Prediction-based dynamic load balancing method in 5G vehicle networking CoMP scenario

Abstract

The present invention relates to the technical field of 5G vehicle networking, and in particular to a prediction-based dynamic load balancing method in the vehicle networking CoMP scenario. The present invention includes counting the initial load status of each base station, and using the lightly loaded base station as a cooperative base station for the fully loaded or overloaded base station; for the overloaded base station, the user is first diverted, and some edge users are diverted to the adjacent lightly loaded base station for service, and then the user's location distribution is analyzed; for the fully loaded or lightly loaded base station, the user's location distribution is directly analyzed; the central user is directly served by this base station, and the edge user must first be predicted, and the CoMP technology is selectively used to provide services to the user. The present invention shows through simulation analysis that the base station overload rate is significantly reduced during both non-peak and peak periods, and the user's rate is also improved to a certain extent, achieving the purpose of load balancing between base stations, while ensuring the user's communication quality and facilitating the user's life.

Description

Technical Field

[0001] The present invention relates to the field of 5G vehicle networking technology, and in particular to a prediction-based dynamic load balancing method in a vehicle networking CoMP scenario. Background Art

[0002] With the rapid increase in the number of cars, the "tidal effect" has become more and more obvious in vehicle travel time, and urban vehicle congestion has gradually become a social problem. The resulting problems such as network freezes and poor communication quality in the Internet of Vehicles have brought certain troubles to users. 5G and future wireless communication networks are characterized by dense distribution of communication users. With the increase in the types of services and the randomness of the distribution of mobile users in time and space, some base stations are easily overloaded. In addition, the resources of some base stations are not fully utilized. Therefore, load balancing is particularly important.

[0003] In the research of existing load balancing algorithms, according to different task scheduling strategies, they are mainly divided into two categories: static load balancing algorithm and dynamic load balancing algorithm. The basis of task allocation of static load balancing algorithm is mainly pre-set strategy, but it will not make real-time adjustments to time-varying base stations, such as polling algorithm, weighted polling algorithm, etc., while dynamic load balancing algorithm will use the current base station as a reference when dividing tasks, such as minimum connection method, weighted minimum connection method, etc.

[0004] Static load balancing usually involves mapping or scheduling problems. There are many research results on static load balancing in academia, such as the paper (C Kim, Hisao Kameda. Optimal Static Load Balancing of Multi-Class Jobs in a Distributed Computer System [C], 10th International Conference on Distributed Computing Systems, Paris France, 1990: 562-569.) which improves the study of the classic static load balancing strategy and designs a load balancing strategy based on minimizing the average response time. This strategy can achieve better balancing effects in heterogeneous environments; the paper (Asser NTantawi, Donald F Towsley. Optimal Static Load Balancing in Distributed Computer Systems [J], Journal of the ACM (JACM), 1985, 32 (2): 445-465) and the paper (Chonggun Kim, H. An algorithm for optimal static load balancing in distributed computer systems [J], IEEE Transaction on Computers, 1992, 41(3): 381-384) used mathematical models to describe the single-task type and multi-task type static load balancing problems respectively, gave its formal representation, analyzed the performance of the algorithm, and verified the feasibility of the algorithm through simulation experiments.

[0005] The traditional static load balancing strategy can no longer meet the dynamic changing needs of current applications due to its own design defects, and the applicable scenarios have certain limitations. Therefore, people propose to use dynamic load balancing strategies to solve practical problems. For example, in reference

[14] (S. Mitra and S. DasBit. A load balancing strategy using dynamic channel assignment and channel borrowing in cellular mobile environment [C]. IEEE International Conference on Personal Wireless Communications. Conference Proceedings, Hyderabad. 2000: 278-282), dynamic channel assignment and channel borrowing technology are combined to balance the load conditions among base stations; in reference

[15] (JANSEN T, BALAN I, STEFANSKI S, et al. Weighted performance based handover parameter optimization in LTE [C] / / International Workshop on Self-Organizing Networks. IEEE VTC 2011-Spring Workshop IEEE. Yokohama: IEEE Press, 2011: 1-5), overloaded base stations are transformed back into non-overloaded base stations through "handover parameter adjustment" to achieve load balancing.

