Resource allocation optimization method based on STAR-RIS assisted NOMA edge computing

Abstract

The embodiment of the invention provides a resource allocation optimization method based on STR-RIS assisted NOMA edge computing. The method is applied to the field of mobile edge computing, and comprises the following steps: establishing an STAR-RIS-assisted edge computing network architecture in an NOMA environment, establishing a communication model, a queue model and a computing model about the edge computing network architecture, and establishing a user computing total cost model according to the established communication model, queue model and computing model. And establishing an optimization problem with the purpose of minimizing the total user calculation cost, converting the optimization problem into a solution of a sub-optimal user grouping scheme and a solution of an optimal unloading power allocation scheme with the purpose of minimizing the total user calculation cost, and carrying out resource allocation optimization of edge calculation according to a solution result. According to the method, the users can be guided to use the optimal power transmission task at the beginning of any time slot, so that the long-term total calculation cost of all the users is minimized.

Description

Technical Field

[0001] This application relates to the field of mobile edge computing, and in particular to a resource allocation optimization method for STAR-RIS assisted NOMA edge computing. Background Art

[0002] Mobile Edge Computing (MEC) enhances the computing power of the Radio Access Network (RAN) by deploying computing resources at the base station, enabling intelligent terminals to offload tasks, thereby reducing the load and latency of the core network. With the explosion of intelligent devices in the 6G era, computationally intensive tasks (such as face recognition, AR / VR, online AI, etc.) pose higher requirements for local computing power, and low-power devices in the Internet of Things also struggle to process massive amounts of data. The introduction of MEC effectively alleviates the problems of high latency and high power consumption, improving the user experience and computing efficiency.

[0003] However, the MEC system is limited by the imperfection of the task offloading link. When the device is located at the edge of the area or blocked by the Line-of-Sight (LoS), the offloading rate decreases, resulting in higher latency and energy consumption, and at the same time, the resource utilization rate of the MEC server decreases.

[0004] Therefore, there is an urgent need for an efficient task offloading power allocation strategy to improve the performance of the MEC system. Summary of the Invention

[0005] This application provides a resource allocation optimization method for STAR-RIS assisted NOMA edge computing to improve the computing efficiency and communication performance of the mobile edge computing system.

[0006] In a first aspect, this application provides a resource allocation optimization method for STAR-RIS assisted NOMA edge computing, the method comprising:

[0007] Construct a STAR-RIS assisted edge computing network architecture in the NOMA environment, the edge computing network architecture comprising: a base station, a STAR-RIS and a plurality of users, and define the communication methods between the base station, the STAR-RIS and the users;

[0008] Establish a communication model, a queue model and a computing model for the edge computing network architecture, and construct a user computing total cost model according to the established communication model, queue model and computing model;

[0009] An optimization problem is established with the goal of minimizing the total user computing cost, and the optimization problem with the goal of minimizing the total user computing cost is split into a first sub-problem and a second sub-problem. The first sub-problem includes: determining an optimal or sub-optimal user grouping scheme, and the second sub-problem includes: determining an optimal offloading power allocation scheme with the goal of minimizing the total user cost;

[0010] Solve the first sub-problem and the second sub-problem, and optimize the resource allocation of edge computing according to the optimal or sub-optimal solution of the first sub-problem and the optimal solution of the second sub-problem.

[0011] Optionally, constructing a total user computing cost model according to the established communication model, queue model, and computing model includes:

[0012] Determine the queue length of the user task in the current time slot through the queue model;

[0013] Determine the local computing power and signal transmission power of the user in the current time slot through the computing model, and construct a total user computing cost model according to the queue length of the user task in the current time slot, the local computing power and signal transmission power of the user in the current time slot;

[0014] The total user computing cost model satisfies the following formula:

[0015]

[0016] Among them, C t represents the computing cost of user k at time slot t, represents the local computing power of user k at time slot t, represents the signal transmission power of user k at time slot t, represents the queue length of user k at time slot t, ω1 is the power weight coefficient, ω2 is the queue length weight coefficient, and ω1 ∈ [0, 1], ω2 ∈ [0, 1], ω1 + ω2 = 1.

