Deterministic transmission method in reconfigurable intelligent surface-assisted network

Abstract

The invention discloses a deterministic transmission method in a reconfigurable intelligent surface (RIS)-assisted network, and the method comprises the steps: constructing an RIS-assisted multi-base-station NOMA communication system for an imperfect channel scene under a complex communication condition. Through an RIS auxiliary transmission channel model, a queue model and an energy consumption model, a resource allocation optimization problem taking minimization of system energy loss as a target function is constructed. In order to solve the long-term random optimization problem, a Lyapunov optimization framework is introduced and converted into a deterministic optimization problem in a decomposable single time slot. Aiming at a single time slot non-convex optimization problem, the single time slot non-convex optimization problem is split into optimization sub-problems about a user association relationship, a base station power allocation variable, an RIS reflection matrix and a base station position, and a convex optimization method and an alternating iterative optimization algorithm are adopted to solve the sub-problems.

Description

Technical Field

[0001] The present invention belongs to the technical field of wireless communication, and particularly relates to a deterministic transmission method in a reconfigurable intelligent surface assisted network. Background Art

[0002] With the rapid development of mobile communication technology and the popularization of intelligent terminals, users' demands for high-bandwidth and low-latency applications such as high-quality video, virtual reality (VR), and augmented reality (AR) are increasing day by day. However, in traditional cellular networks, especially in high-density population scenarios, due to network congestion and uneven resource allocation, the quality of user experience is often difficult to guarantee. For example, during large-scale events or peak hours, the video buffering time of users may increase by 30%-50%, and problems such as video interruption or image quality degradation may even occur. To solve the above problems, by using the dynamic management mechanism of the queue buffer pool, deterministic transmission can be effectively guaranteed, and at the same time, the reconfigurable intelligent surface (RIS) technology is introduced to effectively improve the network performance. The cooperation of the two can not only improve the transmission efficiency of the system, but also further improve the quality of user experience.

[0003] In a communication network, by introducing non-orthogonal multiple access (NOMA) technology and RIS, the capacity and energy efficiency of the network system can be significantly enhanced. The NOMA technology allows multiple users to share the same spectrum resource through power domain multiplexing, thus greatly improving the spectrum utilization rate; while the RIS optimizes the wireless channel environment by dynamically adjusting the phase and amplitude of the reflection unit, thereby improving the signal transmission quality. Research shows that in a communication network, by introducing RIS and NOMA technologies, the throughput of the communication system can be greatly improved, and the energy efficiency is also significantly improved.

[0004] In high-dynamic traffic demand scenarios, in order to ensure the smooth change of the user's channel quality and avoid fluctuations in video quality, deterministic transmission has become one of the key technologies. Deterministic transmission dynamically maintains the transmission queue and monitors the data backlog in real time, and optimizes the data scheduling strategy by combining the network state and user requirements, so as to ensure the timeliness and reliability of data transmission. Specifically, deterministic transmission ensures the reliable transmission of data within a specified time through precise resource allocation and traffic scheduling, meeting the requirements of low latency and high reliability. This mechanism can not only effectively alleviate network congestion, but also significantly improve the quality of user experience.

[0005] In the prior art, corresponding research has been carried out on the transmission problems in multi-user NOMA communication networks. However, these studies mainly focus on static or low-dynamic traffic scenarios, failing to fully consider the signal loss problems brought about by long-distance transmission, as well as the problems of degraded quality of user experience caused by data backlog and transmission delay. Moreover, existing studies usually assume that the base station is deployed at a fixed position, resulting in a constant data transmission rate, which is difficult to meet the requirements of high-dynamic traffic scenarios. During peak hours, data backlog and transmission delay increase significantly, seriously affecting the quality of user experience. Summary of the Invention

[0006] The object of the present invention is to provide a deterministic transmission method in a reconfigurable intelligent surface-assisted network, which can dynamically adjust the position of the base station and provide flexible communication services according to user needs, aiming to solve the problems of insufficient queue stability, large fluctuations in channel quality, and uneven resource allocation faced by the prior art in high-dynamic traffic scenarios.

[0007] The present invention provides a deterministic transmission method in a reconfigurable intelligent surface-assisted network, including:

[0008] Step 1: Define a RIS-assisted multi-user NOMA network communication system;

[0009] Step 2: Establish a channel model, a data transmission queue model, and an energy consumption model for RIS-assisted transmission;

[0010] Step 3: Establish a system resource allocation and transmission queue optimization problem model based on the models in Step 2, and decompose the long-term optimization problem into a deterministic optimization problem for a single time slot through the Lyapunov optimization framework;

[0011] Step 4: Split the deterministic optimization problem for a single time slot into optimization sub-problems regarding user association relationship, base station power allocation variables, RIS reflection matrix, and base station position, and use an alternating iterative optimization algorithm to solve the four types of sub-problems, and output the optimal user association relationship, base station power allocation, RIS reflection matrix, and base station position.

[0012] A deterministic transmission method in a reconfigurable intelligent surface-assisted network of the present invention constructs a transmission channel model, an energy consumption model, and a transmission queue model. Based on optimization variables such as user association relationship, base station power allocation, RIS reflection matrix, and base station location, an optimization problem model with the goal of minimizing long-term energy consumption is established. The optimization problem involves long-term goals and short-term constraints, and directly solving it has a high computational complexity. Therefore, the Lyapunov optimization framework is introduced to decompose the long-term optimization problem into a series of single-slot deterministic optimization problems. By constructing a Lyapunov function, the long-term queue stability and delay constraints are transformed into single-slot optimization goals, thus significantly reducing the complexity of problem solving. On this basis, the resource allocation problem within a single slot can be modeled as a mixed integer non-linear programming problem, which belongs to the NP-hard problem. To solve the optimization problem, the original problem is decomposed into four sub-problems: user association relationship optimization, base station power allocation optimization, RIS reflection coefficient optimization, and base station location optimization, and iterative solutions are carried out under the alternating optimization framework. Specifically, by introducing a queue management mechanism, a resource allocation optimization framework for deterministic transmission is proposed, which further optimizes the system energy efficiency while meeting the quality of service requirements; and using Lyapunov optimization, the long-term optimization problem is transformed and solved; the highly coupled optimization problem is split by the alternating optimization method to obtain the global optimal solution; for the optimization sub-problems, the integer relaxation method is used to convert the discrete variables of the problem into continuous variables and combined with fractional programming to solve the user association relationship matrix; fractional programming and quadratic substitution are used to solve the base station power allocation; the semi-definite relaxation and successive convex approximation (SCA) algorithms are used to transform the non-convex objective function and constraints into convex functions and convex constraints to solve the RIS reflection coefficient matrix; the SCA algorithm is used to solve the base station location. Description of the Drawings

