A resource allocation method based on MEC computing task offloading in Internet of Vehicles

Abstract

This invention relates to the field of task offloading and resource optimization in vehicle-to-everything (V2X) networks, and particularly to a resource optimization method for task offloading based on MEC (Multi-access Edge Computing) in V2X networks. With the assistance of mobile edge computing technology, this invention addresses the problem of V2X networks needing to handle a large number of low-latency tasks under limited resources and dynamic topologies. It studies a task offloading and resource allocation strategy for V2X networks that minimizes system overhead. A mathematical model is established considering the differences between roadside units and vehicles as service nodes, as well as constraints on task latency, communication distance, and computing resources. This mixed-integer non-convex problem is decomposed into three sub-problems for joint solution. By using variable substitution, the computing resource allocation sub-problem is transformed into a convex optimization problem to obtain the optimal offloading ratio. This invention can effectively determine offloading decisions, allocate channel resources and computing resources, and reduce system overhead.

Description

Technical Field

[0001] This invention relates to the field of task offloading and resource optimization in vehicle-to-everything (V2X) networks, and particularly to a resource allocation method based on MEC (Multi-access Edge Computing) task offloading in V2X networks. Background Technology

[0002] The continuous advancement of next-generation wireless and industrial technologies has provided essential technical support for the development of the Internet of Vehicles (IoV). Not only has the number of vehicles increased significantly, but vehicle performance has also been gradually optimized to become more intelligent. Emerging automotive applications are emerging, such as cooperative autonomous driving, intelligent traffic control, and cooperative environmental perception. Consequently, the IoV will face the challenge of processing massive amounts of data under resource constraints, high latency requirements, and topology changes.

[0003] Mobile Edge Computing (MEC) distributes the computing platform from the mobile core network to the edge of the mobile access network, reducing end-to-end latency in mobile service delivery and enhancing user experience. Cellular-vehicle-to-everything (C-V2X) technology relies on cellular networks to enable vehicles to communicate with their surroundings. Combining MEC and C-V2X technologies, tasks can be offloaded to roadside units (RSUs) via vehicle-to-infrastructure (V2I) communication or to other vehicles via vehicle-to-vehicle (V2V) communication, effectively utilizing idle resources and reducing node load.

[0004] Currently, existing technologies have yielded many valuable research results on task offloading and resource allocation in conjunction with MEC in vehicle-to-everything (V2X) networks. However, few studies have explored partial offloading under V2X offloading. Furthermore, most solutions do not consider the differences between RSUs and different vehicles as service nodes, or only consider whether a task is offloaded without studying the specific offloading targets. In addition, some solutions do not consider the limitations of vehicle mobility and communication distance, as well as resource allocation after offloading.

[0005] Therefore, under the constraints of real-world scenarios, a well-designed offloading strategy combined with reasonable resource allocation can reduce the latency of task computation and balance the energy consumption of each node in the system. Summary of the Invention

[0006] To address the aforementioned problems in existing technologies, this invention proposes a resource allocation method for offloading computing tasks based on MEC (Multi-access Edge Computing) in vehicle-to-everything (V2X) networks. With the assistance of mobile edge computing technology, this method addresses the issue of V2X networks needing to handle a large number of low-latency tasks under limited resources and dynamic topologies. Based on the principle of minimizing system overhead, it determines a partial task offloading and resource allocation strategy for V2X networks, thereby solving the task offloading and resource optimization problem in the vehicle edge computing system. The method determines the task offloading strategy by calculating the minimum total overhead in the V2X system and allocates system resources according to the task offloading strategy, including:

[0007] S1: Construct a resource allocation framework based on MEC computing task offloading;

[0008] S2: Determine the unloading mode of the task vehicle based on the resource allocation framework for MEC-based task unloading. The unloading modes include unloading the task vehicle to the service vehicle and unloading the task vehicle to the roadside unit.

[0009] S3: Determine the unloading decision based on the unloading mode and the initial unloading ratio of the task vehicle;

[0010] S4: Allocate channel resources to the selected service nodes based on the offloading decision; calculate the total system overhead based on the offloading decision.

[0011] S5: Update the uninstallation ratio based on the total system overhead;

[0012] S6: Repeat steps S3-S5 to obtain the optimal offload ratio, and perform task offloading and resource allocation on the allocated channel resources according to the optimal offload ratio.

[0013] Preferably, the resource allocation framework based on MEC computing task offloading includes:

[0014] Obtain vehicle information and classify vehicles into task vehicles (TV) and service vehicles (SV) based on the vehicle information; task vehicles (TV) are vehicles that have tasks to be unloaded, and service vehicles (SV) are vehicles with available resources.

[0015] Service vehicles (SV) and roadside units (RSU) are used as service nodes (SN); when a task vehicle is performing task unloading, the task is divided into local task unloading and unloading to the service node.

[0016] Calculate the latency and energy consumption of unloading local tasks;

[0017] Offloading to the service node includes offloading to the roadside unit (RSU) and offloading to the service vehicle (SV); calculate the latency and energy consumption of offloading to the roadside unit (RSU); calculate the latency and energy consumption of offloading to the service vehicle (SV).

