Intelligent metasurface assisted perception communication calculation integrated resource management method

Abstract

The invention provides an intelligent metasurface-assisted sensing communication calculation integrated resource management method, which comprises the following steps of: constructing an intelligent reflecting surface-assisted full-duplex integrated sensing communication calculation integrated system, and connecting a local cache with a full-duplex base station through a backhaul link; determining a signal model, a communication model, a perception model, a calculation model and a cache model, and establishing an optimization problem of system utility maximization according to the models; according to an optimization problem, firstly, an optimal cache strategy is designed, and then the sum of communication throughput and total calculation bits (unloading and local calculation bits) is maximized through joint optimization of a transmitting beam former and a receiving beam former, calculation of user local calculation frequency, calculation of user transmission power and phase shift of a passive reflection unit of an intelligent reflection surface. The effectiveness of the method and the advantages of the method compared with other reference algorithms are verified through simulation results.

Description

Technical Field

[0001] The present invention relates to the technical field of integrated perception communication computing, and in particular to an intelligent super-surface assisted perception communication computing integrated resource management method. Background Art

[0002] With the rapid development of future wireless systems, integrated sensing and communication (ISAC) technology has become a hot research topic in academia and industry. By enabling radar sensing and wireless communication to use the same frequency bandwidth, radio signals and infrastructure in the ISAC system, the spectrum efficiency is significantly improved. At the same time, the application of massive multiple-input multiple-output (MIMO), millimeter-wave and terahertz communication technologies to radar sensing has promoted the realization of ISAC technology, making ISAC technology have higher resolution in many important application scenarios. Therefore, ISAC is envisioned as a promising solution in future wireless networks. However, in order to cope with the growing computing needs of smart applications, more and more intelligent algorithms have been proposed for data processing, which inevitably becomes a heavy computing load. In order to effectively solve these problems, a widely acclaimed solution is mobile edge computing-caching (MECC), which achieves better network performance and improves user service quality by partially or completely offloading computing data to the edge of the network and pre-fetching user data using a local cache connected to the mobile edge computing (MEC) server.

[0003] Due to the complex wireless propagation environment, when the direct link between the base station and the user is blocked by obstacles such as buildings, the network performance will be significantly degraded. To solve this problem, the intelligent reflecting surface (IRS) or reconfigurable smart surface (RIS) has become another key technology in future wireless networks. IRS, which is composed of a large number of low-cost passive components, can provide a virtual line-of-sight link to bypass obstacles and expand the wireless coverage, thereby improving the system performance accordingly. There have been many studies on integrated sensing, communication and computing (ISCC) and semi-passive IRS-assisted ISAC systems. When studying the waveform design of ISCC or ISAC systems, most of the works consider half-duplex (HD) systems. Compared with HD systems, full-duplex (FD) systems can achieve higher spectral efficiency. Therefore, in this paper, an advanced semi-passive IRS-assisted FD ISCC system is studied.

[0004] The invention patent with application number 202111650080.1 discloses a method for secure communication in a full-duplex wireless energy-carrying network assisted by an intelligent reflective surface, including the following steps: adding full-duplex wireless energy-carrying users to the wireless communication network model assisted by an intelligent reflective surface, and constructing a full-duplex user-cooperative intelligent reflective surface-assisted wireless energy-carrying network model; establishing a mathematical model of the optimization problem based on the energy of the transmitter and the user, the phase shift of the intelligent reflective surface, and the confidentiality rate at the legitimate user; maximizing the confidentiality rate of the legitimate receiver by jointly optimizing the transmitting end transmit beamforming vector, the phase shift matrix of the intelligent reflective surface, and the power of the artificial noise signal transmitted by the legitimate receiver. The above invention can effectively improve the service quality of the legitimate receiver and ensure the security of the physical layer of the network, but it does not consider combining perception, communication and computing. Summary of the invention

[0005] In view of the technical problem that the existing resource management methods do not consider the combination of FD systems and ISCC, the present invention proposes an intelligent metasurface-assisted perception, communication and computing integrated resource management method, which meets the worst-case radar SINR constraints, transmission power budget, transmission energy budget, storage capacity budget and IRS reflection coefficient unit modulus constraints, considers the system utility maximization problem, can effectively improve system performance and achieve higher spectrum efficiency.

[0006] In order to achieve the above object, the technical solution of the present invention is implemented as follows: a smart metasurface assisted perception communication computing integrated resource management method, the steps of which are as follows:

[0007] Step 1: Deploy multiple passive reflection units and active sensing elements on the IRS, use a full-duplex base station for uplink and downlink signal transmission, and connect a local cache to the full-duplex base station through a backhaul link to build a full-duplex integrated sensing communication computing integrated system assisted by an intelligent reflection surface;

[0008] Step 2: Determine the signal model, communication model, perception model, computing model and cache model according to the full-duplex integrated perception communication computing integrated system assisted by the reflective surface, and establish an optimization problem for maximizing the system utility according to the phase shift of the passive reflection unit, the transmit beamforming vector and the receive beamforming vector at the full-duplex base station, the local computing frequency and transmission power of all computing users, as well as the communication throughput, the total computing bits and the total backhaul cost of content delivery;

[0009] Step 3: According to the optimization problem, optimize the caching strategy to minimize the total return cost of content delivery of the caching model and obtain the content placement indicator for each file; use the weighted minimum mean square error method to transform the objective function of the maximization problem composed of the computing model and the communication model, and use the block coordinate ascent method to divide the coupled variables into several blocks for alternate optimization to obtain the phase shift of the passive reflection unit of all intelligent reflection surfaces, the transmit beamforming vector and the receive beamforming vector at the full-duplex base station, and the local computing frequency and transmission power of all computing users.

[0010] Preferably, the intelligent reflective surface-assisted full-duplex integrated perception communication computing integrated system further comprises K randomly distributed communication users, L randomly distributed computing users and a perception target with a radar; the full-duplex base station is equipped with two uniform linear arrays, and the IRS is equipped with a uniform linear array composed of M passive reflective elements and M a An active sensing element for receiving target echo signals. The full-duplex base station is equipped with N t transmit antennas and N r The full-duplex base station simultaneously receives uplink offload signals and transmits downlink ISAC signals for multi-user communication and target perception with the help of IRS. The full-duplex base station is connected to a local cache with limited storage capacity and connected to the MEC server through a high-capacity backhaul link, which is used for edge computing to store user data and perform computing tasks.