[0013] Figure 1 It is a flowchart of a deterministic transmission method in a reconfigurable intelligent surface-assisted network of the present invention;

[0014] Figure 2 It is a model diagram of an RIS-assisted multi-user NOMA network communication system;

[0015] Figure 3 It is a curve graph of the system energy consumption during the algorithm iteration process in an embodiment of the present invention;

[0016] Figure 4a It is the deployment location of the base station at time slot t = 3 in an embodiment of the present invention;

[0017] Figure 4b It is the deployment location of the base station at time slot t = 7 in an embodiment of the present invention;

[0018] Figure 5It is a curve graph of the backlog of the user transmission queue during the time slot change process in the embodiment of the present invention. Detailed implementation manners

[0019] As Figure 1 shown, a deterministic transmission method in a reconfigurable intelligent surface assisted network of the present invention includes:

[0020] Step 1: Define a RIS-assisted multi-user NOMA network communication system.

[0021] What the present invention studies is a network model for deterministic transmission in a RIS-assisted multi-user NOMA network. On the premise of ensuring system resource allocation constraints, queue stability constraints, and queue long-term average delay constraints, the long-term average energy consumption in the communication network is minimized by optimizing the base station location, the association relationship between the base station and users, the base station power allocation, and the RIS reflection matrix.

[0022] The system model is as Figure 2 shown. In this model, K mobile base stations use RIS to provide wireless communication services for N users. It is assumed that the users are in a static or low-speed moving state, and both the base stations and the users are equipped with single antennas. RIS optimizes the wireless channel environment by dynamically adjusting the phase and amplitude of the reflection units, thereby significantly improving the network performance.

[0023] In the considered network model, the base station set is defined as The user set is defined as It is assumed that the user set served by base station k is The location of user n is expressed as l n =[x n , y n , 0] T , and the location of RIS is expressed as l r =[x r , y r , z r T . The moving time is divided into T time slots, and the time slot set is expressed as

[0024] At time slot t, the location of base station k is expressed as l k (t)=[x k (t), y k (t), H] T . Each user can only access a single base station within any time slot. It is assumed that RIS consists of Q reflection elements, and the reflection coefficient of the q-th element is expressed as l q ∈[0, 1] and θ q ∈[0, 2π) respectively represent the reflection amplitude and phase shift of the q-th reflection element.​

[0025] Set The diagonal reflection matrix is expressed as

[0026] Step 2: Establish the channel model, data transmission queue model, and energy consumption model for RIS-assisted transmission, specifically as follows:

[0027] Step 2.1: Establish the channel model for RIS-assisted transmission. Assume that the direct link channel from the base station to the user is a LoS channel. Then, the channel gain from base station k to user n is expressed as:

[0028]

[0029] where represents the channel gain of the direct link from base station k to user n; ρ0 is the path loss at a reference distance of 1 meter, l k (t) represents the location of base station k; l n represents the location of user n; c1 represents the path loss exponent of the direct link from the base station to the user; represents the LoS component.

[0030] For the communication link from the base station to the RIS and the communication link from the RIS to the user, assume it is a Rice fading channel, consisting of a LoS component and a non-LoS component. The channel gain corresponding to the transmission link is expressed as:

[0031]

[0032]

[0033] where represents the channel gain from base station k to the RIS, represents the channel gain from the RIS to user n; c2 represents the path loss exponent of the link from the base station to the RIS, c3 represents the path loss exponent of the link from the RIS to the user; l r represents the location of the RIS; K1 and K2 are the Rice factors of the Rice fading channels in the corresponding links, respectively, used to characterize the relative strength of the LoS component and the non-NLoS component in the channel; and are the LoS components; and represent the non-LoS components, which are modeled as circularly symmetric complex Gaussian random variables with zero mean and unit variance to characterize the random scattering effect in the wireless channel.

[0034] Based on the above channel model, the comprehensive channel gain from base station k to user n is expressed as:

[0035]

[0036] Among them, Θ represents the diagonal reflection matrix of the RIS. Since the decoding order of the users is affected by the effective channel gain, it is assumed that the channel gains of the users served by base station k satisfy the following order: g k,1 (t) ≤ g k,2 (t) ≤ … ≤ g k,n (t). Through NOMA technology, for user n served by base station k, the signal-to-noise ratio after SIC is expressed as:

[0037]

[0038] Among them, p k,n (t) represents the transmit power allocated by base station k to user n; c k,n (t) is a binary variable representing the association relationship between base station k and user n. If base station k provides communication services for user n within time slot t, then c k,n (t) = 1; otherwise, c k,n (t) = 0. σ 2 represents the Gaussian white noise. and represent the intra-group interference and inter-group interference respectively:

[0039]

[0040]

[0041] Among them, K is the total number of base stations, and N is the total number of users.

[0042] Within time slot t, the total data transmission rate R n (t) that user n can achieve can be modeled as the sum of the transmission rates of all serving base stations:

[0043]

[0044] R k,n (t) = log2(1 + SINR k,n (t)) (5a)

[0045] Among them, R k,n (t) represents the data transmission rate that user n can achieve through base station k.

