Intelligent metasurface-assisted vehicle-mounted edge computing task unloading method

Abstract

The invention relates to the technical field of edge computing, in particular to an intelligent metasurface-assisted vehicle-mounted edge computing task unloading method, and aims to solve the problems that a wireless unloading transmission link of an intelligent metasurface-assisted vehicle-mounted edge computing scene is high in path loss, prone to being blocked and the like. An intelligent metasurface is used for improving the channel quality of a wireless unloading transmission link, specifically, an intelligent metasurface-assisted vehicle-mounted edge calculation unloading model is established, the problem that the transmission link is blocked when a vehicle moves is solved, meanwhile, the intelligent joint optimization problem of intelligent metasurface phase shift, amplitude, unloading decision and resource allocation is established, and the intelligent metasurface-assisted vehicle-mounted edge calculation unloading model is established. Finally, the dynamic coupling non-convex problem is solved through a DDPG algorithm, the calculation unloading cost is reduced under the condition that link blocking is reduced, and therefore the cost composed of delay and energy consumption of a vehicle-mounted edge calculation system is minimized.

Description

Technical Field

[0001] The present invention relates to the field of edge computing technology, and in particular to a method for offloading vehicle-mounted edge computing tasks assisted by an intelligent metasurface. Background Art

[0002] In recent years, although a lot of research has been done on mobile edge computing (MEC), there are still few studies in ultra-dense heterogeneous cellular network scenarios, whether in terms of computing offloading, resource allocation, or optimization of latency and energy consumption. The rise of reconfigurable intelligent surface (RIS) technology provides a new solution for vehicle edge computing (VEC) computing offloading and resource allocation. RIS is a new type of wireless communication auxiliary technology. It consists of a group of reflective elements. It changes the propagation environment of wireless signals by adjusting the phase shift and amplitude of the reflective unit, thereby enhancing the signal reception quality, expanding the coverage and increasing the communication rate. It has the advantages of low cost, convenient deployment and high spectrum efficiency. With the rapid development of mobile Internet, the network space has become more complex, and the emerging new applications and new scenarios have different requirements for network performance. Among them, system latency is an important evaluation indicator that affects performance. For example, in the scenario of autonomous driving, vehicles need to sense and respond to environmental changes in a complex traffic environment in a timely manner. This requires real-time communication between vehicles, pedestrians and road facilities, and the latency requirement reaches several milliseconds. At the same time, due to the diversity, complexity and intelligence of mobile applications, when the edge network becomes unstable due to energy exhaustion of terminals in the network, more powerful devices are needed to execute these complex applications. Summary of the invention

[0003] The purpose of the present invention is to provide a method for offloading vehicle edge computing tasks assisted by an intelligent metasurface, aiming to solve the problems of high path loss, susceptibility to blockage and poor communication quality faced by wireless offloading transmission links in vehicle edge computing scenarios, and to minimize the total cost of latency and energy consumption.

[0004] To achieve the above object, the present invention provides a method for offloading vehicle edge computing tasks assisted by an intelligent metasurface, comprising the following steps:

[0005] Step 1: Build an edge computing scenario including base stations, smart metasurfaces, and vehicles;

[0006] Step 2: Construct vehicle movement model;

[0007] Step 3: Construct a communication model based on smart metasurface;

[0008] Step 4: Build a local computing model;

[0009] Step 5: Build an offloading calculation model;

[0010] Step 6: Build cost model and optimization objectives;

[0011] Step 7: Optimize the dynamic coupling problem through DDPG to solve the offloading decision.

[0012] Optionally, the intelligent metasurface acts between the edge base station and the vehicle-mounted user, responsible for improving the signal coverage and reducing the impact of blocking the direct link between the vehicle and the base station. All vehicles adopt a partial offloading strategy, with part of the computing tasks executed locally and part offloaded to the mobile edge computing layer for execution.

[0013] Optionally, in step 2, the occlusion of user signals at different locations is different. When the vehicle is driving dynamically, its speed satisfies independent and identical distribution and obeys the Gauss-Markov random process. The speed change v of the vehicle n The relational expression of (t+1) is as follows:

[0014]

[0015] Among them, v n (t) represents the speed of vehicle n at time slot t, ψ∈[0,1] represents the memory depth parameter of the past speed of vehicle n during its motion, and denote the asymptotic mean and asymptotic standard deviation of the speed of vehicle n, respectively. represents an uncorrelated random Gaussian process with zero mean and unit variance.

