Method and system for multi-objective task offloading in internet of vehicles based on delay-aware energy saving

Abstract

The application discloses a kind of based on time delay perception energy-saving internet of vehicles multi-objective task unloading method and system, method is as follows step: S1, obtains information data;S2, calculates energy consumption and time delay;S3, determine the joint optimization problem of time delay, energy consumption;S4, determine optimization objective function;S5, while bypassing local optimal solution using SA-PSO algorithm converges to the node on solution space, to obtain global optimal solution;S6, to joint optimization problem is solved, according to different task unloading position, is divided into different types, and the optimal unloading proportion and optimal transmission power under different types are solved;S7, select the end device or corresponding node with most remaining energy, and select the corresponding optimal unloading proportion and optimal transmission power, and carry out task unloading.The application effectively shortens task completion time, improves task completion rate.

Description

Technical Field

[0001] This invention belongs to the field of vehicle-to-everything (V2X) task offloading technology. Considering different task offloading methods, this invention selects different processing modes for different service types, models a multi-objective optimization problem involving latency and energy consumption, and minimizes device latency and energy consumption by adjusting the task offloading ratio and uplink terminal transmission power. Specifically, it relates to a multi-objective task offloading method and system for V2X based on latency-aware energy saving. Background Technology

[0002] With the exponential growth in the amount of data that needs to be transmitted and processed in the Internet of Vehicles (IoV), transmitting data collected from vehicles to cloud computing platforms for processing leads to a sharp increase in network bandwidth load and a significant increase in network latency. This cannot meet the processing requirements of latency-sensitive tasks in the IoV environment. Cloud-fog converged networks are considered a promising network architecture and have gained widespread industry recognition, proving that they can further improve computing power, expand communication coverage, and reduce transmission latency.

[0003] However, while improving user latency performance, it inevitably leads to increased energy consumption, and energy shortages are becoming a key obstacle restricting the development of connected vehicles. In particular, the near-future dominance of electric vehicles or hybrid electric vehicles has prompted greater attention to energy consumption issues in connected vehicles. In reality, each vehicle typically requires different computing resources due to the varying size of its computational tasks. During task offloading, limited computing resources are shared among vehicles, and each vehicle needs to determine whether to offload. Therefore, jointly optimizing task offloading decisions to minimize the average cost within the system is intuitively important for improving overall system performance.

[0004] The increasing number of connected devices is accompanied by a rapid increase in network traffic and data center load. This growth leads to a significant increase in power consumption in communication networks and data centers, while posing a huge challenge to the low-latency, high-reliability data forwarding requirements in vehicular communications. With the advancement of information technology, latency-sensitive applications such as road environment augmented reality, dynamic human-vehicle interaction, and driving safety warnings are emerging in large numbers. Frequent changes in wireless network access, dynamic allocation of system resources, and rapid changes in nodes are making the requirements for low-latency, high-reliability data forwarding in vehicular communications increasingly stringent. Summary of the Invention

[0005] To address the issue of low power consumption for latency-sensitive services in existing technologies, this invention proposes a latency-aware energy-saving multi-objective task offloading method and system for vehicle-to-everything (V2X) networks. This invention designs a network architecture and task offloading process based on vehicle-cooperative edge computing, analyzes the impact of communication and computing resource allocation on task completion time, and constructs a task completion time model. An optimization model is established with the objectives of minimizing average completion time and maximizing success rate, and a solution is designed based on particle swarm optimization (PSO) and annealing algorithms. Simulation results show that, compared with existing strategies, this invention effectively shortens task completion time and improves task completion rate using simulated annealing-particle swarm optimization (SA-PSO).

[0006] To achieve the above objectives, the present invention adopts the following solution:

[0007] A multi-objective task offloading method for vehicle-to-everything (V2X) based on time-delay-aware energy saving comprises the following steps:

[0008] S1. Obtain information data;

[0009] Preferably, the information data includes: the number of terminal devices N, the number of fog nodes m, and the average task data size θ of device i. i The average arrival rate λ of the task on terminal device i i Terminal i selects to unload ratio β i (0≤β i ≤1), terminal task arrival rate The uninstallation ratios for the four uninstallation modes—local processing, D2D, fog computing, and cloud computing—are respectively... Channel bandwidth B, signal-to-noise ratio Channel rate R i ;

[0010] The average computing power of fog node i is The computing power u of cloud nodes C ;

[0011] S2. Performance analysis, calculating energy consumption and latency;

[0012] S3. Determine the joint optimization problem of time delay and energy consumption;

[0013] S4. Determine the objective function for optimization;

[0014] S5. The SA-PSO algorithm is used to quickly converge to a node in the solution space while bypassing local optima, thereby obtaining the global optimal solution.

