A method for collaborative task offloading in vehicle edge networks

Abstract

The present invention discloses a method for collaborative task unloading in a vehicle edge network, including: obtaining vehicle information, base station information and task information on the current road; determining the best vehicle cloud and the best vehicle cloud leader under each base station; determining the best transmission path from the task vehicle to the processing base station and the best vehicle cloud leader under the base station; establishing a calculation model for transmission and processing delay and a calculation model for transmission and processing energy consumption respectively; and taking the minimization of transmission and processing delay and energy consumption as optimization goals, determining the proportion of tasks assigned to the best vehicle cloud leader, the proportion of tasks assigned to the task vehicle itself and the proportion of tasks assigned to the processing base station; transmitting the tasks to the processing base station and the best vehicle cloud leader through the best transmission path according to the task proportion allocation result; the best vehicle cloud leader transmits the tasks to other vehicle cloud members, and the processing base station transmits the tasks to other base stations; after the tasks are processed, returning the results to the task vehicle.

Description

Technical Field

[0001] The present invention belongs to the technical field of vehicle task offloading, and in particular relates to a method for collaborative task offloading in a vehicle edge network. Background Art

[0002] In the current era, the emergence of intelligent vehicles has led to the emergence of various vehicle applications. These applications usually generate various delay-sensitive computing tasks, such as monitoring the position of vehicles on the road, analyzing traffic information, etc. Traditional task offloading methods are not enough to support the high timeliness of these applications. In the vehicle network system, vehicle information is shared, and more vehicles choose to cooperate to process tasks. There are many ways to coordinate task offloading, but it is difficult for them to integrate the load balancing between the local, strong computing power base station and the highly mobile vehicle cloud. As time changes, the dynamic nature of the entire vehicle network cannot be ignored. Among them, the construction of the vehicle cloud also generates many related dynamic construction problems. In previous studies, most of the vehicle clouds are based on static vehicles to build and participate in task offloading, while ignoring that dynamic vehicles with powerful computing resources can also join the construction of the vehicle cloud, and the offloading efficiency is low. In addition, in previous studies, the local processing center of the vehicle is mainly reserved for its own high-priority tasks, but this undoubtedly wastes a lot of vehicle computing resources. Summary of the invention

[0003] The purpose of the present invention is to provide a method for collaborative task offloading in a vehicle edge network, which adds local computing resources to the collaborative offloading of tasks, can increase the availability of a vehicle group, and achieve efficient task offloading.

[0004] The technical solution provided by the present invention is:

[0005] A method for collaborative task offloading in a vehicle edge network comprises the following steps:

[0006] Step 1: Obtain vehicle information, base station information, and task information on the current road; determine the best vehicle cloud and best vehicle cloud leader under each base station;

[0007] Step 2: determine the best transmission path from the task vehicle to the processing base station and the best vehicle-cloud leader under the base station;

[0008] Wherein, the task vehicle is a vehicle that generates a task, and the processing base station is a subordinate base station of the task vehicle;

[0009] Step 3: Establish the calculation model of transmission and processing delay and the calculation model of transmission and processing energy consumption respectively; and take the minimum transmission and processing delay and energy consumption as the optimization goal, determine the proportion of tasks assigned to the best vehicle-cloud leader, the proportion of tasks assigned to the task vehicle itself, and the proportion of tasks assigned to the processing base station;

[0010] Step 4: According to the task proportion allocation result, the task is transmitted to the processing base station and the best car cloud leader through the best transmission path; the best car cloud leader transmits the task to other car cloud members, and the processing base station transmits the task to other base stations;

[0011] Step 5: The vehicle cloud members and the base stations assigned to the tasks process the tasks respectively and return the results to the task vehicles.