[0006] None of the above load balancing methods considers how to reduce or eliminate the interference problem between base stations.

[0007] Coordinated Multi-Point (CoMP) transmission technology uses multiple base stations to coordinate scheduling or joint processing to suppress or even eliminate interference between base stations. Nowadays, there are many studies on load balancing algorithms based on CoMP. For example, reference

[22] (L. Feng, M. Yin, N. Xiang, W. Li and X. Qiu. Load-balancing based on Base-Station CoMP with guaranteed call blocking rate [C]. International Symposium on Wireless Personal Multimedia Communications. 2014: 271-276) modeled and discussed different load balancing modes based on CoMP, defined the call blocking rate as the service quality indicator guaranteed in the load balancing process, and compared it with load balancing based on different base station clustering modes; reference

[23] (S. Bassoy, M. Jaber, M. A. Imran and P. Xiao. Load Aware Self-Organizing User-Centric Dynamic CoMP Clustering for 5G Networks [J]. IEEE Access, 2016, 4: 2895-2906) proposed a self-organizing, user-centric clustering algorithm with a control-data layer separation architecture for CoMP scenarios; Reference

[24] (GB Figueiredo, X. Wang, CC Meixner, M. Tornatore and B. Mukherjee. Loadbalancing and latency reduction in multi-user CoMP over TWDM-VPONs [C]. IEEE International Conference on Communications. 2016: 1-6) proposed a new network architecture that supports CoMP operations, using virtualized base stations and cloud wireless access networks to distribute load on base station controllers to minimize coordination delays, thereby achieving load balancing of base stations.

[0008] These algorithms are implemented under the CoMP technology, but none of them performs load balancing for the Internet of Vehicles scenario, and cannot effectively solve the load balancing problem for the Internet of Vehicles CoMP scenario proposed in the present invention. There are some prediction-based load balancing algorithms, such as the paper (Luis Miguel Campos, Isaac D. Scherson. Rate of change loadbalancing in distributed and parallel systems. Parallel Computing, 26 (2000), pp. 1213-1230), which uses the load change rate of the processor to dynamically predict the load of the processor. Although this algorithm is simple, the accuracy of this prediction algorithm will be reduced when the processor load changes very quickly, is highly random, and has no stable and continuous state.

[0009] In view of this, the present invention innovatively proposes a load balancing algorithm based on vehicle prediction to solve the load imbalance problem in Poisson distribution scenarios, which can not only achieve the purpose of base station load balancing, but also overcome the performance degradation problem of edge users caused by interference. Compared with previous load balancing algorithms, the algorithm proposed in the present invention can more effectively improve the user experience rate and better balance the load between base stations. Summary of the invention

[0010] The purpose of the present invention is to provide a prediction-based dynamic load balancing method in a 5G vehicle network CoMP scenario to solve the technical problems pointed out in the above background technology.

[0011] In order to achieve the above object, the present invention adopts the following technical solutions:

[0012] The prediction-based dynamic load balancing method in the 5G vehicle networking CoMP scenario includes the following steps:

[0013] S1. Count the initial load status of each base station, which is divided into three states: light load, full load and overload. The lightly loaded base station is used as a cooperative base station for the fully loaded or overloaded base station;

[0014] S2. Analyze the three different states of the base station as follows:

[0015] For overloaded base stations: users are first diverted to nearby lightly loaded base stations until the overloaded base stations are balanced to full load or light load, and then the user location distribution is analyzed;

[0016] For base stations in full or light load states: directly analyze the location distribution of users;

[0017] Analyzing the location distribution of users means dividing users into central users and marginal users;

[0018] S3. Central users are directly served by the base station. For edge users, predictions are made first and their load conditions are continuously counted. CoMP technology is then selectively used to provide services to users.

[0019] S4. Determine the network rate of the edge user. If the network rate of the edge user exceeds the set threshold, it is considered that the base station reaches a load balancing state;

[0020] S5. Repeat the above steps until the network rates of all users are not lower than the set threshold and the number of users served by all base stations is not greater than the upper limit of the number of users that can be served by the base stations.

