CN110891293A - Multi-attribute network selection method based on vehicle track prediction - Google Patents

Abstract

The invention discloses a multi-attribute network selection method based on vehicle trajectory prediction in the technical field of vehicle networking information transmission, and aims to solve the problem in the prior art that a vehicle-mounted mobile terminal can receive network signals covered by multiple different base stations. Medium vehicles have high mobility and fast moving speed. If the time to access a certain network is too short, it will easily cause frequent switching of the network and cause technical problems of unstable signal. The method includes the following steps: predicting the trajectory and speed of the vehicle; predicting the dwell time of the vehicle in different networks based on the trajectory and the speed; comparing the dwell time with a preset threshold, and extracting the dwell time greater than A network with a preset threshold is used as a candidate network; an access network is selected from the candidate networks.

Description

A multi-attribute network selection method based on vehicle trajectory prediction

technical field

The invention relates to a multi-attribute network selection method based on vehicle trajectory prediction, and belongs to the technical field of Internet of Vehicles information transmission.

Background technique

There are many different wireless communication technologies in the Internet of Vehicles, including wireless local area network (WLAN), 4/5G network, dedicated short-range communication technology (DSRC) and satellite communication network, etc. Different networks have different network attributes. The bandwidth and spectrum resources in the wireless network are limited. In hotspot areas covered by multiple wireless networks at the same time, the vehicle-mounted mobile terminal needs to select a suitable network according to the service requirements, in order to meet the premise of real-time information exchange and sharing of the vehicle-mounted mobile terminal in the Internet of Vehicles. In this way, the utilization rate of wireless resources is saved as much as possible, and the access efficiency of the wireless network is improved. The vehicle-mounted mobile terminal can receive network signals covered by multiple different base stations. Due to the high mobility and fast movement speed of vehicles in the Internet of Vehicles, if the time to access a certain network is too short, frequent network switching will occur, resulting in signal inconsistency. The problem of stability, the ping-pong effect.

SUMMARY OF THE INVENTION

In view of the deficiencies of the prior art, the purpose of the present invention is to provide a multi-attribute network selection method based on vehicle trajectory prediction, so as to solve the problem that the vehicle-mounted mobile terminal in the prior art can receive network signals covered by multiple different base stations. In the networked vehicles, the mobility is large and the movement speed is fast. If the time to access a certain network is too short, it is easy to cause frequent switching of the network, which will cause technical problems of unstable signal.

For solving the above-mentioned technical problems, the technical scheme adopted in the present invention is:

A multi-attribute network selection method based on vehicle trajectory prediction, comprising the following steps:

Predict the trajectory and speed of the vehicle;

based on the trajectory and the speed of the vehicle, predict the dwell time of the vehicle in different networks;

comparing the dwell time with a preset threshold, and extracting a network whose dwell time is greater than the preset threshold as a candidate network;

An access network is selected from the candidate networks.

Preferably, the trajectory and speed of the vehicle are predicted, including:

Based on the Markov chain, predict the trajectory of the vehicle;

Based on the predicted trajectory, predict the road segment that the vehicle will travel through;

obtain vehicle speed information on the road section;

Based on the vehicle speed information, the vehicle speed when the vehicle passes through the road section is predicted.

Preferably, based on the Markov chain, the trajectory of the vehicle is predicted, including:

Obtain the one-step transition probability of the vehicle at any intersection based on the historical trajectory data of the vehicle;

Based on the one-step transition probability, obtain the one-step transition probability matrix of the vehicle at all intersections;

Based on the one-step transition probability matrix, the trajectory of the vehicle is predicted.

Preferably, the one-step transition probability matrix, its expression is as follows:

in,

In the formula, P is the one-step transition probability matrix of the vehicle, M is the number of intersections, p _ij is the one-step transition probability from intersection i to intersection j, and N(i,j) is calculated based on historical trajectory data. The number of times of transferring to intersection j, N(i) is the number of times passing through intersection i based on the statistics of historical trajectory data.

Preferably, the trajectory of the vehicle is predicted based on the one-step transition probability matrix, including:

Obtain the row number i in the one-step transition probability matrix P based on the current position of the vehicle;

Compare the elements of the i-th row in the one-step transition probability matrix P, and extract the element with the largest one-step transition probability according to the comparison result;

Extract the column number of the element with the largest transition probability in one step as the current position of the vehicle;

Iterate the above process to obtain the trajectory of the vehicle.