[0016] Optionally, in the communication model based on the intelligent metasurface, the auxiliary offloading link H between the vehicle user and the base station n (t), auxiliary offloading link between base station and smart metasurface Direct link between vehicle users and base stations The relational expression of the signal Y(t) received by the base station through the complete link assisted by the smart metasurface at time slot t is as follows:

[0017]

[0018] Among them, β represents the path loss exponent, α l represents the reference channel power gain at unit distance, (x n (t),y n (t)) represents the position of vehicle n, (x BS (t),y BS (t)) represents the location of the base station, represents the small-scale fading coefficient, ζ R,B represents the Rice coefficient between the smart metasurface and the base station, (x R (t),y R (t)) represents the position of the smart metasurface, denote the channel gains of the line-of-sight link and the non-line-of-sight link between the smart metasurface and the base station, respectively; ζ n,R represents the Rice coefficient between vehicle n and the smart metasurface, represents the channel gain of the line-of-sight link and the non-line-of-sight link between vehicle n and the smart metasurface respectively; the signal-to-noise ratio γ between the vehicle user and the base station at time slot t n,B (t), the data transmission rate γ between the vehicle user and the base station at time slot t n The relational expression of (t) is as follows:

[0019]

[0020] Among them, ε n,B ∈{0,1} indicates whether the smart metasurface assists in offloading transmission, σ 2 Represents the noise power.

[0021] Optionally, during the execution of step 4, the computing tasks generated by the vehicle user are partially executed locally and partially offloaded to the mobile edge computing server for execution; the local computing delay of the vehicle user Local energy consumption of onboard users The relational expression is as follows:

[0022]

[0023] in, represents the local computing power of vehicle n, that is, the number of CPU cycles per second, and κ represents the effective correlation coefficient, which is related to the chip structure of vehicle n.

[0024] Optionally, the offloading calculation in step 5 needs to go through three steps. Vehicle n needs to transmit the computing task to the mobile edge computing server, and then the mobile edge computing server allocates computing resources to perform the offloading task, and finally returns the calculation result to vehicle n through the base station. The transmission delay of the computing task offloading to the mobile edge computing server Execution delay of computing tasks on mobile edge computing servers Total delay of unloading The expression is as follows:

[0025]

[0026] in, represents the computing capacity allocated to vehicle n by the mobile edge computing server and the offloading computing energy consumption of vehicle n Calculate the total latency of the task The total energy consumption of the computing task The expression is as follows:

[0027]

[0028] Optionally, during the execution of step 6, the total cost Γ(t) of the vehicle user is expressed as follows:

[0029]

[0030] Among them, χ1, χ2 represent weight factors and χ1+χ2=1;

[0031] In order to ensure cost minimization under the maximum tolerable delay, the expression of the joint optimization problem of minimizing the total cost of computational offloading of on-board edge computing assisted by intelligent metasurface is as follows:

[0032]

[0033] stC1:

[0034] C2:

[0035] C3:

[0036] C4:χ1+χ2=1

[0037] C5:

[0038] C6:

[0039] C7:

[0040] Among them, C1 represents the value range of the unloading variable, C2 represents the amplitude range of the intelligent metasurface reflection unit, C3 represents the phase shift range of the intelligent metasurface reflection unit, C4 represents the weight factor, C5 represents the constraint on the computing power of each vehicle, C6 represents that the sum of the computing resources allocated to each task cannot exceed the computing power of the mobile edge computing server, and C7 represents that the total delay of the computing task is less than the maximum tolerable delay.

[0041] Optionally, the cost consisting of the latency and energy consumption of the mobile edge computing system is minimized by jointly optimizing the offloading decision, resource allocation, smart metasurface amplitude, and phase-shift reflection coefficient parameters through the DDPG algorithm.

[0042] The present invention provides a method for offloading vehicle edge computing tasks assisted by an intelligent metasurface. Aiming at the problems of high path loss and susceptibility to obstruction faced by wireless offloading transmission links in intelligent metasurface-assisted vehicle edge computing scenarios, an intelligent metasurface is used to improve the channel quality of wireless offloading transmission links. Specifically, an intelligent metasurface-assisted vehicle edge computing offloading model is established to overcome the problem of transmission link obstruction generated when the vehicle moves. At the same time, an intelligent joint optimization problem of intelligent metasurface phase shift, amplitude, offloading decision and resource allocation is established. Finally, a DDPG algorithm is used to solve the dynamic coupling non-convex problem, thereby reducing the computing offloading cost while reducing link obstruction, thereby minimizing the cost consisting of mobile edge computing system delay and energy consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0044] Figure 1 It is a schematic flow chart of the steps of an intelligent metasurface-assisted vehicle-mounted edge computing task offloading method of the present invention.