[0046] Step 2.2: Establish a data transmission queue model:

[0047] Considering the problem that large-scale video transmission of users is likely to cause network traffic congestion under the condition of limited system resources, the present invention constructs a data transmission queue model for each user at the transmitting end. This model effectively alleviates the impact on the system caused by instantaneous traffic peaks through a queue buffering mechanism, thereby ensuring that the system can provide long-term stable communication service quality guarantee for multiple users. During the process of the base station providing communication services for users, a transmission mechanism based on queue offloading of data is adopted. The data is not directly transmitted in real time by the base station, but is first offloaded to the transmission queue for caching. In each time slot, the serving base station transmits the cached data to the user at a steady rate according to the queue scheduling strategy, so as to achieve deterministic transmission and ensure service stability.

[0048] In time slot t, define the amount of data arrived by user n as A n (t), and define the amount of data in the dynamic data transmission queue at the transmitting end as B n (t). Then, in time slot t + 1, the size of the data in the data transmission queue depends on the following three factors: the data backlog B n (t) in the previous time slot, the actually randomly arrived amount of data, and the amount of data D n (t) actually transmitted to user n. Therefore, the amount of data in the queue of user n in time slot t + 1 can be expressed as:

[0049] B n (t + 1) = max{B n (t) - D n (t), 0} + A n (t) (6)

[0050] It should be noted that the actually transmitted amount of data D n (t) is jointly determined by the maximum transmission capacity of user n's node and the current queue backlog situation. Among them, the maximum transmission capacity refers to the transmission rate of the user node per unit time. If the backlog amount of data B n (t) in the current queue is less than the transmission capacity R n (t) of the queue, then the actually transmitted amount is equal to the backlog amount of data in the current queue; otherwise, the actually transmitted amount is equal to the amount of data corresponding to the maximum transmission capacity R n (t) of the queue. Therefore, the actually transmitted amount of user n in time slot t is expressed as:

[0051] D n (t) = min{B n (t), τR n (t)} (6a)

[0052] Among them, A n (t) represents the amount of data arrived by user n in time slot t, B n(t) represents the data volume of the dynamic data transmission queue of the sender user n in time slot t, D n (t) represents the data volume actually transmitted to user n, R n (t) is the total data transmission rate of user n, representing the transmission capacity of the queue; τ represents the length of a unit time slot.

[0053] According to Little's Law, in a long-term stable system, the long-term average time delay is proportional to the long-term average queue backlog and inversely proportional to the long-term average task arrival rate. Define the time delay of the transmission queue as τ n , which is expressed as:

[0054]

[0055] Among them, represents the long-term average data generation volume, and T represents the total number of time slots.

[0056] Step 2.3: Establish an energy consumption model. When the base station k moves horizontally within time slot t, the mobile power is modeled as:

[0057]

[0058] Among them, C1, C2, and C3 respectively represent the characteristic parameters related to the movement of the base station. Within time slot t, d k (t) = ||l k (t + 1) - l k (t)|| represents the horizontal movement distance of the base station k;

[0059] According to the energy calculation formula E = pt, the total energy consumption E tot (t) of the system within time slot t is expressed as:

[0060]

[0061]

[0062]

[0063] Among them, represents the communication energy consumption between the base station k and the user n within time slot t, represents the mobile energy consumption of the base station k within time slot t.

[0064] Step 3: Establish a system resource allocation and transmission queue optimization problem model according to the model in Step 2. Through the Lyapunov optimization framework, decompose the long-term optimization problem into a deterministic optimization problem for a single time slot, specifically:

[0065] Step 3.1: Considering the RIS-assisted multi-user NOMA network communication system, by optimizing the position of the mobile base station the RIS reflection matrix Θ, the base station power allocation the user association relationship joint optimization of four optimization variables to minimize the system energy consumption. For the above analysis, a long-term dynamic optimization model is constructed. The following system resource allocation and transmission queue optimization problem model is established:

[0066]

[0067]

[0068]

[0069]

[0070]

[0071]

[0072]

[0073]

[0074] where E tot (t) represents the total system energy consumption in time slot t. The above optimization objective function is to minimize the long-term average energy consumption of the system, represents the maximum power of the base station. represents the set of base stations, represents the set of time slots, represents the set of users, θ q (t) represents the phase shift of the q-th reflecting element in time slot t; represents the maximum distance of the base station movement between adjacent time slots, represents the safety distance between any two base stations, represents the average data volume of the dynamic data transmission queue at the sending end.

[0075] Constraint (10a) represents the limit of the transmission power of the base station. Constraint (10b) represents that each user can only be associated with one base station at the same time. Constraint (10c) represents the phase shift constraint of the RIS element. Constraint (10d) represents the maximum movement distance limit of the base station in a single time slot. Constraint (10e) represents that the distance between any two base stations should not be less than the safety interval distance. Constraint (10f) represents the long-term delay constraint of the transmission queue of the users served by the base station. Constraint (10g) represents that the transmission queues of the users served by the base station need to satisfy strong stability.

[0076] Different from the traditional single-slot optimization problem, the objective function in the above optimization problem P1 comprehensively considers the association relationship optimization variable, the base station power allocation optimization variable, the reflection matrix optimization variable, and the base station location to minimize the long-term average energy consumption of the system. By analyzing the queue backlog expression and the queue delay expression, it can be found that there are significant inter-slot coupling characteristics in the time domain of the system. The research of this invention is a long-term optimization model of a stochastic network, and the optimization methods in deterministic networks cannot be directly used. Therefore, for the long-term optimization model, an algorithm based on Lyapunov optimization theory is proposed to decouple the control decisions made in consecutive slots while meeting the requirements of long-term stability and reliability. Specifically, the Lyapunov method can transform a multi-level stochastic optimization problem into multiple slot deterministic problems. By iteratively optimizing the resource allocation and location adjustment in each slot, a time-average solution can be obtained with lower complexity.

[0077] The key point of the Lyapunov algorithm is to transform the long-term optimization problem into a single-slot optimization problem. This method first defines the sum of the squares of all virtual queue and actual queue backlogs as a quadratic Lyapunov function. And the defined Lyapunov function can exactly measure the congestion degree of the network. The smaller the Lyapunov function is, the smaller the queue backlogs are and the smaller the network congestion degree is. On the contrary, the larger the Lyapunov function is, the larger the queue backlogs are and the larger the network congestion degree is. Define the difference between the Lyapunov functions in the previous and subsequent slots as the Lyapunov drift. If the Lyapunov drift is minimized through resource allocation optimization in each slot, the stability of the network queue can be guaranteed.