[0011] Preferably, the objective function of the optimization problem of maximizing the system utility is to maximize the difference between the sum of the communication throughput and the total computing bits and the total backhaul cost of the content delivery.

[0012] Preferably, the optimization problem is:

[0013]

[0014] in, represents the achievable communication rate from the full-duplex base station to the kth communication user, represents the achievable offloading rate of the lth computing user at the full-duplex base station, represents the computing rate of the lth computing user, is the total return cost, r tar represents the radar signal-to-interference-noise ratio of the target, Γ tar represents the predefined target perception performance threshold, represents the maximum energy budget that can be consumed by the lth computing user; and They represent the transmit beamforming vector w of the kth communication user at the full-duplex base station. k and the lth calculated user receive beamforming vector ul K and L represent the total number of communication users and computing users respectively, f=[f 1 ,…,f L ] H represents the local computing frequency vector of the computing user, p = [p 1 ,…,p L ] H represents the calculation of the user's transmission power vector, φ=[φ 1 ,…,φ M ] H represents the set of phase shifts of all passive reflection units; e represents the content placement indicator e v The set composed of BS represents the maximum transmission power at the full-duplex base station, T is the channel coherence time, φ m represents the phase shift of the mth passive reflection unit, represents the index set of passive reflective units, ζ represents a constant related to the hardware structure of the computing user, and p l is the transmission power of the lth calculated user, f l represents the local computing frequency of the lth computing user, q v Indicates the length of the vth file, e v Indicates the content placement indicator of the vth file, Tp l It represents the energy consumed by the lth computing user within the channel coherence time, and F is the cache capacity.

[0015] Preferably, the method for optimizing the cache strategy is: within a channel coherence time T, the total return cost is The optimization problem Re-write as:

[0016]

[0017] Among them, ρ v represents the pricing of the backhaul cost of content delivery of the vth file; V represents the total number of files, e v ∈[0,1] represents the content placement indicator, To calculate the probability representation of the vth file requested by the user, R 0,l Backhaul data rate for content acquisition; optimization issues For content placement indicators v The set composed of e=[e 1 ,…e V ] T It is convex and can be solved directly by a convex optimization solver or KKT optimality condition.

[0018] Preferably, the method of transforming the objective function of the maximization problem composed of the calculation model and the communication model by using the weighted minimum mean square error method is: Reformulating it as a problem of maximizing the sum of the communication throughput and the total computation bits for a given placement, the optimization problem is rewritten as:

[0019]

[0020] By introducing the auxiliary variable α = {α 1,k ,α 2,l},β={β 1,k ,β 2,l}, the optimization problem The objective function of is equivalently written in two variant forms as follows:

[0021]

[0022] in, represents a real number, [·] * represents conjugation, and is the intermediate variable; α 1,k ,α 2,l , β 1,k ,β 2,l All are auxiliary variables; r k com is the signal-to-interference-to-noise ratio of the kth communication user, r l off Receive the offloading signal y for the lth computing user BS The signal-to-noise ratio, H represents the transpose of the matrix, and the intermediate variable Intermediate variables h PU,k and G t denote the channel vectors between the kth communication user and the IRS and between the transmitting antenna of the full-duplex base station and the IRS respectively; e l,k and g PU,l denote the channel vectors between the kth communication user and the lth computing user and between the lth computing user and the IRS, respectively. represents the phase shift matrix of IRS, φ m ∈[0,2π) represents the phase shift of the mth passive reflection unit, represents the noise power of the additive white Gaussian noise at the receiver of the kth communication user; the intermediate variable g AU,l represents the channel vector between the lth computing user and the active sensing element of the IRS, G r represents the channel vector between the receiving antenna of the full-duplex base station and the IRS, represents the noise power of the additive white Gaussian noise at the full-duplex base station receiver, H SI represents the residual self-interference channel matrix at the full-duplex base station;

[0023] Optimization Problem Equivalently expressed as:

[0024]

[0025] in, They are all intermediate variables, and B represents the channel bandwidth.

[0026] Preferably, the method of dividing the coupled variables into several blocks for alternate optimization using the block coordinate ascent method is:

[0027] When other optimization variables are fixed, the optimization of auxiliary variables is an unconstrained convex problem. Derivative the objective function to obtain the optimal solution of the auxiliary variables;

[0028] Solving for the set of IRS phase shifts φ based on other given variables will result in the optimization problem It is expressed as:

[0029]

[0030] By introducing some intermediate variables, the objective function and constraint C2 are rewritten as:

[0031]

[0032] Among them, some intermediate variables are expressed as:

[0033]

[0034] in, and M represents the receiving steering vector of the active sensing element of IRS and the transmitting steering vector of the passive reflecting element of IRS respectively; a is the total number of active sensing elements; θ represents the angle of the sensing target relative to the IRS; η RT Represents the reflection coefficient of the perceived target;

[0035] The optimization problem Rewritten as:

[0036]

[0037] Solving optimization problems using ADMM Since the constraint C3 is a non-convex constant amplitude constraint, a copy variable ψ of the phase shift set φ is introduced to align the decoupling, and the optimization problem The equivalent is:

[0038]

[0039] Among them, the replica variable ψ=[ψ 1 ,…,ψ M ] T ; Based on ADMM optimization of the enhanced Lagrangian problem, the optimization problem Re-write as:

[0040]

[0041] in, represents the square of the L2 norm, ρ>0 represents the penalty coefficient, and λ represents the dual variable; the optimization problem is solved by alternately updating each variable

[0042] When the replica variable ψ and the dual variable λ are fixed, the optimization problem Simplified to:

[0043]

[0044] Due to the constraints Optimization Problem is non-convex, using the MM framework, by 0 Convex constraints by linearizing point pairs and given as:

[0045]

[0046] Among them, φ 0 represents the solution in the current iteration;

[0047] Replace constraints φ H T 0 φ, optimization problem Re-write as:

[0048]

[0049] in,(·) * denotes conjugation;

[0050] Optimization Problem It is a typical second-order cone procedure, which is directly solved using the existing convex optimization solver to obtain the optimal solution for phase shift;

[0051] When the phase shift φ and the dual variable λ are fixed, the optimization problem Simplified to:

[0052]

[0053] Since the replica variable ψ has a unit modulus term, the quadratic term in the objective function with respect to the replica variable ψ is a constant, i.e. Optimization Problem Simplified to:

[0054]

[0055] When all elements of the replica variable ψ are consistent with the linear coefficient (ρ -1 When the elements of φ+λ) are aligned, the optimization problem is realized The optimal solution of the replica variable is expressed as:

[0056] After obtaining the phase shift φ and the replica variable ψ, the dual variable λ is updated by gradient ascent:

[0057] λ (i+1) =λ (i) +ρ -1 (φ-ψ);

[0058] Among them, λ (i) , (i+1) denote the dual variables obtained at the i-th iteration and the i+1-th iteration respectively.

