A Mobile Edge Node Offloading Method Applied to Vehicular Ad Hoc Networks

Abstract

The present invention belongs to the field of Internet of Vehicles and mobile edge computing, and specifically relates to a mobile edge node unloading method applied to the Internet of Vehicles. The present invention is aimed at the application scenario of mobile edge computing in the Internet of Vehicles, and proposes a computing task unloading strategy optimization method for multiple mobile terminals and multi-address edge nodes. A joint optimization strategy is adopted to allow the vehicle mobile terminals in the same Internet of Vehicles system scenario to reduce the total time consumed by the system to complete the task in the process of unloading computing tasks with the edge nodes distributed in the scenario. Based on the convex optimization principle, the present invention takes the computing task offloading strategy and the vehicle mobile terminal transmission power as the two optimization ideas, and constructs a mathematical model of task unloading time consumption based on mobile edge computing. Taking into account the randomness of computing task release and the parallelism of task processing, and under the premise of ensuring communication quality, the coordinates of the vehicle mobile terminal that releases the computing task, the task size and the edge node status are input into the above model, and then a computing task unloading strategy close to the global minimum time consumption is solved. Compared with the common nearest node unloading method, the present invention proposes a method for solving the mobile edge computing offloading strategy that effectively reduces time consumption by adjusting the propagation power and reasonably allocating the offloading weight. At the same time, through the conversion method of non-convex constraints, it can effectively achieve a significant reduction in the time consumption of mobile edge computing task offloading in complex vehicle network environments, and has broad application prospects.

Description

Technical Field

[0001] The present invention discloses an edge node computing offloading method based on convex optimization principle, which belongs to the field of Internet of Vehicles and mobile edge computing, and specifically relates to mobile edge node offloading of computing tasks applied in Internet of Vehicles. Background Art

[0002] In recent years, the rapid development of 5G technology has brought unprecedented technological change and industrial integration to a growing number of industries with its advantages of high speed, low latency, and massive connectivity. The emerging concept of the Internet of Vehicles (IoV) has rapidly become a significant and influential research field with the growing influence of 5G, and its application prospects are becoming increasingly broad. The intelligent connected vehicle industry reached 165.6 billion yuan in 2019 and is expected to grow to 1,312.04 billion yuan by 2024. Traditional connected vehicle communications are evolving towards the Internet of Things (IoV), a technology that will play a crucial role in next-generation intelligent transportation systems and autonomous driving. Several IoV communication modes, including V2X (vehicle-to-vehicle), vehicle-to-infrastructure, vehicle-to-network, and vehicle-to-person, have become the subject of extensive research. Among them, V2N (vehicle-to-network) communication is a vehicle-to-network communication mode. Traditional communication technologies cannot meet the low latency, low fault tolerance, and short-term communication requirements of intelligent transportation and autonomous driving. The advent of 5G makes it possible to meet these requirements. To reduce the excessive latency and increased error rates associated with long communication distances, mobile edge computing (MEC) began to emerge in the 4G era and was integrated into the network architecture in the 5G era. The European Telecommunications Standards Institute (ETSI) formally defined its concept in 2014 and has since standardized it. Unlike traditional centralized computing, mobile edge computing offloads computing services from central servers, allowing centralized scheduling. Physically, edge computing nodes further shorten the distance between the information source and the computing service provider, reducing latency and error rates while also alleviating the computing pressure on central nodes, ensuring efficient and high-quality information transmission. Furthermore, shorter transmission distances require less channel gain at the communication end, further reducing energy costs.

[0003] In the face of multi-connection, high-concurrency IoV communication scenarios and the computing tasks of massive mobile vehicle terminals, ensuring the timeliness of communication and computing is of paramount importance. Mobile vehicle terminals have relatively fast movement speeds and large data computing tasks, requiring low-latency, reliable communication to ensure the safety of the mobile process. Simply offloading tasks to edge nodes based on proximity or channel quality can cause edge nodes near densely populated areas or with abundant computing resources to become overloaded, or cause excessive task processing time due to the number of connections instantly filling up the channel, making it difficult to meet the system's low-latency, reliable communication requirements.