[0021] Furthermore, the determination method of the lightly loaded, fully loaded and overloaded base stations is as follows:

[0022] Set a threshold value A th As the upper limit of the number of users served by the base station, let the number of users served by the base station be A now ,but:

[0023] When A now >A th , then the base station is considered to be in an overload state; when A now =A th , then the base station is considered to be in a fully loaded state; when A now th , it is determined that the base station is in a light load state.

[0024] Furthermore, in S2, users are divided into central users and edge users in the following manner:

[0025] For each base station, the user locations are divided based on 3r / 4. Users with a radius greater than 3r / 4 are classified as edge users, and users with a radius less than or equal to 3r / 4 are classified as central users, where r is the radius that a base station can cover.

[0026] Furthermore, it is characterized in that, in S3, prediction is first performed for edge users, and the prediction model is as follows:

[0027] Assuming that the number of vehicles arriving at the base station roughly follows a Poisson distribution with strength λ, the probability that k vehicles arrive at the base station at a certain moment is:

[0028]

[0029] Where λ is a time-related parameter;

[0030] Assume that there are two cases for λ: during the rush hour, the value of λ is λ1, and during the off-peak hour, the value of λ is λ2, where λ1>λ2; ​

[0031] The probability that K vehicles arrive at the base station is expressed as:

[0032]

[0033] The calculation is performed on the premise of obtaining the probability of a vehicle arriving at the base station, and the K value when P(X=K) is the maximum is taken as the predicted value of the number of vehicles arriving at the base station at the next moment;

[0034] When the position and speed of the vehicle itself are known, the number of vehicles leaving the base station can be predicted. The number of vehicles at the base station at a certain moment can be predicted based on the arrival and departure of the vehicles.

[0035] Furthermore, in S3, edge users must be predicted first, their load conditions must be continuously counted, and CoMP technology must be selectively used to provide services to users, mainly including:

[0036] When a base station is initially in a light load state:

[0037] When the predicted base station is in a light-load state, the CoMP technology is used to provide services to users; when the predicted base station is in a fully-loaded state, the CoMP technology is used to provide services to users together with the adjacent light-loaded base station; when the predicted base station is in an overloaded state, the load is re-distributed and balanced;

[0038] When a base station is initially in full load state:

[0039] When the predicted base station is in a light-load or full-load state, the CoMP technology is combined with the adjacent light-load base station to provide services to users at the same time; when the predicted base station is in an overload state, the load is re-diverted and balanced.

[0040] The present invention has at least the following beneficial effects:

[0041] Aiming at the problem of load balancing between base stations, the present invention combines the prediction algorithm with the CoMP technology to form a new load balancing algorithm to balance the load conditions between base stations and make effective use of resources. After simulation analysis and verification, after adopting the algorithm proposed in this article for load balancing, the overload rate of the base station is significantly reduced, whether in peak or non-peak periods, and the average rate of users is also improved to a certain extent, which improves the load balancing problem of the base station to a certain extent and facilitates the life of users. These works together provide a reference for the load balancing problem of the base station. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying any creative work.

[0043] Figure 1 It is the relationship diagram between vehicles and base stations at different times;

[0044] Figure 2 The position relationship between base stations at different locations and vehicles;

[0045] Figure 3 A diagram showing the relationship between the user and the serving base station;

[0046] Figure 4 This is the overall flow chart of the algorithm;

[0047] Figure 5 It is the sub-flow chart of light load state;

[0048] Figure 6 It is the sub-flowchart of full load state;

[0049] Figure 7 It is the sub-flowchart of overload state;

[0050] Figure 8 The overload user count map for all users in each of the 19 cells;

[0051] Fig. 9 It is the ratio diagram of the number of overloaded base stations;

[0052] Fig.10 This is the average rate diagram of users at the edge of each cell in 19 cells;

[0053] Fig.11 This is the average rate diagram of 19 cell edge users;