[0018] Preferably, the formula for calculating the total system overhead is:

[0019]

[0020] stC1:

[0021] C2:

[0022] C3:

[0023] C4:

[0024] C5:

[0025] C6:

[0026] C7:

[0027] in, This represents the set of vehicles that need to be unloaded. C i Let C represent the cost of a single task vehicle, and λ represent the total system cost. t λ represents the time delay factor. e T represents the energy consumption factor, one of two factors that determines the importance of latency and energy consumption in a system; i E represents system latency. i Indicates system energy consumption; g i This indicates unloading the decision factor when g i =1 indicates that the selected unloading target is RSU, when g i =0 indicates that the unloaded object is SV; k i,j It is the vehicle matching factor, indicating whether TVi selects SVj as the service target. When k i,j =1 indicates that TVi is selected to be unloaded to SVj; if TVi and SVj do not match, then k i,j =0; f i rsu This represents the computing resources allocated by RSU to TVi, f rsu-max ξ represents the maximum computing power of RSU. i T represents the task unloading ratio. i max θ represents the maximum tolerable latency for the TVi task. i,k Indicates whether subchannel k is assigned to TVi, when θ i,k When θ = 1, subchannel k is assigned to TVi, when θ i,k When = 0, sub-channel k is not assigned to TVi; Indicates t i The distance between TVi and RSU at any given moment. Indicates t i The distance between time TVi and SVj;

[0028] Preferably, the task vehicle makes an unloading decision based on the resource allocation framework for MEC-based task unloading, and determines the unloading mode of the task vehicle, including: defining the task matching degree M as equal to the task computation delay of the i-th task vehicle TVi selecting the j-th service vehicle SVj as the unloading object. We use a weighted average of distance and other factors to find the most suitable service target for SV, aiming to maximize the degree of matching.

[0029]

[0030] stC2:

[0031] C7:

[0032] in, Indicates the initial unloading ratio, α t and α d N represents the weight values ​​for calculating time and distance; S P represents the set of vehicles that provide services. i P j Indicates the vehicle's location, D V2V This indicates the communication distance between the mission vehicle and the service vehicle.

[0033] Furthermore, the process of finding the most suitable task unloading object for the task vehicle includes: using a dynamic tabu length tabu search pairing algorithm based on task matching degree to find a suitable service object for the service vehicle (SV), including:

[0034] S1: Set the initial solution and initial tabu list, and the initial tabu list is empty;

[0035] S2: Calculate the neighborhood of the current solution, set the calculated neighborhood as tabu objects, fill the tabu objects into the initial tabu table, and obtain tabu optimal solution and non-tabu optimal solution;

[0036] S3: Determine whether the task matching degree M, jointly determined by the tabu optimal solution and the non-tabu optimal solution, is better than the existing optimal solution M'. If it is better than the existing optimal solution, update the optimal solution; otherwise, do not update the optimal solution and return to S2. When the difference between M and M' is less than the minimum value ε or the maximum number of iterations is reached, stop updating.

[0037] S4: Based on the optimal solution, set up the task vehicles. Divided into sets and Among them, set This represents the set of vehicles that have selected the V2I offloading mode. This represents the set of vehicles that have selected the V2V offloading mode.

[0038] Preferably, allocating channel resources to the selected service node based on the offloading decision includes:

[0039] S1: A collection of vehicles that need to be unloaded. Sort in ascending order by the TV task's maximum tolerance time, Timax;

[0040] S2: Assign K colors to the sorted first K TVs, and allocate channels to the first K TVs according to each color corresponding to a channel;

[0041] S3: Use the graph coloring algorithm to color the TVs with unassigned channels and update the channel allocation matrix;

[0042] S4: Based on the expected delay With T i max The relationship is used to determine the TV color of the unassigned channel in sequence. If Less than T i max Then TVi is assigned color k, meaning channel k is allocated to TV; if Greater than T i max Then TVi cannot be assigned color k, meaning the k channel will not be assigned to TVi.

[0043] The formula for calculating the expected latency is:

[0044]

[0045] Where, r i rsu To determine the transmission rate of TVi on channel k in the V2I offloading mode from the task vehicle to the roadside unit, To determine the transmission rate of TVi on channel k in V2V offloading mode from the mission vehicle to the service vehicle, c represents the initial unloading ratio of vehicle i. i s represents the number of CPUs required for vehicle i to calculate its task. i g represents the size of the vehicle task in vehicle i. i This indicates unloading the decision factor, f. i av This represents the local computing power of vehicle i. It means that N S Let k represent the set of service vehicles. i,j Indicates the vehicle matching factor. This indicates the computational power of SVj.

[0046] Preferably, the process of task offloading and resource allocation based on the optimal offloading ratio in channel resource allocation includes: after determining the offloading mode and channel allocation, the task vehicle TV is divided into two new vehicle sets. and The optimization objective is divided into C V2I and C V2V Solve for the optimization objective respectively;

[0047] Compare the local computation time and the time unloaded to SV, and Divided into two new sets and When T i local ≥T i off-sv At that time, the mission vehicle TVi belongs to When T i local <T i off-sv TVi belongs to

[0048] Compare the local computation time with the time of unloading to the roadside unit (RSU), and... Divided into two new sets and When T i local ≥T i off-rsu At that time, TVi belongs to When T i local <T i off-rsu TVi belongs to

[0049] Among them, T i local T represents the latency calculated locally by the mission vehicle's TVi. i off-sv T represents the unloading delay of the task to the service vehicle. i off-rsu This indicates the unloading delay of the task to the roadside unit.

[0050] Furthermore, the process of solving the optimization objective includes: once the unloading mode is determined to be V2V unloading from the task vehicle to the service vehicle, C V2V The optimization objective is:

[0051]

[0052]

[0053]

[0054]

[0055]

[0056]

[0057]

[0058] Among them, C V2V This indicates the total system overhead of unloading from the task vehicle to the service vehicle in the unloading mode, c. a Represents a set The number of CPUs required to calculate the vehicle's task, c b Represents a set The number of CPUs required to calculate the vehicle's task; Represents a set The maximum delay for unloading tasks from vehicles. Represents a set Maximum delay for unloading vehicle tasks; s a Represents a set The size of the vehicle task, s b Represents a set The size of the vehicle task, p a Represents a set The calculated power of the vehicle, p b Represents a set The calculated power of the vehicle, ξ a Represents a set The unloading ratio, ξ b Represents a set The unloading ratio, κ v This represents the vehicle's calculated energy consumption coefficient. Represents a set The computational power of the vehicle itself during the mission, λ t λ represents the time delay factor. e Indicates the energy consumption factor. Represents a set The transmission rate when the mission vehicle TVi communicates with the service vehicle SVj. Represents a set The transmission rate when TVi communicates with SVj; For the computational power of SVj, k i,j The vehicle matching factor indicates whether TVi selects SVj as the service target, k a,j Represents a set Does TVi select SVj as the service target? b,j Represents a set Does TVi select SVj as the service target when k i,j k a,j or k b,j When k equals 1, it indicates that the task vehicle selects to unload to the service vehicle; when k equals 1, it indicates that the task vehicle selects to unload to the service vehicle. i,j k a,j or k b,j When the value is 0, there is no match between the mission vehicle TVi and the service vehicle SVj. Represents a set The distance between TVi and SVj Represents a set The distance between TVi and SVj, D V2V Indicates the communication distance between the mission vehicle and the service vehicle;