[0012] Preferably, in the step 1, determining the best vehicle cloud and the best vehicle cloud leader under each base station by using a k-means clustering algorithm comprises the following steps:

[0013] Step 1: Randomly determine K vehicles as K cluster centers;

[0014] Step 2: Calculate the comprehensive similarity between each vehicle and each cluster center, and add the vehicle to the cluster center cluster with the largest comprehensive similarity;

[0015] Step 3: Calculate the average similarity of all vehicles in the cluster and the vehicle serving as the cluster center, and respectively calculate the difference between the average similarity and the comprehensive similarity corresponding to each vehicle, take the vehicle corresponding to the minimum difference as the new cluster center, and determine whether the new and old cluster centers have changed;

[0016] If the cluster center remains unchanged, the cluster center is used as the candidate car cloud leader, and the corresponding cluster is used as the candidate car cloud; if the cluster center changes, steps 2-3 are repeated;

[0017] Step 4: Calculate the average similarity between all vehicles in each candidate vehicle cloud and the task vehicle, take the vehicle cloud with the highest average similarity as the best vehicle cloud, and take the cluster center of the best vehicle cloud as the best vehicle cloud leader.

[0018] Preferably, the comprehensive similarity is calculated as follows:

[0019]

[0020] Among them, sim i,k is the comprehensive similarity between the vehicle and the cluster center, is the position similarity between the vehicle and the cluster center, is the speed similarity between the vehicle and the cluster center, is the speed direction similarity between the vehicle and the cluster center; β1, β2, and β3 represent the exponential proportion of each similarity.

[0021] Preferably, in step 2, the method for determining the best path from the task vehicle to the processing base station is:

[0022] If the task vehicle is far away from the processing base station and cannot send signals directly to the processing base station, the vehicle closest to the processing base station is taken as the destination vehicle. After determining the best path between the task vehicle and the destination vehicle, the task vehicle transmits the signal to the destination vehicle via the best path, and the destination vehicle transmits the signal to the processing base station.

[0023] Preferably, in step 2, the method for determining the best path from the task vehicle to the best vehicle cloud leader is:

[0024] Step A: All vehicles under the same base station are taken as vertices to form a set V, and the channels between vehicles that can directly communicate are taken as edges to construct a weighted undirected graph, and the weight of the edge connecting two points is represented by the communication time of the two vehicles;

[0025] Step B: taking the task vehicle as the source point s, marking the point s, and taking the best vehicle cloud leader as the destination point t;

[0026] Use point k to represent the current point s;

[0027] Step C, determine the connection weights of all points that can directly communicate with point k, and select point j from the set J of all points that can directly communicate with point k, and mark point j;

[0028] Among them, the points that have been marked will no longer be selected to enter the set J;

[0029] Point j satisfies: d(s, j) = min{d(s, p)}(p∈J);

[0030] d(s,p) satisfies: d(s,p)=min{d(s,p),d(s,k)+w(k,p)};

[0031] Where w(k,p) represents the weight of the edge between points k and p that can communicate directly, d(s,k) represents the sum of the weights corresponding to the current shortest path between points s and k; d(s,p) represents the sum of the weights corresponding to the current shortest path between points s and p;

[0032] Step D: point j is taken as the new point k, and step C is repeated until all points are marked;

[0033] Step E: Take the first point on the path from s to t corresponding to d(s,t), record it as relay node m, and determine whether point m is the destination point t;

[0034] If not, add relay point m to the target path, update the state information of the weighted undirected graph after the task is transmitted to point m, and then repeat steps B to E with relay point m as the new task vehicle;

[0035] If point m is the destination point t, the target path is obtained.

[0036] Preferably, in step 3, the calculation model of the transmission and processing delay is:

[0037]

[0038] Among them, θ l,-1 Represents the task φ l The proportion of tasks assigned to the car-cloud, θ l,0 Represents the task φ l The proportion of tasks assigned to the task vehicle, i.e., itself, θ l,j Represents the task φ l Assigned to processing base station M j The proportion of tasks; S l,k and C l,k Respectively represent the task φ l Assign to Car Cloud l Member Vehicle V k The amount of tasks and the required CPU computing power; S l Represents the task φ l The scale of C l Represents the task φ l Required CPU computing power; Indicates the delay required for unit task transmission between base stations; F l Represents vehicle V l That is, its own CPU computing power, f j Represents base station M j CPU computing power, F k Representative of Cheyunzhong member V k CPU computing power, Represents mission vehicle V l The communication rate to the upper base station, Represents mission vehicle V l The communication rate to the leader of its selected car cloud, On behalf of the leader in car cloud V l′ To Cheyun member V k The communication rate, j max Indicates the number of base stations on the current road.