[0054] Fig.12 It is the average rate graph of all users in the cell;

[0055] Fig.13 This is the average rate diagram of users in 19 cells. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0057] 1. System Model

[0058] 1.1 Vehicle Distribution Based on CoMP

[0059] Normally, vehicles on the road are independent individuals, and their entry and exit times, driving speeds, etc. are all determined by the driver's own wishes, and are also affected by road traffic rules, road conditions, and the behavior of other nearby drivers. Therefore, vehicles are random in their driving process, and when a vehicle is about to reach a base station, it often causes an unbalanced load on the base station. During the driving process of a vehicle, the relevant base stations that process information for the vehicle's communication equipment will change, causing some base stations to be overloaded, while some base stations are lightly loaded or even unloaded, and the base station's resources are not fully utilized. During the driving process of a vehicle, its service base station will also change to a certain extent with its movement position and time.

[0060] If we ignore the length of the vehicle and regard the vehicle as a point mass, and assume that 9 vehicles are driving towards the base station at a speed of V km / h at the same time, and use different colors to represent the position relationship between the vehicle and the base station at different times, we can use Figure 1 To indicate the change in vehicle position at a certain interval.

[0061] As each vehicle is driving, its service base station will also change with its location. Figure 2 It is a relationship diagram between a vehicle and its service base station during driving, where the dotted line represents the service status before driving and the solid line represents the service status after driving.

[0062] The present invention divides users into central users and edge users according to their distance from the center. The present invention divides user positions based on 3r / 4 as the boundary, and the division rules are as follows:

[0063]

[0064] Wherein, r is the radius that a base station can cover. The central user will only affect the load condition of the current base station, and the current base station will directly serve such users. The edge user will affect three base stations including the current base station adjacent to the current base station according to the load condition of the current base station. For the edge user, the present invention adopts CoMP coordinated multi-point transmission technology to provide services, which can not only achieve the purpose of load balancing, but also reduce the interference between adjacent base stations. The relationship between the user and the service base station is as follows: Figure 3 shown.

[0065] 1.2 Prediction Model

[0066] Poisson distribution is a discrete random probability distribution with a wide range of applications in life and production practice. It is mainly used to describe the probability distribution mathematical model of a random event X occurring k times per unit time or space. In actual cases, when a random event, such as a call received by a telephone exchange, a passenger arriving at a bus stop, particles emitted by a radioactive material, etc., occurs randomly and independently at a fixed average instantaneous rate λ (or density), then the number of occurrences of this event per unit time (area or volume) approximately follows the Poisson distribution P(λ).

[0067] In the Internet of Vehicles environment, although the distribution of vehicles is random, it is not completely without rules. In life, the number of vehicles arriving at the base station is often affected by the "tidal effect", and there will be a "peak" at a certain time of the day. Therefore, the present invention assumes that the number of vehicles arriving at the base station roughly follows a Poisson distribution with a strength of λ, then the probability of k vehicles arriving at the base station at a certain moment is;

[0068]

[0069] Wherein λ is a time-related parameter. In the present invention, it is assumed that λ has two conditions: during the rush hour, the value of λ is λ1, while during the non-peak hour, the value of λ is λ2, where λ1>λ2;

[0070] Then the probability that K vehicles arrive at the base station can be expressed as

[0071]

[0072] The calculation is performed on the premise of obtaining the probability of a vehicle arriving at the base station, and the K value when P(X=K) is the maximum is taken as the predicted value of the number of vehicles arriving at the base station at the next moment.

[0073] Under the condition that the position and speed of the vehicle itself are known, it is easy to predict the number of vehicles leaving the base station, and then the number of vehicles at the base station at a certain moment can be predicted based on the arrival and departure of the vehicles.

[0074] 1.3 Load Balancing

[0075] At different times, the service status of the base station is slightly different. The present invention can classify the base station into three states: light load, full load and overload according to the number of users served by the base station or the user rate.