[0059] Preferably, when other variables are fixed, the beamforming vector w of the full-duplex base station is k The optimization method is: Re-expressed as:

[0060]

[0061] By introducing some intermediate variables, the objective function and constraint C2 are rewritten as:

[0062]

[0063] Among them, Tr(·) represents the trace operation, and the intermediate variable is:

[0064]

[0065] Optimization Problem Re-expressed as:

[0066]

[0067] Due to the rank-one constraint of constraint C9, the optimization problem It is non-convex, and the optimal solution of the problem is obtained by relaxing the constraint C9 using SDR technology. Solve it directly through the convex optimization solver, and use the Gaussian randomization method to recover the corresponding rank-one solution to obtain the beamforming vector

[0068] Preferably, when other variables are fixed, the receive beamforming vector u l The optimization is: Re-expressed as:

[0069]

[0070] Rewrite the objective function as:

[0071]

[0072] Among them, the intermediate variables are:

[0073]

[0074] Based on the above transformation, the optimization problem Re-expressed as:

[0075]

[0076] Optimization Problem Decompose into L independent sub-problems, each sub-problem is:

[0077]

[0078] Optimization Problem It is a typical unconstrained convex quadratic problem. By setting the derivative of the objective function to zero, the optimal solution of the receive beamforming vector is obtained:

[0079]

[0080] Preferably, the method for optimizing the transmission power and local computing resources of all computing users given other variables is: Re-expressed as

[0081]

[0082] The objective function and constraint C2 are rewritten as:

[0083]

[0084] Among them, the intermediate variables are:

[0085]

[0086] Optimization Problem Rewritten as:

[0087]

[0088] Optimization Problem Relative to the local computing frequency and transmission power,f,p,is concave, so it is a convex optimization problem. The existing solver is directly used to optimize the transmission power and local computing frequency of each computing user.

[0089] Compared with the prior art, the present invention has the following beneficial effects:

[0090] This paper studies the resource management problem in a full-duplex (FD) integrated sensing communication computing (ISCC) assisted by an intelligent reflecting surface (IRS)---FD ISCC system. An FD base station (BS) is used to provide services for uplink and downlink transmission, and a local cache is connected to the BS through a backhaul link for storing data; at the same time, active sensing elements are deployed on the IRS to receive target echo signals. On this basis, in order to evaluate the overall performance of the system under consideration, we propose an optimization problem to maximize the system utility function on the basis of ensuring the sensing quality, which is expressed as the difference between the sum of the communication throughput and the total computing bits (offloaded and local computing bits) and the total backhaul cost of content delivery, while satisfying the worst-case radar SINR constraint, transmission power budget, transmission energy budget, storage capacity budget and IRS reflection coefficient unit modulus constraint, which can effectively improve system performance and achieve higher spectrum efficiency. Due to the highly non-convex coupling of optimization variables, which makes the problem difficult to solve, an effective algorithm is designed to solve this non-convex problem: first, the optimal caching strategy is designed; then an algorithm based on weighted minimum mean square error (WMMSE), multiplier substitution method (ADMM), maximization-minimization (MM) framework, semi-definite relaxation (SDR) technology and some complex transformations is developed to jointly solve the optimization variables; simulation results verify the utility performance of the proposed algorithm. Finally, the complexity of the present invention is discussed, and the effectiveness of the present invention and its advantages over other benchmark algorithms are verified by simulation results. BRIEF DESCRIPTION OF THE DRAWINGS

[0091] 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.

[0092] Figure 1 It is a framework diagram of the IRS-assisted FD ISCC system of the present invention.

[0093] Figure 2It is a curve diagram of the convergence of the present invention.

[0094] Figure 3 This is a comparison diagram of the relationship between the total number of digits and the number N of IRS reflection elements in the present invention.

[0095] Figure 4 The total number of bits of the present invention and the maximum transmission power P of the BS BS Relationship comparison chart.

[0096] Figure 5 is the sum of the digits of the present invention and the target perception requirement Γ tar Relationship comparison chart.

[0097] Figure 6 is the utility function of the present invention and the return data rate R 0,l Relationship comparison chart.

[0098] Figure 7 It is a comparison diagram of the relationship between the system utility function and the skewness factor ∈ of the present invention. DETAILED DESCRIPTION

[0099] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0100] An intelligent metasurface-assisted perception, communication and computing integrated resource management method, the steps of which are as follows:

[0101] Step 1: Build an IRS-assisted FD ISCC system: deploy multiple passive reflection units and active sensing elements on the IRS, use a full-duplex base station for uplink and downlink signal transmission, and connect a local cache to the full-duplex base station through a backhaul link to build a full-duplex integrated sensing communication computing integrated system assisted by an intelligent reflection surface.

[0102] like Figure 1 As shown in the figure, an IRS-assisted FD ISCC system is considered, which contains an FD BS (full-duplex base station) equipped with two uniform linear arrays (ULA), a semi-passive IRS equipped with ULAs composed of M passive reflective elements, K randomly distributed communication users (CM-UEs), L randomly distributed computational users (CP-UEs) and a point sensing target. The FD BS is equipped with N t transmit antennas (TX) and N rThe FD BS has a receive antenna (RX), while each CM-UE and CP-UE is equipped with a single antenna. The FD BS simultaneously receives uplink offload signals and transmits downlink ISAC signals for multi-user communication and target perception with the help of IRS. In addition, the FD BS is connected to a local cache with limited storage capacity and connected to the MEC server through a high-capacity backhaul link, which is used for edge computing to store user data and perform computing tasks. For convenience, let denotes the index set of IRS elements, communication users, and computing users respectively. In order to solve the serious path loss, it is assumed that in addition to M passive reflection elements, there are M a An active sensing element is used to receive the target echo signal. In addition, due to the unfavorable environmental propagation, it is assumed that there are obstacles between the base station and the user and the target, and the direct link is blocked.

[0103] Step 2: Determine the signal model, communication model, perception model, computing model and cache model based on the full-duplex integrated perception communication computing integrated system assisted by the reflective surface, and establish an optimization problem for maximizing the system utility based on the phase shift of the passive reflection unit, the transmit beamforming vector and the receive beamforming vector at the FD BS, the local computing frequency and transmission power of all computing users, as well as the communication throughput, total computing bits and total backhaul cost of content delivery.

