equivalence condition of the original problem;Based on the above proof, a collaborative task offloading and resource allocation method suitable for heterogeneous edge is proposed, which realizes the benefit resource allocation of single user approximating the optimal resource allocation of multiple users in the original problem.The experimental results show that the application can adapt to high dynamic, delay-sensitive and large computing task demand vehicle networking scene, and meet the diversified computing needs of users.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of edge computing resource allocation of Internet of Vehicles, and relates to a vehicle-vehicle team-base station heterogeneous edge computing resource cooperative allocation method and system, in particular to a vehicle-vehicle team-base station heterogeneous edge computing resource cooperative allocation method and system based on Internet of Vehicles. BACKGROUND

[0002] With the rapid development of the fifth generation, the sixth generation of communication technology and the Internet of Things technology, vehicles equipped with vehicle-mounted systems have the ability of communication, storage and calculation, which brings new opportunities and challenges to the prospect of Internet of Vehicles. As an important part of intelligent transportation systems, the vigorous development of Internet of Vehicles brings a large number of new applications, such as automatic driving, road virtual reality, online immersive games and roadside information collection and processing. Based on the characteristics of rapid driving of vehicles and the goal of road safety, the service providing ability of calculation-intensive and time-delay-sensitive becomes one of the factors restricting the development of Internet of Vehicles. Although the traditional cloud computing has strong computing power, due to the large propagation delay caused by its long transmission distance and the instability of connection caused by complex backhaul network, it cannot meet the needs of time-delay-sensitive applications. Edge computing uses the computing resources of base stations and roadside nodes to reduce data transmission distance and reduce computing delay. However, due to the still large coverage range (>300m) of base stations and roadside nodes, the transmission delay cannot meet the needs of some time-delay-sensitive tasks. And when the number of computing tasks increases, the limited edge computing capacity cannot meet all service needs in time, resulting in an increase in waiting computing delay. In addition, the high-speed driving of vehicles will cause frequent switching of edge servers, increase the completion delay of connected edge computing tasks, and reduce the task completion rate. A vehicle team is composed of multiple intelligent vehicles that drive synchronously in a short distance and have certain computing capacity, which realizes the centralized distribution of the remaining resources of vehicles and has stable and strong computing capacity in a period of time. Moreover, the vehicle team and the surrounding vehicles have the characteristics of relatively synchronous high-speed driving, which is conducive to realizing short-distance, long-connection and high-efficiency transmission, thereby reducing the transmission delay and the switching frequency of edge servers. Therefore, the cooperative multi-layer edge resources of vehicle team and base station can more stably, quickly and continuously provide computing services to surrounding vehicles.

[0003] For the service demand of computing-intensive and time-sensitive in vehicle networking, the existing researches are based on the computing task offloading path, offloading object and offloading proportion respectively. The computing task is dispersed to the surrounding vehicles, vehicle platoon, roadside node and base station for calculation through multi-hop transmission between vehicles, multi-hop transmission between roadside nodes and optimization of task segmentation proportion. However, multi-hop transmission between vehicles and roadside nodes will bring problems such as increase of propagation delay, transmission instability and the like; the uncertainty of service-providing vehicles and the high-speed driving of vehicles increase the complexity of the backhaul path and increase the backhaul failure rate; in addition, due to the limited computing capacity of individual vehicles, how to efficiently allocate the remaining resources of the vehicles also faces severe challenges. For the research on taking the vehicle platoon as an edge service node, due to the characteristics of simplifying the number of vehicle platoons, dynamicity, resource providing difference and multiple task options within the service range, the vehicle platoon does not play its maximum role in the edge computing scene. In order to solve the above problems, the present application proposes a multi-layer computing resource collaborative allocation method of user-vehicle platoon-base station in the scene of multiple vehicle platoons with random distribution and different services. SUMMARY

[0004] In view of the deficiencies of the prior art, the present application proposes a vehicle platoon-base station heterogeneous edge computing resource collaborative allocation method based on vehicle networking. By coordinating the dynamic and differentiated computing resources of multiple vehicle platoons, the allocation strategy between multiple vehicle users and vehicle platoons with different computing capabilities is optimized, and by optimizing the task segmentation proportion of different vehicle platoons and base stations, the differentiated resource collaboration of multiple vehicle platoons and base stations is further realized.

[0005] The present application provides a vehicle platoon-base station heterogeneous edge computing resource allocation method based on vehicle networking. First, the definitions of the maximum benefit resource allocation problem and the maximum benefit resource allocation problem are introduced, and based on this, the optimal semi-closed solution of the resource allocation problem facing multiple users under the heterogeneous edge is derived; then the maximum benefit resource allocation problem facing a single user is constructed, and its optimal solution is derived, which is defined as the benefit resource, and it is proved that the problem and the equivalent condition of the original problem; based on the above proof, a collaborative task offloading and resource allocation method suitable for the heterogeneous edge is proposed, which realizes the benefit resource allocation of a single user approaching the optimal resource allocation of multiple users in the original problem.

[0006] The technical scheme adopted by the method of the present application is: a vehicle platoon-base station heterogeneous edge computing resource allocation method based on vehicle networking, comprising the following steps:

[0007] Step 1: build a collaborative computing mechanism of multiple vehicle users in a heterogeneous edge network of multiple vehicle platoons and the current base station, and establish a user-multiple vehicle platoon-base station heterogeneous resource allocation benefit problem;

[0008] The vehicle user sends a computing task to any one of the vehicle fleets within the service range, and the vehicle fleet, after receiving the computing task differentiated by the delay requirements of different vehicle users, implements joint and cooperative edge computing with the base station by sending part of the computing task to the current base station.