[0015] S6. Solve the joint optimization problem, considering different task offloading locations, which are divided into the following three types: task offloading from end device to end device, task offloading from end device to fog node, and task offloading from end device to fog node to cloud node. Solve to obtain the optimal offloading ratio and optimal transmission power for each type.

[0016] S7. Select the terminal device or corresponding node with the most remaining energy, and choose the optimal offloading ratio and optimal transmit power to perform task offloading. Remaining energy is the system's original energy minus the energy consumed by the system.

[0017] Preferably, step S2 specifically includes:

[0018] S2.1 Calculate the time delay used;

[0019] S2.2 Calculate the energy consumption of the entire system;

[0020] Preferably, in step S2, the calculation of latency and system energy consumption is as follows:

[0021] (1) Calculate the average delay

[0022] a) Transmission Latency: User offloading tasks to cloud nodes, fog nodes, or transmitting them to other device nodes via D2D are collectively referred to as helper nodes. When a task is offloaded to a helper node, considering that the calculated result of the task is relatively small and helper nodes usually have a higher transmission rate, the downlink latency from the helper node back to the user is ignored, and only the user's uplink transmission latency is considered.

[0023]

[0024]

[0025] Where B represents the channel bandwidth. The signal-to-noise ratio (SNR) is positively correlated with the transmit power, while the transmission delay... With channel rate R i The transmission power is negatively correlated, therefore the greater the transmission power, the longer the transmission delay. The smaller the value, the lower the transmission latency. With the unloading ratio β i They are positively correlated, therefore the unloading ratio β i The larger the value, the longer the transmission delay. The larger the value, the more important it is to optimize the unloading ratio and the transmission power. Therefore, the key parameters to focus on are the unloading ratio and the transmission power, so as to minimize the latency and energy consumption.

[0026] b) Local processing latency: Assuming the average task load rate of end device i This indicates the workload rate of terminal device i, excluding system tasks. The smaller the value, the stronger the processing speed of terminal device i. Assume the average computing power of terminal device i is... The local computation latency T of the end device can be obtained. j1 :

[0027]

[0028] c) D2D processing latency: Assume the average transmission rate of the wireless port of end device i is u. D Combining the M / M / 1 queuing theory, the average processing time T for D2D unloading tasks is... D for

[0029]

[0030] d) Fog computing processing latency: The system allows multiple end devices to offload tasks to the fog node layer, at which point the maximum request rate of the fog node is... Assume the average computing power of fog node i is The total service request rate of the fog nodes is:

[0031]

[0032] The total time delay of the fog nodes is:

[0033]

[0034] e) Cloud computing processing latency: Considering the limited computing power of the fog node layer, when the computing power of the fog node layer is insufficient to handle the current task offloading, the offloading ratio will be [percentage missing]. The task is offloaded to a cloud node with greater computing power. Since the fog node and the cloud node are connected via wired fiber optic cable, a fixed latency T will be generated. 0 Assuming the server performance in the cloud node can handle a sufficiently large volume of unloading tasks, tasks unloaded to the cloud node will be processed immediately, thus ignoring waiting time. Based on queuing theory, the execution process of the cloud node can be viewed as an M / M / ∞ queue. Assuming the computing power u of the cloud node... C At this time, the latency of the cloud node layer is:

[0035]

[0036] (2) Calculate the system's energy consumption:

[0037] According to communication theory research, the energy consumption generated during transmission is related to transmission power, load, and transmission time. The transmission power of the end device is defined as... for:

[0038]

[0039] Where, p i For user transmission power;

[0040] In the entire system, since both the fog node layer and the cloud node layer are wired powered, and most end devices are battery powered and subject to significant energy constraints, the focus is on the energy consumption of the end devices. End device energy consumption consists of local computing energy consumption and task transmission energy consumption. Since the amount of task return results is negligible compared to the amount of unloaded tasks, only uplink energy consumption is considered. Based on the latency analysis above, the local energy consumption of the end devices is defined. for:

[0041]

[0042] Where, q i Let be the unit operating power coefficient of the i-th device, which is a fixed constant.