[0104] The optimization problem of maximizing system utility is to maximize the system utility consisting of the difference between the sum of communication throughput and total computing bits and the total backhaul cost of content delivery, and determine the constraints based on the perception model, signal model, communication model, and cache model.

[0105] In the signal model, the present invention first considers the system's downlink synaesthesia, where a dual-function ISAC signal Simultaneously perform downlink multi-user communication and radar sensing. represents the information signal transmitted by the kth communication user, and represents the beamforming vector of the kth communication user. Similarly, s 0 Indicates a dedicated radar signal, represents the corresponding radar signal beamforming vector. Assuming the transmitted information signal and radar signals 0 are all independent random variables with mean 0 and unit variance, and the downlink information signal and radar signals 0 are independent and unrelated, that is Therefore, the dual-function ISAC signal sent by the FD BS can be expressed as:

[0106]

[0107] In addition, the total transmission power constraint of the FD BS is considered as P BS Indicates the maximum transmission power at the BS.

[0108] When the FD BS performs downlink transmission, it receives the uplink offload signal and the target echo at the same time. The offload signal transmitted by the lth computing user can be written as:

[0109]

[0110] in, and p l denote the information signal and transmission power of the lth computing user respectively. For simplicity, it is assumed that each information signal have zero mean and unit variance and are uncorrelated with each other.

[0111] In the communication model, the signal received by the kth communication user can be expressed as:

[0112]

[0113] in, and denote the channel vector between the kth communication user and the IRS and the transmitting antenna of the FD BS and the IRS respectively. and They respectively represent the channel vectors between the kth communication user and the lth calculation user and between the lth calculation user and the IRS, and the channel vectors can be easily obtained by traditional channel estimation methods. represents the additive white Gaussian noise at the user receiver, represents the noise power. The diagonal matrix represents the phase shift matrix of IRS, where represents the phase shift of the mth passive reflection unit. denotes a complex-valued matrix, and the superscript H denotes the transpose of the matrix.

[0114] Therefore, the signal to interference and noise ratio (SINR) of the kth communication user can be written as:

[0115]

[0116] in,

[0117] The achievable communication rate from the FD BS to the kth communication user is expressed as:

[0118]

[0119] Here, B represents the channel bandwidth.

[0120] In the perception model, the target echo signal under consideration is modeled, assuming that the radar signal consists of a no-line-of-sight path (NLos), and the transmitting ULA (uniform linear array) at the FD BS and the receiving ULA at the IRS are both half-wavelength antenna spacings, and the uniform linear array is obtained by uniformly arranging antenna elements at fixed intervals (such as half a wavelength) on a straight line. Therefore, the signal received on the active sensing element of the IRS is expressed as:

[0121]

[0122] in, represents the reflection coefficient of the target, which is mainly determined by the path loss and radar cross section. θ represents the angle of the perceived target relative to the IRS. represents the additive white Gaussian noise at the IRS receiver, I Ma denote the noise power and M respectively a ×M a The identity matrix of . and They represent the receiving steering vector of the active sensing element of the IRS and the transmitting steering vector of the passive reflecting element of the IRS respectively. represents the channel vector between the lth computational user and the IRS. In addition, assuming that the angle θ and the reflection coefficient η RT is given in order to design the most appropriate transmit signal to detect that particular target.

[0123] The performance of a radar is closely related to its corresponding signal-to-interference-to-noise ratio to a large extent. In particular, in a MIMO radar system, the detection probability of a point target usually increases with the increase of the output SINR. Therefore, the radar SINR is directly used to characterize the perception performance, and the radar SINR is obtained as:

[0124]

[0125] The computation model includes offload computation and local computation. In offload computation, the offload signal received by the FD BS can be expressed as:

[0126]

[0127] in, represents the channel vector between the receiving antenna of the FD BS and the IRS. represents the additive white Gaussian noise at the FD BS receiver, I Nr Represent the noise power and N r ×N rThe SI power can be greatly reduced by utilizing the self-interference (SI) cancellation technique. Without loss of generality, the residual SI signal is expressed as H SI x DL ,in Denotes the residual SI channel matrix at the FD BS. Due to the severe attenuation of the channel between the FD BS and the IRS, the reflected echo interference and the self-interference introduced by the IRS reflected synaesthesia signal are ignored.

[0128] Technically, a receive beamforming vector is applied on the FD BS. To restore the uninstall signal y BS , get the receiving offloading signal y corresponding to the lth computing user BS SINR:

[0129]

[0130] Among them, the intermediate variable

[0131] Therefore, the achievable offloading rate of the lth computing user at the FD BS is written as:

[0132]

[0133] In local calculation, assuming ε l represents the CPU cycles required for each computing user CP-UE to process one bit of raw data, f l represents the local CPU computing frequency of the lth computing user, and ζ represents a constant related to the user's hardware structure. Therefore, the computing rate and energy consumption at each computing user can be expressed as: and Where T represents the channel coherence time.

[0134] Next, we model the cache strategy of FD BS. The set of files V is represented by Represented by vector q=[q 1 ,…q V ] represents the length of these files. Assume that in the content database V, all files can be sorted according to their popularity. In addition, a probabilistic caching strategy is adopted to randomly cache content with deterministic probability. The caching strategy can be defined as a vector e = [e 1 ,…e V ] T ,in In addition, since the cache of the FD BS is limited by the cache capacity F, there is a local cache constraint In addition, the probability of calculating the vth file requested by the user is expressed as For simplicity, we assume that the request probability of the vth file is consistent among different computing users without loss of generality. For example, We assume that users request content files based on their popularity. The probability of this request can be modeled as a Zipf distribution, i.e., the request probability Among them, ∈ is the skewness factor, which represents the deviation of file popularity.

[0135] In content delivery, if the vth file has been cached on the local storage of the FD BS, the user can directly access the content by request. Otherwise, the content needs to be obtained from the FD BS through the backhaul link. The performance indicator for obtaining content is defined as the backhaul data rate R 0,l , which is generally calculated based on factors such as link bandwidth and channel conditions. Therefore, within a channel coherence time T, the total return cost can be modeled as Among them, ρ v The pricing represents the round-trip cost of content delivery for the vth file, set to 0.01.