[0076] The light load state indicates that the base station is in a low-load working state, and there are excess system resources to continue to accept and process user requests, with a faster response speed and processing efficiency; the full load state indicates that users can work normally at this moment, but the user experience rate is not as good as low load; and the overload state means that the base station is in an overloaded working state, and users will face problems such as network lag and signal congestion.

[0077] The present invention comprehensively uses the number of users served by the base station to determine whether the base station is overloaded, uses the marginal user rate to control the service quality, and sets a threshold A th As the upper limit of the number of users served by the base station, the minimum threshold of user rate V is set at the same time. th , if A th 、V th and the number of users served by the base station A now 、User average experience rate V use There are the following relationships:

[0078] V use <V th (4)

[0079] or

[0080] A now >A th (5)

[0081] At this time, the base station is considered to be in an overloaded state, while other situations are considered to be non-overloaded states. When the equal sign is taken, the base station is exactly in a fully loaded state.

[0082] Where V th It is the minimum value for users to work normally. th It is the maximum number of users that the base station can serve, and can be set according to actual conditions.

[0083] 2. Load Balancing Algorithm

[0084] 2.1 Overall algorithm flow

[0085] In the algorithm design, the present invention is mainly divided into five parts: the first step is to analyze the load status of each base station. The second step is to perform corresponding analysis and processing on base stations with different load status. The third step is to divide the user location and adopt corresponding services for users in different locations. The fourth step is to re-count the load status of the base station. The fifth step is to loop the first step until all base stations are in a non-overloaded state, that is, the user rate is greater than the set threshold and the number of users served by the base station does not exceed the set threshold. The specific steps are as follows:

[0086] Step 1.1 Count the initial load status of each base station, which is divided into three states: light load, full load and overload. The lightly loaded base station is used as a cooperative base station for the adjacent fully loaded or overloaded base stations.

[0087] Step 1.2 Analyze the three different states of the base station:

[0088] Step 1.2.1 Overload state: First perform load balancing, balancing to full load or light load state, and then analyze the user location distribution.

[0089] Step 1.2.2 In full load or light load state, first analyze the user's location distribution and divide them into central and edge users based on 3 / 4r as the boundary.

[0090] Step 1.3: The central users are directly served by the base station, while the edge users are predicted first and their load conditions are continuously counted, and the CoMP technology is selectively used to provide services to the users.

[0091] Step 1.4 determines the network rate of edge users. If the network rate of edge users exceeds the set threshold, it is considered that the base station has reached a load balancing state.

[0092] Step 1.5 repeats Step 1.1 until the network rates of all users are not lower than the set threshold and the number of users served by all base stations is not greater than the upper limit of the number of users that can be served by the base stations.

[0093] The load status of the base station is updated in real time. The load information of the base station is recalculated every certain period of time. The algorithm is in a dynamic cycle. If 9 vehicles are about to arrive at the base station at the same time, then Figure 1 The color and position changes of the vehicle can be regarded as the number of updates and the calculated load conditions. According to the prediction results, the load status of the base station at the next moment can be calculated, and balancing can be performed in advance to ensure the smoothness of the user network.

[0094] The specific process is as follows Figure 4 shown.

[0095] 2.2 Algorithm sub-process

[0096] 2.2.2 Light load sub-process

[0097] If the base station is in a light load state at the beginning, the base station will be listed as a cooperative base station of the adjacent base station, and the load prediction for this base station will be divided into four steps: the first step is to analyze the user's location. The second step is to predict the edge users. The third step is to adopt different service methods for users in different locations according to the prediction results. The fourth step is to start from the first step and repeat until all base stations are in a non-overloaded state.

[0098] The specific steps are as follows:

[0099] Step 2.1 Analyze the location distribution of users and divide them into central and marginal users based on 3r / 4. Users with a location greater than 3r / 4 are classified as marginal users, and users with a location less than or equal to 3r / 4 are classified as central users.

[0100] Step 2.2: Make business predictions for edge users. After the prediction, there will be three states: light load, overload, and full load.

[0101] Step 2.3 The central user is directly served by this base station; when the predicted base station is in a lightly loaded state, the CoMP technology is used to provide services to the user; when the predicted base station is in a fully loaded state, the CoMP technology is used to provide services to the user together with the adjacent lightly loaded base station; when the predicted base station is in an overloaded state, the load is re-diverted and balanced.

