阅读说明：本技术 基于车流量预测的车联网双边拍卖式边缘计算迁移方法 (Vehicle networking bilateral auction type edge calculation migration method based on vehicle flow prediction ) 是由 闫帅 林艳 李帆远 张一晋 束锋 邹骏 于 2020-06-12 设计创作，主要内容包括：本发明提供了一种新的基于车流量预测的车联网双边拍卖式边缘计算迁移方案。该方案首先需要收集道路中行驶车辆的特征信息,并统计当前周期内的车流量用以预测下一周期的车流量信息。然后,利用车辆特征信息和车流量预测信息完成车辆对计算资源需求的报价,进而可采用一种双边拍卖算法实现车载边缘计算服务器的计算资源分配,最后通过将车辆的计算任务卸载至分配的服务器完成边缘计算迁移过程。本发明与未考虑车流量信息的两种分配方案相比,更能够充分利用车联网中的计算资源,且能显著提升整个系统的边缘计算迁移效率。(The invention provides a novel vehicle networking bilateral auction type edge calculation migration scheme based on vehicle flow prediction. According to the scheme, firstly, the characteristic information of the running vehicles in the road needs to be collected, and the traffic flow in the current period is counted to predict the traffic flow information in the next period. Then, the vehicle characteristic information and the vehicle flow prediction information are used for completing the quotation of the vehicle to the computing resource demand, further a bilateral auction algorithm can be adopted for realizing the computing resource distribution of the vehicle-mounted edge computing server, and finally the computing task of the vehicle is unloaded to the distributed server to complete the edge computing migration process. Compared with two distribution schemes without considering traffic flow information, the method can make full use of computing resources in the Internet of vehicles, and can remarkably improve the edge computing migration efficiency of the whole system.)

1. A vehicle networking bilateral auction type edge calculation migration method based on vehicle flow prediction is characterized by comprising the following specific processes:

s1, determining parameters and time consumption required by each stage in the unloading process of the calculation task in the calculation and migration process of the mobile edge of the Internet of vehicles;

s2, setting the duration of the prediction period, counting the traffic flow in the period after the prediction period is finished, and predicting the traffic flow condition of the next period based on the Markov decision process;

and S3, introducing traffic flow prediction information in the mobile edge calculation process, and designing an edge calculation migration scheme based on auction theory.

2. The method of claim 1, wherein: a plurality of road side units are arranged on one side of the road; each road side unit is provided with an on-board edge computing server; the vehicle enters the road section, communicates with the RSU and submits the vehicle characteristic information; the RSU aggregates the collected information submissions to the base station.

3. The method of claim 1, wherein: in the running process of the vehicle, the RSU is responsible for collecting traffic flow information; setting the time length of an information collection period, and counting the number of vehicles passing through the road section in the period by a base station after each period is finished; determining the conversion relation of the vehicles in different road sections according to the acquired traffic flow data of each road section in two adjacent periods, and calculating to obtain a transition probability matrix; and predicting the traffic flow information of the road section in the next period by utilizing the traffic flow information and the probability transition matrix in the period according to the Markov decision process.

4. The method of claim 1, wherein: the vehicle is used as a buyer, and initial quotation is submitted based on the vehicle characteristic information and the traffic flow information; the RSU and the base station periodically count and predict the traffic flow, and continuously correct and update the vehicle quotation; the VEC server acts as a seller to conduct a bilateral auction with the buyer's vehicle, allocating computing resources.

Technical Field

The invention relates to the technical field of wireless communication, in particular to a vehicle networking bilateral auction type edge computing migration scheme based on vehicle flow prediction.

Background

The vehicle networking utilizes equipment such as a sensor, an image processor and a GPS to sense and acquire vehicle characteristic information and traffic condition information, and establishes interactive connection between vehicles, vehicles and infrastructure, vehicles and roadside pedestrians, vehicles and networks, vehicles and clouds and the like by means of a wireless communication technology, so that the traffic operation efficiency and the system intelligence level are improved.

Mobile edge computing is one of the key technologies for vehicle mobile communication in traffic networks, and is widely applied to the problem of allocation of wireless resources. In order to deal with the situation that the computing resources of the vehicle are limited, the vehicle unloads the computing task to an on-board edge server or other vehicles, and the computing capability is expanded to the edge of the network. By moving the edge calculation, the pressure of the shortage of vehicle calculation resources can be relieved, the calculation resources in the network are utilized to the maximum extent, and the system performance is improved.

The invention develops a bilateral auction type edge calculation migration scheme of the Internet of vehicles, introduces traffic flow prediction information in the mobile edge calculation process of the Internet of vehicles, solves a resource allocation scheme by utilizing a bilateral auction algorithm and improves the efficiency of calculation migration.

