An energy efficiency optimization method for a ris-enhanced uav-assisted edge computing system

Abstract

The application discloses an energy efficiency optimization method of a RIS enhanced UAV assisted edge computing system, which comprises the following steps: firstly, establishing a system model, a complex channel communication model, a computing offloading model and an energy consumption model; then, constructing an optimization problem with the target of maximizing system energy efficiency and decomposing the optimization problem into four sub-problems; then, solving the four sub-problems in different ways; finally, using the block coordinate descent method to iteratively update GT radio frequency power, GT offloading weight, RIS phase shift matrix and UAV flight trajectory until the optimization target value converges, and finally obtaining the maximum system energy efficiency. The application relates to the field of GT task offloading technology in an edge computing system, and through the introduction of a UAV load RIS, the K GTs are jointly assisted in computing offloading by a fixed RIS, and the GT radio frequency power, the GT offloading weight, the RIS phase shift matrix and the UAV flight trajectory planning are optimally solved, so that the target of improving the offloading rate and maximizing the system energy efficiency is achieved.

Description

Technical Field

[0001] The present invention relates to an energy efficiency optimization method for a RIS-enhanced UAV-assisted edge computing system, in particular to a UAV-enhanced RIS and fixed RIS combined assisted edge computing offloading method, belonging to the technical field of GT task offloading in edge computing systems. Technical Background

[0002] With the development of mobile communication technology and the widespread adoption of smart devices, a wide range of network services and applications are constantly emerging. Mobile terminals (Ground Terminals, GTs) are placing increasingly stringent demands on network performance, such as quality of service and request latency. Despite the increasing processing power of the central processing units (CPUs) in new mobile devices, they are still unable to process large and complex applications quickly. Furthermore, processing these applications locally presents another challenge: rapid battery drain and wear and tear. These issues severely impact the operational efficiency of applications on user devices and the user experience.

[0003] Existing communication networks face numerous challenges: high data transmission rates and high traffic density in network hotspots, ensuring high application reliability and latency constraints, power consumption constraints and high connection density of connected devices within the network, and network load capacity and security. To address these issues, the industry has proposed Multi-access Edge Computing (MEC) and compute offloading technologies. Compared to cloud computing offloading, edge computing offloading operates at a shorter physical distance from user terminals, has lower communication latency, and can achieve stable communication with greater bandwidth, making it a suitable strategy for compute offloading for mobile terminals or IoT devices.

[0004] Thanks to the development of new material technology, electromagnetic signals can be reflected in a phase-controlled manner through passive devices, and the array collection of such devices is the reconfigurable intelligent surface (RIS) technology that people currently have high hopes for.

[0005] As a passive component, RIS offers low deployment costs and can be easily configured through software to reflect electromagnetic signals. Therefore, deploying RIS can reconfigure wireless network propagation pathways. Current research has shown that RIS can, through software configuration, enable reconfigured wireless network spaces to support technologies such as: ① enhanced Line of Sight (LoS) communications; ② assisted establishment of Non-Line of Sight (NLoS) communications; ③ secure wireless transmission physical layer communications; and ④ assisted establishment and enhancement of Massive Multiple-Input Multiple-Output (mMIMO) networks.

[0006] Currently, the theoretical application of RIS in edge computing offload service systems often involves attaching the RIS to buildings to assist with uplink transmission. However, in actual application scenarios, the wireless transmission link from the GT to the base station (BS), whether RIS-assisted or not, is likely to be blocked by obstacles such as buildings, inevitably causing a significant degradation in communication quality.

[0007] Unmanned aerial vehicles (UAVs) have recently been considered as a means of providing enhanced coverage or relay services for GTs in wireless systems with limited or no infrastructure. Equipping UAVs with RIS can achieve high maneuverability by establishing new uplink communications, assisting edge computing servers in providing edge computing offload services to GTs with limited local processing power, thereby expanding the scope of these offload services. Introducing UAV-mounted RIS on top of fixed RIS can effectively address the limited coverage of fixed RIS and further improve the quality of GT-to-BS uplink communications. However, current technology clearly limits UAVs due to their energy consumption, making them unable to support prolonged hovering or long-distance maneuvers. Summary of the Invention

[0008] The present invention aims to address the defects and shortcomings of the above-mentioned prior art and proposes an energy efficiency optimization method for a RIS-enhanced UAV-assisted edge computing system. The method establishes K GTs that can perform task offloading through a direct link to a base station (BS) and an uplink jointly assisted by a fixed RIS and a UAV payload (RIS). According to the total GT offloaded task volume and the weighted total energy consumption of the system, a method is constructed with the goal of maximizing system energy efficiency and jointly optimizing the following four sub-problems: GT RF power optimization, GT offload weight optimization, RIS phase shift moment optimization array, and UAV flight trajectory optimization. The method utilizes the high maneuverability of UAVs and the technical advantages of RIS in significantly improving communication quality at low cost and low power consumption. At the same time, a reasonable optimization process is used to mitigate the effects of the limited coverage of the fixed RIS and the high flight energy consumption of the UAV payload RIS, thereby balancing the total edge computing offload volume and system energy consumption, and achieving the goals of improving the offloading rate and maximizing the system energy efficiency.

[0009] The technical solution adopted by the present invention to solve the technical problem is: a RIS enhanced UAV assisted edge computing system energy efficiency optimization method, the method comprising the following steps:

[0010] Step 1: Establish a system model including K GTs, a fixed RIS, a UAV-loaded RIS, and a BS, and establish a complex channel communication model, a computational offloading model, and an energy consumption model;

[0011] Step 2: Based on the total amount of GT offloaded tasks and the system's weighted total energy consumption, an optimization problem is constructed to maximize the system's energy efficiency and decomposed into four sub-problems.

[0012] Step 3: Use mathematical analysis to solve the GT RF power optimization problem;

[0013] Step 4: Use CVX tools to solve the GT offloading weight optimization problem;

[0014] Step 5: Use mathematical analysis to solve the RIS phase shift matrix optimization problem;

[0015] Step 6: Solve the UAV flight trajectory optimization problem using the continuous convex approximation method;

[0016] Step 7: Using the block coordinate descent method, iteratively update the GT RF power, GT unloading weight, RIS phase shift matrix, and UAV flight trajectory until the optimization target converges, ultimately obtaining the maximum system energy efficiency.