[0102] Step 2.4 repeats Step 2.1 until the network rates of all users are not lower than the set threshold and the number of users served by all base stations is not greater than the upper limit of the number of users that can be served by the base stations.

[0103] The specific process is as follows Figure 5 shown.

[0104] 2.2.3 Fully loaded sub-process

[0105] If the base station is already fully loaded, the user's location is analyzed directly. This is divided into four main steps: The first step is to divide the user location into central and edge users. The second step is to predict edge users. The third step is to use different service methods for users in different locations based on the prediction results. The fourth step is to start from the first step and repeat until all base stations are not overloaded.

[0106] The specific steps are as follows:

[0107] Step 3.1 Analyze the location distribution of users and divide them into central and marginal users based on 3r / 4. Users with a location greater than 3r / 4 are classified as marginal users, and users with a location less than or equal to 3r / 4 are classified as central users.

[0108] Step 3.2 performs business prediction on edge users, which are divided into two states: overloaded and not overloaded.

[0109] Step 3.3 The central user is directly served by this base station; when the predicted base station is not overloaded, it combines CoMP technology with the adjacent lightly loaded base station to provide services to the user at the same time; when the predicted base station is overloaded, it re-divides and balances the load.

[0110] Step 3.4 repeats Step 3.1 until the network rates of all users are not lower than the set threshold and the number of users served by all base stations is not greater than the upper limit of the number of users that can be served by the base stations.

[0111] The specific process is as follows Figure 6 shown.

[0112] 2.2.4. Overload sub-process

[0113] If the base station is overloaded, users will be diverted first, and some marginal users will be diverted to nearby lightly loaded base stations to form several lightly loaded or fully loaded base stations. Then, business prediction will be performed and users will be processed accordingly based on the prediction results.

[0114] It is divided into four steps:

[0115] Step 4.1 Users are diverted and the base station will be divided into several full-load or light-load states.

[0116] Step 4.2 Perform corresponding prediction processing on base stations in different states. The predicted base station status and service mode refer to Step 2 and Step 3.

[0117] Step 4.3 Determine the current state of the base station and perform corresponding analysis and processing for different states.

[0118] Step 4.4 repeats Step 4.1 until the network rates of all users are no lower than the set threshold and the number of users served by all base stations is no greater than the upper limit of the number of users that can be served by the base stations.

[0119] The specific process is as follows Figure 7 shown.

[0120] 3. Simulation verification

[0121] For the scheme proposed in this article, simulation comparison is carried out on the cell overload rate, average rate of cell users and average rate of cell edge users before adopting the algorithm proposed in the present invention, before and after equalization, and the reliability of the scheme proposed in the present invention is verified through analysis.

[0122] 3.1 Parameter settings

[0123] The simulation scenario is set to 19 cells, and the positions of users in the cells are randomly and evenly distributed. Table 1 gives the simulation parameter settings. Considering the total width of the standard two-way 6-lane road, it is assumed that there are 19 cells and the cell radius is 28.5m. Other parameters are shown in Table 1, where r_cell is the cell radius, P_T is the total transmission power of the small base station, B_base is the total bandwidth of the base station, F is the carrier frequency, which is the frequency adopted by most 5G in the world, N is the number of cells, num_add_mean_nexttime is the average number of new vehicles at the next moment, that is, the Poisson distribution average. During peak hours, the λ value of the Poisson distribution is 8, num_reduce_nexttime is the number of vehicles leaving at the next moment, num_user_threshold is the standard for the number of overloaded users of the base station, and rate_mean_user_threshold is the standard for the average rate of overloaded users of the base station.