[0039] Preferably, in step 3, the calculation model of the transmission and processing energy consumption is:

[0040]

[0041] Among them, θ l,-1 Represents the task φ l The proportion of tasks assigned to the car-cloud, θ l,0 Represents the task φ l The proportion of tasks assigned to the local area, θ l,j Represents the task φ l Assigned to base station M j The proportion of tasks represents the energy consumption of the unit task transmitted between base stations, ξ l and χ l Represents two coefficients related to the task vehicle itself and the CPU, ξ k and χ k Indicates the two coefficients related to the vehicle in the car cloud and its own CPU, and Indicates base station M j Its own CPU related coefficient; S l,k and C l,k Respectively represent the task φ l Assign to Car Cloud l Member Vehicle V k The amount of tasks and the required CPU computing power; l Indicates vehicle V l The transmission power, P l′ Indicates the leader of car cloud V l′ transmission power.

[0042] Preferably, in step three, a differential evolution algorithm is used to obtain the proportion of tasks assigned to the best vehicle cloud corresponding to the task, the proportion of tasks assigned to the task vehicle itself, and the proportion of tasks assigned to each processing base station through multiple iterations of cross genetics.

[0043] The beneficial effects of the present invention are:

[0044] The method for collaborative task offloading in a vehicle edge network provided by the present invention adds local computing resources to the collaborative offloading of tasks, which can increase the availability of the vehicle group and achieve efficient task offloading; it can also reduce the time delay and energy consumption caused by the transmission of various tasks and increase the information resource sharing of the entire vehicle system. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 A schematic diagram of the vehicle task collaborative unloading process according to the present invention.

[0046] Figure 2 This is a flow chart of the method for collaborative task offloading in a vehicle edge network according to the present invention.

[0047] Figure 3 A flowchart for selecting the best car cloud and the best car cloud leader according to the present invention.

[0048] Figure 4 This is a flow chart of determining the optimal transmission path according to the present invention.

[0049] Figure 5 This is a flow chart of obtaining task proportions through the differential evolution algorithm described in the present invention. DETAILED DESCRIPTION

[0050] The present invention is further described in detail below in conjunction with the accompanying drawings so that those skilled in the art can implement the invention with reference to the description.

[0051] In real life, when vehicles are driving on the road, they may generate various tasks. When the vehicle's own computing center (CPU) is busy, the task can be handed over to other vehicles or base stations for processing. Since it may not be particularly fast for one vehicle to process this task, the present invention considers allowing a group of vehicles, base stations and vehicles that generate tasks to perform collaborative processing. However, if this group of vehicles is selected arbitrarily, it may cause longer delays or higher energy consumption, which is not worth the loss. Therefore, in the present invention, vehicles with some of the same characteristics are classified into one category, and they form a car cloud to process tasks.

[0052] The present invention provides a method for collaborative task unloading in a vehicle edge network. When a vehicle generates a task, the vehicle notifies its upper base station, and its upper base station selects some vehicles as a vehicle cloud for it through the k-means algorithm, and determines its vehicle cloud leader. After the task vehicle learns the message sent by the base station, it transmits the task to the corresponding best vehicle cloud leader through the best transmission path. The best vehicle cloud leader allocates tasks according to the idle resources of the vehicle cloud members, and transmits the tasks to the vehicle cloud members according to the allocation results. In addition, the task can also be transmitted to the upper base station of the task vehicle through the best transmission path for processing, and the upper base station can also transmit the task to other base stations for collaborative processing of the task. As for the return of the result after the task processing is completed, because the processing result of the task is generally small, the time and energy consumption are also small and can be ignored.

[0053] like Figure 1-2 As shown, the method for collaborative task offloading in a vehicle edge network provided by the present invention has the following specific implementation process.