[0017] Optionally, after establishing an optimization problem with the goal of minimizing the total user computing cost, the method further includes:

[0018] Based on the queue model and using Lyapunov optimization theory, transform the optimization problem with the goal of minimizing the total user computing cost into a queue stability problem in different time slots.

[0019] Optionally, establishing an optimization problem with the goal of minimizing the total user computing cost includes:

[0020] By jointly optimizing the local computing power, signal transmission power, and phase shift of the passive phase shifter to minimize the total computing cost of all users in the edge computing network;

[0021] Under the constraint of the maximum average queue length, the optimization problem of the total computing cost of all users satisfies the following formula:

[0022]

[0023] s.t.

[0024]

[0025] where s.t. represents the constraint condition, {u m,k} represents the grouping scheme of all users; {π m,k} represents the decoding order set of all users; represents the set of local computing powers allocated by the system to all users, p L,k represents the local computing power of user k; represents the set of signal transmission powers allocated by the system to all users, p O,k represents the signal transmission power of user k; is the set of phase shifts allocated by the system to all components, represents the phase shift of component n; T is the total number of time slots, and t represents any time slot; represents the average queue length of all users, is the preset queue length value; represents the maximum local computing power; represents the maximum signal transmission power; r k represents the achievable rate of user k, is the minimum rate of user k.

[0026] Optionally, using the Lyapunov optimization theory based on the queue model, the optimization problem with the goal of minimizing the total computing cost of users is transformed into a queue stability problem in different time slots, satisfying the following formula:

[0027]

[0028] s.t.

[0029]

[0030] where, V represents the trade-off parameter between user cost and queue backlog, V≥0, represents the total amount of tasks removed by user k from the task queue in time slot t; s.t. represents the constraint condition, {u m,k} represents the grouping scheme of all users; {π m,k} represents the decoding order set of all users; represents the set of local computing powers allocated by the system to all users, pL,k Denote the local computing power of user k; Denote the set of signal transmission powers allocated by the system to all users, p O,k Denote the signal transmission power of user k; Is the set of phase shifts allocated by the system to all components, Denote the phase shift of component n; T is the total number of time slots, and t represents any time slot; Denote the maximum local computing power; Denote the maximum signal transmission power; r k Denote the achievable rate of user k, Is the minimum rate of user k.

[0031] Optionally, solving the first sub-problem and the second sub-problem includes:

[0032] Finding the optimal or sub-optimal solution of the first sub-problem with the goal of maximizing the sum of data transmission rates of all users;

[0033] Describing the task offloading process through a Markov decision process model and using a deep reinforcement learning algorithm to determine the optimal solution of the second sub-problem.

[0034] Optionally, the Markov decision process model consists of a quadruple Composed of, Denote the state space, Denote the action space, Denote the reward, Denote the state transition probability;

[0035] The state space satisfies the following formula:

[0036]

[0037] Where, Denote the set of channel states of all users, Denote the set of queue task amounts of all users, {u m,k} Denote the grouping scheme of all users, {π m,k} The decoding order set of all users;

[0038] The action space satisfies the following formula:

[0039]

[0040] Where, Denote the set of local computing powers allocated by the system to all users, Denote the set of signal transmission powers allocated by the system to all users, Is the set of phase shifts allocated by the system to all components;

[0041] The reward function satisfies the following formula:

[0042] r = -C' k

[0043] where C' k represents the total computing cost of user k;

[0044] The state transition probability satisfies the following formula:

[0045]

[0046] where s(t) represents the state at time t, a(t) represents the action at time t, and s(t + 1) represents the state at time t + 1.

[0047] In a second aspect, the present application provides a resource allocation optimization device for STAR-RIS-assisted NOMA edge computing, and the device includes:

[0048] A first processing module, configured to construct an edge computing network architecture assisted by STAR-RIS in a NOMA environment, where the edge computing network architecture includes: a base station, STAR-RIS, and multiple users, and define the communication methods between the base station, STAR-RIS, and users;

[0049] A second processing module, configured to establish a communication model, a queue model, and a computing model for the edge computing network architecture, and construct a user computing total cost model according to the established communication model, queue model, and computing model;

[0050] A third processing module, configured to establish an optimization problem with the goal of minimizing the user computing total cost, and split the optimization problem with the goal of minimizing the user computing total cost into a first sub-problem and a second sub-problem, where the first sub-problem includes: determining an optimal or sub-optimal user grouping scheme, and the second sub-problem includes: determining an optimal offloading power allocation scheme with the goal of minimizing the user total cost;

[0051] A fourth processing module, configured to solve the first sub-problem and the second sub-problem, and perform resource allocation optimization for edge computing according to the optimal or sub-optimal solution of the first sub-problem and the optimal solution of the second sub-problem.