[0124] Table 1 Simulation parameter settings

[0125]

[0126] 3.2 Simulation Analysis

[0127] 3.2.1 Overload rate change comparison

[0128] like Figure 8 The 19 cells during the peak period are shown in the normal condition (as shown by the black line in the figure), the moment before equilibrium (as shown by the blue line in the figure), and the moment after equilibrium (as shown by the red line in the figure). The number of overloaded users in these three states (hereinafter referred to as the three states) is random in the load conditions of different cells at the same time, which will lead to very different cell load numbers. Therefore, the resulting line graph is not smooth. When predicting the next moment, it is assumed that the number of users arriving at the cell follows the Poisson distribution. The number of users in the cell at the next moment is predicted by adding the number of new users minus the number of users leaving. Because the number of vehicles arriving at the cell during peak hours is often higher than the number of vehicles leaving the base station, the resulting line graph of the number of overloaded users in the cell is roughly Figure 8 state, and use this to calculate the overload rates of the 19 cells in the three states. Fig. 9 As shown, it is obvious that the overload rate of the cell after balancing is significantly reduced by 52.627% compared with before balancing. This phenomenon is largely caused by the use of the algorithm proposed in the present invention, thus verifying that the solution proposed in the present invention has good load balancing ability.

[0129] 3.2.2 Comparison of edge user rates

[0130] The algorithm proposed in this invention can not only reduce the overload rate of the cell, but also improve the user rate to a certain extent. The average rates of edge users in the three states are as follows: Fig.10 As shown in Figure 1, T1 before balancing, T2 before balancing, and T2 after balancing, a threshold is set for the user rate. When the user rate is lower than the assumed rate, that is, the user rate is less than 0.5Mbps, the cell is considered overloaded. The average rate of cell edge users in various states is as follows: Fig.11 As shown, after predicting the load status of the cell at the next moment and using CoMP technology to load balance the edge users, the edge users of the heavily loaded cell can be served by multiple adjacent lightly loaded cells at the same time, and the network rate between cells is effectively compensated, thereby increasing the rate of edge users. After load balancing using the algorithm proposed in the present invention, the average rate of the cell edge users is increased by 0.0367M / S.

[0131] 3.2.3 Comparison of user rates in different cells

[0132] Fig.12The figure shows the change of the average rate of users in each cell under three conditions. T1 represents the rate distribution curve of users in 19 cells under normal conditions during peak hours. In the simulation, the positions of users in the cells are set to be randomly evenly distributed. The rate distribution curves before and after the algorithm proposed by the present invention are shown as T2 before balancing and T2 after balancing, respectively. The average rate of users in all cells is obtained by counting Fig.13 By comparing the bar graph shown, it can be seen that after adopting the algorithm proposed in the present invention, the average rate of the cell users is increased by 0.0628M / S, thus verifying that the algorithm proposed in the present invention can not only improve the average rate of the cell edge users, but also improve the average rate of the entire cell users to a certain extent.

[0133] The above simulations are all hypothesis verifications during peak hours. During non-peak hours, the cell overload rate and user rate changes are similar to those during peak hours. The difference is that there are fewer vehicles during non-peak hours. Under the premise of low overload rate, the algorithm proposed in the present invention is used to reduce the overload rate to 0. However, because the algorithm proposed in the present invention is mainly aimed at load balancing, it makes reasonable allocation of resources between overloaded cells and lightly loaded cells to reduce the overload rate of each cell. At the same time, it also has a certain degree of impact on the improvement of user rate. After the edge users are balanced by CoMP technology, the overload rate of the cell changes significantly. However, during non-peak hours, the number of users distributed at the edge of the cell is relatively small. The balancing effect of the proposed algorithm also shows certain limitations. Therefore, during non-peak periods when the number of incoming vehicles is very small, the algorithm proposed in the present invention has very limited improvement on the edge and full cell rates. After simulation verification, the cell user rate was 0.9673M / S before balancing and 1.0564M / S after balancing, with an overall improvement of 0.0012M / S. The cell edge user rate was 0.6833M / S before balancing and 0.6845M / S after load balancing using the algorithm proposed in the present invention, with a relative improvement of 0.0023M / S. Although the effect is not as significant as during peak periods, it has also improved the overload problem of the base station to a certain extent.