[0059] Once the unloading mode is determined to be unloading the task vehicle to the roadside unit V2I, the optimization objective C is optimized. V2I for:

[0060]

[0061]

[0062]

[0063]

[0064]

[0065]

[0066]

[0067]

[0068]

[0069] Among them, C V2I The unloading mode represents the total system overhead of unloading tasks from vehicles to roadside units, c. c Represents a set The number of CPUs required to calculate the vehicle's task, c d Represents a set The number of CPUs required to calculate the vehicle's task; Represents a set The maximum delay for unloading tasks from vehicles. Represents a set Maximum delay for unloading vehicle tasks; s c Represents a set The size of the vehicle task, s d Represents a set Medium vehicle task size, p c Represents a set The calculated power of the vehicle, p d Represents a set The calculated power of the vehicle, p com This represents the calculated power of the roadside unit (RSU). Represents a set The uninstallation ratio Represents a set The unloading ratio, κ v This represents the vehicle's calculated energy consumption coefficient. Represents a set The local computing power of the vehicle Represents a set The local computing power of the vehicle, f rsu-max Indicates that λ t Indicates that λ e express, f represents the maximum computing power of RSU. rsu-max λ represents the maximum computational power of RSU. t λ represents the time delay factor. e Indicates the energy consumption factor. Indicates RSU to the set The computing power for vehicle allocation in China Indicates RSU to the set The computing power for vehicle allocation in China Represents a set The transmission rate when the TVi of the medium-duty vehicle communicates with the RSU. Represents a set The transmission rate during TVi-RSU communication; For set The computing power of vehicles performing medium-duty missions. For set The computational power of RSU in China; Represents a set The distance between TVi and RSU Represents a set The distance between TVi and RSU, D V2I This indicates the communication distance from TVi to RSU.

[0070] Preferably, both V2V and V2I optimization problems can be solved using the CVX toolkit.

[0071] Furthermore, the optimization objective is calculated and the calculation result is determined to be converged. If converged, the optimal unloading ratio and the optimal resource allocation result obtained based on the optimal unloading ratio are output. If not converged, the unloading ratio is updated according to the optimization objective. The calculation is iterated until the result converges to the optimal solution. Then, the optimal unloading ratio and the optimal resource allocation result obtained based on the optimal unloading ratio are output. After that, the task vehicle performs task unloading and resource allocation on the allocated channel resources according to the optimal unloading ratio.

[0072] The beneficial effects of this invention are as follows: This invention addresses the resource allocation problem of MEC-based computation task offloading in vehicular network systems. Considering the differences between roadside units and vehicles as service nodes, as well as task latency and communication distance, a mathematical model is established based on computational resource constraints. The problem is decomposed into three sub-problems for joint solution. The concept of vehicle matching degree is proposed, and a dynamic tabu length tabu search algorithm is used to determine reasonable offloading targets. A graph coloring algorithm is employed for channel allocation to reduce channel interference. An optimal offloading strategy, channel resource, and computational resource allocation scheme are presented. By using variable substitution, the computational resource allocation sub-problem is transformed into a convex optimization problem, yielding the optimal offloading ratio. Compared to existing technical solutions, this invention reduces task computation latency, balances the energy consumption of each node in the system, and improves performance in terms of energy consumption and latency. Attached Figure Description

[0073] Figure 1 This is a system model diagram of the MEC-based partial task offloading and resource allocation method of the present invention;

[0074] Figure 2 This is a flowchart illustrating the implementation of the MEC-based task offloading and resource allocation method in the Internet of Vehicles (IoV) of this invention. Detailed Implementation

[0075] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0076] This invention proposes a resource allocation method for MEC-based computing task offloading in vehicle-to-everything (V2X) networks, such as... Figure 1 As shown, this method, with the assistance of mobile edge computing technology, addresses the problem of vehicular networks (V2X) needing to handle a large number of low-latency tasks under limited resources and dynamic topologies. Based on the principle of minimizing system overhead, it determines a task offloading and resource allocation strategy for V2X, thereby solving the task offloading and resource optimization problem in the vehicle edge computing system. The task offloading strategy is determined by calculating the minimum total overhead in the V2X system, and system resources are allocated according to the task offloading strategy, such as... Figure 2 As shown, it includes:

[0077] S1: Construct a resource allocation framework based on MEC computing task offloading;

[0078] S2: Determine the offloading mode of the task vehicle based on the resource allocation framework for MEC-based task offloading. The offloading modes include offloading the task vehicle to the service vehicle (V2V) and offloading the task vehicle to the roadside unit (V2I).

[0079] S3: Determine the unloading decision based on the unloading mode and the initial unloading ratio of the task vehicle;

[0080] S4: Allocate channel resources to the selected service nodes based on the offloading decision; calculate the total system overhead based on the offloading decision.

[0081] S5: Update the uninstallation ratio based on the total system overhead;

[0082] S6: Repeat steps S3-S5 to obtain the optimal offload ratio, and perform task offloading and resource allocation on the allocated channel resources according to the optimal offload ratio.