[0017] Furthermore, the system model, complex channel communication model, computation offloading model, and energy consumption model in step 1 specifically include:

[0018] The system model includes: GT set A collection of reflector units for the UAV payload RIS (RISu) and the fixed RIS (RISb) The total duration of a single UAV flight, that is, the single edge computing offloading service time T, the UAV flight altitude z, and a sufficiently small time slot t ut =N -1 T, time slot set The coordinates of the UAV, i.e., RISu, in the nth time slot are p[n] = [x[n], y[n], z] T , GTk's coordinate p G,k =[x G,k ,y G,k ,0] T , the coordinates of RISb are p B =[x B ,y B ,z B ] T , the coordinates of BS are p BS =[x BS ,y BS ,z BS ] T , UAV speed in a single time slot v[n]=(t ut ) -1 ‖p[n]-p[n-1]‖, UAV maximum flight speed V ma , the UAV’s flight starting point p s =[x s ,y s ,z] T , the UAV's flight endpoint p e =[x e ,y e ,z] T , the phase shift matrix when RISu serves GTk in the nth time slot The phase shift matrix when RISb serves GT k in the nth time slot is in The mth time slot from GTk to RISu in the nth time slot U The reflection angle of each reflection unit, the reflection coefficient β of RISu U ∈[0,1], where The mth time slot from GTk to RISb in the nth time slot B The reflection angle of each reflection unit, the reflection coefficient β of RISb B ∈[0,1];

[0019] The complex channel communication model includes: the ground distance D0 of the receiving power reference point in the large-scale propagation model, the received power (C0) of the antenna at a distance D0 from the unit power transmitter in free space 2, in the nth time slot, GTk to the mth time slot in RISu U The single-reflection channel gain of a reflector unit Among them is the geometric distance D between GTk and RISu in the nth time slot GU,k [n]=‖p[n]-p G,k [n]‖, the path loss exponent α of this single-shot channel GU,k , the Ricean factor δ of this single-shot channel GU,k , the line-of-sight (LoS) component of this channel The non-line-of-sight (NLoS) component of this channel GTk to RISb in the nth time slot B The single-reflection channel gain of a reflector unit Among them is the geometric distance D between GTk and RISb GB,k , the path loss index α of this single-shot channel GB,k , the Ricean factor δ of this single-shot channel GB,k , the LoS component of this channel NLoS component of this channel The mth time slot in RISu in the nth time slot U Single-ray channel gain from a reflector unit to the BS Among them is the geometric distance D between RISu and BS US [n], the path loss exponent α of this single-shot channel US , the Ricean factor δ of this single-shot channel US , the LoS component of this channel NLoS component of this channel The mth RISb in the nth time slot B Single-ray channel gain from a reflector unit to the BS Among them is the geometric distance D between RISb and BS BS , the path loss index α of this single-shot channel BS , the Ricean factor δ of this single-shot channel BS , the LoS component of this channel NLoS component of this channel Single-shot channel gain from GT k to BS in the nth time slot Among them is the geometric distance D between GTk and BS GS,k , the path loss index α of this single-shot channel GS,k , the Ricean factor δ of this single-shot channel GS,k , the LoS component of this channel NLoS component of this channel RIS uses a uniform square planar array model with an azimuth angle of the receiving angle of the GTk RF signal received at RISu. Pitch angle Azimuth of the receiving angle of the GTk RF signal received at RISb Pitch angle Azimuth of the deviation angle of the radio frequency signal transmitted from RISu to BS Pitch angle Azimuth of the deviation angle of the radio frequency signal transmitted from RISb to BS Pitch angle The calculation process is ; Distance d between RIS reflector units, carrier wavelength λ, LoS component LoS component LoS component LoS component Vector Vector Vector Vector GTk to BS link gain h via RISu assistance GUS,k [n]=(h GU,k [n]) H Θ U,k [n]h US [n], GTk to BS link gain h via RISb assistance GBS,k [n]=(h GB,k [n]) H Θ B,k [n]h BS [n]; The total uplink channel gain h of GTk reaching BS through different paths k [n]=h GS,k [n]+h GUS,k [n]+h GBS,k [n];

[0020] The calculation unloading model includes: the unloading duration t of GTk in the nth time slot k [n] = η k [n]t ut , the GT offloading weight η of GTk in the nth time slot k [n], minimum GT unloading weight value η min ≤(K) -1 , the uplink signal-to-noise ratio (SNR) of GTk reaching BS through different paths in the nth time slot k [n] = γ k [n]=σ -2 P k [n]|hk [n]| 2 , complex Gaussian white noise power σ 2 , the average GT RF power P of GTk in the nth time slot k [n], the rated minimum GT RF power P of GT to ensure effective data transmission min , Maximum GT RF power P max , TDMA communication bandwidth B, GTk total uplink communication rate C in the nth time slot k [n] = Blog2(1+γ k [n]), Unload BitQ k [n]=C k [n]t k [n];

[0021] The energy consumption model includes: blade power P0 in the hovering state, blade angular velocity U tip , fuselage drag ratio d0, air density ρ, rotor volume s, rotor area G, induced power P1 in hovering state, induced speed v0 when the rotor flies forward, RF energy consumption e in the nth time slot k [n]=P k [n]t ut , the flight energy consumption of UAV in the nth time slot GTk radio frequency energy consumption e in the nth time slot k [n]=P k [n]t ut .

[0022] Furthermore, the optimization problem constructed in step 2 with the goal of maximizing system energy efficiency specifically includes: GT unloading weight RIS phase shift matrix UAV flight trajectory Maximizing the computational offloading energy efficiency of a single UAV flight is formulated as problem (P0):

[0023]

[0024]

[0025] C2:p[0]=p s ,p[N]=p e

[0026]

[0027]

[0028]

[0029]

[0030] in, ε e It is an energy weight parameter that represents the importance of GT RF energy to the overall computational offloading energy efficiency of the system.

[0031] Furthermore, the four sub-problems in step 2 specifically include: when the GT unloading weight η, the RIS phase shift matrix Θ, and the UAV flight trajectory p are fixed, there is the first sub-problem (P1), which is the GT RF power optimization problem in step 3:

[0032]

[0033] stC6

[0034] When the GT RF power P, RIS phase shift matrix Θ, and UAV flight trajectory p are fixed, there is a second sub-problem (P2), which is the GT offloading weight optimization problem in step 4:

[0035]

[0036] stC4,C5

[0037] When the GT RF power P, GT unloading weight η, and UAV flight trajectory p are fixed, there is a third sub-problem (P3), which is the RIS phase shift matrix optimization problem in step 5:

[0038]

[0039] stC3

[0040] When the GT RF power P, GT unloading weight η, and RIS phase shift matrix Θ are fixed, there is a fourth sub-problem (P4), which is the UAV flight trajectory optimization problem in step 6:

[0041]

[0042] stC1,C2

[0043] The above are the four sub-questions in step 2.