[0043] Define the decision matrix M as whether task j chooses the k-th decision method.

[0044]

[0045] Restriction: For any row k of matrix M, we have That is, each task can only choose one task uninstallation method.

[0046] The objective of this invention is to determine the optimal vehicle task unloading strategy, minimizing the total time delay and energy consumption during task uploading, computation, and unloading. In this multi-objective optimization problem, a linear weighted sum method is employed, introducing a time delay weight coefficient R. T Energy consumption weighting coefficient R E The weighting coefficients can be set according to the current status of the on-board equipment and the task requirements, and must meet the requirements of R. T +R E =1. Define the objective function:

[0047]

[0048] When the car has sufficient battery power, the time delay is the primary optimization objective, with a weighting coefficient set to R. T >R E When the battery is low, the main focus is on optimizing energy consumption and setting R. T <R E .

[0049] Preferably, the optimization problem in step S3 is described as follows:

[0050] Assume the computation rate of the fog node is u at this time. F The transmit power of the fog node is The cloud node's transmit power is Based on the definitions of energy consumption, time delay, and cost functions, the problem of minimizing these functions can be expressed as:

[0051] min{F}(12)

[0052]

[0053]

[0054]

[0055]

[0056]

[0057]

[0058]

[0059] Among them, (12-1) indicates that the local computing capacity does not exceed the local computing power; (12-2) indicates that the amount of tasks offloaded through D2D does not exceed the amount of tasks that the helper node can handle; (12-3) indicates that the amount of tasks offloaded to the fog node is less than the amount of tasks that the fog node can handle; (12-4) indicates that the transmit power of the fog node is greater than the completion power of the fog node to ensure the stability of the fog node; (12-5) indicates that the amount of tasks offloaded to the cloud node is less than the computing power of the cloud node; (12-6) indicates that the transmit power of the end device is not greater than its maximum transmit power; and (12-7) indicates that the task offload ratio is between 0 and 1, and the sum of the offload ratios to each helper node and the local offload ratio is 1.

[0060] Preferably, in step S5, the simulated annealing particle swarm optimization (SA-PSO) algorithm is described as follows:

[0061] (1) Particle Swarm Optimization Algorithm

[0062] Assume that in a D-dimensional search space, a swarm of M particles "flies" (i.e., searches). The position of the i-th particle is denoted as x. i (t)=[x i1 (t),x i2 (t),…x iD [(t)], the velocity of the i-th particle is expressed as v i (t)=[v i1 (t),v i2 (t),…v iD (t)]. The best position found by each particle during flight, compared with its historical best fitness value, is called the individual extreme value P. best=(p i1 ,p i2 ,…p iD The best position found by searching the entire population is called the global optimum G. best =(g i1 ,g i2 ,…g iD The velocity and position update formulas during the particle search process are as follows:

[0063] v id =wv id (t)+c1r1(p id (t)-x id (t))+c2r2(g id (t)-x id (t))(13)

[0064] x id (t+1)=x id (t)+v id (t+1)(14)

[0065] In the formula: i = 1, 2, 3... N; d = 1, 2, 3... D; w is the inertia weight factor; c1 and c2 are learning factors, also called acceleration factors, which are generally negative; r1 and r2 are two random numbers between [0, 1].

[0066] (2) Simulated Annealing Algorithm

[0067] Simulated annealing algorithm requires setting an initial temperature based on the initial state of the population during the initial iteration phase. In each iteration, the movement of particles inside the simulated solid under decreasing temperature is analyzed. The Mitropolis criterion is used to determine whether a new solution generated by disturbance replaces the global optimum, and its expression is as follows:

[0068]

[0069] Among them, E i (k) represents the internal energy of the i-th particle in the k-th iteration, i.e., the fitness value of the current particle; E g T represents the internal energy of the current population optimum. i This represents the current temperature. The temperature decreases linearly with each iteration; the optimization process is an alternating process of continuously finding new solutions and slowly cooling down. E i (k) completely determines the new state E it will generate next. i (k+1), compared to the previous E i (0) to E i (k-1) is irrelevant; this process is a Markov process.