[0083] Furthermore, a resource allocation framework based on MEC task offloading is constructed, including: acquiring vehicle information and dividing vehicles into task vehicles (TV) and service vehicles (SV) based on the vehicle information; task vehicles (TV) are vehicles with tasks that need to be offloaded, and service vehicles (SV) are vehicles with idle resources; using service vehicles (SV) and roadside units (RSU) as service nodes (SN); when a task vehicle is performing task offloading, the task is divided into local task offloading and offloading to a service node; calculating the latency and energy consumption of local task offloading; offloading to a service node includes offloading to a roadside unit (RSU) and offloading to a service vehicle (SV); calculating the latency and energy consumption of offloading to a roadside unit (RSU); calculating the latency and energy consumption of offloading to a service vehicle (SV); the specific process is as follows:

[0084] Consider a vehicle communication network consisting of several vehicles traveling in the same direction, roadside RSUs, and a macro base station. The vehicles are traveling in the same direction on the road, and the set of vehicles in this system is... The set of vehicles that need to be unloaded is: This type of vehicle is called a Task Vehicle (TV), and TVi represents the i-th task vehicle in the set of task vehicles; the set of vehicles with available resources is... These are called service vehicles (SVs), and SVj represents the j-th service vehicle in the set of service vehicles. Each vehicle has only one attribute, namely...

[0085] Tasks can be computed locally by the TV using its own resources, or they can be offloaded to the MEC server via RSU in the form of V2I, or offloaded to a surrounding SV in the form of V2V for computation.

[0086] The parameters of vehicle i are expressed as follows P i Indicates vehicle position P i =(x i ,y i ), v i θ i and f i These represent the vehicle's speed, its travel angle relative to a certain reference line, and its computing power, respectively. Additionally, the task information that the vehicle needs to unload is represented as follows: Where s i ,c i ,T i max These represent the size of the vehicle task, the number of CPUs required for computation, and the maximum latency, respectively. The RSU and SV together form the Service Node (SN) to accept TV offload requests. The set of available sub-channels under each RSU is... The bandwidth of each sub-channel is BHz.

[0087] The system uses OFDM to allocate orthogonal channels for V2I mode, and V2V mode multiplexes the uplink transmission signals of V2I mode. N is introduced. T The channel connection matrix C of ×K describes the channel allocation. θ i,k Indicates whether subchannel k is assigned to TVi. When θ i,k When θ = 1, subchannel k is assigned to TVi, when θ i,k When the signal-to-interference-plus-noise ratio (SINR) is 0, subchannel k is not allocated to TVi. Therefore, the signal-to-interference-plus-noise ratio (SINR) and transmission rate of TVi when communicating with the RSU via subchannel k are as follows:

[0088]

[0089]

[0090] Similarly, the signal-to-interference-plus-noise ratio (SINR) and transmission rate when TVi and SVj communicate are as follows:

[0091]

[0092]

[0093] in, and p represents the channel gain of TVi when communicating with RSU and SVj on channel k. i For the transmission power, σ 2 This refers to noise interference during communication. i rsu and These represent the transmission rates of TVi offloaded to RSU and SV, respectively. Due to the mobility of vehicles and the limited communication range of V2X (Vehicle to Everything), the communication distance of the mission vehicle in V2I mode and the communication distance in V2V mode are respectively represented by D. V2I and D V2V It is necessary to consider whether the communication distance between dynamic nodes meets the requirements.

[0094] The positions of TVi and SVj after time t are respectively and Among them are:

[0095]

[0096]

[0097] The position of RSU remains unchanged, therefore the task completion time T in TVi is calculated. i The distances between the inner TV and the SN are as follows:

[0098]

[0099]

[0100] As a node with excessive computing tasks, TV needs to request offloading services. To achieve resource utilization balance between TV and SN and improve system resource utilization, TV adopts partial offloading, dividing the task into two divisible parts, one locally offloaded and the other offloaded to a specific SN. Offloading tasks to multiple SNs is not considered at this stage. Therefore, the latency and energy consumption of the task vehicle TVi for local computation are as follows:

[0101]

[0102]

[0103] Where f i tv κ represents TVi's own computing power. v This is the vehicle's calculated energy consumption coefficient, which is related to chip performance. Here, κ... v =10 -28 .

[0104] Because the RSU has stronger computing power than the SV, when the RSU acts as the SN, it can accept task requests from multiple TVs simultaneously, while the SV chooses a non-CPU preemptive computing method to serve the TVs. Therefore, when TVi chooses to offload to the RSU, the offloading latency and energy consumption are as follows:

[0105]

[0106]

[0107] Where T i tran-rsu and T i com-rsu These are the transmission and computation delays offloaded to the RSU, f i rsu This indicates the computing resources allocated by RSU to TVi. and These are the transmission power consumption offloaded to the RSU, p rsu This represents the computational power of RSU. Task return latency is ignored here.

[0108] When TVi chooses to offload to SVj, the offloading latency and energy consumption are as follows:

[0109]

[0110]

[0111] in and These are the transmission and computation delays offloaded to the RSU, respectively, excluding the return transmission delay. and These are the transmission and computing energy consumption offloaded to the RSU, respectively. This refers to the computational power of SVj.

[0112] To represent the specific SN selected by TVi unloading, an unloading decision factor g is introduced. i When g i =1 indicates that the selected unloading target is RSU. When g i When k = 0, it indicates that the unloaded object is SV. In this case, it is necessary to further determine the specific SV. i,j =1 indicates that TVi is selected to be unloaded to SVj; if TVi and SVj do not match, then k i,j =0. Therefore, the latency and energy consumption of the task unloading process can be expressed as:

[0113]

[0114]

[0115] Therefore, system latency and energy consumption can be expressed as follows:

[0116]

[0117] T i =max{T i local ,T i off}

[0118] Based on the above information, the optimization objective is to minimize the total system overhead, expressed as:

[0119]

[0120] stC1:

[0121] C2:

[0122] C3:

[0123] C4:

[0124] C5:

[0125] C6:

[0126] C7:

[0127] in, This represents the set of vehicles that need to be unloaded. C i Let C represent the cost of a single task vehicle, and λ represent the total system cost. t λ represents the time delay factor. e T represents the energy consumption factor. i E represents system latency. i Indicates system energy consumption; g i Indicates unloading decision factors, k i,j f represents the vehicle matching factor. i rsu This represents the computing resources allocated by RSU to TVi, f rsu-max ξ represents the maximum computing power of RSU. i T represents the task unloading ratio. i max θ represents the maximum tolerable latency for the TVi task. i,k Indicates whether subchannel k is assigned to TVi. This represents the distance between TVi and RSU. This represents the distance between TVi and SVj.