[0009] Step 2: define the maximum benefit resource allocation problem, and calculate the maximum benefit of all vehicle fleets and derive the optimal number of serviceable users for each vehicle fleet

[0010] Step 3: based on step 2, define the resource allocation problem at the optimal number of users as a maximum benefit resource allocation problem; and The expression is brought into the solution to obtain the optimal semi-closed solution of the maximum benefit resource allocation of multiple users.

[0011] Step 4: according to the maximum benefit resource allocation problem in step 3, calculate the maximum benefit computing resource allocation value for a single user, defined as the optimal benefit resource of a single user; and prove that the optimal benefit resource of a single user is equivalent to the maximum benefit resource allocation in step 3.

[0012] Step 5: accumulate the remaining computing resources of all vehicle fleets, denoted as total remaining resources of vehicle fleets; and calculate the upper bound benefit under the current remaining resources according to the remaining computing resources and the number of unallocated tasks.

[0013] Step 6: according to the driving speed and position of the vehicle fleet and the user and the maximum time delay of the user computing task, calculate the range of users that can be served by each vehicle fleet within the guaranteed communication connection time; based on the optimal benefit resource of a single user obtained in step 4, sort the computing tasks generated by the users within the serviceable range according to the ratio of their completion benefit to allocated resources from large to small.

[0014] Step 7: complete the task pre-allocation of users-vehicle fleets through a three-time task allocation strategy.

[0015] Step 8: according to the pre-allocation results of users-vehicle fleets and the optimal benefit resource of a single user, calculate the total amount of computing resources required by each vehicle fleet; when the available computing resources of the vehicle fleet meet the required resource amount, the computing task on the vehicle fleet is completed at a ratio of 1; when the required resource amount is greater than the available computing resource amount of the vehicle fleet, introduce a new variable p as the task segmentation ratio, i.e. p proportion of the computing task is calculated on the vehicle fleet, and (1-p) proportion of the computing task is calculated on the base station, and update the user-multi-vehicle fleet-base station heterogeneous resource allocation benefit problem in step 1.

[0016] Step 9: The updated user-multi-vehicle team-base station heterogeneous resource allocation benefit problem is divided into two sub-problems, the vehicle team resource allocation value and the base station resource allocation value under the current task allocation strategy are solved, the task segmentation ratio is calculated, and the maximum benefit at this time is calculated;

[0017] Step 10: Compare the benefit at this time with the historical best benefit, when the benefit at this time is greater than the historical best benefit, update the best benefit value and the corresponding task allocation strategy, vehicle team resource allocation value, base station resource allocation value and task segmentation ratio; otherwise, according to the current task allocation strategy, the calculation task allocation strategy obtained by backtracking in step 11 is updated in the order of total vehicle team resource amount from small to large and in the unit of single calculation task allocation, and the upper limit benefit of the current allocation strategy is calculated according to step 5, when the upper limit benefit is less than the historical best benefit, the task allocation is ended;

[0018] Step 11: According to the current task allocation strategy and the corresponding vehicle team resource allocation amount and base station resource allocation amount, the backtracking is performed on the allocated calculation tasks in the order of total vehicle team resource amount from large to small and task allocation from back to front, and the upper limit benefit of the remaining allocation strategy is calculated according to step 5, when the upper limit benefit is greater than the historical best benefit, return to step 7; otherwise, the best user-multi-vehicle team-base station multi-layer computing resource collaborative allocation result is obtained.

[0019] The technical scheme adopted by the system of the application is: a vehicle team base station heterogeneous edge computing resource allocation system based on vehicle networking, comprising:

[0020] One or more processors;

[0021] A storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the vehicle team base station heterogeneous edge computing resource allocation method based on vehicle networking.

[0022] Compared with the prior art, the application has the following advantages:

[0023] 1. The application models the user-vehicle team-base station heterogeneous edge computing scene, optimizes the user-multi-vehicle team task allocation strategy, vehicle team computing resource allocation, base station computing resource allocation and task segmentation ratio in the vehicle team and the base station, reduces the completion time delay of the calculation task under the premise of ensuring the completion rate of all calculation tasks, and realizes the maximization of the overall resource utilization benefit.

[0024] 2. The application solves the maximum benefit resource allocation function according to the task completion benefit model of the heterogeneous edge computing, and proves that when the number of calculation tasks is sufficient, the user-multi-vehicle team task allocation strategy based on the maximum benefit resource allocation function can maximize the resource use efficiency of the multi-vehicle team.

[0025] 3.The application provides a multi-user-multi-vehicle fleet-base station heterogeneous computing resource cooperative allocation method under a heterogeneous edge network, based on a maximum benefit resource allocation function, to realize dynamic allocation of computing tasks and multi-vehicle fleet and base station resources, improve multi-layer computing resource utilization efficiency of the vehicle fleet and base station, ensure maximization of overall benefits, and effectively improve the continuity of cross-base station computing task services. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1 is a vehicle fleet-base station heterogeneous edge network structure diagram for multiple vehicle users according to an embodiment of the application.