[0070] This invention also discloses a time-delay-aware energy-saving multi-objective task offloading system for vehicle networking based on the above method, which includes the following:

[0071] Information and data acquisition module: Acquires information and data;

[0072] Calculation module: calculates energy consumption and latency;

[0073] Joint Optimization Problem Module: Determining the joint optimization problem of latency and energy consumption;

[0074] Optimization objective function module: Determines the optimization objective function;

[0075] Global optimal solution module: The SA-PSO algorithm is used to converge to a certain node in the solution space while bypassing local optima, thereby obtaining the global optimal solution;

[0076] Optimal offloading ratio and optimal transmit power solution module: solves the joint optimization problem, classifies it into different types according to different task offloading locations, and solves the optimal offloading ratio and optimal transmit power for different types;

[0077] Selection and Unloading Module: Select the end device or corresponding node with the most remaining energy, and select the optimal unloading ratio and optimal transmission power to unload the task.

[0078] Compared with the prior art, the present invention has the following technical effects:

[0079] This invention addresses two types of tasks: latency-sensitive and large-scale machine-type communication. It proposes a multi-objective optimization problem for latency and energy consumption, establishing a mathematical model of dynamic service flows. This model considers request characteristics, data stream size, and the impact of weather and geographical factors, comprehensively taking into account factors such as cost, supported service capacity, energy efficiency, latency performance, and service connectivity. Based on users' different resource processing needs, this invention refines user group tags and, based on a dynamic random service model for users with different priorities, utilizes a particle swarm optimization algorithm with simulated annealing to derive the optimal weighted sum of objectives, reducing algorithm runtime. Attached Figure Description

[0080] Figure 1 This is a flowchart of a multi-objective optimization method for cloud-fog fusion networks based on a trade-off between time delay and energy consumption, according to the present invention.

[0081] Figure 2 This is a network model diagram of a cloud-fog fusion network based on a trade-off between time delay and energy consumption, according to the present invention.

[0082] Figure 3 This is a block diagram of a multi-objective optimization system for cloud-fog fusion networks based on a trade-off between time delay and energy consumption, according to the present invention. Detailed Implementation

[0083] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, unless otherwise specified, the following embodiments and features described therein can be combined with each other.

[0084] like Figure 1 As shown, this embodiment is a multi-objective task offloading method for vehicle networking based on latency-aware energy saving. Based on the cloud-fog-edge model, the model minimizes latency and energy consumption. Taking into account network lifetime, the optimal offloading ratio and optimal transmission power are obtained.

[0085] This embodiment specifically includes the following steps:

[0086] S1, Operators obtain information and data.

[0087] S2. Performance analysis, calculating the energy consumption and latency of the system.

[0088] S3. Determine the joint optimization problem.

[0089] S4. Determine the objective function for optimization.

[0090] S5. Optimization problem solving.

[0091] There are three scenarios for unloading tasks to different nodes;

[0092] 1) End-to-end task offloading (abbreviated as end-to-end task offloading)

[0093] 2) End device-fog node task offloading (abbreviated as end-fog task offloading)

[0094] 3) Task offloading from end device to fog node to cloud node (abbreviated as end-fog-cloud task offloading)

[0095] S6. Solve for end-to-end task unloading.

[0096] S7, Solver-Fog Task Unloading.

[0097] S8, Task unloading from the solver-fog-cloud.

[0098] S9. The optimal solution is obtained;

[0099] S10, Execute the task to uninstall.

[0100] This embodiment applies to the cloud-fog-edge network model, and the model diagram is shown below. Figure 2As shown, in this network, end devices specifically include security cameras, smart door locks, smartphones, computers, smart TVs, and other devices in smart homes. End devices can connect to each other and can transmit tasks to fog nodes. Fog nodes are connected to remote cloud nodes via optical fibers.

[0101] This embodiment makes the following assumptions:

[0102] (1) The task generated by the terminal device can be divided into two subtasks. One subtask is executed locally, and the other subtask is offloaded to the helper node or fog node for computation.

[0103] (2) Since the calculation results are often small, the transmission time of the calculation results during task unloading is ignored.

[0104] (3) Assuming that D2D task unloading has no waiting time, it can satisfy latency-sensitive task unloading.