Preferably, based on the trajectory and vehicle speed, the dwell time of the vehicle in different networks is predicted, including:

Obtaining coverage of different networks on the road section based on the base station attribute;

Based on the predicted vehicle speed when the vehicle passes through the road section and the acquired coverage of the different networks on the road section, the residence time of the vehicle in the different networks is obtained.

Preferably, selecting an access network from the candidate networks includes:

Selecting the attribute of the candidate network and obtaining the attribute vector of the candidate network, the attribute includes at least any two of transmission rate, transmission delay, transmission power, and interruption probability;

Based on the attribute vector, use AHP to obtain the weighted attribute vector of the candidate network;

Based on the weighted attribute vector, obtain the comprehensive index value of the candidate network;

The candidate network with the largest comprehensive index value is selected as the access network.

Preferably, based on the attribute vector, the AHP method is used to obtain the weighted attribute vector of the candidate network, including:

Define the relative importance of each two attributes in any candidate network;

Based on the relative importance and the attribute vector, the weighted attribute vector of the candidate network is obtained.

Preferably, based on the weighted attribute vector, the comprehensive index value of the candidate network is obtained, including:

Define the relative dependence of each two candidate networks on any attribute;

Based on the relative dependence degree and the weighted attribute vector, the comprehensive result matrix is obtained;

The comprehensive index value is the sum of all elements of the row vector of any row in the comprehensive result matrix.

Compared with the prior art, the present invention has the beneficial effects: the present invention adopts Markov chain to predict the vehicle trajectory, and then calculates the dwell time according to the predicted trajectory, and excludes the network whose dwell time is lower than the preset threshold. . The advantage of this approach is that networks that are prone to ping-pong effects are excluded in advance, and networks that do not produce ping-pong effects are selected in a centralized manner, which not only ensures the stability of subsequent network connections, but also improves network selection efficiency. Then, AHP is used to select the network. This method does not divide the influence of each attribute of the network on the final selection, and the weight of each layer will affect the final selection. It is ensured that the selected optimal network is not only better than the other candidate networks in one or several attributes, but is overall better than the rest of the candidate networks after weighted comparison in many aspects.

Description of drawings

1 is a schematic diagram of an intersection where a vehicle is located in an embodiment of the present invention;

2 is a schematic diagram of network coverage where a vehicle is located in an embodiment of the present invention;

3 is a schematic diagram of a hierarchical structure in an embodiment of the present invention;

FIG. 4 is a schematic flowchart of an embodiment of the present invention.

Detailed ways

The present invention will be further described below in conjunction with the accompanying drawings. The following examples are only used to illustrate the technical solutions of the present invention more clearly, and cannot be used to limit the protection scope of the present invention.

The specific embodiment of the present invention provides a multi-attribute network selection method based on vehicle trajectory prediction. As shown in FIG. 4, it is a schematic flowchart of an embodiment of the present invention. The method includes the following steps:

Step 1, trajectory prediction based on Markov chain

It mainly relies on the state transition probability matrix to predict the possibility of the vehicle to the next state position. The location of vehicles in the Internet of Vehicles is not discrete, but we can predict each intersection as a state point. As shown in FIG. 1 , it is a schematic diagram of the intersection where the vehicle is located in the embodiment of the present invention. The process of the vehicle traveling from the intersection X to the intersection O is called: the vehicle travels along the XO direction. If the vehicle is about to pass the intersection O at the next moment, the vehicle will have three driving possibilities next, namely: the OX direction, the OY direction and the OZ direction. The above processing can convert the trajectory prediction of the vehicle into a point-to-point problem of selecting different driving directions;

Assume that the position information {X _n , n∈N} of the vehicle passing through each intersection is a random sequence of discrete state space, n=1, 2,...,N. According to the properties of Markov chains, we can get:

P(X _n+1 |..., X _n-2 , X _n-1 , X _n )=P(X _n+1 |X _n ) (1)

In the formula, P(X _n+1 |X _n ) is the probability of transferring from X _n to X _n+1 , and the probability of the next state X _n+1 appearing is only related to the current state X _n , and has nothing to do with the previous state. . The one-step transition probability of the vehicle transitioning from state i to state j at time n is:

p _ij (n)=P(X _n+1 =j|X _n =i) (2)