[0052] In a third aspect, the present application provides a resource allocation optimization device for STAR-RIS-assisted NOMA edge computing, including:

[0053] A memory;

[0054] A processor;

[0055] Among them, the memory stores computer-executable instructions;

[0056] The processor executes the computer-executable instructions stored in the memory to implement the resource allocation optimization method for STAR-RIS-assisted NOMA edge computing as described in the first aspect and various possible implementation manners of the first aspect above.

[0057] In a fourth aspect, the present application provides a computer storage medium, on which a computer program is stored, and the computer program is executed by a processor to implement the resource allocation optimization method for STAR-RIS-assisted NOMA edge computing as described in the first aspect and various possible implementation manners of the first aspect above.

[0058] The present application provides a resource allocation optimization method for STAR-RIS-assisted NOMA edge computing. The method constructs a STAR-RIS-assisted edge computing network architecture in a NOMA environment, establishes a communication model, a queue model, and a computing model for the edge computing network architecture, constructs a user computing total cost model according to the established communication model, queue model, and computing model, establishes an optimization problem with the goal of minimizing the user computing total cost, and converts the optimization problem into solving an optimal or sub-optimal user grouping scheme and solving an optimal offloading power allocation scheme with the goal of minimizing the user total cost, and optimizes the resource allocation of edge computing according to the solution results. The method can guide users to use the optimal power to transmit tasks at the beginning of any time slot, so as to minimize the long-term total computing cost of all users. Description of the Drawings

[0059] The drawings here are incorporated into the specification and constitute a part of this specification, showing embodiments consistent with the present application, and are used together with the specification to explain the principles of the present application.

[0060] Figure 1 It is a schematic flowchart of the resource allocation optimization method for STAR-RIS-assisted NOMA edge computing provided by an embodiment of the present application;

[0061] Figure 2 It is a schematic diagram of the structure of a STAR-RIS-assisted MEC system provided by an embodiment of the present application;

[0062] Figure 3 It is a schematic diagram of the structure of a resource allocation optimization device for STAR-RIS-assisted NOMA edge computing provided by an embodiment of the present application;

[0063] Figure 4 It is a schematic diagram of the structure of a resource allocation optimization device for STAR-RIS-assisted NOMA edge computing provided by an embodiment of the present application.

[0064] Through the above-mentioned drawings, specific embodiments of the present application have been shown, and will be described in more detail hereinafter. These drawings and the written description are not intended to limit the scope of the concept of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by reference to specific embodiments. Detailed Description of the Invention

[0065] To make the objectives, technical solutions and advantages of the present application clearer, the technical solutions in the present application will be clearly and completely described below with reference to the accompanying drawings in the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative efforts shall fall within the scope of protection of the present application.

[0066] In the description and claims of the present invention and the above-mentioned drawings, the terms "first", "second", "third", "fourth", etc. (if any) are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present invention described herein can be implemented in an order different from those illustrated or described herein.

[0067] In the embodiments of the present application, words such as "exemplary" or "for example" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as "exemplary" or "for example" in the present application should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Rather, the use of words such as "exemplary" or "for example" is intended to present relevant concepts in a specific manner.

[0068] To optimize the MEC system, the Simultaneously Transmitting and Reflecting Reconfigurable Intelligent Surface (STAR-RIS), as a programmable artificial electromagnetic surface, can dynamically adjust the channel conditions and improve the spectral efficiency. Compared with traditional transceivers, STAR-RIS does not require high-power radio frequency links or additional MEC hardware, and only relies on a low-complexity control circuit, with the advantages of low cost, low power consumption and easy deployment, and is expected to facilitate the wide application of MEC.