[0136] In order to evaluate the overall performance of the considered system, a system utility maximization problem is discussed, which is expressed as the gap between the sum of the communication throughput and the total computation bits (offloaded and local computation bits) and the total backhaul cost of content delivery. Mathematically, the optimization problem can be expressed as follows:

[0137]

[0138] Among them, Γ tar represents the predefined target perception performance threshold, represents the maximum energy budget that can be consumed by computing user l. denote the set of transmit beamforming vectors and receive beamforming vectors at the FD BS, respectively, and f = [f 1 ,…,f L ] T represents the local computing frequency vector of the computing user, p = [p 1 ,…,p L ] T represents the calculation of the user's transmission power vector, φ=[φ 1 ,…,φ M ] T Represents the set of phase shifts of all passive reflective units. e represents the content placement indicator e v The set composed of BS represents the maximum transmission power at the FD BS, T is the channel coherence time, r tar is the radar SINR. m represents the phase shift of the mth passive reflection unit, p lis the transmission power of the lth calculated user, f l represents the local computing frequency of the lth computing user, q v Indicates the length of the vth file, e v Indicates the content placement indicator for the vth file. l It represents the energy consumed by the lth computing user within the channel coherence time.

[0139] Step 3: According to the optimization problem, minimize the total return cost of content delivery of the cache model to obtain the content placement indicator for each file; use the weighted minimum mean square error method to transform the objective function of the maximization problem composed of the computing model and the communication model, and use the block coordinate ascent method to divide the coupled variables into several blocks for alternate optimization to obtain the phase shift of the passive reflection unit of all smart reflection surfaces, the transmit beamforming vector and the receive beamforming vector at the FD BS, and the local computing frequency and transmission power of all computing users.

[0140] It is obvious that due to the logarithmic nature of the objective function, the coupling of the optimization variables, the radar signal-to-noise ratio constraint C2 and the unit modulus constraint C3, the formulated optimization problem is a highly non-convex optimization problem and is difficult to solve. In order to solve these difficulties, the present invention proposes to use weighted minimum mean square error (WMMSE), BCA (block coordinate ascent method), MM (maximization minimization), ADMM (multiplier substitution method) and SDR (semi-definite relaxation) methods to convert the problem into several tractable sub-problems and solve them iteratively.

[0141] In the present invention, the cache strategy is first optimized, and then the WMMSE method is proposed to solve the problem of maximizing the sum of communication throughput and total computing bits. Next, the coupled variables are divided into several blocks for alternate optimization using the BCA method, and the specific steps are given below.

[0142] In order to optimize the designed cache strategy, the optimization problem Re-write as:

[0143]

[0144] Optimization Problem For the set of content placement indicators e = [e 1 ,…e V ] T It is convex and can be solved directly by a convex optimization solver or KKT optimal conditions to ensure the global optimal solution and improve the solving efficiency.

[0145] After optimizing the optimal caching strategy, the optimization problem It can be reformulated as a problem of maximizing the sum of the communication throughput and the total computation bits for a given placement, rewriting the optimization problem as:

[0146]

[0147] To make it easier to solve the optimization problem First, the weighted minimum mean square error (WMMSE) method is used to transform the objective function. Specifically, by introducing the auxiliary variable {α 1,k},{α 2,l} and {β 1,k},{β 2,l}, the original objective function is equivalently written into two variant forms as follows:

[0148]

[0149] in, represents a real number, [·] * represents conjugation, and is an intermediate variable.

[0150] Therefore, the original optimization problem It can be equivalently expressed as:

[0151]

[0152] Among them, the auxiliary variable α={α 1,k ,α 2,l},β={β 1,k ,β 2,l}. Next, we use the block coordinate ascent method (BCA) to solve the problem

[0153] Based on the above WMMSE transformation, when other optimization variables are fixed, the optimization of auxiliary variables is an unconstrained convex problem. By taking the derivative of the objective function, we can get the optimal solution of the auxiliary variable:

[0154]

[0155] Based on other given variables, the optimization of the phase shift φ of the IRS is studied, which is transformed into Re-expressed as:

[0156]

[0157] By introducing some new coefficients, the objective function and constraint C2 are rewritten as:

[0158]

[0159] Some of these coefficients are defined as follows:

[0160]

[0161] According to equations (23) and (24), the problem Rewritten as:

[0162]

[0163] Solving the problem using ADMM Since constraint C3 is a non-convex constant amplitude constraint, a replica variable ψ of φ is introduced to align and decouple. It can be equivalently written as:

[0164]

[0165] Where ψ=[ψ 1 ,…,ψ M ] T To solve the problem Based on ADMM optimization, the enhanced Lagrangian problem is solved. Re-write as:

[0166]

[0167] in, represents the square of the L2 norm, ρ>0 represents the penalty coefficient, and λ represents the dual variable. The problem is solved by updating each variable alternately.

[0168] When the replica variable ψ and the dual variable λ are fixed, the problem Simplified to:

[0169]

[0170] Due to the constraints question is non-convex. Using the MM framework, by 0 Convex constraints by linearizing point pairs and given as:

[0171]

[0172] Among them, φ 0 Represents the solution in the current iteration.

[0173] Next, replace the constraints φ H T 0 φ, Problem can be rewritten as:

[0174]

[0175] in,(·) * Indicates conjugation.

[0176] question It is a typical second-order cone program (SOCP) and can be directly solved using existing convex optimization solvers, such as CVX, to obtain the optimal solution for phase shift. And improve the solving efficiency.

[0177] When the phase shift φ and the dual variable λ are fixed, the problem Simplified to:

[0178]

[0179] Since the replica variable ψ has a unit modulus term, the quadratic term in the objective function with respect to the replica variable ψ is a constant, i.e. So the question can be simplified to:

[0180]

[0181] It can be seen that when all elements of the replica variable ψ are consistent with the linear coefficient (ρ -1 When the elements of φ+λ) are aligned, the problem can be solved The maximization of , that is, its optimal solution is expressed as:

[0182]

[0183] After obtaining the phase shift φ and the replica variable ψ, the dual variable λ will be updated by gradient ascent:

[0184] λ (i+1) =λ (i) +ρ -1 (φ-ψ) (34)

[0185] Among them, λ (i) , (i+1) denote the i-th iteration and the i+1-th iteration respectively.