[0045] Figure 2 It is a schematic diagram of a model of an intelligent metasurface-assisted vehicle-mounted edge computing system according to a specific embodiment of the present invention.

[0046] Figure 3 It is a schematic diagram of constructing the DDPG algorithm training framework of the present invention.

[0047] Figure 4 It is a comparison chart of total costs under different numbers of vehicles when comparing algorithms in a specific embodiment of the present invention.

[0048] Figure 5 It is a total cost comparison chart under different MEC server computing capabilities when comparing algorithms in a specific embodiment of the present invention. DETAILED DESCRIPTION

[0049] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0050] See also Figure 1 The present invention provides a method for offloading vehicle edge computing tasks assisted by an intelligent metasurface, comprising the following steps:

[0051] S1: Build edge computing scenarios including base stations, smart metasurfaces, and vehicles;

[0052] S2: Construct vehicle mobility model;

[0053] S3: Construct a communication model based on smart metasurfaces;

[0054] S4: Build a local computing model;

[0055] S5: build an offloading calculation model;

[0056] S6: Build cost model and optimization objectives;

[0057] S7: Optimize the dynamic coupling problem through DDPG to solve the offloading decision.

[0058] The following is a further description in conjunction with specific embodiments and execution steps (the subsequent terms are described using English abbreviations):

[0059] In step S1, the intelligent metasurface (RIS)-assisted vehicle edge computing (VEC) system model in the construction scene consists of RIS, base station (BS) and N mobile vehicle users (US) with single antennas, where RIS consists of K passive reflective elements and the reflective unit set is K = {1, 2, ..., k, ..., K}, and the vehicle set is represented by N = {1, 2, ..., n, ..., N}. The structure of the RIS-assisted VEC system model is as follows: Figure 2 As shown in the figure, the communication environment at different locations on the lane is different. When the vehicle moves, the direct communication link between the vehicle and the base station will be blocked by the building. RIS can improve the communication quality. Assume that the entire service time of the vehicle is T = {1, 2, ..., t, ..., T}, the wireless channel state remains unchanged in each time slot t, and the channel state of different time slots is time-varying. Assume that the computing task generated by vehicle n can be expressed as D n = {I n ,C n ,τ n}, where I n Indicates the size of the input data, C n Indicates the number of CPU cycles, τ n Indicates task D n The maximum acceptable delay. All vehicles adopt a partial unloading strategy, δ n ∈[0,1] represents task D n The offloading decision variables, the computing tasks can be partially executed locally and partially offloaded to the mobile edge computing (MEC) layer for execution;

[0060] Step 2: Construct vehicle movement model;

[0061] The occlusion of user signals at different locations is different. When the vehicle is driving dynamically, its speed satisfies the independent and identical distribution and obeys the Gauss-Markov random process. The speed change v of the vehicle n The relational expression of (t+1) is as follows:

[0062]

[0063] Among them, v n (t) represents the speed of vehicle n at time slot t, ψ∈[0,1] represents the memory depth parameter of the past speed of vehicle n during its motion, and denote the asymptotic mean and asymptotic standard deviation of the speed of vehicle n, respectively. represents an uncorrelated random Gaussian process with zero mean and unit variance;

[0064] Step S3 is the process of building a communication model based on RIS;

[0065] Assuming that the link between US, BS and RIS follows Rayleigh fading, the auxiliary offloading link H between US and BS n (t), auxiliary offloading link between BS and RIS Direct link between US and BS The relational expression is as follows:

[0066]

[0067] Among them, β represents the path loss exponent, α l represents the reference channel power gain at unit distance, (x n (t),y n (t)) represents the position of vehicle n, (x BS (t),y BS (t)) represents the location of BS, represents the small-scale fading coefficient, ζ R,B represents the Rice coefficient between RIS and BS, (x R (t),y R (t)) indicates the position of RIS, Denote the channel gains of the line-of-sight link and the non-line-of-sight link between RIS and BS, respectively. n,R represents the Rice coefficient between vehicle n and RIS, denote the channel gains of the line-of-sight link and the non-line-of-sight link between vehicle n and RIS, respectively;