[0078] In the optimization problem P1, there is a long-term delay constraint on the transmission queue. In order to transform the problem into a deterministic optimization problem within a single slot, the long-term delay constraint needs to be processed and transformed first. The main processing idea is: by introducing virtual queues, the queue long-term delay constraint in the optimization problem P1 is transformed into a queue stability constraint, so that the problem can be transformed into the form of a Lyapunov optimization problem.

[0079] To meet the time delay constraint of the data transmission queue, virtual data transmission queues are defined based on Lyapunov optimization technology. Assume that the queue is set to zero in the initial slot, that is Therefore, the virtual queue can be updated according to the following expression:

[0080]

[0081] where represents the time delay budget for each slot.

[0082] Define the set of column vectors It includes the actual data transmission queue backlog of the user and the corresponding virtual queue backlog. Then, regarding the current virtual queue composite vector, at time slot t, define the corresponding Lyapunov function L(Z(t)) as a scalar of the queue backlog:

[0083]

[0084] Within time slot t, define the conditional Lyapunov drift ΔL(Z(t)) as:

[0085]

[0086] Considering that the optimization objective in the model is to maximize the long-term system throughput, therefore, to simultaneously achieve maximizing the system throughput and queue stability, define the Lyapunov drift-plus-penalty function, which can be expressed as:

[0087]

[0088] Among them, V is the key control parameter used to achieve a dynamic balance between queue stability and system throughput.

[0089] By minimizing the Lyapunov drift-plus-penalty function, the long-term optimization problem can be transformed into a single-time-slot optimization problem. However, the expression form of Δ V L(Z(t)) is very complex and cannot be directly solved using convex optimization methods. To transform it into a simple form, transform the problem of minimizing Δ V L(Z(t)) into the problem of minimizing a simple upper bound form of Δ V L(Z(t)). Then, organize the optimization objective function into a form regarding the optimization variables:

[0090] target = W(t) + f(t) + VE tot (t) = W(t) + f(C, P, Θ, l, t) (15)

[0091] Among them, W(t) belongs to the constant term and has no impact on the optimization result. Based on the above analysis, only the optimal solution or sub-optimal solution of the deterministic optimization problem needs to be solved within each time slot to obtain the optimal solution or sub-optimal solution within a continuous long time period or even an infinite long time.

[0092] Step 3.2: According to the Lyapunov optimization algorithm and the above derivation results, transform the long-term optimization problem into a deterministic optimization problem within a single time slot, and f(C, P, Θ, l, t) can be expressed as:

[0093]

[0094]

[0095]

[0096]

[0097] Among them, the constraint conditions also include formulas (10a)-(10e). denotes the minimum value of the total data transmission rate that user n can achieve within time slot t, denotes the maximum value of the total data transmission rate that user n can achieve within time slot t, denotes the maximum value of the data volume arrived by user n.

[0098] Step 4: Split the deterministic optimization problem of a single time slot in Step 3 into optimization sub-problems regarding user association relationship, base station power allocation variables, RIS reflection matrix, and base station location. Use the alternating iteration optimization algorithm to solve the four types of sub-problems, and output the optimal user association relationship, base station power allocation, RIS reflection matrix, and base station location. Specifically:

[0099] Step 4.1: Split the optimization problem of a single time slot into optimization sub-problems regarding user association relationship, base station power allocation variables, RIS reflection matrix, and base station location. Initialize the base station location, user location, channel gain, user association relationship, RIS reflection matrix, and the maximum number of iterations.

[0100] Step 4.2: Relax the binary user association relationship variables into continuous variables according to the integer relaxation method, construct a fractional programming problem, and solve the optimal user association relationship matrix. Specifically:

[0101] Step 4.2.1: Under the condition of a given feasible base station power allocation P, RIS reflection matrix Θ, and base station location l, simplify the original optimization problem P2 into an optimization problem P3 only regarding the base station-user association relationship. By fixing the known quantities and removing the irrelevant constant terms in the objective function, the original problem can be transformed into a more tractable optimization problem. Simplify the objective function f(C, P, Θ, l, t) into a function f c (C, t). After the above transformation and parameter term definition, the optimization sub-problem P3 can be re-expressed in the following form:

[0102]

[0103]

[0104]

[0105]

[0106]

[0107]

[0108] Step 4.2.2: The sub-problem P3 of user association relationship optimization belongs to a non-convex optimization problem and cannot be directly solved using traditional convex optimization methods. Therefore, it is necessary to process the constraints and the objective function. For the convenience of subsequent analysis and solution, the explicit representation of time slot t will be omitted in the subsequent expressions. Since c k,n is a binary variable, directly solving it using traditional optimization methods has a high complexity. To simplify the problem and facilitate the solution, the binary variable c k,n can be relaxed, changing it from a discrete binary variable to a continuous variable. The specific transformation form is: 0 ≤ c k,n ≤ 1.

[0109] Since the logarithmic function is a monotonically increasing function, the constraints (17d) and (17e) are approximated to obtain the constraints as:

[0110]

[0111]

[0112] In the optimization objective function, the structure of R k,n = log2(1 + SINR k,n ) makes the objective function have non-linear and non-convex characteristics. Especially the combination of the fractional structure of SINR k,n and the logarithmic function increases the complexity of the problem. To handle the fractional structure in the objective function, fractional programming (FP) is introduced. By introducing an auxiliary variable ξ and through a quadratic transformation, the original fractional programming problem can be transformed into a joint optimization problem about ξ and C. The function is introduced as:

[0113] L 1,k,n (C) = c k,n p k,n g k,n (17h)

[0114]

[0115] Using fractional programming and quadratic transformation techniques, the objective function P3 is simplified, and the ratio form in the problem is transformed into a subtraction form, obtaining the approximate optimization problem as:

[0116]

[0117]

[0118]

[0119] Among them, ξ k,n represents the introduced auxiliary variable. The constraint conditions also include formulas (17a) and (17b).