[0186] Next, we study the beamforming vector w of FD BS when other variables are fixed: k Optimization, the problem Re-expressed as:

[0187]

[0188] The objective function and constraint C2 are rewritten by introducing some new coefficients as follows:

[0189]

[0190] Here Tr(·) represents the trace operation, and some coefficients are defined as follows:

[0191]

[0192] Based on the above transformation, the problem Re-expressed as:

[0193]

[0194] Due to the rank-one constraint of constraint C9, the problem It is still non-convex, so we choose to use SDR technology to obtain the optimal solution of the problem by relaxing the constraint condition C9. At the same time, it speeds up the solution and ensures a certain feasibility. Finally, the solution is directly solved by the convex optimization solver, and the corresponding rank-one solution is restored using the Gaussian randomization method.

[0195] Next, we study the receive beamforming vector u when other variables are fixed l Optimization. Re-expressed as:

[0196]

[0197] Rewrite the objective function as:

[0198]

[0199] Some of these coefficients are defined as follows:

[0200]

[0201] Based on the above transformation, the problem Re-expression:

[0202]

[0203] question It can be decomposed into L independent sub-problems, each of which can be written as:

[0204]

[0205] Notice the problem It is a typical unconstrained convex quadratic problem. By setting the derivative of its objective function to zero, its optimal solution can be obtained:

[0206]

[0207] When optimizing the transmission power and local computing resources of computing users, consider optimizing the transmission power and local computing resources of all computing users given other variables. Re-expressed as

[0208]

[0209] Then rewrite the objective function and constraint C2 as:

[0210]

[0211] Some of these coefficients are defined as follows:

[0212]

[0213] question Rewritten as:

[0214]

[0215] question Relative to the local calculation frequency and the transmission power,f,p,is concave, so this is a convex optimization problem, which can directly use the existing solver optimization, such as CVX, to improve the solution efficiency.

[0216] Solving the problem The overall algorithm is summarized in Algorithm 1. After given appropriate initial values, each subproblem is iteratively updated until convergence. It should be noted that the penalty coefficient ρ is reduced at each iteration to enforce the equality constraint. As the penalty parameter is gradually reduced, i.e., ρ→0, the problem The solution to ultimately guarantees the satisfaction of the unit modulus constraint. For any fixed penalty coefficient ρ, the problem The target value achieved is the problem The upper bound of the target value achieved by solving the problem alternately The upper bound can be gradually tightened. Each sub-problem of gets the optimal solution, the objective function is monotonically non-increasing, and the solution obtained by substitution optimization can be guaranteed to converge to the problem In addition, we can see that the original problem is solved by alternating updates of φ,w,u,{f,p}, and each subproblem converges to a stationary point. Note that the objective value is non-decreasing over the iterations, and any limit point of any optimization variable is the original optimization problem In addition, due to the transmit power budget and energy consumption budget, the upper bound of the target value is limited. Therefore, Algorithm 1 can converge to a stable point and a local optimal solution.

[0217]

[0218] N max Indicates the maximum number of iterations.

[0219] In the present invention, simulation results of the IRS-assisted FD ISCC system are given to verify the performance of the proposed system. The BS TX-IRS RE link, BS RX-IRS RE link, IRS RE CP-UEs link, IRS SE CP-UEs link, and IRSRE CM-UEs link are modeled as Rice fading with a Rice factor of 3 dB. The direct link between CP-UEs and CM-UEs is modeled as Rayleigh fading. For all channels, a typical distance-dependent path loss model is used. The large-scale fading model is expressed as:

[0220]

[0221] Where Λ=-30dB represents the reference distance d 0 =1m path loss, d and η represent the propagation distance and path loss index respectively. The path loss indexes between BS-IRS, IRS-UE, IRS-Target, and CP UE-CM UE are η BR =2.2,η RU =2.5,η RT =2.2,η MP =3.9. For the residual SI channel at the FD BS, the model Each entrance of in represents the residual SI power, d t,r >0 indicates the distance between the tth transmitting antenna and the rth receiving antenna. For simplicity, we set And let For all transceiver antenna pairs (t, r) is a unit modulus variable with random phase.

[0222] Some specific system parameters are shown in Table 1. In addition, it is assumed that the BS and IRS are deployed at (-50m, 0) and (0, 6m) respectively. All users are randomly distributed at (d x m,d y m), where d x ∈[10,40], d y ∈[0,1]. Assume that the radar target is 3m away from the IRS and θ=40°. The noise power at the BS, each CM-UE, and the IRS is set to

[0223] Table 1 System parameters

[0224]

[0225]

[0226] In order to verify the performance of the proposed system, the following schemes are also included for comparison:

[0227] Full offloading: Each CP-UE performs computation offloading only through the IRS.

[0228] Fixed phase shift: The IRS is assumed to use a fixed phase shift, which is determined by the maximum channel gain.

[0229] Half-duplex: Assume that the BS uses HD mode, where CM-UEs reception and CP-UEs transmission are implemented in two orthogonal time slots of equal duration. Therefore, the co-channel interference of uplink offload signals to CM-UEs and SI at the BS do not exist.

[0230] Optimizing caching: In each implementation, the caching strategy is solved by solving an optimization problem Make optimizations.

[0231] · Random caching strategy: The probability of BS local random caching depends on the popularity of cached data, that is, if the data is more popular, it is more likely to be cached.

[0232] No cache: Both computation offloading and local computation can be performed without cache.

[0233] Figure 2 The convergence of the proposed algorithm is shown, and the relationship between the achievable sum digits and the number of iterations under different settings is proposed to verify the convergence of Algorithm 1. It is observed that after 30 iterations, the achievable sum digits have a stable trend, which verifies the fast convergence and effectiveness of the algorithm of the present invention. In addition, it can be seen that as the number of users increases, a higher sum digit can usually be obtained.

[0234] Figure 3 The relationship between the total number of bits of the proposed scheme and the benchmark scheme and the number of IRS reflection elements N is described. It can be seen that by increasing the number of IRS reflection elements, the total number of bits of all schemes increases. This is because more reflection elements can customize favorable channels with more degrees of freedom, thereby providing greater passive beamforming gain for all users. In addition, the algorithm proposed in the present invention outperforms full unloading and fixed phase shift in both FD and HD systems. This highlights the optimality of partial unloading and that a more favorable propagation environment can be created by optimizing the phase shift matrix of the IRS. In addition, the performance of the FD scheme, that is, the proposed scheme, is significantly better than that of the traditional HD, which can be explained by the fact that in the same time, HD can make full use of the degrees of freedom introduced by the multiplexing of uplink and downlink users on the same spectrum resources, thereby improving system performance.