[0027] Figure 2 is a cooperative computing schematic diagram of user tasks under a heterogeneous edge network according to an embodiment of the application.

[0028] Figure 3 is a multi-vehicle fleet cooperative resource allocation computing schematic diagram based on backtracking and dynamic programming according to an embodiment of the application.

[0029] Figure 4 is a heterogeneous edge network cooperative task offloading and resource allocation schematic diagram according to an embodiment of the application. DETAILED DESCRIPTION

[0030] In order to facilitate those skilled in the art to understand and implement the present application, the present application will be further described in detail below in conjunction with the drawings and embodiments. It should be understood that the embodiments described herein are only used to illustrate and explain the present application, and are not used to limit the present application.

[0031] The application provides a vehicle user-vehicle team-base station-based heterogeneous edge collaborative computing resource allocation method for vehicle networking edge computing, aiming to maximize the collaborative utilization efficiency of heterogeneous edge resources, ensure the joint allocation of computing resources among multiple vehicle teams, and complete the collaborative service of multiple vehicle teams and the current base station, so as to realize high-quality service of vehicle user computing tasks, reduce the computing task completion delay, and improve the task completion reliability. Specifically, a vehicle team-base station task allocation and collaborative resource offloading method for heterogeneous edge computing resource allocation is provided, wherein the definitions of the maximum benefit resource allocation problem and the maximum benefit resource allocation problem are first introduced, and the optimal semi-closed solution of the original problem is derived based on the definitions; in order to further realize the collaborative task offloading and resource allocation of multiple users, based on the definition of the maximum benefit resource allocation problem, a low-complexity resource allocation problem for a single vehicle user is constructed, and its optimal solution is obtained by derivation and calculation, which is defined as the benefit resource, and it is proved that the problem and the original problem have equivalent conditions; based on the above derivation and proof, a collaborative task offloading and resource allocation method suitable for heterogeneous edge is proposed, which realizes the approximation of the optimal resource allocation for multiple users in the original problem through the benefit resource allocation of a single user. First, based on the current unallocated computing task and the task-resource collaborative backtracking result, the benefit resource obtained by calculation is used for task offloading, and the association variables between the current multiple users and multiple vehicle teams are updated; secondly, the maximum benefit resource allocation problem based on the task segmentation ratio is constructed, and the optimal solution is calculated by using the Lagrange dual; then, based on the optimal vehicle team-base station allocation resource in the current iteration, the task-resource collaborative backtracking under the heterogeneous edge is expanded, and the upper bound value is calculated based on the current backtracking result, and the backtracking is stopped when the upper bound value is greater than the historical optimal solution; finally, the historical optimal solution is updated by dynamic programming based on the maximum benefit resource allocation problem under the heterogeneous edge, and is used to judge whether the maximum system benefit is obtained in the backtracking iteration, and when there is no backtracking target, the system optimal collaborative resource allocation result is calculated. The optimal solution search is carried out based on the theoretical derivation of the maximum benefit resource allocation problem and the maximum benefit resource allocation problem. Since the maximum benefit resource allocation problem for a single user can be equivalent to the original problem under certain conditions, the solution space of the original problem is reduced, and the system delay is reduced. Thanks to the collaborative task offloading among multiple vehicle teams, it can adapt to the high-dynamic, time-delay-sensitive and large-computing-task-demand vehicle networking scene; the collaborative service of multiple vehicle teams and the base station can make up for the lack of vehicle computing resources, provide sufficient computing resources for users in time and effectively, and meet the diversified computing needs of users.

[0032] See Figure 1 The vehicle team-base station heterogeneous edge computing resource allocation method provided by the application comprises the following steps:

[0033] Step 1: Constructing a collaborative computing mechanism for multiple vehicle users in a heterogeneous edge network of multi-vehicle fleets and current base stations, establishing a user-multi-vehicle fleet-base station heterogeneous resource allocation benefit problem;

[0034] First, the vehicle user sends the computing task to any one of the vehicle fleets within the service range, and after receiving the delay requirement differentiated computing tasks of different vehicle users, the vehicle fleet sends part of the computing task to the current base station to realize the joint collaborative edge computing of multi-vehicle fleet and base station through task segmentation.

[0035] The computing capacity of different vehicle fleets is represented as f = mf0, where f0 represents the minimum unit computing resource value that can be segmented within the vehicle fleet, and m represents the number of minimum unit resources in different vehicle fleets; based on this computing model, the computing completion time t i pc can be expressed as:

[0036]

[0037] where N p represents the number of vehicle fleets, x j,i represents the allocation association variable of user task i and vehicle fleet j, x j,i = 1 when the user task is calculated in vehicle fleet j, otherwise x j,i = 0; p j,i represents the computing proportion of user task i in vehicle fleet j; c i represents the total amount of computing resources required by user task i. Similarly, the computing time of the remaining task proportion of this user task i in the base station is

[0038]

[0039] where f b,i represents the computing resource value allocated by the current base station to the task i.