[0105] Specifically, the multi-objective task offloading method for cloud-fog fusion networks based on latency and energy consumption tradeoff in this embodiment is described in detail below:

[0106] S1. The operator obtains information data from the network at the same time.

[0107] The information data includes: the number of end devices N, the number of fog nodes m, and the average task data size θ of device i. i The average arrival rate λ of the task on terminal device i i Terminal i selects to unload ratio β i (0≤β i ≤1), terminal task arrival rate The uninstallation ratios for the four uninstallation modes—local processing, D2D, fog computing, and cloud computing—are respectively... Channel bandwidth B, signal-to-noise ratio Channel rate R i .

[0108] The average computing power of fog node i is The computing power u of cloud nodes C .

[0109] S2. Performance analysis, calculating the energy consumption and latency of the system;

[0110] Furthermore, step S2 specifically includes:

[0111] S2.1 Calculate the user's latency;

[0112] S2.2 Calculate the energy consumption of the entire system;

[0113] Specifically, in step S2.1:

[0114] User latency is actually the maximum of local latency and offload latency, calculated using the following formula to determine the transmission latency of the terminal device. and local computation latency T j1 :

[0115]

[0116]

[0117]

[0118] Where B represents the channel bandwidth. The signal-to-noise ratio is positively correlated with the transmit power, while the transmission delay... With channel rate R i The transmission power is negatively correlated, therefore the greater the transmission power, the longer the transmission delay. The smaller the value, the lower the transmission latency. With the unloading ratio β i They are positively correlated, therefore the unloading ratio β i The larger the value, the longer the transmission delay. The larger the value, the more important the parameters become. Therefore, the key parameters to focus on are the offloading ratio and the transmit power. By optimizing the offloading ratio and the transmit power, the latency and energy consumption can be minimized.

[0119] Furthermore, for end devices with different needs, computing needs to be offloaded to different network layers. For latency-sensitive tasks, D2D should be used first, while for tasks with high quantity, offloading to the cloud-fog converged network should be used first.

[0120] The average processing time T for D2D technology unloading tasks D for:

[0121]

[0122] The sum of service request rates offloaded from multiple end devices to the fog node is

[0123]

[0124] Considering the limited computing power of the fog node layer, when its computing power is insufficient to handle the current task offloading, the offloading ratio will be [percentage missing]. The task is offloaded to a cloud node with greater computing power. Since the fog node and the cloud node are connected via wired fiber optic cable, a fixed latency T will be generated. 0 Assuming the server performance in the cloud node can handle a sufficiently large volume of unloading tasks, tasks unloaded to the cloud node will be processed immediately, thus ignoring waiting time. Based on queuing theory, the execution process of the cloud node can be viewed as an M / M / ∞ queue. Assuming the computing power u of the cloud node... CAt this time, the latency of the cloud node layer is:

[0125]

[0126] In step S2.2:

[0127] According to communication theory research, the energy consumption generated during transmission is related to transmission power, load, and transmission time. The transmission power of the end device is defined as... for:

[0128]

[0129] S3. Determine the joint optimization problem.

[0130] Based on the definitions of energy consumption, time delay, and cost functions, the problem of minimizing these functions can be expressed as:

[0131] min{F}(23)

[0132] S4. Determine the objective function for optimization.

[0133] The objective of this invention is to determine the optimal vehicle task unloading strategy, minimizing the total time delay and energy consumption during task uploading, computation, and unloading. In this multi-objective optimization problem, a linear weighted sum method is employed, introducing a time delay weight coefficient R. T Energy consumption weighting coefficient R E The weighting coefficients can be set according to the current status of the on-board equipment and the task requirements, and must meet the requirements of R. T +R E =1. Define the objective function:

[0134]

[0135] S5. Optimization problem solving.