[0128] Furthermore, based on the resource allocation framework for MEC-based task unloading, the unloading mode of the task vehicle is determined, and an unloading decision is made based on the unloading mode and the initial unloading ratio of the task vehicle, including:

[0129] When considering serving a TV, the SV (Service Provider) should, as far as possible, consider the matching degree between its own computing power and the receiving tasks. The matching degree includes the fit between computing power and receiving tasks, as well as the distance between the SV and the TV. The matching degree M is defined as a weighted average of computing latency and distance. With the goal of maximizing the matching degree, the SV is used to find the most suitable service target.

[0130]

[0131] stC2:

[0132] C7:

[0133] in, This represents the initial unloading ratio. The initial unloading ratio of the mission vehicle is determined by the TV and is denoted as . α t and α d Represents the weight values ​​for calculating time and distance; N represents the set of vehicles that need to be unloaded. S Let k represent the set of service vehicles. i,j P represents the vehicle matching factor. i P represents the position of vehicle i. j Indicates the position of vehicle j; Indicates t i The distance between time TVi and SVj, D V2V This indicates the communication distance between the mission vehicle and the service vehicle.

[0134] This problem is a 0-1 integer programming problem, which can be solved using heuristic algorithms. This invention employs a dynamic tabu length tabu search matching algorithm (DTTS) based on task matching degree to solve this problem. The pre-calculation of the DTTS algorithm is as follows: relax the non-convex constraint C7 to... The constrained mathematical model is transformed into an unconstrained mathematical programming model using a penalty function, where L represents a very large number, preferably set to 10000. The maximum matching degree is expressed as:

[0135]

[0136] Furthermore, the parameter settings for using the tabu search algorithm to solve this problem are as follows:

[0137] (1) Initial solution: The initial solution will affect the convergence speed and result of the tabu algorithm. Therefore, unlike the traditional TS algorithm which randomly generates initial values ​​or the greedy algorithm which searches for initial solutions, the nearest pair is set as the initial solution.

[0138] (2) Neighborhood: The condition that the neighborhood satisfies is Therefore, there are a total of (N) T -N S -2)×N S / 2 neighborhoods.

[0139] (3) Taboo objects: The selected column when the optimal value of any single row of SVj changes is a taboo object.

[0140] (4) Taboo Length: This paper sets a dynamic taboo length, allowing the algorithm to search more unknown domains when the function value decreases significantly, and to focus more on fine-grained local searches when the decrease is smaller; the taboo length formula is set as follows.

[0141]

[0142] Where ω is the coefficient of variation, and ΔM represents the change in the function value.

[0143] (5) Candidate solutions: consist of tabu optimal solutions and non-tabu optimal solutions in the neighborhood solution set.

[0144] (6) Amnesty Criterion: Evaluation value-based criteria within a certain number of iterations.

[0145] Furthermore, a tabu search algorithm is used to find suitable service targets for service vehicles (SVs), including:

[0146] S1: Set the initial solution and initial tabu list, and the initial tabu list is empty;

[0147] S2: Calculate the neighborhood of the current solution, set the calculated neighborhood as tabu objects, fill the tabu objects into the initial tabu table, and obtain tabu optimal solution and non-tabu optimal solution;

[0148] S3: Determine whether the task matching degree M, jointly determined by the tabu optimal solution and the non-tabu optimal solution, is better than the existing optimal solution M'. If it is better than the existing optimal solution, update the optimal solution; otherwise, do not update the optimal solution and return to S2. When the difference between M and M' is less than the minimum value ε or the maximum number of iterations is reached, stop updating.

[0149] S4: Based on the optimal solution, set up the task vehicles. Divided into sets and That

[0150] In, set This represents the set of vehicles that have selected the V2I offloading mode. This represents the set of vehicles that have selected the V2V offloading mode.

[0151] Furthermore, after determining the offloading target, the sub-channel allocation is transformed into a graph coloring model. K channel resources are modeled as K different colors, and all TVs are considered vertices. When two nodes are assigned the same color, it indicates that the two TVs are using the same channel to communicate with their respective serving nodes, resulting in co-channel interference and affecting transmission speed and time. The channel allocation resources for the selected serving node based on the offloading decision include:

[0152] S1: A collection of vehicles that need to be unloaded. According to the TV's maximum tolerance time T i max Sort in ascending order;

[0153] S2: Assign K colors to the sorted first K TVs, and allocate channels to the first K TVs according to each color corresponding to a channel;

[0154] S3: Use the graph coloring algorithm to color the TVs with unassigned channels and update the channel allocation matrix;

[0155] S4: Based on the expected delay With T i max The relationship is used to determine the TV color of the unassigned channel in sequence. If Less than T i max Then TVi is assigned color k, meaning channel k is allocated to TV; if Greater than T i max Then TVi cannot be assigned color k, meaning the k channel will not be assigned to TVi.