[0235] Figure 4 Describes the total number of bits of the proposed scheme and the benchmark scheme and the BS maximum transmission power P BS As expected, the total number of bits for all schemes increases with the maximum transmission power P BS The maximum transmission power P increases monotonically with the increase of BS It provides higher beamforming gain for signal transmission, thus improving the system performance. In addition, it is observed that the proposed scheme outperforms all other schemes. In fact, the proposed resource allocation scheme is able to be significantly improved compared with other schemes by jointly optimizing φ,w,u,f,p. In contrast, for full offloading, due to the influence of co-channel interference, as the maximum transmission power P BS With the increase of , both co-channel interference and residual SI power become more serious, which affects the system performance. For the FD scheme, although orthogonal transmission avoids co-channel interference and residual SI power, the resulting strictly suboptimal use of uplink and downlink time resources leads to a degradation of system performance.

[0236] Figure 5 The sum of the digits of the proposed solution and the baseline solution and the target perception requirement Γ are shown tar It can be observed that the total number of digits of all solutions decreases with Γ tar This is because when Γ tar When it increases, the system needs to reduce the communication rate to utilize more transmission power to guarantee the sensing demand, resulting in a decrease in system performance. In addition, it can be seen that the system performance improves as the number of BS antennas increases. This is due to the additional degrees of freedom provided by the additional antennas, which provide higher beamforming resolution for CM-UE reception and CP-UE transmission, resulting in higher system performance.

[0237] Figure 6 shows the system utility function versus the backhaul data rate R 0,l It can be observed that as the return data rate R 0,l As the rate of return increases, the system utility functions of all schemes decrease, which can be explained by the fact that the increase in the backhaul data rate will increase the total cost of content delivery, thus reducing the system utility function. In addition, the optimal caching scheme outperforms the random caching and no-caching schemes in both FD and HD systems.

[0238] Figure 7 The relationship between the system utility function and the skewness factor ∈ in the Zipf distribution function is shown. Figure 7As can be seen from the figure, as ∈ increases, the system utility functions of all schemes first gradually show a linear increase trend. When ∈ is large, the growth of the optimal cache gradually slows down. In addition, the optimal cache scheme is better than the random cache and no cache schemes in both FD and HD systems. And the no cache scheme is not affected by ∈ and remains unchanged.

[0239] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for integrated resource management of intelligent metasurface-assisted perception, communication and computing, characterized in that: The steps are as follows: Step 1: Deploy multiple passive reflection units and active sensing elements on the IRS, use a full-duplex base station for uplink and downlink signal transmission, and connect a local cache to the full-duplex base station through a backhaul link to build a full-duplex integrated sensing communication computing integrated system assisted by an intelligent reflection surface; Step 2: Determine the signal model, communication model, perception model, computing model and cache model according to the full-duplex integrated perception communication computing integrated system assisted by the reflective surface, and establish an optimization problem for maximizing the system utility according to the phase shift of the passive reflection unit, the transmit beamforming vector and the receive beamforming vector at the full-duplex base station, the local computing frequency and transmission power of all computing users, as well as the communication throughput, the total computing bits and the total backhaul cost of content delivery; Step 3: According to the optimization problem, optimize the cache strategy to minimize the total backhaul cost of content delivery of the cache model and obtain the content placement indicator of each file; The weighted minimum mean square error method is used to transform the objective function of the maximization problem composed of the computational model and the communication model. The coupled variables are divided into several blocks for alternate optimization using the block coordinate ascent method. The phase shifts of the passive reflection units of all smart reflection surfaces, the transmit beamforming vectors and receive beamforming vectors at the full-duplex base station, and the local computational frequency and transmission power of all computational users are obtained.

2. The intelligent metasurface-assisted perception, communication and computing integrated resource management method according to claim 1 is characterized in that: The intelligent reflective surface-assisted full-duplex integrated perception communication computing integrated system also includes K randomly distributed communication users, L randomly distributed computing users and a perception target with a radar; the full-duplex base station is equipped with two uniform linear arrays, and the IRS is equipped with a uniform linear array composed of M passive reflective elements and M a An active sensing element for receiving target echo signals. The full-duplex base station is equipped with N t transmit antennas and N r The full-duplex base station simultaneously receives uplink offload signals and transmits downlink ISAC signals for multi-user communication and target perception with the help of IRS. The full-duplex base station is connected to a local cache with limited storage capacity and connected to the MEC server through a high-capacity backhaul link, which is used for edge computing to store user data and perform computing tasks.

3. The intelligent metasurface-assisted perception, communication and computing integrated resource management method according to claim 1 or 2, characterized in that: The objective function of the optimization problem of maximizing the system utility is to maximize the difference between the sum of the communication throughput and the total computational bits and the total backhaul cost of content delivery.

4. The intelligent metasurface-assisted perception, communication and computing integrated resource management method according to claim 3 is characterized in that: The optimization problem is: C2:r tar ≥Γ tar , in, represents the achievable communication rate from the full-duplex base station to the kth communication user, represents the achievable offloading rate of the lth computing user at the full-duplex base station, represents the computing rate of the lth computing user, is the total return cost, r tar represents the radar signal-to-interference-noise ratio of the target, Γ tar represents the predefined target perception performance threshold, represents the maximum energy budget that can be consumed by the lth computing user; and They represent the transmit beamforming vector w of the kth communication user at the full-duplex base station. k and the lth calculated user receive beamforming vector u l K and L represent the total number of communication users and computing users, respectively, f=[f1,…,f L ] H represents the local computation frequency vector of the computing user, p = [p1,…,p L ] H represents the calculation of the user's transmission power vector, φ=[φ1,…,φ M ] H represents the set of phase shifts of all passive reflection units; e represents the content placement indicator e v The set composed of BS represents the maximum transmission power at the full-duplex base station, T is the channel coherence time, φ m represents the phase shift of the mth passive reflection unit, represents the index set of passive reflective units, ζ represents a constant related to the hardware structure of the computing user, and p l is the transmission power of the lth calculated user, f l represents the local computing frequency of the lth computing user, q v Indicates the length of the vth file, e v Indicates the content placement indicator of the vth file, Tp l It represents the energy consumed by the lth computing user within the channel coherence time, and F is the cache capacity.

5. The intelligent metasurface-assisted perception, communication and computing integrated resource management method according to claim 4 is characterized in that: The method for optimizing the cache strategy is: within a channel coherence time T, the total return cost is The optimization problem Re-write as: Among them, ρ v represents the pricing of the backhaul cost of content delivery of the vth file; V represents the total number of files, e v ∈[0,1] represents the content placement indicator, To calculate the probability representation of the vth file requested by the user, R 0,l Backhaul data rate for content acquisition; optimization issues For content placement indicators v The set composed of e=[e1,…e V ] T It is convex, directly through the convex optimization The solution is obtained by using the optimization solver or KKT optimal conditions.