[0004] In summary, current mobile edge node communications still suffer from imbalanced task offloading, inadequate resource utilization, and a lack of specificity in fusion strategies. Therefore, this paper explores the distributed offloading of edge node computing tasks, fully considering the energy consumption of vehicle terminals and edge computing nodes, and constructing a stable, reliable, and low-latency total time consumption model for the Internet of Vehicles system. Summary of the Invention

[0005] This paper addresses the shortcomings and deficiencies of existing technologies and proposes a stable, low-latency, and low-energy mobile edge offloading method. Based on the principles of convex optimization, this paper designs propagation power and offloading weight subproblems. Using a large-scale fading channel model with line-of-sight link loss, this paper effectively constructs an offloading time model for mobile edge computing. This approach has important implications for optimizing task offloading strategies in mobile edge computing.

[0006] To achieve the above objectives, the technical solutions of the present invention include the following:

[0007] Step 1: Obtain the current coordinates and computing capabilities of each vehicle mobile terminal and edge computing node in the system, set the initial propagation power and calculate the channel rate.

[0008] Step 2: Construct a total time consumption model for the Internet of Vehicles system based on the parallel release of terminal computing tasks and the parallel processing of tasks by edge computing nodes.

[0009] (2-1): Establish a total time consumption model for the Internet of Vehicles system. The total time consumption required to complete each task is calculated using the computing task size, terminal local processing capacity, offload target node processing capacity, and offload weight.

[0010] (2-1-1): Computing terminal The time consumed by local calculation.

[0011] (2-1-2): Computing terminal To the edge node The propagation takes time.

[0012] (2-1-3): Computing terminal Offloading to Node The task calculation time is consuming.

[0013] In this scenario, the time taken to offload tasks from release to calculation is negligible. The action of offloading to the edge node is carried out in parallel with the action of starting calculation locally. At the same time, considering that the size of the data sent back to the terminal after the calculation is completed is much different from the size of the data of the calculation task itself, the time taken to send back is also negligible.

[0014] (2-2): Disassemble the target model and decompose the problem model into two sub-problems: finding the optimal task offloading matrix under a fixed transmission power and finding the optimal transmission power under a fixed task offloading matrix.

[0015] (2-2-1) The objective problem is stated as minimizing the total time consumption of the system.

[0016] (2-2-2) The target problem is decomposed into the optimal task offloading matrix sub-problem under fixed transmission power.

[0017] (2-2-3) The target problem is decomposed into the optimal propagation power sub-problem under a fixed task offloading matrix.

[0018] (2-3): Based on the principle of convex optimization, all functions and constraints in the two subproblems are transformed into convex functions and convex constraints. Non-convex constraints are transformed into convex constraints by introducing auxiliary variables to set the upper bound of the constraint, and the minimum value is indirectly obtained.

[0019] (2-4): The convex functions and convex constraints transformed from the two sub-problems are solved by convex optimization, and the fixed propagation power or fixed task offloading weight matrix is sent into the model for alternating solution to obtain the optimal offloading strategy.

[0020] Step 3: Initialize the model parameters, set the initial propagation power and initial offloading weight matrix with the time consumption of completely local processing as the upper bound of energy consumption, and use the interior point method to alternately solve the two sub-problems to obtain the optimal offloading strategy.

[0021] Step 4: Substitute the vehicle mobile terminal parameters to be tested, the edge computing node parameters, and the solved optimal task offloading matrix and propagation power into the system total time consumption model to obtain the time consumption optimization result.

[0022] Advantages and positive effects of the present invention

[0023] Compared with the prior art, the present invention has the following advantages and positive effects:

[0024] First, the present invention addresses the urgent problem of edge mobile computing offloading in intelligent transportation systems, and proposes a method for solving optimized offloading strategies that can fully utilize the computing resources and system energy of edge computing nodes, which can achieve sufficient, accurate and fast scenario task offloading calculations.

[0025] Second, the present invention innovatively constructs a total time consumption model of the Internet of Vehicles system based on the parallel release of terminal computing tasks and the parallel processing of tasks by edge computing nodes, so that the optimization goal is transformed from individual terminal computing tasks to the total time consumption of the system, and at the same time, the quality of the channel is guaranteed by adding constraints in the model.