[0044] Furthermore, the use of mathematical analysis to solve the GT radio frequency power optimization problem in step 3 specifically includes the following steps:

[0045] Step 3-1: Expand the objective function in the first sub-problem (P1) as follows:

[0046]

[0047] in And record the objective function The current problem is:

[0048]

[0049] stC6

[0050] Step 3-2: Since (P1′) The constraint C6 is for Independence, then the following problem (P1″) is equivalent to (P1′):

[0051]

[0052] stC6

[0053] That is, it is transformed into exist The maximum value of

[0054] Step 3-3: Ask about P k The partial derivative of [n] is:

[0055]

[0056] From constraint C5, we can get: Let's assume but

[0057]

[0058] After calculation, it can be simplified to:

[0059]

[0060] but and The signs of are consistent;

[0061] Step 3-4: Note Easy to be a P k When [n]→0, When P k When [n]→+∞,

[0062] Steps 3-5: Easy to know is a non-decreasing function; hour, is a non-decreasing function, when hour, is a decreasing function with only one zero able to Established, and Part Part

[0063] Step 3-6: Make have to:

[0064] It can be seen that P k [n] appears on the left side of the equal sign, and the right side of the equal sign can be regarded as a constant in this subproblem; Considered as P k [n] is a function of the independent variable, and this transcendental equation can be solved.

[0065] Step 3-7: In summary, when hour, but Satisfied at this time P k [n] takes the value of P min ;when When satisfied P k [n] takes the value

[0066] Furthermore, the use of the CVX tool in step 4 to solve the GT offloading weight optimization problem specifically includes the following steps:

[0067] Step 4-1: Expand the objective function in the second sub-problem (P2) as follows:

[0068]

[0069] And record the objective function The current problem is:

[0070]

[0071] stC4,C5

[0072] Step 4-2: Since (P2′) Constraints C4 and C5 for Independence, then the following problem (P2″) is equivalent to (P2′):

[0073]

[0074] stC4,C5

[0075] Step 4-3: Obviously, the (P2″) optimization problem is a convex optimization problem and can be solved efficiently using the CVX tool.

[0076] Furthermore, the use of mathematical analysis to solve the RIS phase shift matrix optimization problem in step 5 specifically includes the following steps:

[0077] Step 5-1: Expand the objective function in the third sub-problem (P3) as follows:

[0078]

[0079] And record the objective function The current problem is:

[0080]

[0081] stC3

[0082] Step 5-2: Since (P3′) The constraint C3 is for Independence, then the following problem (P3″) is equivalent to (P3′):

[0083]

[0084] stC3

[0085] That is, it is transformed into exist The maximum value of

[0086] Step 5-3: Only h k [n] with Related, obviously For h k [n]| is an increasing function, then the following problem (P3″′) is equivalent to (P3″):

[0087]

[0088] stC3

[0089] Step 5-4: The following problem (P3°) and problem (P3″′) have the same optimization process:

[0090]

[0091] stC3

[0092] Step 5-5: Remember

[0093] Then we have the following channel gain expansion, (·) * This means to find the conjugate of the formula in brackets:

[0094]

[0095]

[0096]

[0097] Among them, due to are all random variables;

[0098] Step 5-6: The factors containing random variables in the above step 5-5 should not enter the optimization process. Then the following channel gain partial expansion actually enters the optimization process:

[0099]

[0100]

[0101]

[0102] in:

[0103]

[0104]

[0105] Step 5-7: The problem (P3°) should be rewritten as the following problem (P3°′) based on the actual part of the optimization process, and record

[0106]

[0107] stC3

[0108] Steps 5-8: Obviously, As the upper bound of this optimization objective function, it is a reachable upper bound; there is a feasible way to reach this upper bound, that is, let:

[0109]

[0110] in:

[0111]

[0112] Make It can make

[0113] Step 5-9: In summary, find the upper bound of the objective function Corresponding to The optimal solution is

[0114] Furthermore, the method of solving the UAV flight trajectory optimization problem using the continuous convex approximation method in step 6 specifically includes the following steps:

[0115] Step 6-1: Expand the objective function in the fourth sub-problem (P4) as follows:

[0116]

[0117] Step 6-2: According to the value obtained from (P3°′) in Step 5-7 above, h k [n] is rewritten as The upper bound of Then we have:

[0118]

[0119] in,

[0120]

[0121] D US [n]=‖p[n]-p BS ‖,v[n]=(t ut ) -1 ‖p[n]-p[n-1]‖;

[0122] Step 6-3: Due to constraint C1, we have So the objective function has the following lower bounds:

[0123]

[0124] Step 6-4: Due to the non-convexity of the objective function, introduce the slack variable g[n]≥‖p[n]-p G,k ‖, j[n]≥‖p[n]-p BS ‖,w[n]≥(t ut ) -1 ‖p[n]-p[n-1]‖, Then we have:

[0125]

[0126] Step 6-5: Remember the objective function

[0127]

[0128] Then the following problem (P4′) has the same optimization process as problem (P4):

[0129]

[0130] stC1,C2

[0131] C7:g[n]≥‖p[n]-p G,k ‖

[0132] C8:j[n]≥‖p[n]-p BS ‖

[0133] C9:w[n]≥(t ut ) -1 ‖p[n]-p[n-1]‖

[0134] Step 6-6: To solve the non-convexity of the objective function and constraints C7, C8, and C9, in the rth iteration, a feasible solution g is given. (r) [n],j (r) [n],w (r) [n], and use the first-order Taylor formula to obtain an optimal approximate solution, which can be expressed as the following problem (P4″):

[0135]

[0136] stC1,C2

[0137] C7′:‖p[n]-p G,k ‖ 2 ≤2g (r) [n]g[n]-(g (r) [n]) 2

[0138] C8′:‖p[n]-p A ‖ 2 ≤2j (r) [n]j[n]-(j (r) [n]) 2

[0139] C9′:(t ut ) -2 ‖p[n]-p[n-1]‖ 2 ≤2w (r)[n]w[n]-(w (r) [n]) 2

[0140] in:

[0141]

[0142] Steps 6-7: In summary, problem (P4″) is now a convex optimization problem that can be solved efficiently using the CVX tool.

[0143] Furthermore, in step 7, the block coordinate descent method is used to iteratively update the GT RF power, GT unloading weight, RIS phase shift matrix, and UAV flight trajectory until the optimization target value converges, and ultimately the maximum system energy efficiency is obtained. Specifically, the following steps are included:

[0144] Step 7-1: Initialize P (0) , η (0) ,Θ (0) , p (0) , iterative precision parameter χ = 10 -3 , the number of iterations l = 0;

[0145] Step 7-2: Given GT unloading weight η (l) , RIS phase shift matrix Θ (l) 、UAV flight trajectoryp (l) , solve the problem (P1) using the analytical method and obtain the GT RF power P (l+1) ;

[0146] Step 7-3: Given the GT RF power P (l+1) , RIS phase shift matrix Θ (l) 、UAV flight trajectoryp (l) , use CVX tools to solve the convex optimization problem (P2) and obtain the GT unloading weight η (l+1) ;

[0147] Step 7-4: Given the GT RF power P (l+1) , GT unloading weight η (l+1) 、UAV flight trajectoryp (l) , solve problem (P3) using analytical method and obtain the RIS phase shift matrix Θ (l+1) ;

[0148] Step 7-5: Given the GT RF power P (l+1) , GT unloading weight η (l+1) , RIS phase shift matrix Θ (l+1) , use the continuous convex approximation method to solve the problem (P4) and obtain the UAV flight trajectory p (l+1) ;

[0149] Step 7-6: Let l = l + 1;

[0150] Step 7-7: Repeat the above steps 7-2 to 7-6 until That is, the optimization objective numerical convergence;

[0151] Step 7-8: Finally, the optimal solution P is obtained within the allowable accuracy range. (l) , η (l) ,Θ (l) , p (l) .