[0040] Based on the Rayleigh fading model, the communication channel gain between the vehicle fleet and the user and the base station is expressed as: where G represents the determined power gain related to the antenna and amplifier parameters, h η ~ CN(0, 1) represents Rayleigh fading, which conforms to complex Gaussian distribution, represents the Euclidean distance between vehicle i and vehicle j, where x i , y i , x j , y j represent the horizontal and vertical coordinate values of vehicle i and vehicle j, respectively; a represents the path attenuation factor; then the data transmission time of the vehicle user and the vehicle fleet and the base station is:

[0041]

[0042]

[0043] where s i denotes the size of user task i; and denote the sending rate of user vehicle i to the platoon and the sending rate of platoon j to the base station, respectively, which are expressed as where denotes the bandwidth resource allocated according to the platoon length, B denotes the total bandwidth resource, n j denotes the number of vehicles constituting platoon j, N u denotes the number of vehicle users, N denotes the power spectral density of Gaussian white noise, p i denotes the sending power of vehicle user i.

[0044] A task completion benefit model of heterogeneous edge computing is designed based on the benefits brought by the reduction of user task completion delay and the resource allocation consumption of the platoon and the base station:

[0045]

[0046] where τ i denotes the maximum allowed delay of user task i, denotes the actual completion delay of user task i, (τ i -t i ) denotes the reduced task completion delay of user task i; f j,i ,f b,i denote the computing resource values allocated by platoon j and the base station to user task i, respectively; denotes the user benefit brought by unit shortened delay, denote the fees of platoon and base station computing resource allocation, respectively; μ denotes the penalty factor of energy consumption, denotes the energy consumption of the platoon, where p n denotes the data sending power of the platoon;

[0047] and a vehicle user-multi-platoon-base station heterogeneous resource allocation benefit problem is established:

[0048]

[0049] s.t.

[0050] t i ≤τ i ,i∈{1,2,...,N e},

[0051]

[0052]

[0053]

[0054] x j,i ∈{0,1},i∈{1,2,...,N e},j∈{1,2,...,N p},

[0055] 0≤ρ j,i ≤1,i∈{1,2,...,N e},j∈{1,2,...,N p},

[0056] where χ={x j,i} represents the set of association variables between vehicle users and vehicle fleets, F b ,F p represent the set of resource allocation of base station and vehicle fleet j to different users, respectively, and ρ represents the set of user task partitioning ratio; F b ,F j represent the total computing resources of base station and vehicle fleet j, respectively; N e represents the total number of tasks generated by N u vehicle users at this time;

[0057] Maximize the revenue, i.e., improve the utilization of computing resources, under the premise of ensuring the task completion delay.

[0058] Step 2: Calculate the maximum revenue of all vehicle fleets and derive the optimal number of serviceable users of each vehicle fleet

[0059] In order to solve the problem in step 1, the maximum revenue resource allocation problem is defined as follows:

[0060]

[0061] s.t.

[0062] t i ≤τ i ,i∈{1,2,...,N e},

[0063] M j ≤N e ,

[0064]

[0065] where Σ represents the sum of the number of users allocated to vehicle fleet j;

[0066] By proving that the problem is a single-peak problem, we obtain the closed-form solution for the optimal number of service users at the peak of the problem, that is, the number of users assigned to fleet j reaches When , the maximum benefit of fleet j can be obtained

[0067] in:

[0068]

[0069]

[0070] This embodiment defines two problems. The first problem is the maximum revenue problem, and the second problem is the maximum benefit problem. The second problem is defined based on the solution of the first problem.

[0071] Step 3: Based on step 2, define the resource allocation problem at the optimal number of users as the maximum benefit resource allocation problem; Substituting this expression into the original problem, we obtain a semi-closed-form solution for the maximum resource allocation per user, which is the optimal solution to the original problem. This optimal solution is derived based on the established associations between users and fleets. However, the allocation variables for multiple users and fleets are multiply coupled, making the problem intractable. Therefore, the original problem is transformed into how to gradually approach this optimal solution.

[0072] Please refer to Figure 2 To solve the original problem, the maximum benefit resource allocation problem is defined in step 2, and the maximum benefit resource allocation problem is defined at the maximum value of the maximum benefit allocation problem. The maximum benefit resource allocation problem is to maximize the total benefit of the multi-fleet system by adjusting the task offloading mechanism and the resource allocation mechanism within each fleet.

[0073] Based on the optimal solution to the maximum benefit problem, the maximum benefit resource allocation problem is defined as: maximizing the efficiency of multi-fleet computing resource allocation, that is, achieving the optimal resource allocation problem for all associated users in each fleet when the overall benefit is maximized, which is the optimal solution to the original problem. Through theoretical derivation, the optimal solution can be obtained as:

[0074]

[0075] Among them, among them, represents the maximum amount of beneficial resources allocated by fleet j to user task i, which is the optimal solution to the original problem; c p 、s p , τ p 、 denote the required computing resources, task size, maximum allowed delay, and data transmission rate of user task p assigned to fleet j, respectively.

[0076] Step 4: Construct a low-complexity problem for single-user resource allocation according to the definition of the maximum benefit resource allocation problem in Step 3; solve its optimal resource allocation value, defined as the optimal benefit resource of a single user;

[0077] Since the closed-form solution in Step 3 is is a variable related to M j However, M j is a coupling variable between multiple vehicle fleets, which is difficult to solve directly. Therefore, according to the definition of the maximum benefit resource allocation problem in Step 3, a low-complexity problem for single-user resource allocation is constructed:

[0078]

[0079] t i ≤τ i ,i∈{1,2,...,N e},

[0080] f j,i ≤F j ,

[0081] Solve the single-user optimal benefit resource

[0082]

[0083] It is proved that when the task size, required computing resource amount, maximum allowed time delay, and transmission rate allocated to vehicle fleet j satisfy the following relationship

[0084]

[0085] The single-user optimal benefit resource is equivalent to the maximum benefit resource Therefore, the subsequent goal is to approximate the optimal solution of the multi-user resource allocation in the original problem by the single-user optimal benefit resource Step 6: Based on the equivalence relationship, the proposed heterogeneous edge computing task-resource collaborative algorithm (Steps 7-11) is used to approximate the optimal solution of the original problem from the benefit resource.