[0026] Third, the present invention establishes a method for solving the optimal strategy for the total system time consumption based on the convex optimization principle, so that the system time consumption can obtain the optimal offloading strategy for computing tasks based on the computing resources and energy consumption of each communication node in the system in a shorter time. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a flowchart of the implementation process of the present invention;

[0028] Figure 2 These are the simulation results under different bandwidths in the present invention;

[0029] Figure 3 is the simulation result under different noises in the present invention;

[0030] Figure 4 These are the simulation results under different energy limits in the present invention;

[0031] Figure 5 The simulation results of three comparative experiments under different bandwidths are shown: the joint optimization method of the present invention and the benchmark experiment of comparative experiment 1 in which only the offloading weight matrix is optimized, and the comparative experiment 2 in which only the propagation power is optimized;

[0032] Figure 6 The simulation results of three comparative experiments under different noise conditions are shown: the joint optimization method of the present invention and the benchmark experiment of comparative experiment 1 in which only the offloading weight matrix is optimized, and the comparative experiment 2 in which only the propagation power is optimized;

[0033] Figure 7 These are the simulation results of two comparative experiments, namely, the joint optimization method of the present invention and the comparative experiment 1 of optimizing only the offloading weight matrix and the comparative experiment 2 of optimizing only the propagation power under different energy upper limits; DETAILED DESCRIPTION

[0034] The implementation process and effects of the present invention will be further described in detail below with reference to the accompanying drawings.

[0035] Reference Figure 1 , the implementation steps of the present invention include the following:

[0036] Step 1: Obtain the current coordinates and computing capabilities of each vehicle mobile terminal and edge computing node in the system, set the initial propagation power and calculate the channel rate. , whose coordinates are Q i = [ x i , y i ]( ∀ i ∈ N ) , the edge computing node is , whose coordinates are q j = [ x j , y j ]( ∀ j ∈ M ) , the path loss is calculated according to the large-scale fading model and is expressed as

[0037]

[0038] Here Represents the channel gain per unit distance, then the terminal To Node The uplink transmission rate can be calculated as follows:

[0039]

[0040] in Indicates terminal The transmission power, Denotes the transmission noise power. The channel rate from each terminal to each node can be calculated by equations (1) and (2).

[0041] Step 2: Construct a total time consumption model for the Internet of Vehicles system based on the parallel release of terminal computing tasks and the parallel processing of tasks by edge computing nodes.

[0042] (2.1): Establish a total time consumption model for the Internet of Vehicles system. The total time consumption required to complete each task is calculated by computing the task size, the terminal's local processing capacity, the offload target node's processing capacity, and the offload weight. Specifically, the time consumption for each task is given by the following formula:

[0043] (2.1.1): Terminal The time consumption of local calculation is given by formula (3)

[0044]

[0045] in Indicates the size of the computing task, Indicates terminal of local computing power, including x ij ∈ [ 0 , 1 ] , indicating that the terminal Offloading to Node The ratio of the task size calculated at to the total task size is , and Equation (3) calculates the time consumed for local processing of tasks excluding those offloaded to edge nodes.

[0046] (2.1.2): Terminal To the edge node The propagation time of is given by formula (4)

[0047]

[0048] (2.1.3): Terminal Offloading to Node The computation time of the task at is given by formula (5)

[0049]

[0050] in Representation node computing power.

[0051] In the scenario, the time taken for local offloading from task publishing to calculation is negligible, and the action of offloading to the edge node is parallel to the action of starting local calculation. At the same time, considering that the size of the data sent back to the terminal after the calculation is completed is much different from the size of the data of the calculation task itself, the time taken for the return is also negligible. Therefore, the terminal The total computation time required to complete the task is given by (6):

[0052]

[0053] in It is a conversion variable to unify the units of propagation time and computation time, taking into account the time slot All tasks generated by terminals are released at the same time, so the time consumption of the entire scenario system is given by formula (7)

[0054]

[0055] (2.2): Decompose the target model into two sub-problems: finding the optimal task offloading matrix under a fixed transmission power and finding the optimal transmission power under a fixed task offloading matrix. This step is specifically implemented as follows:

[0056] (2.2.1) The objective problem is to minimize the total time consumption of the system, which can be expressed as formula (8)

[0057]

[0058] (2.2.2) Solve the optimal task offloading matrix under fixed transmission power. The sub-problem can be expressed as Equation (9)

[0059]

[0060]

[0061]

[0062]

[0063]

[0064]

[0065] Among them, equations (9b) to (9f) are the constraints related to the task offloading matrix, and equation (9d) indicates that tasks with a channel rate less than or equal to the threshold are not offloaded. Equations (9e) and (9f) are the energy constraints of the mobile terminal and the edge node, respectively. A parameter representing the weight of transmission power relative to energy consumption.