[0120] By utilizing the properties of quadratic transformation, P3 and P3.1 are equivalent in terms of the optimal objective and the optimal user association relationship C. Although the objective function itself is a non-convex function, by alternately optimizing the original variable C and the auxiliary variable ξ k,n , the problem P3.1 can be effectively solved. In addition, the sub-problems in the iterative process are all tractable problems, and it can be ensured that the sub-optimal algorithm based on fractional programming can converge to a stationary point.

[0121] Step 4.3: Use fractional programming to transform the objective function into a subtractive form, and transform the non-convex constraints into convex constraints through quadratic transformation. Finally, use the interior point method to solve the optimal base station power allocation scheme, specifically:

[0122] Based on the association between the base station and the user in Step 4.2, under the conditions of the given user association relationship C, the RIS reflection matrix Θ, and the base station location l, the base station power allocation variable P can be further optimized. By fixing the known quantities and removing the irrelevant constant terms in the objective function, the original problem can be transformed into a sub-problem regarding the base station power allocation variable P. Optimize and solve the sub-problem to determine the optimal base station power allocation scheme.

[0123] Simplify the objective function into a function f p (P, t), and set the set to represent the set of users associated with base station k at time slot t. The original optimization problem P2 is simplified into an optimization sub-problem P4 regarding the base station power allocation variable P:

[0124]

[0125]

[0126]

[0127]

[0128] Step 4.3.2: The base station power allocation variable optimization sub-problem P4 is a non-convex optimization problem. Process the constraints and the objective function. For simplicity of description, the explicit representation of time slot t will be omitted in the subsequent expressions, and the data transmission rate formula that can be achieved by user n through base station k is re-expressed:

[0129]

[0130] Since R k,n is a composite function of the base station power allocation variable P and contains a fractional expression of the base station power allocation variable P, in order to simplify the objective function, using the fractional programming theory and quadratic form transformation technology, the ratio form in the objective function is transformed into a subtraction form that is easier to handle. Introduce the function:

[0131] O 1,k,n (P) = p k,n g k,n (20a)

[0132]

[0133] Using the fractional programming theory and quadratic form transformation technology, the ratio form in the optimization sub-problem P4 is transformed into a subtraction form, and the optimization sub-problem P4 is equivalently transformed into:

[0134]

[0136]

[0137]

[0138]

[0139] The constraint conditions also include formula (19a). For the optimization sub-problem P4.1, by alternately optimizing the original variable P and the auxiliary variable λ k,n to solve, when P or λ k,n remains unchanged, the optimization problem P4.1 with respect to the variable λ k,n and P belongs to a convex optimization problem.

[0140] Step 4.4: Relax the RIS reflection coefficient matrix to a positive semi-definite matrix by semi-definite relaxation SDR, use the successive convex approximation (SCA) algorithm to gradually approximate the optimal solution, transform the non-convex objective function and constraints into the form of a convex optimization problem, and recover the feasible solution by the Gaussian randomization method, specifically:

[0141] Step 4.4.1: Given the base station and user association relationship C, the base station power allocation P, and the base station location l, transform the original problem into an optimization problem with respect to the RIS reflection matrix. During the optimization process of the RIS reflection matrix, the influence is limited to the channel gains of the base station to RIS and RIS to user cascaded links, and the channel gain of the direct link from the base station to the user will not be affected. Therefore, simplify the objective function to a function f with respect to the RIS reflection matrix Θ variable Θ$(Θ,t)$, after the above transformation and parameter item definition, the original optimization problem P2 is transformed into an optimization sub-problem P5 about the RIS reflection matrix:

[0142]

[0143]

[0144]

[0145]

[0146] Step 4.4.2: The optimization problem P5 of the RIS reflection matrix belongs to a non-convex optimization problem. To facilitate the processing of the RIS reflection matrix variable Θ, the channel gain expression needs to be rewritten. For simplicity of description, the explicit representation of time slot t will be omitted in the subsequent expressions. The combined channel gain from base station k to user n can be reformulated as:

[0147]

[0148]

[0149] where, represents the cascaded link between the base station - RIS - user without configured RIS; v = [r 1] T , V = vv H , After the transformation of the optimization variable matrix, V needs to satisfy the following constraint conditions: [V] qq = 1;

[0150] The achievable data transmission rate for user n through base station k is reformulated, and the expression is as follows:

[0151]

[0152] Therefore, the optimization problem P5 is transformed into:

[0153]

[0154] rank(V) = 1 (24c)

[0155] [V] qq = 1 (24d)

[0156] For the non-convex rank-one constraint (24c), it is equivalently transformed into the constraint ||V|| * - ||V||2 ≤ 0; ||V|| * = ∑ i δi (V) and ||V||2 = δ1(V) represent the nuclear norm and spectral norm of matrix V respectively, where δ i (V) is the i-th largest singular value of matrix δ i (V); for any we have ||V|| * - ||V||2 ≥ 0 and the equality holds if and only if V is a rank-one matrix; however, the constraint ||V|| * - ||V||2 ≤ 0 still belongs to non-convex constraints. To handle the constraint (24c), the rank-one constraint is added as a penalty term to the objective function, and the optimization problem can be reformulated as:

[0157]

[0158] (24a), (24b), (24d) (25a)

[0159] In the above optimization problem P5.2, is the penalty coefficient. For any given the objective function P5.2 is in the form of the difference of convex functions. The optimal solution is gradually approximated through SCA. This is a standard convex semi-definite programming (SDP) problem, which is solved by the classical convex toolbox to obtain the optimal solution V * , and then through eigenvalue decomposition, the optimal RIS reflection coefficient vector r * is extracted.