[0136] Specifically, the SA-PSO algorithm is as follows:

[0137] Assume that in a D-dimensional search space, a population of M particles "flies" (i.e., searches); the position of the i-th particle is denoted as x. i (t)=[x i1 (t),x i2 (t),…x iD [t], the velocity of the i-th particle is expressed as v i (t)=[v i1 (t),v i2 (t),…v iD [(t)]; The best position found by each particle during flight, compared with the particle's historical best fitness value, is called the individual extreme value P. best =(pi1 ,p i2 ,…p iD The best position found by searching the entire population is called the global optimum G. best =(g i1 ,g i2 ,…g iD The velocity and position update formulas during the particle search process are as follows:

[0138] v id =wv id (t)+c1r1(p id (t)-x id (t))+c2r2(g id (t)-x id (t))(13)

[0139] x id (t+1)=x id (t)+v id (t+1)(14)

[0140] In the formula: i = 1, 2, 3... N; d = 1, 2, 3... D; w is the inertia weight factor; c1 and c2 are learning factors, also known as acceleration factors, which are generally negative; r1 and r2 are two random numbers between [0, 1].

[0141] The simulated annealing algorithm is as follows:

[0142] The simulated annealing algorithm requires setting an initial temperature based on the initial state of the population during the initial iteration phase. In each iteration, the movement of particles inside the simulated solid under decreasing temperature is assessed using the Mitropolis criterion to determine whether a new solution generated by disturbance replaces the global optimum. The expression for this is as follows:

[0143]

[0144] Among them, E i (k) represents the internal energy of the i-th particle in the k-th iteration, i.e., the fitness value of the current particle; E g T represents the internal energy of the current population optimum. i This represents the current temperature; the temperature decreases linearly to a certain extent with each iteration, and the optimization process is an alternating process of continuously finding new solutions and slowly cooling down; E i (k) determines the new state E it will generate next. i (k+1), compared to the previous E i (0) to E i (k-1) is irrelevant; this process is a Markov process.

[0145] There are three scenarios depending on the task's uninstallation location;

[0146] 1) End-to-end task unloading;

[0147] 2) End-to-Fog task offloading;

[0148] 3) Task unloading from terminal to fog to cloud.

[0149] S6. Solve the end-to-end task unloading.

[0150] S6.1 Initialize and optimize system parameters, and calculate the initial minimum weighted sum F according to equation (24);

[0151] S6.2. Based on the system parameters, initialize the particle swarm, setting the initial position x0 and initial velocity v0 for each particle. The initial temperature in the simulated annealing algorithm parameters is set to T. i ;

[0152] S6.3 Calculate the fitness of each particle at the current annealing temperature and compare it with the historical best fitness value of that particle to update the individual extreme value P. best =(p i1 ,p i2 ,…p iD ), and global extremum G best =(g i1 ,g i2 ,…g iD Then, the position and velocity of the particle are iterated, using the following formula:

[0153] v id =wv id (t)+c1r1(p id (t)-x id (t))+c2r2(g id (t)-x id (t))(25)

[0154] x id (t+1)=x id (t)+v id (t+1)(26)

[0155] Calculate the fitness of each particle at the current temperature, and express it with probability p. i (k) Accept new value:

[0156]

[0157] S6.4 Repeat the iteration until the number of iterations reaches the threshold or the update result converges to the preset value.

[0158] S7. Solve for task unloading with only end-fog;

[0159] This step is the same as steps S6.1, S6.2, S6.3, and S6.4.

[0160] S8. Task unloading from the solver-fog-cloud;

[0161] This step is the same as steps S6.1, S6.2, S6.3, and S6.4.

[0162] S9. Obtain the optimal solution: Select the optimal unloading ratio and optimal transmission power under different conditions. When selecting nodes, prioritize the node with the most remaining energy for task unloading to minimize overall energy consumption and latency.

[0163] S10, Execute the task to uninstall.

[0164] like Figure 3 As shown, this embodiment discloses a time-delay-aware energy-saving multi-objective task offloading system for vehicle networking based on the above method, which includes the following:

[0165] Information and data acquisition module: Acquires information and data;

[0166] Calculation module: calculates energy consumption and latency;

[0167] Joint Optimization Problem Module: Deterministic joint optimization problems involving time delay and energy consumption;

[0168] Optimization objective function module: Determines the optimization objective function;

[0169] Global optimal solution module: It uses the SA-PSO algorithm to converge to a certain node in the solution space while bypassing local optima, thereby obtaining the global optimal solution;

[0170] Optimal offloading ratio and optimal transmit power solution module: solves the joint optimization problem, classifies it into different types according to different task offloading locations, and solves the optimal offloading ratio and optimal transmit power for different types;

[0171] Selection and Unloading Module: Select the end device or corresponding node with the most remaining energy, and select the optimal unloading ratio and optimal transmission power to unload the task.