In the formula, p _ij (n) is abbreviated as p _ij . According to the statistics of the historical trajectory data of a large number of vehicles moving, p _ij can be obtained. The specific process is as follows: Suppose N(i,j) is the number of times that intersection i turns to intersection j, and N(i) is the number of times passing a certain intersection i, Then there are:

If the two intersections are not connected, then p _ij =0. The matrix composed of all one-step transition probabilities P _ij is called the one-step transition probability matrix at a certain moment, denoted as P. Assuming that there are M intersections in total, the one-step transition probability matrix P is an M×M second-order matrix, and its expression is as follows:

For each vehicle, a one-step transition probability matrix can be derived from its historical trajectory. Scan the row number in the one-step transition probability matrix P according to the current position, compare the elements in the i-th row corresponding to the previous position, and obtain the column number j of the element with the largest one-step transition probability. The position corresponding to j is is the predicted position of the next state of the vehicle; then scan the row number in the one-step transition probability matrix P again according to the predicted position, and solve it iteratively to obtain the future movement trajectory of the vehicle.

Step 2: Estimation of vehicle residence time in the network

Since the trajectory of the vehicle at the next intersection can be predicted through the vehicle trajectory prediction, it can be considered that the road segment the vehicle will travel through is known; at the same time, the network coverage of the base station can be known according to the attributes of the base station, so that the network coverage of different networks on the road segment can be obtained. coverage. Then, according to the speed information reported by multiple vehicles ahead in the range of the Internet of Vehicles, the estimated speed of the road section ahead can be comprehensively calculated, so as to calculate the residence time of the vehicle in the network ahead.

In order to select an access network from multiple networks covering a road section, a threshold of dwell time can be preset. If the dwell time of a vehicle in a certain network coverage area is higher than the threshold, it means that the vehicle is in the network coverage area. If the stay time is long, it is not easy to produce ping-pong effect, and the network can be used as a candidate network; otherwise, it means that the vehicle stays in the coverage area of the network for a short time and is at the edge of the network, so the network is excluded. As shown in FIG. 2, it is a schematic diagram of the network coverage of the vehicle in the embodiment of the present invention. The vehicle is traveling on the XO road. It is assumed that the probability of the vehicle traveling from the intersection O to the intersection Y is the greatest after the trajectory prediction. The signal coverage of _F , _G , and _H , the lengths of the roads covered by the four networks are _LE , LF, LG, and LH respectively, and the estimated average speed of the OY section is V, then it can be calculated that the vehicle is in the above different networks. medium residence time, select the network with the residence time greater than the pre-threshold as the candidate network;

The dwell times of vehicles in networks E, F, G, and H are in sequence: L _E /V, L _F /V, L _G /V, and L _H /V.

Step 3: Determine the access network according to the Analytic Hierarchy Process (AHP)

For the candidate network, the optimal network can be finally accessed by comparing the attributes of different networks through AHP. First, the problem that affects network selection is decomposed into various attributes of the network, and then according to the network attributes and the degree of dependence of the network on a certain attribute, the relative importance of the attributes is determined by pairwise comparison, so that the degree of influence of each attribute on the network can be obtained. , find out the weight of each attribute in the network. Then, by comparing the degree of influence of each single attribute on different candidate networks, and giving a quantitative representation. Finally, the specific network attribute values are substituted into the calculation, and the optimal network among the candidate networks is obtained for access.

As shown in FIG. 3 , which is a schematic diagram of a hierarchical structure in an embodiment of the present invention, it is assumed that three networks E, F, and G are selected as candidate networks. The attributes that generally affect the network are transmission rate (A ₁ ), transmission delay (A ₂ ), transmit power (A ₃ ) and interruption probability (A ₄ ), among which transmission delay (A ₂ ) and interruption probability ( A ₄ ) is a negative effect, the lower the value, the better, so the attribute value is set as a negative number. The attribute values of networks E, F, and G are shown in Table 1;

Table 1: Network attribute values

网络E、F、G的属性向量依次为Define the attribute vector of the candidate network as

The attribute vectors of networks E, F, and G are in turn:

The steps to determine the optimal access network with different network attributes are as follows:

The first step is to construct the network attribute judgment matrix A. The scale definition of the network attribute judgment matrix is shown in Table 2, and the scale value a _ij is the relative importance of the attribute A _i compared to the attribute A _j . Assuming that the transmission delay A ₂ of the Internet of Vehicles access network is twice as important as the transmission rate A ₁ , there are:

a ₂₁ =2, a ₂₁ =1/2;