[0068] Due to the passive nature of RIS, RIS does not have signal processing capabilities, so the BS-RIS and RIS-US channels are jointly established. The relationship expression of the received signal Y(t) in time slot t is as follows:

[0069]

[0070] The signal-to-noise ratio γ between US and BS at time slot t n,B (t), the data transmission rate γ between US and BS at time slot t n The relational expression of (t) is as follows:

[0071]

[0072] Among them, ε n,B ∈{0,1} indicates whether RIS assists in offloading transmission, σ 2 represents the noise power;

[0073] Step S4: construct a local computing model;

[0074] The computing tasks generated by the US can be partially executed locally and partially offloaded to the MEC server for execution. Local energy consumption in US The relational expression is as follows:

[0075]

[0076] in, represents the local computing power of vehicle n, that is, the number of CPU cycles per second, κ represents the effective correlation coefficient, which is related to the chip structure of vehicle n;

[0077] Step S5: construct an unloading calculation model;

[0078] Offloading calculations requires three steps. Vehicle n needs to transmit the calculation task to the MEC server, and then the MEC server allocates computing resources to perform the offloaded task. Finally, the calculation result is returned to vehicle n through the BS. The transmission delay of offloading the calculation task to the MEC server Execution delay of computing tasks on MEC servers Total delay of unloading The expression is as follows:

[0079]

[0080] in, represents the computing capacity allocated to vehicle n by the MEC server;

[0081] Unloading calculation energy consumption of vehicle n Calculate the total latency of the task The total energy consumption of the computing task The expression is as follows:

[0082]

[0083] Step 6: Build cost model and optimization objectives;

[0084] The total cost Γ(t) of US is expressed as follows:

[0085]

[0086] Wherein, χ1, χ2 represent weight factors and χ1+χ2=1. The objective of the present invention is to minimize the cost while ensuring the maximum tolerable delay;

[0087] The joint optimization problem of minimizing the total cost of computational offloading of VEC assisted by RIS is as follows:

[0088]

[0089] stC1:

[0090] C2:

[0091] C3:

[0092] C4:χ1+χ2=1

[0093] C5:

[0094] C6:

[0095] C7:

[0096] Among them, C1 represents the value range of the unloading variable, C2 represents the amplitude range of the RIS reflection unit, C3 represents the phase shift range of the RIS reflection unit, C4 represents the weight factor, C5 represents the constraint on the computing power of each vehicle, C6 represents that the sum of the computing resources allocated to each task cannot exceed the computing power of the MEC server, and C7 represents that the total delay of the computing task is less than the maximum tolerable delay;

[0097] Step S7: Optimize the dynamic coupling problem through DDPG to solve the offloading decision;

[0098] Specifically, the computation offloading algorithm based on DDPG is modeled as an MDP model, which consists of <S, A, R, P>, which represent the state space, action space, reward value and transition probability respectively;

[0099] The state space is defined as Among them C n Indicates task D n The number of CPU cycles required, I n Indicates the size of the input data, represents the local computing capacity of vehicle n, Indicates the computing power of the MEC server. represents the reflection coefficient of RIS, MV loc Indicates the position of the vehicle;

[0100] Construct the action space. In this model, the action space is defined as where δ n ∈[0,1] represents task D n The uninstall decision variables, represents the computing power allocated to vehicle n by the MEC server, ρ n ∈(0,1) represents the amplitude reflection factor, θ n ∈[0,2π] represents the phase shift coefficient;

[0101] Construct a reward function. The reward function formula is defined as The agent seeks lower latency and energy consumption in its interaction with the environment, thereby minimizing the total cost of the system;

[0102] Build a DDPG algorithm training framework, such as Figure 3 As shown in the figure, unlike the traditional DRL algorithm, the DDPG algorithm uses the Actor-Critic framework to enhance the algorithm's exploration ability by adding behavioral noise to the agent, and uses the experience replay mechanism to enhance the stability of the training process. There are 4 neural networks inside DDPG: the main actor, the main critic, the target actor, and the target critic;

[0103] The Actor network parameters are θ A and θ A′ , Critic network parameters θ C and θ C′ , experience pool size |R|, current state s t As well as other system parameters, update the parameters according to the determined policy gradient;

[0104]

[0105] After the DDPG algorithm training is completed, save θ A When the method of the present invention is applied, the stored θ A Replace the Actor neural network parameters in the current algorithm;

[0106] Furthermore, in order to verify the superiority of the intelligent metasurface-assisted vehicle edge computing task offloading method proposed in the present invention, the present invention also adopts the following four strategies to compare the performance with the proposed RDDPG strategy. Specifically, the present invention is verified in the Python 3.10 environment, and the RDDPG algorithm of the present invention is compared with the Local Execute (LE) algorithm, the Offload Execute (OE) algorithm, the DQN algorithm and the DDPG (Non-RIS-DDPG, NRDDPG) algorithm without RIS assistance;

[0107] 1. Local Execute (LE) algorithm

[0108] The vehicle's computing tasks are performed locally.