Simulation results show that compared with a scheme without traffic flow, the scheme with traffic flow prediction information introduced can remarkably improve a task unloading utility function and a resource utilization utility function.

Disclosure of Invention

The invention provides a vehicle networking bilateral auction type calculation migration scheme based on vehicle flow prediction.

In order to achieve the purpose, the technical scheme adopted by the invention comprises the following steps: determining the consumption time of the vehicle at each stage in the moving edge calculation process; predicting traffic flow information based on a Markov decision process; and introducing the traffic flow prediction result into a mobile edge calculation process, and designing an edge calculation migration scheme by using a bilateral auction algorithm.

Further, the specific process comprises the following steps: s1, collecting characteristic information of a vehicle, and determining consumed time of vehicle movement, task transmission and task calculation in a moving edge calculation process; s2, predicting traffic flow information based on a Markov decision process, setting a fixed time length as a prediction period, counting the traffic flow in the period after the prediction period is finished, and predicting the traffic flow condition of the next period; and S3, introducing traffic flow prediction information in the mobile edge calculation process, and designing an edge calculation migration scheme based on an auction theory.

Has the advantages that: the scheme has the advantages that 1, the scheme realizes the introduction of traffic flow information into the edge calculation process, and considers the influence of the traffic flow information on the edge calculation migration efficiency; 2. compared with a sequential distribution algorithm, the bilateral auction algorithm adopted by the scheme can improve the edge calculation migration efficiency; 3. according to the scheme, the traffic flow prediction information is introduced, and compared with a resource allocation scheme without considering the traffic flow information, the edge calculation migration efficiency can be further remarkably improved.

Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.

Drawings

FIG. 1 is a block diagram of a design of a computational migration scheme based on traffic flow prediction.

Fig. 2 shows a traffic flow transfer relationship.

Figure 3 is a flow chart of the auction algorithm solving.

FIG. 4 is a diagram of VEC server computing resources occupied by vehicles.

FIG. 5 is a vehicle mission offload utility function versus number of vehicles.

FIG. 6 is a VEC server resource utilization utility function versus number of vehicles.

FIG. 7 is a vehicle task offload utility function versus number of VEC servers.

FIG. 8 is a relationship of VEC server resource utilization utility function versus number of VEC servers.

Detailed Description

The present invention is further illustrated by the following figures and specific examples, which are to be construed as merely illustrative and not limitative of the remainder of the disclosure, and all changes and modifications that would be obvious to those skilled in the art are intended to be included within the scope of the present invention and the appended claims are intended to be embraced therein.

The specific method for determining the time consumed by each stage in the moving edge calculation process comprises the following steps: the vehicle access segment communicates with the RSU and submits vehicle characteristic information. Assuming that vehicle i is assigned the jth rsu, the travel time is:

wherein L is_jIs RSU_jDistance from the starting point of the road section,/_iAnd distributing the distance between the vehicle i and the starting point of the road section at the starting moment of the period.

The vehicle uploads the task to the VEC server, and the time for transmitting the task in the channel is calculated as follows:

wherein, B_jIs RSU_jBandwidth of P_iFor transmission power of vehicle i, G is vehicle i and RSU_jThe channel gain in between.

The VEC receives the calculation task and calculates the task, and the time required by calculation is as follows:

wherein, U_cThe constant value indicates the CPU frequency, i.e., the number of CPU cycles required to calculate the unit bit data amount.

The specific method for predicting the traffic flow information comprises the following steps: at the end of each prediction period, the RSU counts the number of vehicles passing through each lane of the road section in the period and records the number as V⁽ⁱ⁾：

Wherein the content of the first and second substances,the number of vehicles in the left-turn lane, the straight lane and the right-turn lane in the ith cycle is respectively.

And determining the conversion relation of the vehicle between different road sections according to the acquired traffic flow data of each road section in two adjacent periods. Taking the first road segment in fig. 2 as an example, the vehicle source of the first driving road segment comprises a left-turn lane (i), a straight-going lane (ii) and a right-turn lane (iii), the new lane selected after driving comprises a left-turn lane (a), a straight-going lane (B) and a right-turn lane (C), and the statistical traffic flow transfer relationship is shown in the following table:

TABLE 1 traffic flow transfer relationship

And (3) calculating to obtain a transition probability matrix p by using the data:

according to the Markov decision process, the traffic flow V of the next period can be predicted by utilizing the traffic flow of the period and the probability transfer matrix p⁽ⁱ⁺¹⁾：

The invention introduces traffic flow prediction information and utilizes an auction algorithm to solve an edge calculation migration scheme, and the specific method comprises the following steps: the vehicle is used as a buyer based on the vehicle characteristic information: calculating the size d of the task data volume_iVehicle speed v_iTask priority N₁The amount of demand on computing resources s_iAnd submitting a quote:

N₁is set to {1, 2, 3 … }, N₁The larger the task, the higher the importance of the task, beta₁Controlling the gap in importance between different priorities, X representing the total number of rounds in which the auction is conducted, Num_iAnd the failure frequency of the vehicle i participating in the auction is represented by {0, 1, 2.., X }.