6. The intelligent metasurface-assisted perception, communication and computing integrated resource management method according to claim 4 or 5, characterized in that: The method of transforming the objective function of the maximization problem composed of the calculation model and the communication model by using the weighted minimum mean square error method is: Reformulating it as a problem of maximizing the sum of the communication throughput and the total computation bits for a given placement, the optimization problem is rewritten as: stC1,C2,C3,C4, By introducing the auxiliary variable α = {α 1,k ,α 2,l },β={β 1,k ,β 2,l }, the optimization problem The objective function of is equivalently written in two variant forms as follows: in, represents a real number, [·] * represents conjugation, and is the intermediate variable; α 1,k ,α 2,l , β 1,k ,β 2,l All are auxiliary variables; is the signal-to-interference-noise ratio of the kth communication user, Receive the offloading signal y for the lth computing user BS The signal-to-noise ratio, H represents the transpose of the matrix, and the intermediate variable Intermediate variables h PU,k and G t denote the channel vectors between the kth communication user and the IRS and between the transmitting antenna of the full-duplex base station and the IRS respectively; e l,k and g PU,l denote the channel vectors between the kth communication user and the lth computing user and between the lth computing user and the IRS, respectively. represents the phase shift matrix of IRS, φ m ∈[0,2π) represents the phase shift of the mth passive reflection unit, represents the noise power of the additive white Gaussian noise at the receiver of the kth communication user; the intermediate variable g AU,l represents the channel vector between the lth computing user and the active sensing element of the IRS, G r represents the channel vector between the receiving antenna of the full-duplex base station and the IRS, represents the noise power of the additive white Gaussian noise at the full-duplex base station receiver, H SI represents the residual self-interference channel matrix at the full-duplex base station; Optimization Problem Equivalently expressed as: stC1,C2,C3,C4, in, They are all intermediate variables, and B represents the channel bandwidth.

7. The intelligent metasurface-assisted perception, communication and computing integrated resource management method according to claim 6 is characterized in that: The method of using the block coordinate ascent method to divide the coupled variables into several blocks for alternate optimization is: When other optimization variables are fixed, the optimization of auxiliary variables is an unconstrained convex problem. Derivative the objective function to obtain the optimal solution of the auxiliary variables; Solving for the set of IRS phase shifts φ based on other given variables will result in the optimization problem It is expressed as: s.t.C2:r tar ≥Γ tar , By introducing some intermediate variables, the objective function and constraint C2 are rewritten as: Among them, some intermediate variables are expressed as: in, and M represents the receiving steering vector of the active sensing element of IRS and the transmitting steering vector of the passive reflecting element of IRS respectively; a is the total number of active sensing elements; θ represents the angle of the sensing target relative to the IRS; η RT Represents the reflection coefficient of the perceived target; The optimization problem Rewritten as: Solving optimization problems using ADMM Since the constraint C3 is a non-convex constant amplitude constraint, a copy variable ψ of the phase shift set φ is introduced to align the decoupling, and the optimization problem The equivalent is: Among them, the replica variable ψ=[ψ1,…,ψ M ] T ; Based on ADMM optimization of the enhanced Lagrangian problem, the optimization problem Re-write as: in, represents the square of the L2 norm, ρ>0 represents the penalty coefficient, and λ represents the dual variable; the optimization problem is solved by alternately updating each variable When the replica variable ψ and the dual variable λ are fixed, the optimization problem Simplified to: Due to the constraints Optimization Problem It is non-convex. We use the MM framework to convexify the constraints by linearizing the convex terms at φ0. and given as: Among them, φ0 represents the solution in the current iteration; Replace constraints φ H T0φ, optimization problem Re-write as: in,(·) * denotes conjugation; Optimization Problem It is a typical second-order cone procedure, which is directly solved using the existing convex optimization solver to obtain the optimal solution for phase shift; When the phase shift φ and the dual variable λ are fixed, the optimization problem Simplified to: Since the replica variable ψ has a unit modulus term, the quadratic term in the objective function with respect to the replica variable ψ is a constant, i.e. Optimization Problem Simplified to: When all elements of the replica variable ψ are consistent with the linear coefficient (ρ -1 When the elements of φ+λ) are aligned, the optimization problem is realized The optimal solution of the replica variable is expressed as: After obtaining the phase shift φ and the replica variable ψ, the dual variable λ is updated by gradient ascent: l (i+1) =λ (i) +r -1 (f-ψ); Among them, λ (i) , (i+1) denote the dual variables obtained at the i-th iteration and the i+1-th iteration respectively.

8. The intelligent metasurface-assisted perception, communication and computing integrated resource management method according to claim 7 is characterized in that: The beamforming vector w of the full-duplex base station when other variables are fixed k The optimization method is: Re-expressed as: C2:r tar ≥Γ tar By introducing some intermediate variables, the objective function and constraint C2 are rewritten as: Among them, Tr(·) represents the trace operation, and the intermediate variable is: Optimization Problem Re-expressed as: Due to the rank-one constraint of constraint C9, the optimization problem It is non-convex, and the optimal solution of the problem is obtained by relaxing the constraint C9 using SDR technology. Solve it directly through the convex optimization solver, and use the Gaussian randomization method to recover the corresponding rank-one solution to obtain the beamforming vector 9. The intelligent metasurface-assisted perception, communication and computing integrated resource management method according to claim 7 is characterized in that: When other variables are fixed, the receive beamforming vector u l The optimization is: Re-expressed as: Rewrite the objective function as: Among them, the intermediate variables are: Based on the above transformation, the optimization problem Re-expressed as: Optimization Problem Decompose into L independent sub-problems, each sub-problem is: Optimization Problem It is a typical unconstrained convex quadratic problem. By setting the derivative of the objective function to zero, the optimal solution of the receive beamforming vector is obtained:

10. The intelligent metasurface-assisted perception, communication and computing integrated resource management method according to any one of claims 7 to 9, characterized in that: Considering that other variables are given, the method to optimize the transmission power and local computing resources of all computing users is: Re-expressed as The objective function and constraint C2 are rewritten as: Among them, the intermediate variables are: Optimization Problem Rewritten as: Optimization Problem Relative to the local computing frequency and transmission power,f,p,is concave, so it is a convex optimization problem. The existing solver is directly used to optimize the transmission power and local computing frequency of each computing user.

Images