[0086] Step 5: Accumulate the remaining computing resources of all vehicle fleets, denoted as the total remaining resources of vehicle fleets;

[0087] Solve the total benefit of the remaining resources of vehicle fleets and base stations

[0088]

[0089] where ζ i represents the currently allocated user task; t'​i , f i , f b，i , p p i respectively represent the task completion time of task i based on the total remaining resources of the vehicle fleet, the vehicle fleet resource allocation, the base station resource allocation, the user data transmission power, the task partitioning ratio and the maximum transmission rate of the vehicle fleet base station.

[0090] Let be the upper bound of the revenue value at the i-th iteration, where represents the user-vehicle fleet offloading allocation at the (i-1)-th iteration, represents the maximum revenue under the user-vehicle fleet allocation.

[0091] Step 6: According to the driving speed and position of the vehicle fleet and the user and the maximum time delay of the user computing task, the range of users that each vehicle fleet can serve within the guaranteed communication connection time is calculated; the computing tasks of the users within the serviceable range are sorted in descending order according to the ratio of the completion revenue to the allocated resources based on the single-user optimal benefit resource solved in step 4;

[0092] Step 7: Through a three-time task allocation strategy, the user-vehicle fleet task pre-allocation is completed.

[0093] Please refer to Figure 3 and Figure 4 , in order to realize the approximation of the single-user benefit resource to the optimal solution of the original problem, the present application proposes a heterogeneous edge computing task-resource collaborative algorithm, which completes the user-vehicle fleet task allocation through multiple iterations of a three-time task allocation strategy. First, according to the order of the serviceable resource amount of the vehicle fleet from small to large, the computing tasks are allocated to the corresponding vehicle fleet in order according to the sorting in step 6, and the remaining computing resources of the vehicle fleet are updated according to the single-user optimal benefit resource of the corresponding computing task. When all vehicle fleets cannot meet the demand of the single-user optimal benefit resource, the initial allocation of the computing task is completed; then, the remaining unallocated computing tasks are classified according to the number of serviceable vehicle fleets, and the allocation strategy of the computing task with a single serviceable vehicle fleet is fixed to complete the secondary allocation, so as to ensure the total completion rate of the task under the current scene; finally, the remaining resources of all vehicle fleets are updated according to the minimum resource amount required to complete the computing task within the maximum time delay, the serviceable vehicle fleet list is dynamically updated according to the remaining resource amount of the vehicle fleet from large to small, and all the still unallocated tasks are allocated to the user-vehicle fleet offloading according to the dynamic vehicle fleet list in a polling manner.

[0094] Step 8: Calculate the total amount of computing resources each team needs to provide according to the pre-allocation results of user-teams and the single-user optimal benefit resources of the computing task. When the available computing resources of the team meet the required resource amount, the computing task on the team is completed with a team completion ratio of 1. When the required resource amount is greater than the available computing resource amount of the team, a new variable p is introduced as the task segmentation ratio, i.e., p proportion of the computing task is calculated in the team, and (1-p) proportion of the computing task is calculated in the base station. The user-multiple team-base station heterogeneous resource allocation benefit problem in step 1 is updated;

[0095] Step 9: Split the updated user-multiple team-base station heterogeneous resource allocation benefit problem into two sub-problems, and each sub-problem is a convex function. The team resource allocation value and the base station resource allocation value under the current task allocation strategy are solved by convex optimization, the task segmentation ratio is calculated, and the best benefit at this time is calculated.

[0096] The two sub-problems are: (1) user-base station resource allocation problem and (2) user-multiple team resource allocation problem. Since the second problem is relatively simple, it can be directly solved by CVX. The solution method of the first problem is explained in detail below.

[0097] First, construct a maximum benefit resource allocation problem based on the task segmentation ratio, and get its optimal solution as:

[0098]

[0099] Where, p j,i represents the task segmentation ratio of user task i served by team j, M j represents the total number of users allocated to the current team j, c p , τ p , s p represent the required computing resources, maximum allowed time delay, and task size of task p allocated to the team, respectively.

[0100] Substitute into sub-problem 1. According to the pre-allocation results of user-teams and the single-user optimal benefit resources, calculate the total amount of computing resources each team needs to provide. When the available computing resources of the team meet the required resource amount, the computing task on the team is completed with a team completion ratio of 1. When the required resource amount is greater than the available computing resource amount of the team, a new variable p is introduced as the task segmentation ratio, i.e., p proportion of the computing task is calculated in the team, and (1-p) proportion of the computing task is calculated in the base station. The user-multiple team-base station heterogeneous resource allocation benefit problem in problem 1 is updated. The Hessian matrix of sub-problem 1 at this time can be obtained as:

[0101]

[0102] Where Since and H j,i (1,1)≤0, H j,i (2,2)≤0, the problem is proved to be convex. Introducing Lagrange variables θ, κ, ω, υ, the original problem is transformed into

[0103]

[0104] Solving the problem by Lagrange dual function, the relationship between f b,i and ρ j,i is obtained:

[0105]

[0106] wherein Thus, the optimal solution F b , F p and the task partition ratio ρ of the vehicle user-vehicle team-base station heterogeneous edge resource allocation in this iteration are obtained.