[0054] Get the basic information of all vehicles on the current road i ,vi ,φ i ,F i}, respectively representing vehicle V i The location coordinates, speed, mission information and computing power provided by the vehicle CPU; obtain basic information of all base stations Represents base station M j The computing power, location coordinates and communication range that the CPU can provide; and obtain basic information of all tasks {S l ,t l ,C l}, to represent the task φ l The scale, the maximum tolerable delay, and the CPU computing power required for the task are considered. After that, the best car cloud (member) and the best car cloud leader under each base station are determined.

[0055] In this embodiment, by using the vehicle position similarity Speed ​​(size) similarity Similarity to speed direction The comprehensive similarity sim i,k The standard k-means clustering algorithm is used to obtain the best car cloud members and car cloud leaders under each base station. Figure 3 As shown, the specific process includes:

[0056] (1) Randomly determine K vehicles as K cluster centers;

[0057] (2) Calculate the comprehensive similarity between each vehicle and each cluster center, and add the vehicle to the cluster center cluster with the largest comprehensive similarity;

[0058] (3) Calculating the average similarity of all vehicles in the cluster and the vehicle serving as the cluster center, and respectively calculating the difference between the average similarity and the comprehensive similarity corresponding to each vehicle, taking the vehicle corresponding to the minimum difference as the new cluster center, and determining whether the new and old cluster centers have changed;

[0059] If the cluster center remains unchanged, the cluster center is used as the candidate car cloud leader and its corresponding cluster is used as the candidate car cloud; if the cluster center changes, steps (2)-(3) are repeated;

[0060] (4) Calculate the average similarity between all vehicles in each candidate vehicle cloud and the task vehicle, take the vehicle cloud with the highest average similarity as the best vehicle cloud, and take the cluster center of the best vehicle cloud as the best vehicle cloud leader.

[0061] Comprehensive similarity sim i,k The calculation method is as follows:

[0062]

[0063] in, is the location similarity, is the speed similarity, is the velocity direction similarity;

[0064]

[0065]

[0066]

[0067]

[0068]

[0069] Among them, β1, β2, and β3 represent the index proportion of each similarity. Each proportion can be determined according to the density of vehicles on the road at that time. When the vehicles on the road are denser, the corresponding speed similarity will be higher, so we need to convert the position similarity The exponent β1 is set higher, otherwise β1, β2, and β3 are set to be balanced. i,k Indicates V i and V k The Euclidean distance of Indicates M j The collection of all vehicles under . (x i ,y i ) and (x k ,y k ) indicates vehicle V i and V k The position coordinates, v i and v k Indicates vehicle V i and V k The velocity vector, |v i The value of | is the velocity vector v i The value range of β1, β2, and β3 is (0, 1), that is, β1, β2, and β3∈(0, 1).

[0070] Determine the optimal transmission path from the task vehicle to the processing base station and the best car cloud leader under the base station. If the task vehicle can directly transmit information to the processing base station (the upper base station of the task vehicle), the task is directly passed to the processing base station. If the task vehicle is far away from the processing base station (the upper base station of the task vehicle) and cannot directly transmit information to the processing base station, the vehicle closest to the processing base station is used as the destination vehicle. After determining the optimal path between the task vehicle and the destination vehicle, the task vehicle transmits the signal to the destination vehicle through the optimal path, and the destination vehicle transmits the signal to the processing base station.

[0071] In this embodiment, by using the transmission time as the weight W a,b The shortest path algorithm obtains the optimal transmission path of the destination vehicle or the best car-cloud leader.

[0072] like Figure 4 As shown in Figure 2, the process of determining the best transmission path is:

[0073] Step A: All vehicles under the same base station are taken as vertices to form a set V, and the channels between vehicles that can directly communicate are taken as edges to construct a weighted undirected graph, and the weight of the edge connecting two points is represented by the communication time of the two vehicles;

[0074] Step B: taking the task vehicle as the source point s, marking the point s, and taking the best vehicle cloud leader as the destination point t;

[0075] Use point k to represent the current point s;

[0076] Step C, determine the connection weights of all points that can directly communicate with point k, and select point j from the set J of all points that can directly communicate with point k, and mark point j;

[0077] Among them, the marked points will not be selected to enter the set J; point j satisfies: d(s, j) = min{d(s, p)}(p∈J);

[0078] d(s, p) satisfies: d(s, p) = min{d(s, p), d(s, k) + w(k, p)};

[0079] Where w(k, p) represents the weight of the edge between points k and p that can communicate directly, d(s, k) represents the sum of the weights corresponding to the current shortest path between points s and k without direct communication restrictions; d(s, p) represents the sum of the weights corresponding to the current shortest path between points s and p.