[0152] Beneficial effects:

[0153] 1. To address the problems of slow offloading rate and narrow coverage area for ground terminals when edge computing offloading is performed only with the assistance of fixed RIS in complex urban communication environments, this invention combines UAV, MEC, and RIS. It utilizes the high maneuverability of UAV and the technical advantages of RIS to enhance communication quality, coverage, low cost, and low power consumption. Through a reasonable optimization process, it overcomes the defects of fixed RIS coverage range and high flight energy consumption of UAV payload RIS, balances the total edge computing offloading amount and system energy consumption, and ultimately further improves the performance of the current edge computing offloading system.

[0154] 2. This invention introduces a UAV payload RIS and combines it with a fixed RIS to assist in computational offloading of K GTs. Based on the complex channel state, the communication model is set up to optimize and solve the appropriate and feasible GT RF power, GT offloading weight, RIS phase shift matrix, and UAV flight trajectory planning with respect to the time variable, thereby achieving the goals of improving the offloading rate and maximizing the system energy efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0155] Figure 1 This is a basic flow chart of a method for optimizing energy efficiency of a RIS-enhanced UAV-assisted edge computing system according to the present invention.

[0156] Figure 2 This is a system communication model diagram of the energy efficiency optimization method of the RIS enhanced UAV-assisted edge computing system in the present invention.

[0157] Figure 3 This is a GT task offloading protocol diagram of the present invention. DETAILED DESCRIPTION

[0158] The present invention will be described in further detail below with reference to the accompanying drawings.

[0159] Figure 1 The following is a basic flow chart of a method for optimizing energy efficiency of a RIS-enhanced UAV-assisted edge computing system provided by the present invention, including the following steps:

[0160] Step 1: Establish a system model including K GTs, a fixed RIS, a UAV-loaded RIS, and a BS, and establish a complex channel communication model, a computational offloading model, and an energy consumption model;

[0161] Step 2: Based on the total amount of GT offloaded tasks and the system's weighted total energy consumption, an optimization problem is constructed to maximize the system's energy efficiency and decomposed into four sub-problems.

[0162] Step 3: Use mathematical analysis to solve the GT RF power optimization problem;

[0163] Step 4: Use CVX tools to solve the GT offloading weight optimization problem;

[0164] Step 5: Use mathematical analysis to solve the RIS phase shift matrix optimization problem;

[0165] Step 6: Solve the UAV flight trajectory optimization problem using the continuous convex approximation method;

[0166] Step 7: Using the block coordinate descent method, iteratively update the GT RF power, GT unloading weight, RIS phase shift matrix, and UAV flight trajectory until the optimization target converges, ultimately obtaining the maximum system energy efficiency.

[0167] The system model, complex channel communication model, computational offloading model, and energy consumption model in step 1 of the present invention specifically include:

[0168] System model includes: GT set A collection of reflector units for the UAV payload RIS (RISu) and the fixed RIS (RISb) The total duration of a single UAV flight, that is, the single edge computing offloading service time T, the UAV flight altitude z, and a sufficiently small time slot t ut =N -1 T, time slot set The coordinates of the UAV, i.e., RISu, in the nth time slot are p[n] = [x[n], y[n], z] T , GTk's coordinate p G,k =[x G,k ,y G,k ,0] T , the coordinates of RISb are p B =[x B ,y B ,z B ] T , the coordinates of BS are p BS =[x BS ,y BS ,z BS ] T, UAV speed in a single time slot v[n]=(t ut ) -1 ‖p[n]-p[n-1]‖, UAV maximum flight speed V max , the UAV’s flight starting point p s =[x s ,y s ,z] T , the UAV's flight endpoint p e =[x e ,y e ,z] T , the phase shift matrix when RISu serves GTk in the nth time slot The phase shift matrix when RISb serves GT k in the nth time slot is in The mth time slot from GTk to RISu in the nth time slot U The reflection angle of each reflection unit, the reflection coefficient β of RISu U ∈[0,1], where The mth time slot from GTk to RISb in the nth time slot B The reflection angle of each reflection unit, the reflection coefficient β of RISb B ∈[0,1];

[0169] The complex channel communication model includes: the ground distance D0 of the received power reference point in the large-scale propagation model, the received power (C0) of the antenna at a distance D0 from the unit power transmitter in free space. 2 , in the nth time slot, GTk to the mth time slot in RISu U The single-reflection channel gain of a reflector unit Among them is the geometric distance D between GTk and RISu in the nth time slot GU,k [n]=‖p[n]-p G,k [n]‖, the path loss index α of this single-shot channel GU,k , the Ricean factor δ of this single-shot channel GU,k , the line-of-sight (LoS) component of this channel The non-line-of-sight (NLoS) component of this channel GTk to RISb in the nth time slot B The single-reflection channel gain of a reflector unit Among them is the geometric distance D between GTk and RISb GB,k , the path loss index α of this single-shot channel GB,k , the Ricean factor δ of this single-shot channel GB,k , the LoS component of this channel NLoS component of this channel The mth time slot in RISu in the nth time slotU Single-ray channel gain from a reflector unit to the BS Among them is the geometric distance D between RISu and BS US [n], the path loss exponent α of this single-shot channel US , the Ricean factor δ of this single-shot channel US , the LoS component of this channel NLoS component of this channel The mth RISb in the nth time slot B Single-ray channel gain from a reflector unit to the BS Among them is the geometric distance D between RISb and BS BS , the path loss index α of this single-shot channel BS , the Ricean factor δ of this single-shot channel BS , the LoS component of this channel NLoS component of this channel Single-shot channel gain from GT k to BS in the nth time slot Among them is the geometric distance D between GTk and BS GS,k , the path loss index α of this single-shot channel GS,k , the Ricean factor δ of this single-shot channel GS,k , the LoS component of this channel NLoS component of this channel RIS uses a uniform square planar array model with an azimuth angle of the receiving angle of the GTk RF signal received at RISu. Pitch angle Azimuth of the receiving angle of the GTk RF signal received at RISb Pitch angle Azimuth of the deviation angle of the radio frequency signal transmitted from RISu to BS Pitch angle Azimuth of the deviation angle of the radio frequency signal transmitted from RISb to BS Pitch angle The calculation process is ; Distance d between RIS reflector units, carrier wavelength λ, LoS component LoS component LoS component LoS component Vector Vector Vector Vector GTk to BS link gain h via RISu assistance GUS,k [n]=(hGU,k [n]) H Θ U,k [n]h US [n], GTk to BS link gain h via RISb assistance GBS,k [n]=(h GB,k [n]) H Θ B,k [n]h BS [n]; The total uplink channel gain h of GTk reaching BS through different paths k [n]=h GS,k [n]+h GUS,k [n]+h GBS,k [n];