[0160] Step 4.5: Adopt the SCA algorithm to gradually transform the non-convex problem into a convex optimization problem for solution. The non-convex objective function and constraints are linearized through first-order Taylor expansion, and the optimal position of the base station is iteratively solved until convergence. Specifically:

[0161] Step 4.5.1: Given the base station and user association relationship C, base station power allocation P, and RIS reflection matrix Θ, the original optimization problem is transformed into a sub-optimization problem regarding the base station position. During the optimization process of the base station position, the moving range and starting point of the base station do not affect the optimization derivation. Therefore, the objective function is simplified to a function f l (l, t) of the base station position. After the above transformation and parameter definition, the original optimization problem P2 is simplified to an optimization sub-problem P6 only regarding the base station position l:

[0162]

[0163]

[0164]

[0165]

[0166]

[0167] Step 4.5.2: The optimization sub-problem P6 of the base station location \(l\) is a non-convex optimization problem. The achievable data transmission rate of user \(n\) through base station \(k\) is reformulated as follows:

[0168]

[0169] For simplicity of description, the explicit representation of time slot \(t\) will be omitted in the subsequent expressions. In dealing with the non-convex objective function, auxiliary variables \(d_u\) k,n , \(d_l\) j,n , \(d_u\) k,r , \(d_l\) k,r , where \(d_u\) k,n represents the upper bound of the distance between the base station and its served user, \(d_l\) j,n represents the lower bound of the distance between the base station and the unassociated user, \(d_u\) k,r represents the upper bound of the distance between the base station and the RIS, \(d_l\) k,r represents the lower bound of the distance between the base station and the RIS; the variable is introduced to reformulate the comprehensive channel gain expression from base station \(k\) to user \(n\) as:

[0170]

[0171] Let The lower bound auxiliary variable of \(g\) k,n is introduced The upper bound auxiliary variable of \(g\) j,n is introduced The auxiliary variable is introduced as the lower bound of the user transmission rate \(R\) k,n ;

[0172]

[0173]

[0174]

[0175] The optimization sub-problem P6 is equivalently transformed into:

[0176]

[0177]

[0178] Among them, the constraint conditions also include formulas (26c) and (26d). For the non-convex characteristics of the above optimization problem P6.1, Taylor expansion can be used for processing, and the sequential convex approximation algorithm is used for solution. This algorithm transforms the original non-convex optimization problem into a convex optimization problem and gradually approaches the global optimal solution through iterative convex optimization methods, thereby obtaining the optimal location l of the base station k .

[0179] Step 4.6: Alternately and iteratively optimize and solve the sub-optimization problems of user association relationship, base station power allocation variables, RIS reflection matrix, and base station location, and output the optimal user association relationship, base station power allocation, RIS reflection matrix, and base station location

[0180] Next, simulation experiments are carried out to verify the convergence and effectiveness of the method of the present invention, and further explore the influence of different parameters on the system energy consumption. Since the optimization problem of the system model has been decoupled into independent sub-problems, it is only necessary to verify the convergence of the alternating optimization algorithm

[0181] As Figure 3 shown, the experiment plots the change curves of the system energy consumption during the iterative process of the optimization algorithm when the number of base stations is 2 and 3 respectively, and the number of users is 6, 10, and 14 respectively. The experimental results show that under different configurations of the number of base stations and the number of users, the total system energy consumption based on the alternating optimization algorithm can quickly converge and reach a stable state. In a dynamic random network, the data transmission volume of users is jointly affected by various factors, including the backlog of user queues and delay constraints, etc. The proposed algorithm can adaptively adjust the data transmission volume, while ensuring the quality of service, minimizing the system energy consumption

[0182] As Figure 4a and 4b shown, the experiment plots the deployment locations of the base stations at time slots t = 3 and t = 7 when the number of base stations is 2 and the number of users is 14. In the experiment, the users randomly generate the transmission data volume in each time slot, and the base stations optimize their locations in real time according to the spatial distribution and task requirements of the users within their service ranges, so as to ensure the quality of communication services

[0183] As Figure 5As shown, the experiment plotted the variation of the backlog in the user data transmission queue with the growth of time slots within a 100-second time window, focusing on studying the stability characteristics of the queue in the dynamic random network scenario. In the experiment, the queue management adopted a dynamic resource allocation strategy based on Lyapunov optimization. In terms of the queue backlog, the system showed a typical dynamic change pattern. As the number of time slots increased, the length of the user queue first increased and then decreased to a stable state. This change trend was mainly caused by the distance. In the initial transient stage (0 - 10 seconds), due to the relatively far initial deployment position of the base station and poor channel quality, the queue backlog accumulated rapidly and reached a peak of 19.44 Mbit at 12 seconds. As the system entered the adjustment stage (10 - 100 seconds), under the control of the optimization algorithm, the base station moved towards the user, the channel conditions continued to improve, and the queue backlog data gradually decreased. Eventually, the queue backlog was stably maintained within the ideal range of 11.49 - 11.82 Mbit.

[0184] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, various changes and modifications can be made to the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A deterministic transmission method in a reconfigurable intelligent surface assisted network, characterized in that Including: Step 1: Define a RIS-assisted multi-user NOMA network communication system; Step 2: Establish a channel model, a data transmission queue model, and an energy consumption model for RIS-assisted transmission; Step 3: Based on the models in Step 2, establish a system resource allocation and transmission queue optimization problem model. Through the Lyapunov optimization framework, decompose the long-term optimization problem into a deterministic optimization problem for a single time slot; Step 4: Split the deterministic optimization problem for a single time slot into optimization sub-problems regarding user association relationships, base station power allocation variables, RIS reflection matrices, and base station positions. Use an alternating iterative optimization algorithm to solve the four types of sub-problems, and output the optimal user association relationships, base station power allocation, RIS reflection matrices, and base station positions.

2. The resource allocation optimization method for deterministic transmission in an RIS-assisted network according to claim 1, wherein In the RIS-assisted multi-user NOMA network communication system, K mobile base stations equipped with single antennas use a RIS with Q reflection elements to provide wireless communication services for N users equipped with single antennas.