[0156] The formula for calculating the expected latency is:

[0157]

[0158] Where, r i rsu To determine the transmission rate of TVi on channel k in the V2I offloading mode from the task vehicle to the roadside unit, To determine the transmission rate of TVi on channel k in V2V offloading mode from the mission vehicle to the service vehicle, c represents the initial unloading ratio of vehicle i.i s represents the number of CPUs required for vehicle i to calculate its task. i g represents the size of the vehicle task in vehicle i. i This indicates unloading the decision factor, f. i av This represents the local computing power of vehicle i. This represents the average computational resources allocated by the roadside unit (RSU) to calculating the unloading task. N S Let k represent the set of service vehicles. i,j Indicates the vehicle matching factor. This indicates the computational power of SVj.

[0159] Furthermore, the unloading ratio is updated based on the total system overhead, and the calculation is iteratively repeated until the optimal unloading ratio is obtained. Task unloading and resource allocation are then performed based on the optimal unloading ratio, including:

[0160] After the SN is determined and the channel is allocated, TV is divided into two new vehicle sets. and Due to the differences in computational capabilities and service properties when SN is SV or RSU respectively, the optimization objective is divided into C V2I and C V2V Solve for the optimization objective separately.

[0161] After the unloading mode was determined and the channel was allocated, the mission vehicle TV was divided into two new vehicle sets. and The optimization objective is divided into C V2I and C V2V Solve by optimization separately;

[0162] Compare the local computation time and the time unloaded to SV, and Divided into two new sets and When T i local ≥T i off-sv At that time, the mission vehicle TVi belongs to When T i local <T i off-sv TVi belongs to

[0163] Compare the local computation time with the time of unloading to the roadside unit (RSU), and... Divided into two new sets and When T i local ≥T ioff-rsu At that time, TVi belongs to When T i local <T i off-rsu TVi belongs to

[0164] Among them, T i local T represents the latency of task vehicle i being computed locally. i off-sv T represents the unloading delay of the task to the service vehicle. i off-rsu This indicates the unloading delay of the task to the roadside unit.

[0165] The process of solving the optimization objective includes: The optimization objective for the V2V part is:

[0166]

[0167] stC4,C5,C7

[0168] Because there are comparison terms involving variables, the nature of this subproblem cannot be directly determined; therefore, Divided into two new sets and When T i local ≥T i off-sv At that time, TVi belongs to When T i local <T i off-sv TVi belongs to The optimization objective can be written as:

[0169]

[0170]

[0171]

[0172]

[0173]

[0174]

[0175]

[0176] Among them, C V2V This indicates the total system overhead of unloading from the task vehicle to the service vehicle in the unloading mode, c.a Represents a set The number of CPUs required to calculate the vehicle's task, c b Represents a set The number of CPUs required to calculate the vehicle's task; Represents a set The maximum delay for unloading tasks from vehicles. Represents a set Maximum delay for unloading vehicle tasks; s a Represents a set The size of the vehicle task, s b Represents a set The size of the vehicle task, p a Represents a set The calculated power of the vehicle, p b Represents a set The calculated power of the vehicle, ξ a Represents a set The unloading ratio, ξ b Represents a set The unloading ratio, κ v This represents the vehicle's calculated energy consumption coefficient. Represents a set The computational power of the vehicle itself during the mission, λ t λ represents the time delay factor. e Indicates the energy consumption factor. Represents a set The transmission rate when the mission vehicle TVi communicates with the service vehicle SVj. Represents a set The transmission rate when TVi communicates with SVj; For the computational power of SVj, k i,j The vehicle matching factor indicates whether TVi selects SVj as the service target, k a,j Represents a set Does TVi select SVj as the service target? b,j Represents a set Should TVi select SVj as its service target? Represents a set The distance between TVi and SVj Represents a set The distance between TVi and SVj, D V2V This indicates the communication distance between the mission vehicle and the service vehicle.

[0177] Once the communication mode is determined to be V2I, the optimization objective C is... V2I for:

[0178]

[0179] st:C3,C4,C5,C7

[0180] This problem is the same as the V2V part, and there are comparison terms, so it will also be... Divided into two new sets: and Furthermore, allocating RSU resources and coupling the variables in the formula makes this problem still difficult to solve. Therefore, let Substituting into the above formula, we get:

[0181]

[0182]

[0183]

[0184]

[0185]

[0186]

[0187]

[0188]

[0189]

[0190] C V2I The unloading mode represents the total system overhead of unloading tasks from vehicles to roadside units, c. c Represents a set The number of CPUs required to calculate the vehicle's task, c d Represents a set The number of CPUs required to calculate the vehicle's task; Represents a set The maximum delay for unloading tasks from vehicles. Represents a set Maximum delay for unloading vehicle tasks; s c Represents a set The size of the vehicle task, s d Represents a set Medium vehicle task size, p c Represents a set The calculated power of the vehicle, p d Represents a set The calculated power of the vehicle, p com This represents the calculated power of the roadside unit (RSU). Represents a set The uninstallation ratio Represents a set The unloading ratio, κ v f represents the vehicle's calculated energy consumption coefficient. c av Represents a set The local computing power of the vehicle Represents a set The local computing power of the vehicle, f rsu-max λ represents the maximum computational power of RSU. t λ represents the time delay factor. e Indicates the energy consumption factor. Indicates RSU to the set The computing power for vehicle allocation in China Indicates RSU to the set The computational power for vehicle allocation in the middle represents the set. The transmission rate when the TVi of the medium-duty vehicle communicates with the RSU. Represents a set The transmission rate during TVi-RSU communication; For set The computing power of vehicles performing medium-duty missions. For set The computational power of RSU in China; Represents a set The distance between TVi and RSU Represents a set The distance between TVi and RSU, D V2I This indicates the communication distance from TVi to RSU.

[0191] Preferably, the V2V optimization problem and optimization conditions are both about the variable ξ. a and ξ b The linear function is a convex optimization problem, which can be solved using the CVX toolkit. The V2I optimization problem is a sum of an exponential function and a linear function, which is also a convex optimization problem and can be solved using the CVX toolkit. The unloading is similar to the update and V2V problems.