[0107] Step 10: Compare the current revenue with the historical best revenue. When the current revenue is greater than the historical best revenue, update the best revenue value and the corresponding task allocation strategy, vehicle team resource allocation value, base station resource allocation value and task partition ratio. Otherwise, according to the current task allocation strategy, update the calculation task allocation strategy obtained in step 11 in the order of the total resource amount of the vehicle team from small to large, and calculate the upper bound revenue of the current allocation strategy according to step 5. When the upper bound revenue is less than the historical best revenue, end the task allocation.

[0108] Based on the offloading variable χ of the current vehicle user-vehicle team and the optimal solution F b and F p , backtrack the heterogeneous task-resource, that is, initialize the already allocated calculation tasks in reverse order from back to front according to the order of the vehicle team, and update the corresponding vehicle team remaining calculation resources. Initialize one calculation task each time, and calculate the upper bound value of the problem at this moment:

[0109]

[0110] wherein represents the offloading association result of the vehicle user task and the vehicle team after the (i-1)th iteration in step 7, is the revenue based on in step 9, represents the remaining revenue calculated according to the current remaining vehicle team total resources in the ith iteration:

[0111]

[0112] st

[0113] t i ≤τ i ,i∈{1,2,...,N e},

[0114]

[0115]

[0116] Among them, F' p ,F' b ,f' i ,f' b,i They represent the allocation set of the remaining fleet resources, the allocation set of the remaining base station resources, the computing resources allocated to user task i from the remaining fleet resources, and the computing resources allocated to user task i from the remaining base station resources; and Represent the remaining computing resources of fleet j and base station after the iteration update. The solution of this problem is used as the upper bound of the system benefit of the original problem. Continue heterogeneous task-resource backtracking until

[0117] Step 11: Based on the current task allocation strategy and the corresponding fleet resource allocation and base station resource allocation, the already allocated computing tasks are backtracked in descending order of fleet total resources and from back to front. The upper bound of the remaining allocation strategy is calculated according to step 5. If the upper bound is greater than the historical best benefit, return to step 7. Otherwise, the optimal user-multi-fleet-base station multi-layer computing resource collaborative allocation result is obtained.

[0118] In step 10 It represents the historical optimal return of the system under the current iteration, and its expression is:

[0119]

[0120] in Indicates that the computing task is in the unassigned state after backtracking.

[0121] ζ i ={e u |x p,u =1∈χ,u∈N e ,p∈N p} represents the user tasks assigned to the fleet in the i-th iteration.

[0122] represents the historical optimal solution of the problem in step 9, that is, the maximum benefit of the system before the i-1th iteration, denotes the output of the backtracking algorithm at the i-1th iteration, Δζ i-1 (e u ) denotes the updated output of the task-fleet offloading association in step 7 based on the current task association. Step 10 is continued until no larger output result.

[0123] The following further illustrates the present application by further proving.

[0124] Proof 1: Proof process of step 2

[0125] Based on the maximum revenue resource allocation problem in step 2, when the number of users M j allocated by the fleet j is less than the total number of users, the original problem becomes

[0126]

[0127] By calculating its derivative equal to 0, i.e. calculating and The relationship between the computing resources f j,i allocated by the fleet j to the user task i, the number of users M j allocated by the fleet j, and the total computing resources F j of the fleet j can be obtained:

[0128]

[0129] Bring the expression of f j,i back to the original problem, and the system revenue can be expressed as:

[0130]

[0131] Since denotes the total number of user tasks offloaded to the fleet j, then and can be expressed as the average attribute of the user tasks allocated to the fleet j multiplied by the number of tasks, i.e. expressed as E1(M j )m and E2(M j )m respectively, where m is the number of users, and E1(M j ) and E2(M j ) represent the average attribute of the offloaded users, whose expressions are and Bring E1(M j )m and E2(M j )m back to the original expression, and the total revenue of the fleet can be expressed as:

[0132]

[0133] Since U j (m) is a concave function, its maximum value is obtained at and Substituting E1(M j ) and E2(M j ) into the expression of E1(M j ) and E2(M j ) gives:

[0134]

[0135] The conclusion of Step 2 is proved.

[0136] Proof Two: Proof process of Step 3

[0137] Based on the proof of Step 2, the total revenue of the vehicle fleet can be expressed as:

[0138]

[0139] and U j (m) can obtain the maximum revenue value at , and and are substituted into the expression of , the relationship between M j , E1(M j ) and E2(M j ) can be obtained, which is expressed as:

[0140]

[0141] Substituting it into the relationship between the computing resources f j,i allocated by the vehicle fleet j to the user task i, the number of users M j allocated by the vehicle fleet j, and the total computing resources F j of the vehicle fleet j, that is, the relationship in the original formula , the optimal solution of the original formula is obtained:

[0142]

[0143] The conclusion in Step 2 is proved.