[0109] 2. Offload Execute (OE) algorithm

[0110] The vehicle's computing tasks are offloaded to the MEC server for execution.

[0111] 3. DQN Algorithm

[0112] A computation offloading strategy based on the DQN algorithm is adopted to solve the RIS-assisted VEC offloading decision. The continuous actions in the environment are discretized and a finite number of discrete values ​​are used to approximate the real values.

[0113] 4. Non-RIS-DDPG (NRDDPG) algorithm without RIS assistance

[0114] In the VEC environment without RIS assistance, the DDPG algorithm is used to solve the computation offloading decision and resource allocation strategy.

[0115] Figure 4The effect of different numbers of vehicles on the system cost is shown, assuming that the processing task size is fixed. As can be seen from the figure, the system cost of the LE algorithm and the OE algorithm remains almost unchanged, and the system cost of LE is much higher than that of OE. This is because the total scale of the computing task remains unchanged, while the computing power of the vehicle is much smaller than that of the MEC server. Compared with the DQN algorithm and the NRDDPG algorithm, the RDDPG algorithm performs better. This is because the channel environment between the vehicle and the BS is improved with the assistance of RIS, and the impact of channel instability caused by interference and obstacles on the quality of offloading transmission is reduced, thereby effectively improving the energy efficiency of offloading. Here, since the action space of the proposed optimization problem is continuous, the RDDPG algorithm converges faster and can continuously interact with the environment to make the best offloading decision. For example, when the number of vehicles is 12, the system overheads of LE, OE, DQN, and RDDPG are 141, 75, 60, and 31, respectively. RDDPG reduces the system cost by 78%, 58.7%, and 48.3% for other benchmark algorithms, respectively.

[0116] Figure 5 The figure shows the impact of different computing capabilities of MEC servers on system cost. As can be seen from the figure, the system cost of the LE algorithm remains basically unchanged, because the computing power of the MEC server has no effect on the local computing model. The system costs of the OE, DQN and RDDPG algorithms gradually decrease with the increase of the computing power of the MEC server, and RDDPG is significantly better than OE and DQN. This is because the RDDPG algorithm comprehensively optimizes the RIS reflection coefficient, offloading decision and MEC resource allocation, so that the total system cost is always minimized. When the computing power of the MEC server is 6GHz, the system overheads of LE, OE, DQN and RDDPG are 140, 53, 25 and 19 respectively. Compared with other benchmark algorithms, the system cost of the RDDPG algorithm is reduced by 86.4%, 64.2% and 24% respectively. It is worth noting that due to the lack of RIS assistance, the quality of the wireless offloading transmission channel of NRDDPG is unstable due to the influence of the dynamic vehicle environment, which is manifested as fluctuations in system cost.

[0117] What is disclosed above is only a preferred embodiment of the present invention, and it certainly cannot be used to limit the scope of rights of the present invention. Ordinary technicians in this field can understand that all or part of the processes of the above embodiment and equivalent changes made according to the claims of the present invention still fall within the scope of the invention.

Claims

1. A method for offloading vehicle edge computing tasks assisted by an intelligent metasurface, characterized in that: The following steps are involved: Step 1: Build an edge computing scenario including base stations, smart metasurfaces, and vehicles; Step 2: Construct vehicle movement model; Step 3: Construct a communication model based on smart metasurface; Step 4: Build a local computing model; Step 5: Build an offloading calculation model; Step 6: Build cost model and optimization objectives; Step 7: Optimize the dynamic coupling problem through DDPG to solve the offloading decision.

2. The intelligent metasurface-assisted vehicle-mounted edge computing task offloading method according to claim 1, characterized in that: The intelligent metasurface acts between the edge base station and the vehicle-mounted users, responsible for improving the signal coverage and reducing the impact of the blocking of the direct link between the vehicle and the base station. All vehicles adopt a partial offloading strategy, with some computing tasks executed locally and some offloaded to the mobile edge computing layer for execution.