[0170] The calculation of the unloading model includes: the unloading duration t of GTk in the nth time slot k [n] = η k [n]t ut , the GT offloading weight η of GTk in the nth time slot k [n], minimum GT unloading weight value η min ≤(K) -1 , the uplink signal-to-noise ratio (SNR) of GTk reaching BS through different paths in the nth time slot k [n] = γ k [n]=σ -2 P k [n]|h k [n]| 2 , complex Gaussian white noise power σ 2 , the average GT RF power P of GTk in the nth time slot k [n], the rated minimum GT RF power P of GT to ensure effective data transmission min , Maximum GT RF power P max , TDMA communication bandwidth B, GTk total uplink communication rate C in the nth time slot k [n] = Blog2(1+γ k [n]), Unload BitQ k [n]=C k [n]t k [n];

[0171] The energy consumption model includes: blade power P0 in hovering state, blade angular velocity U tip , fuselage drag ratio d0, air density ρ, rotor volume s, rotor area G, induced power P1 in hovering state, induced speed v0 when the rotor flies forward, RF energy consumption e in the nth time slot k [n]=P k [n]t ut, the flight energy consumption of UAV in the nth time slot GTk radio frequency energy consumption e in the nth time slot k [n]=P k [n]t ut .

[0172] The optimization problem constructed in step 2 of the present invention with the goal of maximizing system energy efficiency specifically includes: GT unloading weight RIS phase shift matrix UAV flight trajectory Maximizing the computational offloading energy efficiency of a single UAV flight is formulated as problem (P0):

[0173]

[0174]

[0175] C2:p[0]=p s ,p[N]=p e

[0176]

[0177]

[0178]

[0179]

[0180] in, ε e It is an energy weight parameter that represents the importance of GT RF energy to the overall computational offloading energy efficiency of the system.

[0181] The four sub-problems in step 2 of the present invention specifically include: when the GT unloading weight η, the RIS phase shift matrix Θ, and the UAV flight trajectory p are fixed, there is a first sub-problem (P1), which is the GT RF power optimization problem in step 3 of the present invention:

[0182]

[0183] stC6

[0184] When the GT RF power P, RIS phase shift matrix Θ, and UAV flight trajectory p are fixed, there is a second sub-problem (P2), which is the GT offloading weight optimization problem in step 4 of the present invention:

[0185]

[0186] stC4,C5

[0187] When the GT RF power P, GT unloading weight η, and UAV flight trajectory p are fixed, there is a third sub-problem (P3), which is the RIS phase shift matrix optimization problem in step 5 of the present invention:

[0188]

[0189] stC3

[0190] When the GT RF power P, GT unloading weight η, and RIS phase shift matrix Θ are fixed, there is a fourth sub-problem (P4), which is the UAV flight trajectory optimization problem in step 6 of the present invention:

[0191]

[0192] stC1,C2

[0193] The above are the four sub-problems in step 2 of the present invention.

[0194] In step 3 of the present invention, solving the GT radio frequency power optimization problem using a mathematical analysis method specifically includes the following steps:

[0195] Step 3-1: Expand the objective function in the first sub-problem (P1) as follows:

[0196]

[0197] in And record the objective function The current problem is:

[0198]

[0199] stC6

[0200] Step 3-2: Since (P1′) The constraint C6 is for Independence, then the following problem (P1″) is equivalent to (P1′):

[0201]

[0202] stC6

[0203] That is, it is transformed into exist The maximum value of

[0204] Step 3-3: Ask about P k The partial derivative of [n] is:

[0205]

[0206] From constraint C5, we can get: Let's assume but

[0207]

[0208] After calculation, it can be simplified to:

[0209]

[0210] but and The signs of are consistent;

[0211] Step 3-4: Note Easy to be P k When [n]→0, When P k When [n]→+∞,

[0212] Steps 3-5: Easy to know is a non-decreasing function; hour, is a non-decreasing function, when hour, is a decreasing function with only one zero able to Established, and Part Part

[0213] Step 3-6: Make have to:

[0214] It can be seen that P k [n] appears on the left side of the equal sign, and the right side of the equal sign can be regarded as a constant in this subproblem; Considered as P k [n] is a function of the independent variable, and this transcendental equation can be solved.

[0215] Step 3-7: In summary, when hour, but Satisfied at this time P k [n] takes the value of Pmin ;when When satisfied P k [n] takes the value

[0216] In step 4 of the present invention, solving the GT offloading weight optimization problem using the CVX tool specifically includes the following steps:

[0217] Step 4-1: Expand the objective function in the second sub-problem (P2) as follows:

[0218]

[0219] And record the objective function The current problem is:

[0220]

[0221] stC4,C5

[0222] Step 4-2: Since in (P2′), Constraints C4 and C5 for Independence, then the following problem (P2″) is equivalent to (P2′):

[0223]

[0224] stC4,C5

[0225] Step 4-3: Obviously, the (P2″) optimization problem is a convex optimization problem and can be solved efficiently using the CVX tool.