Table 2: Definition of Network Attribute Judgment Matrix Scale

According to the influence degree of the four attributes of the standard layer on the network access of the target layer in Fig. 3, a network attribute judgment matrix A=(a _ij ) _N×N , i,j=1,2,...N is established, then there are:

为网络属性判断矩阵A的标准化矩阵，

中的元素记为a′_ij，即：In the formula, N is the number of selected network attributes, and in this embodiment, N=4.

is the standardized matrix of the network attribute judgment matrix A,

The elements in are denoted as a′ _ij , namely:

The second step is to calculate and obtain the candidate network access feature vector W=(a ₁ ,a ₂ ,...,a _i ,...,a _N ), that is, the weight representation of each attribute in the network, the i-th element in W a _i is:

The third step is to calculate the maximum eigenvalue λ _max of the candidate network access feature vector W, and the formula is as follows:

In the formula, the vector AW is the product of the network attribute judgment matrix A and the candidate network access feature vector W, and (AW) _i is the ith element of the vector AW.

The fourth step is to perform a consistency check. The average random consistency index RI value is usually defined, as shown in Table 3;

Table 3: Consistency metrics

Order 1 2 3 4 5 6 7 8 9 10 RI value 0 0 0.52 0.89 1.12 1.26 1.36 1.41 1.46 0.46

where CI is the consistency index and CR is the consistency ratio. Consistency is satisfied if CR<0.1. If the consistency check is not met, the value of the network attribute judgment matrix needs to be adjusted until the consistency check is met.

The fifth step is to construct a network judgment matrix, the scale definition of which is shown in Table 4, and the scale value b _ij is the relative dependence degree of two networks on a certain attribute. Assuming that the network E is twice as dependent on the transmission rate as the network F, there are:

b _EF = 2, otherwise b _EF = 1/2;

Table 4: Definition of Network Judgment Matrix Scale

scale value b_ij of equal importance 1 slightly important 3 important 5 Very important 7 extremely important 9 Median 2,4,6,8

According to the dependence of the three candidate networks of the decision layer on the first attribute (transmission rate) of the standard layer in Fig. 3, a network judgment matrix B ₁ =(b _ij ) _3×3 for pairwise comparison is constructed, where i,j correspond to Indicates E, F, and G networks. Then, the eigenvector θ ₁ of the network judgment matrix B ₁ for the transmission rate is obtained. In the same way, according to the dependence of the three candidate networks on the other three attributes (transmission delay, transmission power, and interruption probability), the judgment matrices B ₂ , B ₃ , and B ₄ are constructed in turn, and the judgment matrices for the three eigenvectors θ ₂ , θ ₃ , θ ₄ of each attribute. Finally, the four eigenvectors are synthesized into a matrix R, that is, R=(θ ₁ , θ ₂ , θ ₃ , θ ₄ ) _3×4 .

In the sixth step, the consistency check is performed with reference to the third and fourth steps above. If each candidate network satisfies the consistency check, go to the seventh step; if not all of the candidate networks meet the consistency check, then adjust the scale definition of the network judgment matrix until all of them meet the consistency check.

The seventh step is to calculate the weighted attribute vectors δ _E , δ _F , δ _G , the expressions are as follows:

In the formula, * represents the Hadamard product. The matrix synthesized by δ _E , δ _F , and δ _G is K=(δ _E ,δ _F ,δ _G ) _4×3 .

The eighth step is to perform Hadamard product calculation on the transpose of matrix R and matrix K to obtain the comprehensive result matrix S. The expression is as follows:

S=R*K ^T (12)

Add all the elements of the row vector of each row of the matrix S to obtain the comprehensive index values SE, SF, and SG of the _E , _F , and _G networks in turn, and compare the three numerical values of _SE , _SF , and _SG . , the network corresponding to the maximum value is the optimal network.

The present invention adopts Markov chain to predict the vehicle trajectory, and then calculates the dwell time according to the predicted trajectory, and excludes the network whose dwell time is lower than the preset threshold. The advantage of this approach is that networks that are prone to ping-pong effects are excluded in advance, and networks that do not produce ping-pong effects are selected in a centralized manner, which not only ensures the stability of subsequent network connections, but also improves network selection efficiency. Then, AHP is used to select the network. This method does not divide the influence of each attribute of the network on the final selection, and the weight of each layer will affect the final selection. It is ensured that the selected network is not only better than the other candidate networks in one or several attributes, but is overall better than the rest of the candidate networks after weighted comparison in many aspects.