3. The intelligent metasurface-assisted vehicle-mounted edge computing task offloading method according to claim 2, characterized in that: In step 2, the occlusion of user signals at different locations is different. When the vehicle is driving dynamically, its speed satisfies independent and identical distribution and obeys the Gauss-Markov random process. The speed change v of the vehicle n The relational expression of (t+1) is as follows: Among them, v n (t) represents the speed of vehicle n at time slot t, ψ∈[0,1] represents the memory depth parameter of the past speed of vehicle n during its motion, and denote the asymptotic mean and asymptotic standard deviation of the speed of vehicle n, respectively. represents an uncorrelated random Gaussian process with zero mean and unit variance.

4. The intelligent metasurface-assisted vehicle-mounted edge computing task offloading method according to claim 3, characterized in that: In the communication model based on the intelligent metasurface, the auxiliary offloading link H between the vehicle user and the base station n (t), auxiliary offloading link between base station and smart metasurface Direct link between vehicle users and base stations The relational expression of the signal Y(t) received by the base station through the complete link assisted by the smart metasurface at time slot t is as follows: Among them, β represents the path loss exponent, α l represents the reference channel power gain at unit distance, (x n (t),y n (t)) represents the position of vehicle n, (x BS (t),y BS (t)) represents the location of the base station, represents the small-scale fading coefficient, ζ R,B represents the Rice coefficient between the smart metasurface and the base station, (x R (t),y R (t)) represents the position of the smart metasurface, denote the channel gains of the line-of-sight link and the non-line-of-sight link between the smart metasurface and the base station, respectively; ζ n,R represents the Rice coefficient between vehicle n and the smart metasurface, represents the channel gain of the line-of-sight link and the non-line-of-sight link between vehicle n and the smart metasurface respectively; the signal-to-noise ratio γ between the vehicle user and the base station at time slot t n,B (t), the data transmission rate γ between the vehicle user and the base station at time slot t n The relational expression of (t) is as follows: Among them, ε n,B ∈{0,1} indicates whether the smart metasurface assists in offloading transmission, σ 2 Represents the noise power.

5. The intelligent metasurface-assisted vehicle-mounted edge computing task offloading method according to claim 4, characterized in that: During the execution of step 4, the computing tasks generated by the vehicle user are partially executed locally and partially offloaded to the mobile edge computing server for execution; the local computing delay of the vehicle user Local energy consumption of onboard users The relational expression is as follows: in, represents the local computing power of vehicle n, that is, the number of CPU cycles per second, and κ represents the effective correlation coefficient, which is related to the chip structure of vehicle n.

6. The method for offloading vehicle edge computing tasks assisted by a smart metasurface as claimed in claim 5, characterized in that: The offloading calculation in step 5 requires three steps. Vehicle n needs to transmit the computing task to the mobile edge computing server, and then the mobile edge computing server allocates computing resources to perform the offloading task. Finally, the calculation result is returned to vehicle n through the base station. The transmission delay of the computing task offloading to the mobile edge computing server Execution delay of computing tasks on mobile edge computing servers Total delay of unloading The expression is as follows: in, represents the computing capacity allocated to vehicle n by the mobile edge computing server and the offloading computing energy consumption of vehicle n Calculate the total latency of the task The total energy consumption of the computing task The expression is as follows:

7. The intelligent metasurface-assisted vehicle-mounted edge computing task offloading method according to claim 6, characterized in that: During the execution of step 6, the total cost Γ(t) of the vehicle user is expressed as follows: Among them, χ1, χ2 represent weight factors and χ1+χ2=1; In order to ensure cost minimization under the maximum tolerable delay, the expression of the joint optimization problem of minimizing the total cost of computational offloading of on-board edge computing assisted by intelligent metasurface is as follows: s.t.C1: C2: C3: C4:χ1+χ2=1 C5: C6: C7: Among them, C1 represents the value range of the unloading variable, C2 represents the amplitude range of the intelligent metasurface reflection unit, C3 represents the phase shift range of the intelligent metasurface reflection unit, C4 represents the weight factor, C5 represents the constraint on the computing power of each vehicle, C6 represents that the sum of the computing resources allocated to each task cannot exceed the computing power of the mobile edge computing server, and C7 represents that the total delay of the computing task is less than the maximum tolerable delay.

Images