[0192] Furthermore, the optimization objective is calculated and the calculation result is determined to be converged. If converged, the optimal unloading ratio and the optimal resource allocation result obtained based on the optimal unloading ratio are output. If not converged, the unloading ratio is updated according to the optimization objective. The calculation is iterated until the result converges to the optimal solution. Then, the optimal unloading ratio and the optimal resource allocation result obtained based on the optimal unloading ratio are output. After that, the task vehicle performs task unloading and resource allocation on the allocated channel resources according to the optimal unloading ratio.

[0193] This invention addresses the resource allocation problem for MEC-based computational task offloading in vehicular network systems. Considering the differences between roadside units and vehicles as service nodes, as well as task latency and communication distance, a mathematical model is established based on computational resource constraints. The problem is decomposed into three sub-problems for joint solution. The concept of vehicle matching degree is proposed, and a dynamic tabu length tabu search algorithm is used to determine reasonable offloading targets. A graph coloring algorithm is employed for channel allocation to reduce channel interference. The optimal offloading strategy, channel resource, and computational resource allocation scheme are presented. By using variable substitution, the computational resource allocation sub-problem is transformed into a convex optimization problem, yielding the optimal offloading ratio. Compared to existing technologies, this invention reduces task computation latency, balances energy consumption across nodes in the system, and improves performance in terms of energy consumption and latency.