Defining a traffic flow priority N for a vehicle₂，N₂The value of (1) is {1, 2, 3}, the degree of vehicle congestion on three lanes of left turn, straight going and right turn is distinguished, and the larger the value is, the more congested the lane is, namely the traffic flow is larger.

After introducing the traffic flow prediction information, further correcting the vehicle quoted price:

the VEC server acts as a seller and conducts a bilateral auction with the buyer's vehicle, the auction process being as shown in fig. 3. And calculating the initial quotation of the vehicle according to the vehicle characteristic information and the traffic flow prediction information. The vehicle offers are sorted in descending order, and the time that the vehicle occupies the VEC server is calculated in turn to determine whether a match is achieved with the server. And the unmatched vehicles fail to pass the auction round, and enter the next auction round after the quotation is updated.

Before implementing the invention, a performance evaluation index needs to be set. The method sets a utility function of vehicle task unloading and a VEC server resource utility function as performance evaluation indexes.

Considering the unloading proportion condition of the vehicle task, defining a utility function U for unloading the vehicle task₁Comprises the following steps:

wherein x is_iE {0, 1} is an allocation decision variable, if the base station allocates a certain VEC server for the vehicle i, x_i1, otherwise x_i0. Definition of R ═ d_i·N₁·N₂For a task d of the vehicle_iThe benefit value which can be brought to the vehicle after being calculated. T is_waitIndicating the length of time the vehicle waits from entering the road segment to uploading data to the VEC server,for all vehicles T_waitAverage value of (a).

The whole time process is divided into a plurality of delta t, and the time of each vehicle occupying the server is integral multiple of the delta t. The case where a plurality of vehicles occupy the same server will be described with reference to fig. 4. In the figure, Δ t is 1 second, the horizontal axis represents time, and the vertical axis represents the amount of computing resources held by one VEC server. The left side represents the time occupation situation and the resource occupation situation of the vehicle on the server, and the right side is the window function addition of the vehicle occupation time, so that the VEC server resource utilization situation can be reflected. Thus defining VEC server resource utility function U₂Comprises the following steps:

wherein s is_jnRepresents the computing resources that the jth VEC server is occupied for the nth Δ T period, and T represents the total time of resource allocation, i.e., the sum of all Δ T.

Consider the performance comparison of the following three algorithms: the first algorithm is as follows: regardless of the task priority factor of the vehicle, the computing resources of the VEC server are allocated nearby, simply in the order in which the vehicles appear in the traffic system, referred to as a sequential allocation algorithm. And the second algorithm is to adopt a bilateral auction algorithm to distribute computing resources. And thirdly, introducing traffic flow prediction information on the basis of the second algorithm as a basis for resource allocation.

FIG. 5 illustrates a vehicle mission offload utility function versus a number of vehicles. With the increasing number of vehicles, no matter which algorithm, the utility function U of unloading the vehicle task₁In a downward trend. This is consistent with the actual situation. Since the VEC server can bear a certain number of vehicle calculation tasks at most, the number of tasks to be unloaded increases as the number of vehicles increases, the possibility of unloading the tasks decreases, and the unloading proportion of the vehicle tasks decreases. FIG. 6 shows a VEC server resource utilization utility function versus number of vehicles. VEC server resource utility function U as the number of vehicles increases₂In an upward trend. This is because the amount of tasks increases while the amount of VEC server resources does not change, so the utilization ratio of VEC server resources inevitably increases. FIG. 7 shows a vehicle task offload utility function versus the number of VEC servers. With the increasing number of VEC servers, the utility function U for unloading the vehicle task₁In an upward trend. This is because increasing the number of VEC servers provides more options for task offloading of vehicles, and thus the offloading rate is increased, while the number of vehicles remains substantially unchanged. FIG. 8 shows a VEC server resource utilization utility function versus the number of VEC servers. VEC server resource utility function U₂Increasing as the number of VEC servers increases. This indicates that the increase in the number of VEC servers can increase the probability of offloading of vehicle tasks, making more efficient use of the computing resources of the VEC servers.

Comparing three different algorithms, it can be found that the performance of the auction type calculation migration scheme based on the traffic flow prediction is optimal, and the task auction type calculation migration scheme without the traffic flow prediction is the second time. And with the increase of the number of vehicles and the number of VEC servers, compared with other two algorithms, the auction-type calculation migration scheme for introducing the traffic prediction information has higher and higher superiority.