[0080] Initial regulations: d(a, b) = +∞(a∈V, b∈V, a≠b), d(a, a) = 0(a∈V), w(a, b) is the weight of the edge connecting points a and b, and a and b are any two points in the graph.

[0081] Step D: point j is taken as the new point k, and step C is repeated until all points are marked;

[0082] Step E: Take the first point on the path from s to t corresponding to d(s, t), record it as relay node m, and determine whether point m is the destination point t;

[0083] If not, add relay point m to the target path, update the state information of the weighted undirected graph after the task is transmitted to point m, and then repeat steps B to E with relay point m as the new task vehicle;

[0084] If point m is the destination point t, the target path can be obtained.

[0085] The maximum distance over which two infrastructures can communicate without distortion is γ0 represents the signal-to-interference-noise ratio threshold at which signal transmission is not distorted.

[0086]

[0087]

[0088] After the optimal path of task transmission, the differential evolution algorithm is used to solve the task φ l Transmission and processing delay∑ l∈φ T l Energy consumption∑ l∈φ E l Minimize the optimal task allocation weight θ l .

[0089]

[0090]

[0091]

[0092]

[0093]

[0094] Among them, θ l,-1 Represents the task φ l The proportion of tasks assigned to the car-cloud, θ l,0 Represents the task φ l The proportion of tasks assigned to the local area, θ l,j Represents the task φ l Assigned to base station M j When calculating the task transmission rate R at the sending and receiving ends i,jAccording to Shannon's theorem, we need to first calculate the signal-to-interference-plus-noise ratio (SINR) γ during the transmission process. i,j . Among them B i,j represents the channel bandwidth between the sender and the receiver, P i Indicates the transmit power of the sender, H i,j represents the channel gain between the sender and the receiver, c represents the speed of light, G i,j Indicates the carrier frequency of the sender, d i,j represents the Euclidean distance between the sender and the receiver, τ represents the loss coefficient associated with the path, and αk i,j z i,j B i,j represents the ambient noise power, α represents the Boltzmann constant, k i,j represents the relative Kelvin temperature of the receiver, z i,j represents the noise index of the receiver, Indicates the signal interference power generated by other vehicles.

[0095] Then calculate the corresponding single task φ l The resulting delay T l and energy consumption E l ,in Indicates the delay required for unit tasks to be transmitted between base stations. represents the energy consumption of the unit task transmitted between base stations, ξ l and χ l Represents two coefficients related to the CPU of the processing center. l,k and C l,k Respectively represent the task φ l Assign to Car Cloud l Member Vehicle V k When obtaining the delay corresponding to a solution in the solution space, since the assigned tasks can be sent in parallel, the maximum value of the delay assigned to each processing center represents the task φ l When solving the energy consumption, it is necessary to find the sum of the energy generated by each transmission and processing. l∈φ T l and energy consumption∑ l∈φ E l The minimization of the two objective functions is the convergence condition, and the differential evolution algorithm is used to iterate multiple times through crossover genetics to reach the optimal solution.

[0096] like Figure 5 As shown, the process of the differential evolution algorithm is:

[0097] (1) Randomly initialize N groups of task allocation strategies to form an initialization population X. Each group of strategies specifies a mutation factor S and a crossover probability factor K for a single individual.