[0134] 4. Summary

[0135] From the above, it can be seen that the present invention proposes a load balancing algorithm based on a prediction model for the load balancing problem between base stations. The prediction algorithm is combined with the CoMP technology to form a new load balancing algorithm to balance the load conditions between base stations and make effective use of resources. After simulation analysis and verification, after using the algorithm proposed by the present invention for load balancing, the overload rate of the base station has been significantly reduced, whether in peak or non-peak periods, and the average rate of users has also been improved to a certain extent, which improves the load balancing problem of the base station to a certain extent and facilitates the life of users. These works together provide a reference for the load balancing problem of the base station.

[0136] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions only describe the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention. These changes and improvements fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the attached claims and their equivalents.

Claims

1. A prediction-based dynamic load balancing method in the CoMP scenario of Internet of Vehicles, characterized in that: The following steps are involved: S1. Count the initial load status of each base station, which is divided into three states: light load, full load and overload. The lightly loaded base station is used as a cooperative base station for the fully loaded or overloaded base station; S2. Analyze the three different states of the base station as follows: For overloaded base stations: users are first diverted to nearby lightly loaded base stations until the overloaded base stations are balanced to full load or light load, and then the user location distribution is analyzed; For base stations in full or light load states: directly analyze the location distribution of users; Analyzing the location distribution of users means dividing users into central users and marginal users; S3. Central users are directly served by the base station. For edge users, predictions are made first and their load conditions are continuously counted. CoMP technology is then selectively used to provide services to users. S4. Determine the network rate of the edge user. If the network rate of the edge user exceeds the set threshold, it is considered that the base station reaches a load balancing state; S5. Repeat the above steps until the network rate of all users is not lower than the set threshold, and the number of users served by the base station of all base stations is not greater than the upper limit of the number of users that the base station can serve; In S3, prediction must be performed first for edge users. The prediction model is as follows: Assuming that the number of vehicles arriving at the base station follows a Poisson distribution with strength λ, the probability that k vehicles arrive at the base station at a certain moment is: Where λ is a time-related parameter; Assume that there are two cases for λ: during the rush hour, the value of λ is λ1, and during the off-peak hour, the value of λ is λ2, where λ1>λ2; The probability that K vehicles arrive at the base station is expressed as: Calculate the probability of the vehicle arriving at the base station and take P ( The K value when X=K) is the maximum is the predicted value of the number of vehicles arriving at the base station at the next moment; If the position and speed of the vehicle itself are known, the number of vehicles leaving the base station can be predicted. The number of vehicles at the base station at a certain moment can be predicted based on the arrival and departure of the vehicles. In S3, edge users must first be predicted and their load conditions continuously counted, and CoMP technology is selectively used to provide services to users, including: When a base station is initially in a light load state: When the predicted base station is in a light-load state, the CoMP technology is used to provide services to users; when the predicted base station is in a fully-loaded state, the CoMP technology is used to provide services to users together with the adjacent light-loaded base station; when the predicted base station is in an overloaded state, the load is re-distributed and balanced; When a base station is initially in full load state: When the predicted base station is in a light-load or full-load state, the CoMP technology is combined with the adjacent light-load base station to provide services to users at the same time; when the predicted base station is in an overload state, the load is re-diverted and balanced.

2. The prediction-based dynamic load balancing method in the CoMP scenario of the Internet of Vehicles according to claim 1 is characterized in that: The determination method of the lightly loaded, fully loaded and overloaded base stations is as follows: Set a threshold value A th As the upper limit of the number of users that the base station can serve, let the number of users served by the base station be A now ,but: When A now >A th , then the base station is considered to be in an overload state; when A now =A th , then the base station is considered to be in a fully loaded state; when A now th , it is determined that the base station is in a light load state.​ 3. The prediction-based dynamic load balancing method in the CoMP scenario of the Internet of Vehicles according to claim 1 is characterized in that: Among them, in S2, users are divided into central users and edge users in the following way: For each base station, 3r / 4 is used to divide the user location, which is greater than 3r / 4 is classified as marginal users, less than or equal to 3r / 4 is the central user, where r is the radius that a base station can cover.

Images