[0066] (2.2.3) Solve the optimal propagation power under a fixed task offloading matrix. The sub-problem can be expressed as Equation (10)

[0067]

[0068]

[0069] The formula (10) is not entirely about the convex function and convex constraints of the propagation power. After adding the auxiliary variable 、 , transform equation (10) into equation (11) and solve:

[0070]

[0071]

[0072] Step 3: Initialize the model parameters, set the initial propagation power and initial offloading weight matrix with the time consumption of completely local processing as the upper bound of energy consumption, and use the interior point method to alternately solve the two sub-problems to obtain the optimal offloading strategy. This step is specifically implemented as follows:

[0073] (3.1) Set the initial propagation power that satisfies the constraints , initialize the number of iterations and the convergence threshold , all current tasks are run locally, that is The total system delay is .

[0074] (3.2) In In the iteration, Bring it into the sub-problem of optimizing weight unloading matrix and use interior point method to find Minimum delay unloading matrix under power setting .

[0075] (3.3) Bring it into the solution of the propagation power sub-problem and obtain the propagation power with the minimum delay under the current edge offloading strategy .

[0076] (3.4) Use and Substitute into formula (8) to calculate the total delay of the current system ,calculate Or when the maximum number of iterations is reached, the loop ends and the final result is output, otherwise , return to step (3.2).

[0077] Step 4: Substitute the vehicle mobile terminal parameters to be tested, the edge computing node parameters, and the solved optimal task offloading matrix and propagation power into the system total time consumption model to obtain the system total time consumption optimization result.

Claims

1. A method for offloading traffic from a mobile edge node based on convex optimization principles and convex function upper bound approximation, the method comprising the following steps: (1) Obtain the current coordinates and computing capabilities of each vehicle mobile terminal and edge computing node in the system, set the initial propagation power, and calculate the channel rate; (2) Construct a total time consumption model for the Internet of Vehicles system based on the parallel release of terminal computing tasks and the parallel processing of tasks by edge computing nodes; use the obtained channel rate to add constraints to the time consumption model to ensure communication quality; Based on the total system time consumption model, the original model is decomposed into two sub-problem models from the perspectives of optimizing the offloading weight matrix and optimizing the transmission power. The two sub-problem models are solved alternately to obtain the approximate optimal solution for the model. 1) Establish a total time consumption model for the connected vehicle system. All computing tasks are published and processed in parallel. The maximum value of all completed tasks is used as the total system time consumption. The total time consumption required to complete each task is calculated using the computing task size, the terminal's local processing capacity, the offload target node's processing capacity, and the offload weight. 2) Decompose the target model into two sub-problems: finding the optimal task offloading matrix under a fixed transmission power and finding the optimal transmission power under a fixed task offloading matrix. By alternately solving these two sub-problems, we continuously approach the approximate optimal solution of the problem. 3) Based on the principle of convex optimization, all functions and constraints in the two subproblems are transformed into convex functions and convex constraints. Specifically, for non-convex constraints, an auxiliary variable is introduced to set the upper bound of the constraint, and the minimum value of the non-convex constraint is indirectly obtained by minimizing the upper bound. 4) The convex functions and convex constraints transformed from the two subproblems are solved by convex optimization. Fixed propagation power or fixed task offloading weight matrix is fed into the model for alternating solutions to obtain the optimal offloading strategy. (3) Initialize the model parameters, set the initial propagation power and the initial offloading weight matrix with the time consumption of completely local processing as the upper bound of energy consumption, and use the interior point method to alternately solve the two sub-problems to obtain the optimal offloading strategy; 1) Set the initial propagation power that satisfies the constraints , initialize the number of iterations and the convergence threshold , calculate the initial channel rate, taking all current tasks running locally, that is, The total system time consumption is ; 2) In In the iteration, Bring it into the sub-problem of optimizing weight unloading matrix and use interior point method to find Minimum delay unloading matrix under power setting ; 3) Bring it into the solution of the propagation power sub-problem and obtain the propagation power with the minimum delay under the current edge offloading strategy ; 4) Use and Bring in the system total time consumption model to calculate the total time consumption of the current system ,calculate ; 5) Repeat the steps until satisfied Or when the maximum number of iterations is reached, the loop ends and the final result is output; (4) The parameters of the vehicle mobile terminal to be tested, the parameters of the edge computing nodes, and the solved optimal task offloading matrix and propagation power are brought into the system total time consumption model to obtain the time consumption optimization result.

Images