[0098] (2) Initialize the population V, V = X, and update each individual V in V i =V l +S×(V m -V n )(V l ,V m ,V n Randomly selected, and l,m,n≠i);

[0099] (3) For each set of genes of all individuals in population V, a random value is given. Compared with the crossover probability factor, if the given random value is greater than K, the genome of population V is replaced with the genome of population X, otherwise it remains unchanged;

[0100] (4) Calculate the fitness function of population V and population X, that is, calculate the processing delay and energy consumption corresponding to each individual strategy in population V and population X. The processing delay and energy consumption are linearly weighted to obtain the comprehensive consumption of each individual in each population. Compare the sum of the comprehensive consumption of all individuals in population X and population V, and select the population with the lowest comprehensive consumption as the new population X;

[0101] (5) Determine whether the difference in comprehensive consumption between the new and old populations X is less than a threshold. If so, go to (6); otherwise, go to (2);

[0102] (6) Calculate the comprehensive consumption of each individual in population X, and take the individual with the smallest comprehensive consumption as the optimal task allocation strategy.

[0103] According to the task proportion allocation result, the task is transmitted to the processing base station and the best car cloud leader through the best transmission path; the best car cloud leader transmits the task to other car cloud members, and the processing base station transmits the task to other base stations;

[0104] The vehicle cloud members and the base stations assigned to the tasks process the tasks respectively and return the results to the task vehicles.

[0105] Although the embodiments of the present invention have been disclosed as above, they are not limited to the applications listed in the specification and the implementation modes, and they can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and the illustrations shown and described herein.

Claims

1. A method for collaborative task offloading in a vehicle edge network, characterized in that: The steps include: Step 1: Obtain vehicle information, base station information, and task information on the current road; determine the best vehicle cloud and best vehicle cloud leader under each base station; Step 2: determine the best transmission path from the task vehicle to the processing base station and the best vehicle-cloud leader under the base station; Wherein, the task vehicle is a vehicle that generates a task, and the processing base station is a subordinate base station of the task vehicle; Step 3: Establish the calculation model of transmission and processing delay and the calculation model of transmission and processing energy consumption respectively; and take the minimum transmission and processing delay and energy consumption as the optimization goal, determine the proportion of tasks assigned to the best car-cloud leader, the proportion of tasks assigned to the task vehicle itself, and the proportion of tasks assigned to the processing base station; Step 4: According to the task proportion allocation result, the task is transmitted to the processing base station and the best car cloud leader through the best transmission path; the best car cloud leader transmits the task to other car cloud members, and the processing base station transmits the task to other base stations; Step 5: The vehicle cloud members and the base station assigned to the task process the task respectively and return the result to the task vehicle; In step 2, the method for determining the best path from the task vehicle to the best vehicle cloud leader is: Step A: All vehicles under the same base station are taken as vertices to form a set V, and the channels between vehicles that can directly communicate are taken as edges to construct a weighted undirected graph, and the weight of the edge connecting two points is represented by the communication time of the two vehicles; Step B: taking the task vehicle as the source point s, marking the point s, and taking the best vehicle cloud leader as the destination point t; Use point k to represent the current point s; Step C, determine the connection weights of all points that can directly communicate with point k, and select point j from the set J of all points that can directly communicate with point k, and mark point j; Among them, the points that have been marked will no longer be selected to enter the set J; Point j satisfies: d(s, j) = min{d(s, p)}(p∈J); d(s,p) satisfies: d(s,p)=min{d(s,p),d(s,k)+w(k,p)}; Where w(k,p) represents the weight of the edge between points k and p that can communicate directly, d(s,k) represents the sum of the weights corresponding to the current shortest path between points s and k; d(s,p) represents the sum of the weights corresponding to the current shortest path between points s and p; Step D: point j is taken as the new point k, and step C is repeated until all points are marked; Step E: Take the first point on the path from s to t corresponding to d(s,t), record it as relay node m, and determine whether point m is the destination point t; If not, add relay point m to the target path, update the state information of the weighted undirected graph after the task is transmitted to point m, and then repeat steps B to E with relay point m as the new task vehicle; If point m is the destination point t, the target path is obtained.