3. The resource allocation optimization method for deterministic transmission in an RIS-assisted network according to claim 1, wherein The specific content of Step 2 is as follows: Step 2.1: Establish a channel model for RIS-assisted transmission. Assume that the direct link channel from the base station to the user is a LoS channel. Then, the channel gain from base station k to user n is expressed as: Among them, represents the channel gain of the direct link from base station k to user n; ρ0 is the path loss at a reference distance of 1 meter, and l k (t) represents the location of base station k; l n represents the location of user n; c1 represents the path loss exponent of the direct link from the base station to the user; represents the LoS component; For the communication link from the base station to the RIS and the communication link from the RIS to the user, assume they are Rice fading channels, consisting of a LoS component and a non-LoS component. The channel gains corresponding to the transmission links are expressed as: wherein, represents the channel gain from base station k to the RIS, represents the channel gain from the RIS to user n; c2 represents the path loss exponent of the link from the base station to the RIS, and c3 represents the path loss exponent of the link from the RIS to the user; l r represents the position of the RIS; K1 and K2 are the Rice factors of the Rice fading channels in the corresponding links, respectively, used to characterize the relative strength of the LoS component and the non-NLoS component in the channel; and are the LoS components; and represent the non-LoS components, which are modeled as circularly symmetric complex Gaussian random variables with zero mean and unit variance to characterize the random scattering effect in the wireless channel; Therefore, the comprehensive channel gain from base station k to user n is expressed as: where, Θ represents a diagonal reflection matrix; since the decoding order of users is affected by the effective channel gain, it is assumed that the channel gains of the users served by base station k satisfy the following order: g k,1 (t) ≤ g k,2 (t) ≤ … ≤ g k,n (t). Through NOMA technology, for user n served by base station k, the signal-to-noise ratio after SIC is expressed as: Among them, p k,n (t) represents the transmission power allocated by base station k to user n; c k,n (t) is a binary variable representing the association relationship between base station k and user n. If base station k provides communication services for user n within time slot t, then c k,n (t) = 1; otherwise c k,n (t) = 0; σ 2 represents additive white Gaussian noise; and represent intra-group interference and inter-group interference respectively; where K is the total number of base stations, and N is the total number of users; During time slot t, the total data transmission rate R n (t) that user n can achieve can be modeled as the sum of the transmission rates of all serving base stations: R k,n R(t) = log2(1 + SINR k,n (t)) (5a) where R k,n (t) represents the achievable data transmission rate of user n through base station k; Step 2.2: Establish a data transmission queue model. During the process of the base station providing communication services for users, a transmission mechanism based on queue offloading of data is adopted. First, the data is offloaded to the transmission queue for caching. In each time slot, the serving base station, according to the queue scheduling strategy, transmits the cached data to the user at a steady rate. The amount of queue data of user n in time slot t + 1 is expressed as: B n (t + 1)= max{B n (t)- D n (t), 0}+ A n (t) (6) D n D(t) = min{B n D(t), τR n D(t)} (6a) Among them, A n (t) represents the amount of data arrived by user n within time slot t, B n (t) represents the amount of data in the dynamic data transmission queue of the sending-end user n in time slot t, D n (t) represents the amount of data actually transmitted to user n, R n (t) is the total data transmission rate of user n, representing the transmission capacity of the queue; τ represents the length of a unit time slot; Define the time delay of the transmission queue corresponding to user n as τ n , denoted as: Among them, represents the long-term average generated data volume, and T represents the total number of time slots; Step 2.3: Establish an energy consumption model. When base station k moves horizontally in time slot t, the mobile power is modeled as: wherein, C1, C2, and C3 respectively represent characteristic parameters related to the movement of the base station; within time slot t, d k (t) = ||l k (t + 1) - l k (t)|| represents the horizontal movement distance of base station k; According to the energy calculation formula E = pt, the total energy consumption E of the system within time slot t tot (t) is expressed as: Among them, represents the communication energy consumption between base station k and user n within time slot t, represents the mobile energy consumption of base station k within time slot t.

4. The resource allocation optimization method for deterministic transmission in the RIS-assisted network according to claim 1, wherein, The specific content of Step 3 is as follows: Step 3.1: Establish the following system resource allocation and transmission queue optimization problem model: Among them, E tot (t) represents the total system energy consumption in time slot t; the objective function is to minimize the long-term average energy consumption of the system, and the optimization variables are the user association relationship The position of the movable base station Base station power allocation and the RIS reflection matrix Θ; represents the maximum power of the base station; represents the set of base stations, represents the set of time slots, represents the set of users, θ q (t) represents the phase shift of the q-th reflecting element at time slot t; represents the maximum distance of the base station movement between adjacent time slots, represents the safety distance between any two base stations, represents the average data volume of the dynamic data transmission queue at the sending end; Step 3.2: According to the Lyapunov optimization algorithm, transform the long-term optimization problem into a deterministic optimization problem within a single time slot. Its mathematical expression is: Among them, the constraint conditions also include formulas (10a)-(10e), is the defined virtual data transmission queue, and the initial time slot is set to zero, that is V is a key control parameter used to achieve a dynamic balance between queue stability and system throughput; represents the minimum value of the total data transmission rate that user n can achieve within time slot t, represents the maximum value of the total data transmission rate that user n can achieve within time slot t, represents the maximum value of the data volume arriving at user n.

5. The resource allocation optimization method for deterministic transmission in the RIS-assisted network according to claim 1, wherein The specific content of Step 4 is as follows: Step 4.1: Split the optimization problem for a single time slot into optimization sub-problems regarding user association relationships, base station power allocation variables, RIS reflection matrices, and base station positions. Initialize the base station positions, user positions, channel gains, user association relationships, RIS reflection matrices, and the maximum number of iterations; Step 4.2: According to the integer relaxation method, relax the binary user association relationship variables into continuous variables, construct a fractional programming problem, and solve the optimal user association relationship matrix; Step 4.3: Use fractional programming to transform the objective function into a subtractive form, and through a quadratic transformation, transform the non-convex constraints into convex constraints. Finally, use the interior point method to solve the optimal base station power allocation scheme; Step 4.4: Relax the RIS reflection coefficient matrix into a positive semi - definite matrix through semi - definite relaxation (SDR), use the successive convex approximation algorithm to gradually approximate the optimal solution, transform the non - convex objective function and constraints into the form of a convex optimization problem for solution, and recover the feasible solution through the Gaussian randomization method; Step 4.5: Adopt the successive convex approximation algorithm to gradually transform the non - convex problem into a convex optimization problem for solution. Linearize the non - convex objective function and constraints through the first - order Taylor expansion, and iteratively solve the optimal position of the base station until convergence; Step 4.6: Use the alternating iterative optimization algorithm to solve the optimization sub - problems of user association relationship, base - station power allocation variables, RIS reflection matrix, and base - station position, and output the optimal user association relationship, base - station power allocation, RIS reflection matrix, and base - station position.