[0194] The above-described embodiments further illustrate the purpose, technical solution, and advantages of the present invention. It should be understood that the above-described embodiments are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made to the present invention within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A resource allocation method for MEC-based computing task offloading in vehicle-to-everything (V2X) networks, characterized in that, The task offloading strategy is determined by calculating the minimum total overhead in the vehicle-to-everything (V2X) system, and system resources are allocated according to the task offloading strategy, including: S1: Construct a resource allocation framework based on MEC computing task offloading; the construction of a resource allocation framework based on MEC computing task offloading includes: Obtain vehicle information and classify vehicles into task vehicles (TV) and service vehicles (SV) based on the vehicle information; designate service vehicles (SV) and roadside units (RSU) as service nodes (SN); when a task vehicle is performing task unloading, the task is divided into local task unloading and unloading to service nodes. Calculate the latency and energy consumption of unloading local tasks; Offloading to the service node includes offloading to the roadside unit (RSU) and offloading to the service vehicle (SV); calculate the latency and energy consumption of offloading to the roadside unit (RSU); calculate the latency and energy consumption of offloading to the service vehicle (SV); The total system overhead is calculated based on the latency and energy consumption of offloading tasks locally, offloading to Roadside Units (RSUs), and offloading to Service Vehicles (SVs). The formula for calculating the total system overhead is: in, This represents the set of vehicles that need to be unloaded. C i Let C represent the cost of a single task vehicle, and λ represent the total system cost. t λ represents the time delay factor. e T represents the energy consumption factor. i E represents system latency. i Indicates system energy consumption; g i Indicates unloading decision factors, k i,j f represents the vehicle matching factor. i rsu Indicates RSU as TV i Allocated computing resources, f rsu -max ξ represents the maximum computing power of RSU. i T represents the task unloading ratio. i max TV i The maximum tolerable delay for the task; θ i,k Indicates whether subchannel k is assigned to TV. i , TV i Distance to RSU TV i With SV j The distance between them; S2: The task vehicle determines its unloading mode based on the resource allocation framework for task unloading computation using MEC, and makes an unloading decision based on the unloading mode and the initial unloading ratio of the task vehicle. The unloading modes include task vehicle unloading to service vehicle (V2V) and task vehicle unloading to roadside unit (V2I). Determining the unloading mode of the task vehicle includes defining the task matching degree M, which is equal to the TV of the i-th task vehicle. i Select the j-th service vehicle SV j The task to be unloaded is calculated as a weighted sum of latency and distance, where distance refers to the distance between the task vehicle and the service vehicle; the maximum task matching degree is calculated, and the most suitable task unloading object is found for the task vehicle based on the maximum task matching degree; the formula for calculating the maximum task matching degree is: in, α represents the initial unloading ratio of vehicle i. t and α d These represent the weight values ​​for calculating time and distance, respectively; N S P represents the set of vehicles that provide services. i P represents the position of vehicle i. j Indicates the position of vehicle j; D V2V c represents the communication distance between the mission vehicle and the service vehicle. i This represents the number of CPUs required for vehicle i to calculate its task. SV j Computational power; The process of finding the most suitable task unloading object for the task vehicle includes: using a dynamic tabu length tabu search pairing algorithm based on task matching degree to find a suitable service object for the service vehicle (SV), including: When using the tabu search algorithm to solve this problem, the parameters are set as follows: (1) Initial solution: Set the nearest pair as the initial solution; (2) Neighborhood: The neighborhood satisfies the following conditions. Therefore, there are a total of (N) T -N S -2)×N S / 2 neighborhoods; (3) Taboo objects: For any SV j The selection column when the optimal value is changed in a single row is a taboo object; (4) Taboo Length: The formula for setting the taboo length q is: Where ω is the coefficient of variation, and ΔM represents the change in the function value; (5) Candidate solutions: consist of tabu optimal solutions and non-tabu optimal solutions in the neighborhood solution set; S21: Set the initial solution and initial tabu list, and the initial tabu list is empty; S22: Calculate the neighborhood of the current solution, set the calculated neighborhood as tabu objects, fill the tabu objects into the initial tabu table, and obtain tabu optimal solution and non-tabu optimal solution; S23: Determine whether the task matching degree M, jointly determined by the tabu optimal solution and the non-tabu optimal solution, is better than the existing optimal solution M′. If it is better than the existing optimal solution, update the optimal solution; otherwise, do not update the optimal solution and return to S2. When the difference between M and M′ is less than the minimum value ε or the maximum number of iterations is reached, stop updating. S24: Based on the optimal solution, set up the task vehicle group. Divided into sets and Among them, set This represents the set of vehicles that have selected the V2I offloading mode. This represents the set of vehicles that have selected the V2V offloading mode; S3: Allocate channel resources to the selected service node according to the offloading decision; the allocation of channel resources to the selected service node according to the offloading decision includes: S31: A collection of vehicles that need to be unloaded. According to the maximum tolerance time T of the mission vehicle TV. i max Sort in ascending order; S32: Assign K colors to the sorted first K TVs, and complete the channel allocation for the first K TVs according to each color corresponding to a channel; S33: Use graph coloring algorithm to color the TVs of unassigned channels and update the channel allocation matrix; S34: Based on the expected delay With TV i Maximum tolerable delay T of the task i max The relationship is used to determine the TV color of the unassigned channel in sequence. If Less than T i max TV i It is designated as color k, meaning channel k is assigned to TV; if Greater than T i max TV i It cannot be assigned to color k, meaning the k channel should not be allocated to TV. i ; The formula for calculating the expected latency is: Where, r i rsu TV in V2I unloading mode for unloading from task vehicle to roadside unit i The transmission rate in channel k, TV in V2V offloading mode from task vehicle to service vehicle i At the transmission rate of channel k, c i s represents the number of CPUs required for vehicle i to calculate its task. i f represents the size of the vehicle task in vehicle i. i av This represents the local computing power of vehicle i. This indicates that the roadside units (RSUs) are evenly distributed to the TVs. i Computing resources, N S This refers to a set of vehicles used for service. SV j Computational power; S4: Update the offload ratio; perform task offloading and resource allocation on the channel allocation resources according to the updated offload ratio; update the offload ratio and perform task offloading and resource allocation on the allocated channel allocation resources according to the offload ratio, including: After the unloading mode is determined and the channel is allocated, the task vehicle TV is divided into vehicle sets. and The optimization objective is divided into C V2I and C V2V Solve for the optimization objective respectively; Compare the local computation time and the time unloaded to SV, and Divided into two new sets and When T i local ≥T i off-sv At that time, the mission vehicle TV i belong When T i local <T i off-sv TV i belong Compare the local computation time with the time of unloading to the roadside unit (RSU), and... Divided into two new sets and When T i local ≥T i off-rsu At that time, TV i belong When T i local <T i off-rsu TV i belong Among them, T i local TV indicating mission vehicle i The latency calculated locally, T i off-sv T represents the unloading delay of the task to the service vehicle. i off-rsu This indicates the unloading delay for the task to be unloaded to the roadside unit; The process of solving the optimization objective includes: when the unloading mode is determined to be V2V unloading from the task vehicle to the service vehicle, C V2V The optimization objective is: Among them, C V2V This indicates the total system overhead of unloading from the task vehicle to the service vehicle in the unloading mode, c. a Represents a set The number of CPUs required to calculate the vehicle's task, c b Represents a set The number of CPUs required to calculate the vehicle's task; Represents a set The maximum delay for unloading tasks from vehicles. Represents a set Maximum delay for unloading tasks from vehicles; s a Represents a set The size of the vehicle task, s b Represents a set The size of the vehicle task, p a Represents a set The calculated power of the vehicle, p b Represents a set The calculated power of the vehicle, ξ a Represents a set The unloading ratio, ξ b Represents a set The unloading ratio, κ v This represents the vehicle's calculated energy consumption coefficient. Represents a set The computing power of the vehicle itself during the mission. Represents a set Medium Mission Vehicle TV i With service vehicle SV j Transmission rate during communication Represents a set China TV i With SV j Transmission rate during communication; For SV j The computing power, k i,j For vehicle matching factors, representing TV i Do you want to select SV? j As a service recipient, k a,j Represents a set China TV i Do you want to select SV? j As a service recipient, k b,j Represents a set China TV i Do you want to select SV? j As a service recipient, Represents a set China TV i With SV j The distance between them Represents a set China TV i With SV j The distance between them; Once the unloading mode is determined to be unloading the task vehicle to the roadside unit V2I, the optimization objective C is optimized. V2I for Among them, C V2I The unloading mode represents the total system overhead of unloading tasks from vehicles to roadside units, c. c Represents a set The number of CPUs required to calculate the vehicle's task, c d Represents a set The number of CPUs required to calculate the vehicle's task; Represents a set The maximum delay for unloading tasks from vehicles. Represents a set Maximum delay for unloading tasks from vehicles; s c Represents a set The size of the vehicle task, s d Represents a set Medium vehicle task size, p c Represents a set The calculated power of the vehicle, p d Represents a set The calculated power of the vehicle, p com This represents the calculated power of the roadside unit (RSU). Represents a set The uninstallation ratio Represents a set The unloading ratio, κ v This represents the vehicle's calculated energy consumption coefficient. Represents a set The local computing power of the vehicle Represents a set The local computing power of the vehicle, f rsu-max This indicates the maximum computing power of RSU. Indicates RSU to the set The computing power for vehicle allocation in China Indicates RSU to the set The computing power for vehicle allocation in China Represents a set Medium Mission Vehicle TV i Transmission rate when communicating with RSU Represents a set China TV i Transmission rate when communicating with RSU; For set The computing power of vehicles performing medium-duty missions. For set The computational power of RSU in China; Represents a set China TV i Distance to RSU Represents a set China TV i Distance to RSU, D V2I TV i Communication distance to RSU; Both the V2V and V2I optimization problems were solved using the CVX toolkit. The optimization objective is calculated and the calculation result is determined to be converged. If converged, the optimal unloading ratio and the optimal resource allocation result obtained based on the optimal unloading ratio are output. If not converged, the unloading ratio is updated and the calculation is iterated until the result converges to the optimal solution. Then, the optimal unloading ratio and the optimal resource allocation result obtained based on the optimal unloading ratio are output. After that, the task vehicle performs task unloading and resource allocation on the allocated channel resources according to the optimal unloading ratio.

Images