[0226] In step 5 of the present invention, solving the RIS phase shift matrix optimization problem using a mathematical analysis method specifically includes the following steps:

[0227] Step 5-1: Expand the objective function in the third sub-problem (P3) as follows:

[0228]

[0229] And record the objective function The current problem is:

[0230]

[0231] stC3

[0232] Step 5-2: Since (P3′) The constraint C3 is for Independence, then the following problem (P3″) is equivalent to (P3′):

[0233]

[0234] stC3

[0235] That is, it is transformed into exist The maximum value of

[0236] Step 5-3: Only h k [n] with Related, obviously For h k [n]| is an increasing function, then the following problem (P3″′) is equivalent to (P3″):

[0237]

[0238] stC3

[0239] Step 5-4: The following problem (P3°) and problem (P3″′) have the same optimization process:

[0240]

[0241] stC3

[0242] Step 5-5: Remember

[0243] Then we have the following channel gain expansion, (·) * This means to find the conjugate of the formula in brackets:

[0244]

[0245]

[0246]

[0247] Among them, due to are all random variables;

[0248] Step 5-6: The factors containing random variables in the above step 5-5 should not enter the optimization process. Then the following channel gain partial expansion actually enters the optimization process:

[0249]

[0250]

[0251]

[0252] in:

[0253]

[0254]

[0255] Step 5-7: The problem (P3°) should be rewritten as the following problem (P3°′) based on the actual part of the optimization process, and record

[0256]

[0257] stC3

[0258] Steps 5-8: Obviously, As the upper bound of this optimization objective function, it is a reachable upper bound; there is a feasible way to reach this upper bound, that is, let:

[0259]

[0260] in:

[0261]

[0262] Make It can make

[0263] Step 5-9: In summary, find the upper bound of the objective function Corresponding to The optimal solution is

[0264] In step 6 of the present invention, solving the UAV flight trajectory optimization problem using the continuous convex approximation method specifically includes the following steps:

[0265] Step 6-1: Expand the objective function in the fourth sub-problem (P4) as follows:

[0266]

[0267] Step 6-2: According to the value obtained from (P3°′) in Step 5-7 above, h k [n] is rewritten as The upper bound of Then we have:

[0268]

[0269] in,

[0270]

[0271]

[0272] D GU,k [n]=‖p[n]-p G,k ‖,

[0273] D US [n]=‖p[n]-p BS ‖,v[n]=(t ut ) -1 ‖p[n]-p[n-1]‖;

[0274] Step 6-3: Due to constraint C1, we have So the objective function has the following lower bounds:

[0275]

[0276] Step 6-4: Due to the non-convexity of the objective function, introduce the slack variable g[n]≥‖p[n]-p G,k ‖, j[n]≥‖p[n]-p BS ‖,w[n]≥(t ut ) -1 ‖p[n]-p[n-1]‖, Then we have:

[0277]

[0278] Step 6-5: Remember the objective function

[0279]

[0280] Then the following problem (P4′) has the same optimization process as problem (P4):

[0281]

[0282] stC1,C2

[0283] C7:g[n]≥‖p[n]-p G,k ‖

[0284] C8:j[n]≥‖p[n]-p BS ‖

[0285] C9:w[n]≥(tut ) -1 ‖p[n]-p[n-1]‖

[0286] Step 6-6: To solve the non-convexity of the objective function and constraints C7, C8, and C9, in the rth iteration, a feasible solution g is given. (r) [n],j (r) [n],w (r) [n], and use the first-order Taylor formula to obtain an optimal approximate solution, which can be expressed as the following problem (P4″):

[0287]

[0288] stC1,C2

[0289] C7′:‖p[n]-p G,k ‖ 2 ≤2g (r) [n]g[n]-(g (r) [n]) 2

[0290] C8′:‖p[n]-p A ‖ 2 ≤2j (r) [n]j[n]-(j (r) [n]) 2

[0291] C9′:(t ut ) -2 ‖p[n]-p[n-1]‖ 2 ≤2w (r) [n]w[n]-(w (r) [n]) 2

[0292] in:

[0293]

[0294] Steps 6-7: In summary, problem (P4″) is now a convex optimization problem that can be solved efficiently using the CVX tool.

[0295] In step 7 of the present invention, the block coordinate descent method is used to iteratively update the GT radio frequency power, GT unloading weight, RIS phase shift matrix, and UAV flight trajectory until the optimization target value converges, and ultimately the maximum system energy efficiency is obtained. Specifically, the following steps are included:

[0296] Step 7-1: Initialize P (0) , η (0) ,Θ (0) , p (0), iterative precision parameter χ = 10 -3 , the number of iterations l = 0;

[0297] Step 7-2: Given GT unloading weight η (l) , RIS phase shift matrix Θ (l) 、UAV flight trajectoryp (l) , solve the problem (P1) using the analytical method and obtain the GT RF power P (l+1) ;

[0298] Step 7-3: Given the GT RF power P (l+1) , RIS phase shift matrix Θ (l) 、UAV flight trajectoryp (l) , use CVX tools to solve the convex optimization problem (P2) and obtain the GT unloading weight η (l+1) ;

[0299] Step 7-4: Given the GT RF power P (l+1) , GT unloading weight η (l+1) 、UAV flight trajectoryp (l) , solve problem (P3) using analytical method and obtain the RIS phase shift matrix Θ (l+1) ;

[0300] Step 7-5: Given the GT RF power P (l+1) , GT unloading weight η (l+1) , RIS phase shift matrix Θ (l+1) , use the continuous convex approximation method to solve the problem (P4) and obtain the UAV flight trajectory p (l+1) ;

[0301] Step 7-6: Let l = l + 1;

[0302] Step 7-7: Loop S7.2 to S7.6 until That is, the optimization objective numerical convergence;

[0303] Step 7-8: Finally, the optimal solution P is obtained within the allowable accuracy range. (l) , η (l) ,Θ (l) , p (l) .

[0304] like Figure 2As shown, the present invention proposes an energy efficiency optimization method for a RIS-enhanced UAV-assisted edge computing system. For practical reasons, the channel for K GTs in an urban area to offload tasks to an edge computing server near the BS via uplink is relatively complex. In the state where a fixed RIS (RISb) and a UAV payload RIS (RISu) jointly assist in communication, there are possible LoS channels and NLoS channels according to the different locations of multiple GTs. Based on this consideration, the uplink is divided into a GT direct to BS link, a GT to BS link assisted by RISu, and a GT to BS link assisted by RISb. The single-shot channel state between any two of the GT, RISu, RISb, and BS is represented by different path loss exponents α and Ricean factors δ, so as to describe the complex channel environment in the urban area as much as possible.

[0305] Figure 3 The GT task offloading protocol diagram of the present invention is shown in the figure. The total flight time of the UAV, that is, the single edge computing offloading service time T, is evenly divided into N sufficiently small time slots t ut =N -1 T. Since the number of result bits returned from the edge computing server to the GT is too small, the result return time is negligible. Each time slot is divided into K time periods, providing computing offloading services for K GTs. In the nth time slot, the offloading time of GTk is t k [n] is unloaded by its GT weight η k [n]Decision.

[0306] The above description is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a technician familiar with this technical field within the technical scope disclosed by the present invention should be covered by the scope of protection of the present invention.