[0144] Proof Three: Proof process of Step 4

[0145] According to the equal proportion formula, when the computing tasks allocated to the same vehicle fleet satisfy:

[0146]

[0147] the optimal solution of the original problem can be simplified as: When i = b is satisfied (since it is an equal proportion formula, changing the subscript value does not change), the optimal solution of the original problem is: That is, the optimal benefit resource for the single user is equivalent to the optimal solution of the original problem, and the conclusion in step 4 is proved.

[0148] It should be understood that the above description of the preferred embodiments is more detailed and is not considered as limiting the scope of patent protection of the present application. Those skilled in the art can make substitutions or modifications without departing from the scope of the present application, and all fall within the scope of protection of the present application. The scope of protection of the present application should be subject to the appended claims.

Claims

1. A method for allocating heterogeneous edge computing resources of fleet base stations based on the Internet of Vehicles, characterized in that: The following steps are involved: Step 1: Construct a collaborative computing mechanism for multiple vehicle users in a heterogeneous edge network of multiple fleets and the current base station, and establish a user-multiple fleet-base station heterogeneous resource allocation benefit problem; Vehicle users send computing tasks to any fleet within the service range. After receiving computing tasks with differentiated latency requirements from different vehicle users, the fleet sends some of the computing tasks to the current base station to achieve joint collaborative edge computing between multiple fleets and base stations. Step 2: Define the maximum benefit resource allocation problem and calculate the maximum benefit of all fleets And derive the optimal number of serviceable users for each fleet Step 3: Based on step 2, define the resource allocation problem at the optimal number of users as the maximum benefit resource allocation problem; Substitute the expression into the solution and obtain the optimal semi-closed solution for the maximum benefit resource allocation for multiple users; Step 4: Based on the maximum benefit resource allocation problem in step 3, calculate the maximum benefit computing resource allocation value for a single user, which is defined as the single-user optimal benefit resource. And prove the equivalence condition between the single-user optimal benefit resource and the maximum benefit resource allocation in step 3. Step 5: Accumulate the remaining computing resources of all fleets and record it as the total remaining resources of the fleet; calculate the upper bound of the current remaining resources based on the remaining computing resources and the number of unassigned tasks; Step 6: Based on the driving speeds and locations of the fleet and users, as well as the maximum latency of the user computing tasks, calculate the range of users that each fleet can serve while ensuring communication connection time. Based on the optimal single-user benefit resources obtained in Step 4, sort the computing tasks generated by users within the serviceable range from largest to smallest based on the ratio of their completed benefits to allocated resources. Step 7: Complete the task pre-allocation between users and fleets through the three-step task allocation strategy; Step 8: Based on the user-fleet pre-allocation results and the single-user optimal benefit resources of the computing task, calculate the total amount of computing resources required by each fleet. When the fleet's available computing resources meet the required resources, the proportion of computing tasks completed by the fleet is 1. When the required resources are greater than the fleet's available computing resources, a new variable ρ is introduced as the task split ratio, that is, computing tasks of proportion ρ are calculated in the fleet, and computing tasks of proportion (1-ρ) are calculated at the base station. The user-multi-fleet-base station heterogeneous resource allocation benefit problem in step 1 is updated. Step 9: Split the updated user-multi-fleet-base station heterogeneous resource allocation benefit problem into two sub-problems. Solve the fleet resource allocation value and base station resource allocation value under the current task allocation strategy, calculate the task split ratio, and calculate the maximum benefit at this time. Step 10: Compare the current profit with the historical best profit. If the current profit is greater than the historical best profit, update the best profit value and the corresponding task allocation strategy, fleet resource allocation value, base station resource allocation value, and task split ratio. Otherwise, based on the current task allocation strategy, in ascending order of fleet total resources, and with the allocation of a single computing task as the unit, update the computing task allocation strategy backtracked from step 11, and calculate the upper bound profit of the current allocation strategy according to step 5. If the upper bound profit is less than the historical best profit, end the task allocation. Step 11: Based on the current task allocation strategy and the corresponding fleet resource allocation and base station resource allocation, the allocated computing tasks are backtracked in the order of fleet total resources from large to small and task allocation from back to front. The upper bound of the remaining allocation strategy is calculated according to step 5. If the upper bound is greater than the historical best benefit, return to step 7; otherwise, the optimal user-multi-fleet-base station multi-layer computing resource collaborative allocation result is obtained.