[0172] Other aspects of this embodiment can be found in the embodiments described above.

[0173] The above description is merely a detailed explanation of preferred embodiments and principles of the present invention. For those skilled in the art, there may be changes in specific implementation methods based on the ideas provided by the present invention, and these changes should also be considered within the scope of protection of the present invention.

Claims

1. A multi-objective task offloading method for vehicle-to-everything (V2X) networks based on time-delay-aware energy saving, characterized by: Follow these steps: S1. Obtain information data; S2, calculate energy consumption and latency; S3. Determine the joint optimization problem of time delay and energy consumption; specifically, in this step, a linear weighted sum method is used, introducing a time delay weight coefficient R. T Energy consumption weighting coefficient R E The weighting coefficients are set based on the current status of the onboard equipment and the task requirements, and must meet the R... T +R E =1; Define the objective function: in, For local power consumption of terminal devices; Transmit power to end devices; When the car has a full battery, the time delay is the primary optimization objective, and the weighting coefficient is set to... When the battery is low, optimize energy consumption and set [the appropriate settings]. ; The joint optimization problem for this step is described as follows: Let the computation rate of the fog node be u. F The transmit power of the fog node is The cloud node's transmit power is Based on the definitions of energy consumption and time delay functions, the problem of minimizing the time delay and energy consumption functions can be expressed as: Among them, the uninstallation ratios for the four uninstallation modes—local processing, D2D, fog computing, and cloud computing—are respectively... Average arrival rate of tasks on terminal device i The average computing power of fog node i is The average task load rate of terminal device i is The average transmission rate of the wireless port of terminal device i is u. D The total service request rate of the fog nodes is Equation (12-1) represents the size of the local computing load, which does not exceed the local computing capacity, p i Equation (12-2) indicates that the amount of tasks offloaded through D2D does not exceed the amount of tasks that the helper nodes can handle; Equation (12-3) indicates that the sum of the service request rates of the fog nodes is less than the computing rate of the fog nodes; Equation (12-4) indicates that the sum of the service request rates of the fog nodes is less than the transmission power of the fog nodes; Equation (12-5) indicates that the amount of tasks offloaded to the cloud nodes is less than the computing capacity of the cloud nodes; Equation (12-6) indicates that the transmission power of the end device is not greater than its maximum transmission power; Equation (12-7) indicates that the task offload ratio is between 0 and 1, and the sum of the ratios of offloaded to each helper node and local offload is 1. S4. Determine the objective function for optimization; S5. By using the SA-PSO algorithm to converge to a certain node in the solution space while bypassing local optima, the global optimal solution can be obtained. S6. Solve the joint optimization problem, classify it into different types according to different task offloading locations, and solve for the optimal offloading ratio and optimal transmission power under different types; S7. Select the terminal device or corresponding node with the most remaining energy, and select the optimal offloading ratio and optimal transmission power to perform task offloading.

2. The method for offloading multi-objective tasks in vehicle networking based on time-delay awareness and energy saving as described in claim 1, characterized in that: step In S1, the information data includes: the number of end devices N, the number of fog nodes m, and the average task data size of end device i. Terminal device i selects the offload ratio Terminal task arrival rate Channel bandwidth B, signal-to-noise ratio Channel rate R i .