Claims

1. A method for optimizing energy efficiency of a RIS-enhanced UAV-assisted edge computing system, characterized by: The method comprises the following steps: Step 1: Establish a system model including K GTs, a fixed RIS, a UAV-loaded RIS, and a BS, and establish a complex channel communication model, a computational offloading model, and an energy consumption model; Step 2: Based on the total GT offload task volume and the system's weighted total energy consumption, construct an optimization problem with the goal of maximizing system energy efficiency. This problem is then decomposed into four sub-problems. The optimization problem with the goal of maximizing system energy efficiency specifically includes: By jointly optimizing GT RF power GT unloading weight RIS phase shift matrix UAV flight trajectory Maximizing the computational offloading energy efficiency of a single UAV flight is formulated as problem (P0): in, ε e The energy weight parameter represents the importance of GT RF energy to the overall computational offloading energy efficiency of the system; The four sub-questions specifically include: When the GT unloading weight η, RIS phase shift matrix Θ, and UAV flight trajectory p are fixed, there is the first sub-problem (P1), which is the GT RF power optimization problem in step 3: When the GT RF power P, RIS phase shift matrix Θ, and UAV flight trajectory p are fixed, there is a second sub-problem (P2), which is the GT offloading weight optimization problem in step 4: When the GT RF power P, GT unloading weight η, and UAV flight trajectory p are fixed, there is a third sub-problem (P3), which is the RIS phase shift matrix optimization problem in step 5: When the GT RF power P, GT unloading weight η, and RIS phase shift matrix Θ are fixed, there is a fourth sub-problem (P4), which is the UAV flight trajectory optimization problem in step 6: The above are the four sub-questions in step 2; Step 3: Use mathematical analysis to solve the GT RF power optimization problem; Step 4: Use CVX tools to solve the GT offloading weight optimization problem; Step 5: Use mathematical analysis to solve the RIS phase shift matrix optimization problem; Step 6: Solve the UAV flight trajectory optimization problem using the continuous convex approximation method; Step 7: Using the block coordinate descent method, repeatedly solve the GT RF power, GT unloading weight, RIS phase shift matrix, and UAV flight trajectory until the numerical convergence, and finally obtain the maximum system energy efficiency.

2. The energy efficiency optimization method of a RIS-enhanced UAV-assisted edge computing system according to claim 1, characterized in that: The system model, complex channel communication model, computation offloading model, and energy consumption model in step 1 specifically include: Step 1-1: The system model includes: GT set A collection of reflector units for the UAV payload RIS (RISu) and the fixed RIS (RISb) The total duration of a single UAV flight, that is, the single edge computing offloading service time T, the UAV flight altitude z, and a sufficiently small time slot t ut =N -1 T, time slot set The coordinates of the UAV, i.e., RISu, in the nth time slot are p[n] = [x[n], y[n], z] T , GTk's coordinate p G,k =[x G,k ,y G,k ,0] T , the coordinates of RISb are p B =[x B ,y B ,z B ] T , the coordinates of BS are p BS =[x BS ,y BS ,z BS ] T , UAV speed in a single time slot v[n]=(t ut ) -1 ‖p[n]-p[n-1]‖, UAV maximum flight speed V max , the UAV’s flight starting point p s =[x s ,y s ,z] T , the UAV's flight endpoint p e =[x e ,y e ,z] T , the phase shift matrix when RISu serves GTk in the nth time slot Phase shift matrix when RISb serves GTk in the nth time slot in The mth time slot from GTk to RISu in the nth time slot U The reflection angle of each reflection unit, the reflection coefficient β of RISu U ∈[0,1], where The mth time slot from GTk to RISb in the nth time slot B The reflection angle of each reflection unit, the reflection coefficient β of RISb B ∈[0,1]; Step 1-2: The complex channel communication model includes: the ground distance D0 of the received power reference point in the large-scale propagation model, the received power (C0) of the antenna at a distance D0 from the unit power transmitter in free space. 2 , in the nth time slot, GTk to the mth time slot in RISu U The single-reflection channel gain of a reflector unit Among them is the geometric distance D between GTk and RISu in the nth time slot GU,k [n]=‖p[n]-p G,k [n]‖, the path loss exponent α of this single-shot channel GU,k , the Ricean factor δ of this single-shot channel GU,k , the line-of-sight (LoS) component of this channel The non-line-of-sight (NLoS) component of this channel GTk to RISb in the nth time slot B The single-reflection channel gain of a reflector unit Among them is the geometric distance D between GTk and RISb GB,k , the path loss index α of this single-shot channel GB,k , the Ricean factor δ of this single-shot channel GB,k , the LoS component of this channel NLoS component of this channel The mth time slot in RISu in the nth time slot U Single-ray channel gain from a reflector unit to the BS Among them is the geometric distance D between RISu and BS US [n], the path loss exponent α of this single-shot channel US , the Ricean factor δ of this single-shot channel US , the LoS component of this channel NLoS component of this channel The mth RISb in the nth time slot B Single-ray channel gain from a reflector unit to the BS Among them is the geometric distance D between RISb and BS BS , the path loss index α of this single-shot channel BS , the Ricean factor δ of this single-shot channel BS , the LoS component of this channel NLoS component of this channel Single-shot channel gain from GTk to BS in the nth time slot Among them is the geometric distance D between GTk and BS GS,k , the path loss index α of this single-shot channel GS,k , the Ricean factor δ of this single-shot channel GS,k , the LoS component of this channel NLoS component of this channel RIS uses a uniform square planar array model with an azimuth angle of the receiving angle of the GTk RF signal received at RISu. Pitch angle Azimuth of the receiving angle of the GTk RF signal received at RISb Pitch angle Azimuth of the deviation angle of the radio frequency signal transmitted from RISu to BS Pitch angle Azimuth of the deviation angle of the radio frequency signal transmitted from RISb to BS Pitch angle The calculation process is Distance d between RIS reflectors, carrier wavelength λ, LoS component LoS component LoS component LoS component Vector Vector Vector Vector GTk to BS link gain h via RISu assistance GUS,k [n]=(h GU,k [n]) H Θ U,k [n]h US [n], GTk to BS link gain h via RISb assistance GBS,k [n]=(h GB,k [n]) H Θ B,k [n]h BS [n]; The total uplink channel gain h of GTk reaching BS through different paths k [n]=h GS,k [n]+h GUS,k [n]+h GBS,k [n]; Step 1-3: The calculation of the unloading model includes: the unloading duration t of GTk in the nth time slot k [n] = η k [n]t ut , the GT offloading weight η of GTk in the nth time slot k [n], minimum GT unloading weight value η min ≤(K) -1 , the uplink signal-to-noise ratio (SNR) of GTk reaching BS through different paths in the nth time slot k [n] = γ k [n]=σ -2 P k [n]|h k [n]| 2 , complex Gaussian white noise power σ 2 , the average GT RF power P of GTk in the nth time slot k [n], the rated minimum GT RF power P of GT to ensure effective data transmission min , Maximum GT RF power P max , TDMA communication bandwidth B, GTk total uplink communication rate C in the nth time slot k [n] = Blog2(1+γ k [n]), unloading bit Q k [n]=C k [n]t k [n]; Step 1-4: The energy consumption model includes: blade power P0 in the hovering state, blade angular velocity U tip , fuselage drag ratio d0, air density ρ, rotor volume s, rotor area G, induced power P1 in hovering state, induced speed v0 when the rotor flies forward, RF energy consumption e in the nth time slot k [n]=P k [n]t ut , the flight energy consumption of UAV in the nth time slot GTk radio frequency energy consumption e in the nth time slot k [n]=P k [n]t ut .