2. The method for allocating heterogeneous edge computing resources for fleet base stations based on the Internet of Vehicles according to claim 1 is characterized by: In step 1, the computing capacity of different fleets is expressed as f = mf0, where f0 represents the minimum unit computing resource value that can be allocated within the fleet, and m represents the minimum unit resource quantity in different fleets; then the computing completion time of user task i in fleet j is for: where N p represents the number of fleets, x j,i It represents the variable associated with the assignment of user task i to fleet j. When the user task is calculated in fleet j, x j,i =1, otherwise x j,i =0;ρ j,i represents the calculation ratio of user task i in fleet j; c i Indicates the total amount of computing resources required for user task i; Similarly, the remaining task ratio of user task i is calculated in the calculation time of the base station where f b,i Indicates the computing resource value currently allocated by the base station to task i; Based on the Rayleigh fading model, the communication channel gain between the fleet and the user and base station is expressed as: where G represents the determined power gain related to the antenna and amplifier parameters, h η ~CN(0,1) represents Rayleigh fading, which conforms to the complex Gaussian distribution. represents the Euclidean distance between vehicle i and vehicle j, where x i ,y i ,x j ,y j denote the horizontal and vertical coordinate values ​​of vehicle i and vehicle j respectively; α denotes the path attenuation factor; the data transmission time between the vehicle user and the fleet and the base station is: where s i represents the size of user task i; and They represent the sending rate of user vehicle i to the fleet and the sending rate of fleet j to the base station, respectively. Their expressions are: in Indicates the bandwidth resources allocated according to the length of the vehicle group, B is the total bandwidth resources, n j represents the number of vehicles that make up fleet j, N u represents the number of vehicle users, represents the power spectral density of Gaussian white noise, p i represents the transmission power of vehicle user i; Based on the benefits brought by the reduction of user task completion latency and the resource allocation consumption of the fleet and base stations, a task completion profit model for heterogeneous edge computing is designed: where τ i represents the maximum allowed delay of user task i, represents the actual completion delay of user task i, (τ i -t i ) represents the reduced task completion delay of user task i; f j,i ,f b,i denote the computing resource values ​​allocated by fleet j and base station to user task i respectively; It represents the user benefit brought by shortening the delay. They represent the cost of resource allocation for the fleet and base station respectively; μ represents the penalty factor for energy consumption, represents the energy consumption of the fleet, where p n Indicates the data transmission power of the fleet; Establish the vehicle user-multiple fleet-base station heterogeneous resource allocation benefit problem: where χ={x j,i } represents the set of associated variables between vehicle users and fleets, F b ,F p represents the resource allocation set of the base station and the fleet to different users, ρ represents the user task partitioning ratio set; F j represents the total computing resources of fleet j; N e Indicates N u The total number of tasks generated by a vehicle user at this time; Under the premise of ensuring the task completion delay, maximize the benefits, that is, improve the utilization of computing resources.

3. The method for allocating heterogeneous edge computing resources for fleet base stations based on the Internet of Vehicles according to claim 2 is characterized by: In step 2, define the maximum benefit resource allocation problem: in, represents the sum of the number of users assigned to fleet j; By proving that the problem is a single-peak problem, we obtain the closed-form solution for the optimal number of service users at the peak of the problem, that is, the number of users assigned to fleet j reaches When , the maximum benefit of fleet j can be obtained in:

4. The method for allocating heterogeneous edge computing resources for fleet base stations based on the Internet of Vehicles according to claim 3 is characterized by: In step 3, the optimal solution for computing resources allocated to each user is for: in, represents the maximum amount of beneficial resources allocated by fleet j to user task i, which is the optimal solution to the original problem; c p 、s p , τ p 、 denote the required computing resources, task size, maximum allowed delay, and data transmission rate of user task p assigned to fleet j, respectively.

5. The method for allocating heterogeneous edge computing resources of base stations in a fleet based on the Internet of Vehicles according to claim 4 is characterized by: In step 4, a low-complexity problem for single-user resource allocation is constructed: Solving the optimal benefit resource for a single user Prove that when the task size, required computing resources, maximum allowable delay and transmission rate assigned to fleet j satisfy the following relationship: Single user optimal benefit resource Equivalent to the most efficient resource Therefore, the subsequent goal is to transform the single user's optimal benefit resource Approaching the optimal solution for multi-user resource allocation in the original problem 6. The method for allocating heterogeneous edge computing resources for fleet base stations based on the Internet of Vehicles according to claim 5 is characterized by: In step 5, the total remaining resource benefits of the fleet and base station are solved Among them i Indicates the currently assigned user tasks; t i '、f i ', f′ b,i 、p、 ρ i They represent the task completion time of task i based on the total remaining resources of the fleet, the fleet resource allocation, the base station resource allocation, the user data transmission power, the task split ratio and the maximum transmission rate of the fleet base station respectively; remember is the upper bound of the return value under the i-th iteration, where represents the user-fleet unloading allocation at (i-1) iterations, represents the maximum revenue under this user-fleet assignment.

7. The method for allocating heterogeneous edge computing resources for fleet base stations based on the Internet of Vehicles according to claim 1 is characterized by: In step 7, computing tasks are first assigned to the corresponding fleets in the order of the fleet's available resource capacity, as sorted in step 6. The fleet's remaining computing resources are updated based on the maximum benefit resource allocation estimate for the corresponding computing task. When all fleets are unable to provide computing services according to the maximum benefit resource estimate, the initial allocation of computing tasks is completed. Then, the remaining unassigned computing tasks are classified according to the number of fleets that can be served. For computing tasks with a single number of service fleets, their allocation strategy is fixed and secondary allocation is completed to ensure the overall completion rate of tasks in the current scenario. Finally, the remaining resources of all fleets are updated based on the minimum amount of resources required to complete the computing task within the maximum delay. The list of available fleets is dynamically updated from large to small based on the fleet's remaining resource capacity, and all unassigned tasks are pre-allocated based on the dynamic fleet list in a round-robin manner.

8. A fleet base station heterogeneous edge computing resource allocation system based on the Internet of Vehicles, characterized by: include: one or more processors; A storage device for storing one or more programs, which, when executed by the one or more processors, enables the one or more processors to implement the method for allocating heterogeneous edge computing resources of a fleet base station based on the Internet of Vehicles as described in any one of claims 1 to 7.

Images