2. The method for collaborative offloading of tasks in a vehicle edge network according to claim 1, characterized in that: In the step 1, the best car cloud and the best car cloud leader under each base station are determined by the k-means clustering algorithm, including the following steps: Step 1: Randomly determine K vehicles as K cluster centers; Step 2: Calculate the comprehensive similarity between each vehicle and each cluster center, and add the vehicle to the cluster center cluster with the largest comprehensive similarity; Step 3: Calculate the average similarity of all vehicles in the cluster and the vehicle serving as the cluster center, and respectively calculate the difference between the average similarity and the comprehensive similarity corresponding to each vehicle, take the vehicle corresponding to the minimum difference as the new cluster center, and determine whether the new and old cluster centers have changed; If the cluster center remains unchanged, the cluster center is used as the candidate car cloud leader, and the corresponding cluster is used as the candidate car cloud; if the cluster center changes, steps 2-3 are repeated; Step 4: Calculate the average similarity between all vehicles in each candidate vehicle cloud and the task vehicle, take the vehicle cloud with the highest average similarity as the best vehicle cloud, and take the cluster center of the best vehicle cloud as the best vehicle cloud leader.

3. The method for collaborative offloading of tasks in a vehicle edge network according to claim 2, characterized in that: The calculation method of the comprehensive similarity is: Among them, sim i,k is the comprehensive similarity between the vehicle and the cluster center, is the position similarity between the vehicle and the cluster center, is the speed similarity between the vehicle and the cluster center, is the speed direction similarity between the vehicle and the cluster center; β1, β2, and β3 represent the exponential proportion of each similarity.

4. The method for collaborative offloading of tasks in a vehicle edge network according to claim 3, characterized in that: In step 2, the method for determining the best path from the task vehicle to the processing base station is: If the task vehicle is far away from the processing base station and cannot send signals directly to the processing base station, the vehicle closest to the processing base station is taken as the destination vehicle. After determining the best path between the task vehicle and the destination vehicle, the task vehicle transmits the signal to the destination vehicle via the best path, and the destination vehicle transmits the signal to the processing base station.

5. The method for collaborative offloading of tasks in a vehicle edge network according to claim 1 or 4, characterized in that: In step 3, the calculation model of the transmission and processing delay is: Among them, θ l,-1 Represents the task φ l The proportion of tasks assigned to the car-cloud, θ l,0 Represents the task φ l The proportion of tasks assigned to the task vehicle, i.e., itself, θ l,j Represents the task φ l Assigned to processing base station M j The proportion of tasks; S l,k and C l,k Respectively represent the task φ l Assigned to Cheyun l Member Vehicle V k The amount of tasks and the required CPU computing power; S l Represents the task φ l The scale of C l Represents the task φ l Required CPU computing power; Indicates the delay required for unit task transmission between base stations; F l Represents vehicle V l That is, its own CPU computing power, f j Represents base station M j CPU computing power, F k Representative of Cheyunzhong member V k CPU computing power, Represents mission vehicle V l The communication rate to the upper base station, Represents mission vehicle V l The communication rate to the leader of its selected car cloud, On behalf of the leader in car cloud V l′ To Cheyun member V k The communication rate, j max Indicates the number of base stations on the current road.

6. The method for collaborative offloading of tasks in a vehicle edge network according to claim 5, characterized in that: In step 3, the calculation model of the transmission and processing energy consumption is: Among them, θ l,-1 Represents the task φ l The proportion of tasks assigned to the car-cloud, θ l,0 Represents the task φ l The proportion of tasks assigned to the local area, θ l,j Represents the task φ l Assigned to base station M j The proportion of tasks represents the energy consumption of the unit task transmitted between base stations, ξ l and χ l Represents two coefficients related to the task vehicle itself and the CPU, ξ k and χ k Indicates the two coefficients related to the vehicle in the car cloud and its own CPU, and Indicates base station M j Its own CPU related coefficient; S l,k and C l,k Respectively represent the task φ l Assigned to Cheyun l Member Vehicle V k The amount of tasks and the required CPU computing power; l Indicates vehicle V l The transmission power, P l ′ Indicates the leader of car cloud V l ′ transmission power.

7. The method for collaborative offloading of tasks in a vehicle edge network according to claim 6, characterized in that: In the step three, a differential evolution algorithm is used to obtain the task ratio assigned to the best vehicle cloud corresponding to the task, the task ratio assigned to the task vehicle itself, and the task ratio assigned to each processing base station through multiple iterations of crossover genetics.

Images