3. The energy efficiency optimization method of a RIS-enhanced UAV-assisted edge computing system according to claim 1, characterized in that: In step 3, solving the GT radio frequency power optimization problem using a mathematical analysis method specifically includes the following steps: Step 3-1: Expand the objective function in the first sub-problem (P1) as follows: in And record the objective function The current problem is: Step 3-2: Since (P1′) The constraint C6 is for Independence, then the following problem (P1″) is equivalent to (P1′): That is, it is transformed into exist The maximum value of Step 3-3: Ask about P k The partial derivative of [n] is: From constraint C5, we can get: Let's assume but After calculation, it is simplified to: but and The signs of are consistent; Step 3-4: Note Easy to be a P k When [n]→0, When P k When [n]→+∞, Steps 3-5: Easy to know is a non-decreasing function; hour, is a non-decreasing function, when hour, is a decreasing function with only one zero able to Established, and Part Part Step 3-6: Make have to: See P k [n] appears on the left side of the equal sign, and the right side of the equal sign is considered a constant in this subproblem; Considered as P k [n] is a function of the independent variable. Solve this transcendental equation and we have the solution Step 3-7: In summary, when hour, but Satisfied at this time P k [n] takes the value of P min ;when When satisfied P k [n] takes the value 4. The energy efficiency optimization method for a RIS-enhanced UAV-assisted edge computing system according to claim 1, characterized in that: In step 4, using the CVX tool to solve the GT offloading weight optimization problem specifically includes the following steps: Step 4-1: Expand the objective function in the second sub-problem (P2) as follows: And record the objective function The current problem is: Step 4-2: Since (P2 ′ )middle Constraints C4 and C5 for Independent, then the following problem (P2″) is equivalent to (P2 ′ ): Step 4-3: Obviously, the (P2″) optimization problem is a convex optimization problem and can be solved efficiently using CVX tools.

5. The energy efficiency optimization method of a RIS-enhanced UAV-assisted edge computing system according to claim 1, characterized in that: In step 5, solving the RIS phase shift matrix optimization problem using a mathematical analysis method specifically includes the following steps: Step 5-1: Expand the objective function in the third sub-problem (P3) as follows: And record the objective function The current problem is: Step 5-2: Due to (P3 ′ )middle The constraint C3 is for Independent, then the following problem (P3″) is equivalent to (P3 ′ ): That is, it is transformed into exist The maximum value of Step 5-3: Only h k [n] with Related, obviously For h k [n]| is an increasing function, then the following problem (P3″′) is equivalent to (P3″): Step 5-4: The following problem (P3°) and problem (P3″′) have the same optimization process: Step 5-5: Remember Then we have the following channel gain expansion, (·) * This means to find the conjugate of the formula in brackets: Among them, due to are all random variables; Step 5-6: The factors containing random variables in the above step 5-5 should not enter the optimization process. Then the following channel gain partial expansion actually enters the optimization process: in: Step 5-7: Then the problem (P3°) should be rewritten as the following problem (P3°) based on the actual part of the optimization process. ′ ), and remember Steps 5-8: Obviously, As the upper bound of this optimization objective function, it is a reachable upper bound; there is a feasible way to reach this upper bound, that is, let: in: Make It can make Step 5-9: In summary, find the upper bound of the objective function Corresponding to The optimal solution is 6. The energy efficiency optimization method of a RIS-enhanced UAV-assisted edge computing system according to claim 1, characterized in that: The method of solving the UAV flight trajectory optimization problem using the continuous convex approximation method in step 6 specifically includes the following steps: Step 6-1: Expand the objective function in the fourth sub-problem (P4) as follows: Step 6-2: According to the value obtained from (P3°′) in Step 5-7 above, h k [n] is rewritten as The upper bound of Then we have: in, Step 6-3: Due to constraint C1, we have So the objective function has the following lower bounds: Step 6-4: Due to the non-convexity of the objective function, introduce the slack variable g[n]≥‖p[n]-p G,k ‖, j[n]≥‖p[n]-p BS ‖,w[n]≥(t ut ) -1 ‖p[n]-p[n-1]‖, Then we have: Step 6-5: Remember the objective function Then there are the following problems (P4 ′ ) has the same optimization process as problem (P4): Step 6-6: To solve the non-convexity of the objective function and constraints C7, C8, and C9, in the rth iteration, a feasible solution g is given. (r) [n],j (r) [n],w (r) [n], and use the first-order Taylor formula to obtain an optimal approximate solution, which can be expressed as the following problem (P4″): in: Steps 6-7: In summary, the problem (P4″) is now a convex optimization problem that can be solved efficiently using CVX tools.

7. The energy efficiency optimization method of a RIS-enhanced UAV-assisted edge computing system according to claim 1, characterized in that: In step 7, the block coordinate descent method is used to iteratively update the GT RF power, GT unloading weight, RIS phase shift matrix, and UAV flight trajectory until the optimization target value converges, and ultimately the maximum system energy efficiency is obtained. Specifically, the following steps are included: Step 7-1: Initialize P (0) , η (0) ,Θ (0) , p (0) , iterative precision parameter η=10 -3 , the number of iterations l = 0; Step 7-2: Given GT unloading weight η (l) , RIS phase shift matrix Θ (l) 、UAV flight trajectoryp (l) , solve the problem (P1) using the analytical method and obtain the GT RF power P (l+1) ; Step 7-3: Given the GT RF power P (l+1) , RIS phase shift matrix Θ (l) 、UAV flight trajectoryp (l) , use CVX tools to solve the convex optimization problem (P2) and obtain the GT unloading weight η (l+1) ; Step 7-4: Given the GT RF power P (l+1) , GT unloading weight η (l+1) 、UAV flight trajectoryp (l) , solve problem (P3) using analytical method and obtain the RIS phase shift matrix Θ (l+1) ; Step 7-5: Given the GT RF power P (l+1) , GT unloading weight η (l+1) , RIS phase shift matrix Θ (l+1) , use the continuous convex approximation method to solve the problem (P4) and obtain the UAV flight trajectory p (l+1) ; Step 7-6: Let l = l + 1; Step 7-7: Repeat the above steps 7-2 to 7-6 until That is, the optimization objective numerical convergence; Step 7-8: Finally, the optimal solution P is obtained within the allowable accuracy range. (l) , η (l) ,Θ (l) , p (l) .

Images