3. The multi-objective task offloading method for vehicle networking based on time-delay perception energy saving as described in claim 2, characterized in that: In step S2, the average delay is calculated as follows: a) Transmission latency: User unloading tasks to cloud nodes, fog nodes, or transmitting them to other device nodes via D2D are collectively referred to as helper nodes; when a task is unloaded to a helper node, only the user's uplink transmission latency is considered. : Where B represents the channel bandwidth. The signal-to-noise ratio (SNR) is positively correlated with the transmit power, while the transmission delay... With channel rate R i The transmission power is negatively correlated, therefore the greater the transmission power, the longer the transmission delay. The smaller the value, the lower the transmission latency. Relative to uninstallation ratio Positively correlated, therefore the uninstallation ratio The larger the value, the longer the transmission delay. The larger; b) Local processing latency: Assume the average computing power of terminal device i is... The local computation latency T of the terminal device is obtained. j1 : c) D2D processing latency: Based on the M / M / 1 queuing theory, the average processing time for D2D offloading tasks is: d) Fog computing processing latency: Allows multiple end devices to offload tasks to the fog node layer, at which point the maximum request rate of the fog node is... Let the average computing power of fog node i be . Then the total service request rate of the fog nodes is: The total time delay of the fog nodes is: e) Cloud computing processing latency: When the computing power of the fog node layer is insufficient to handle the current task offloading, the offloading ratio will be [percentage missing]. The task is offloaded to a cloud node with greater computing power. Since the fog node and the cloud node are connected via wired fiber optic cable, a fixed latency T will be generated. 0 Assume that the server performance in the cloud node can handle the workload of offloading tasks. Therefore, tasks offloaded to the cloud node will be processed immediately, so the waiting time is ignored. According to queuing theory, the execution process of the cloud node can be regarded as a queuing process. Queue, assuming the computational power u of the cloud node c At this time, the latency of the cloud node layer is: 。 4. The multi-objective task offloading method for vehicle networking based on time-delay perception energy saving as described in claim 3, characterized in that: In step S2, the energy consumption is calculated as follows: Energy consumption during transmission is related to transmission power, load, and transmission time, as well as the transmission power of the end device. for: Local power consumption of terminal devices for: Where, q i Let be the unit operating power coefficient of the i-th device, which is a fixed constant; m jk Specifically as follows: Restriction: For any row k of matrix M, we have This means that each task can only choose one task uninstallation method.

5. The multi-objective task offloading method for vehicle networking based on time-delay perception energy saving as described in claim 4, characterized in that: In step S5, the simulated annealing particle swarm optimization (SA-PSO) algorithm is described as follows: (1) Particle Swarm Optimization Algorithm Suppose that in a D-dimensional search space, a swarm of M particles "flies" (i.e., searches); the position of the i-th particle is denoted as... The velocity of the i-th particle is expressed as The best position found by each particle during its flight, compared to its historical best fitness value, is called the individual extreme value. The best position found by searching the entire population is called the global optimum. The velocity and position update formulas during the particle search process are as follows: In the formula: i=1,2,3...N; d=1,2,3...D; w is the inertia weight factor; c1 and c2 are learning factors, also known as acceleration factors, which are negative; r1 and r2 are two random numbers between [0,1]. (2) Simulated annealing algorithm The simulated annealing algorithm requires setting an initial temperature based on the initial state of the population during the initial iteration phase. In each iteration, the movement of particles inside the simulated solid under decreasing temperature is analyzed, and the Mitropolis criterion is used to determine whether a new solution generated by disturbance replaces the global optimum. The expression is as follows: Among them, E i (k) represents the internal energy of the i-th particle in the k-th iteration, i.e., the fitness value of the current particle; E g T represents the internal energy of the current population optimum. i This represents the current temperature; the temperature decreases linearly to a certain extent with each iteration, and the optimization process is an alternating process of continuously finding new solutions and slowly cooling down; E i (k) determines the new state E it will generate next. i (k+1), compared to the previous E i (0) to E i (k-1) is irrelevant.

6. A multi-objective task offloading method for vehicle networking based on time-delay awareness and energy saving as described in any one of claims 1-5, characterized in that: In step S6, based on different task unloading locations, there are three types: task unloading from end device to end device, task unloading from end device to fog node, and task unloading from end device to fog node to cloud node.

7. A time-delay-aware energy-saving multi-objective task offloading system for vehicle networking based on the method of any one of claims 1-6, characterized in that: Includes the following modules: Information and data acquisition module: Acquires information and data; Calculation module: calculates energy consumption and latency; Joint Optimization Problem Module: Deterministic joint optimization problems involving time delay and energy consumption; Optimization objective function module: Determines the optimization objective function; Global optimal solution module: It uses the SA-PSO algorithm to converge to a certain node in the solution space while bypassing local optima, thereby obtaining the global optimal solution; Optimal offloading ratio and optimal transmit power solution module: solves the joint optimization problem, classifies it into different types according to different task offloading locations, and solves the optimal offloading ratio and optimal transmit power for different types; Selection and Unloading Module: Select the end device or corresponding node with the most remaining energy, and select the optimal unloading ratio and optimal transmission power to unload the task.

Images