The above are only the preferred embodiments of the present invention. It should be pointed out that for those skilled in the art, without departing from the technical principle of the present invention, several improvements and modifications can also be made. These improvements and modifications It should also be regarded as the protection scope of the present invention.

Claims (9)

1. A multi-attribute network selection method based on vehicle track prediction is characterized by comprising the following steps:

predicting the track and the speed of the vehicle;

predicting residence time of the vehicle in different networks based on the track and the vehicle speed;

comparing the residence time with a preset threshold value, and extracting the network with the residence time larger than the preset threshold value as a candidate network;

and selecting an access network from the candidate networks.

2. The vehicle trajectory prediction-based multi-attribute network selection method of claim 1, wherein predicting the trajectory and speed of the vehicle comprises:

predicting a trajectory of the vehicle based on the Markov chain;

predicting a road section on which the vehicle is going to pass based on the predicted track;

acquiring vehicle speed information on the road section;

and predicting the speed of the vehicle when the vehicle passes through the road section based on the speed information.

3. The vehicle trajectory prediction-based multi-attribute network selection method of claim 2, wherein predicting the trajectory of the vehicle based on a markov chain comprises:

acquiring one-step transition probability of the vehicle at any intersection based on the historical track data of the vehicle;

acquiring a one-step transition probability matrix of the vehicle at all intersections based on the one-step transition probability;

and predicting the track of the vehicle based on the one-step transition probability matrix.

4. The vehicle trajectory prediction-based multi-attribute network selection method according to claim 3, wherein the expression of the one-step transition probability matrix is as follows:

wherein,

wherein P is a one-step transition probability matrix of the vehicle, M is the number of intersections, and P_ijThe transition probability of the intersection i to the intersection j is one step, N (i, j) is the number of times of transition from the intersection i to the intersection j counted based on the historical track data, and N (i) is the number of times of passing through the intersection i counted based on the historical track data.

5. The vehicle trajectory prediction-based multi-attribute network selection method of claim 4, wherein predicting the trajectory of the vehicle based on the one-step transition probability matrix comprises:

acquiring a row number i in the one-step transition probability matrix P based on the matching of the current position of the vehicle;

comparing the ith row elements in the one-step transition probability matrix P, and extracting the element with the maximum one-step transition probability according to the comparison result;

extracting the train number of the element with the maximum one-step transition probability as the current position of the vehicle;

and iterating the process to obtain the track of the vehicle.

6. The vehicle trajectory prediction-based multi-attribute network selection method of claim 2, wherein predicting vehicle residence times in different networks based on the trajectory and vehicle speed comprises:

acquiring coverage ranges of different networks on the road section based on the base station attribute;

and acquiring the residence time of the vehicle in different networks based on the predicted speed of the vehicle passing through the road section and the acquired coverage areas of the different networks on the road section.

7. The vehicle trajectory prediction-based multi-attribute network selection method according to any one of claims 1 to 6, wherein selecting an access network from the candidate networks comprises:

selecting the attributes of the candidate network and obtaining the attribute vector of the candidate network, wherein the attributes comprise at least any two items of transmission rate, transmission time delay, transmission power and interruption probability;

based on the attribute vector, calculating a weighted attribute vector of the candidate network by using an analytic hierarchy process;

based on the weighted attribute vector, calculating a comprehensive index value of the candidate network;

and selecting the candidate network with the maximum comprehensive index value as an access network.

8. The method of claim 7, wherein the step of using an analytic hierarchy process to find the weighted attribute vector of the candidate network based on the attribute vector comprises:

defining the relative importance degree of every two attributes in any candidate network;

and calculating the weighted attribute vector of the candidate network based on the relative importance degree and the attribute vector.

9. The method of claim 7, wherein the obtaining a composite index value for the candidate network based on the weighted attribute vector comprises:

defining the relative dependence degree of each two candidate networks on any attribute;

solving a comprehensive result matrix based on the relative dependence degree and the weighted attribute vector;

and the comprehensive index value is the sum of all elements of the row vector of any row in the comprehensive result matrix.

Images