6. The resource allocation optimization method for deterministic transmission in the RIS-assisted network according to claim 5, wherein The specific content of step 4.2 is as follows: Step 4.2.1: Given the feasible base - station power allocation P, RIS reflection matrix Θ, and base - station position l, simplify the original optimization problem P2 into an optimization sub - problem P3 only about the base - station and user association relationship C: Step 4.2.2: The user association relationship optimization sub-problem P3 belongs to a non-convex optimization problem. Process the constraints and the objective function. For simplicity of description, the explicit representation of time slot t will be omitted in the subsequent expressions. For the binary variable c k,n perform relaxation processing, relax it from a discrete binary variable to a continuous variable, and the specific conversion form is: 0 ≤ c k,n ≤ 1; Make an approximation for constraints (17d) and (17e) to obtain the constraints: Introduce the function: L 1,k,n (C) = c k,n p k,n g k,n (17h) Using fractional programming and quadratic transformation techniques, simplify the objective function P3, transform the ratio form in the problem into a subtraction form, and obtain the approximate optimization problem: where ξ k,n represents an introduced auxiliary variable, and the constraint conditions also include formulas (17a) and (17b).

7. The resource allocation optimization method for deterministic transmission in an RIS-assisted network according to claim 5, characterized in that, The specific content of step 4.3 is as follows: Step 4.3.1: Given the user association relationship C, RIS reflection matrix Θ, and base - station position l, simplify the original optimization problem P2 into an optimization sub - problem P4 about the base - station power allocation P: Among them, represents the set of users associated with base station k within time slot t; Step 4.3.2: The optimization sub - problem P4 of the base - station power allocation variable belongs to a non - convex optimization problem. Process the constraints and the objective function. First, re - express the data transmission rate formula that user n can achieve through base - station k: Introduce the function: O 1,k,n (P) = p k,n g k,n (20a) Using fractional programming theory and quadratic form transformation techniques, transform the ratio form in the optimization sub - problem P4 into a subtraction form, and the optimization sub - problem P4 is equivalently transformed into: For the optimization sub-problem P4.1, the solution is obtained by alternately optimizing the original variable P and the auxiliary variable λ k,n When P or λ k,n remains unchanged, the optimization problem P4.1 with respect to the variable λ k,n and P belongs to a convex optimization problem.

8. The resource allocation optimization method for deterministic transmission in the RIS-assisted network according to claim 5, wherein The specific content of step 4.4 is as follows: Step 4.4.1: Given the base - station and user association relationship C, base - station power allocation P, and base - station position l, transform the original optimization problem P2 into an optimization sub - problem P5 about the RIS reflection matrix: Step 4.4.2: The optimization problem P5 of the RIS reflection matrix belongs to a non - convex optimization problem. To facilitate the processing of the RIS reflection matrix variable Θ, rewrite the channel gain expression, and re - express the comprehensive channel gain from base - station k to user n as: Among them, represents the cascaded link between the base station-RIS-user without RIS configured; v = [r 1] T , V = vv H , After the conversion of the optimization variable matrix, V needs to satisfy the following constraint conditions: rank(V) = 1; [V] qq = 1; Restate the data transfer rate achievable by user n through base station k, and the expression is as follows: Therefore, the optimization problem P5 is transformed into: rank(V) = 1 (24c) [V] qq = 1 (24d) For the non-convex rank-one constraint (24c), it is equivalently transformed into the constraint ‖V‖ * −‖V‖2 ≤ 0; ‖V‖ * = ∑ i δ i (V) and ||V||2 = δ1(V) represent the nuclear norm and spectral norm of matrix V respectively, where δ i (V) is the i-th largest singular value of matrix δ i (V); for any there is ||V|| * −||V||2 ≥ 0 and the equality holds if and only if V is a rank-one matrix; the constraint ||V|| * −||V||2 ≤ 0 still belongs to non-convex constraints. To handle the constraint (24c), the rank-one constraint is added to the objective function as a penalty term, and the optimization problem can be reformulated as: wherein, is the penalty coefficient, for any given The objective function P5.2 is in the form of the difference of convex functions, and the optimal solution is gradually approximated by the successive convex approximation algorithm and solved by the classical convex toolbox to obtain the optimal solution V * , and then the optimal RIS reflection coefficient vector r * is extracted through eigenvalue decomposition.

9. The resource allocation optimization method for deterministic transmission in an RIS-assisted network according to claim 5, characterized in that, The specific content of step 4.5 is as follows: Step 4.5.1: Given the base - station and user association relationship C, base - station power allocation P, and RIS reflection matrix Θ, simplify the original optimization problem P2 into an optimization sub - problem P6 only about the base - station position l: Step 4.5.2: The optimization sub - problem P6 of the base - station position l belongs to a non - convex optimization problem. Re - express the data transmission rate that user n can achieve through base - station k: Introduce auxiliary variables du k,n , dl j,n , du k,r , dl k,r , where du k,n represents the upper bound of the distance between the base station and its served user, and dl j,n represents the lower bound of the distance between the base station and the unassociated user, du k,r represents the upper bound of the distance between the base station and the RIS, and dl k,r represents the lower bound of the distance between the base station and the RIS; introduce the variable to reformulate the comprehensive channel gain expression from base station k to user n: Let introduce the lower-bound auxiliary variable of g k,n ; introduce the upper-bound auxiliary variable of g j,n ; introduce the auxiliary variable as the lower bound of the user transmission rate R k,n ; The optimization sub - problem P6 is equivalently transformed into: For the non-convex characteristic of the above optimization problem P6.1, Taylor expansion is adopted for processing, and the sequential convex approximation algorithm is used for solving. This algorithm transforms the original non-convex optimization problem into a convex optimization problem and gradually approaches the global optimal solution through iterative convex optimization methods, thereby obtaining the optimal location \(l\) of the base station k .

Images