[0069] On the other hand, Non-Orthogonal Multiple Access (NOMA) is regarded as a major wireless access technology for 6G. Through NOMA technology, users can reuse the same frequency domain resources, thus saving energy and spectrum resources. In addition, the NOMA-based MEC system is superior to the Orthogonal Multiple Access (OMA)-based MEC system in effectively reducing latency and energy consumption. When the channel gain differences among users are large or the frequency resources are limited, the advantages of NOMA are more obvious. Deploying STAR-RIS and designing its phase shift can help NOMA technology meet the usage requirements under such conditions.

[0070] In addition, task offloading is one of the key research issues in the field of edge computing, which is directly related to the improvement of user experience. Task offloading means moving latency-sensitive or computationally intensive tasks in terminal devices to edge servers or cloud servers with richer computing resources for processing. Therefore, it is particularly urgent to guide the terminal to obtain a real-time task offloading solution. Traditional offloading strategy optimization schemes require very precise mathematical models to characterize the network environment, which is very difficult. With the gradual maturity of "intelligence", algorithms based on Reinforcement Learning (RL) can be used to learn dynamic task offloading strategies without prior knowledge of the statistical knowledge of the MEC system. Traditional RL algorithms use tabular methods to approximate the state function. As the number of agents increases, the state space will explode, making this method infeasible. Deep Reinforcement Learning (DRL) uses deep neural networks to approximate the function, which has stronger scalability and can effectively make strategic decisions under uncertain conditions without relying on any mathematical models.

[0071] In the design of existing MEC systems, the combined use of NOMA, STAR-RIS, and DRL technologies is lacking, and the advantages of the synergistic effects of NOMA, STAR-RIS, and DRL technologies cannot be fully exploited.

[0072] To address the above problems, this application proposes an optimization method for resource allocation based on STAR-RIS assisted NOMA edge computing. This method models the STAR-RIS assisted MEC task offloading system in a NOMA environment, establishes an optimization problem with the goal of minimizing the long-term total computing cost of users; and proposes a scheme for obtaining user grouping based on the closed-form solution of the optimal decoding order; uses the Markov decision process to describe the task offloading process, and then solves the power offloading decision based on the deep reinforcement learning model. This method can guide users to use the optimal power to transmit tasks at the beginning of any time slot, so as to minimize the long-term total computing cost of all users.

[0073] The following uses specific embodiments to detail the technical solutions of this application and how the technical solutions of this application solve the above technical problems. These several specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of this application will be described below with reference to the accompanying drawings.

[0074] Figure 1 It is a schematic flowchart of an optimization method for resource allocation based on STAR-RIS assisted NOMA edge computing provided by an embodiment of this application. As Figure 1 shown, the optimization method for resource allocation based on STAR-RIS assisted NOMA edge computing provided by this embodiment includes:

[0075] S1: Construct an edge computing network architecture assisted by STAR-RIS in a NOMA environment. The edge computing network architecture includes: a base station, a STAR-RIS, and multiple users, and defines the communication methods between the base station, the STAR-RIS, and the users.

[0076] Specifically, Figure 2 It is a schematic diagram of the STAR-RIS assisted MEC system structure provided by an embodiment of this application. As Figure 2 shown, there is a base station (abbreviated as BS) equipped with an MEC server, a STAR-RIS, and K users in the system. All users can access the BS through NOMA for data offloading. Assume that the STAR-RIS consists of N elements, and n represents any one of them. There are I users in the reflection area and J users in the transmission area of the STAR-RIS, and K = I + J. Use i and j to represent any one transmission area and reflection area user respectively. The set representing all users is The set representing the users in the reflection area is The set representing the users in the transmission area is And Suppose the direct communication link between the BS and the user is blocked by obstacles and communication is impossible. Therefore, the task offloading signals of all users can only be transmitted to the BS through reflection or transmission by the STAR-RIS.

[0077] S2: Establish a communication model, a queue model, and a computing model for the edge computing network architecture, and construct a total user computing cost model based on the established communication model, queue model, and computing model.

[0078] Specifically, the process of establishing the communication model is as follows:

[0079] Let the total time slot set of the system be denoted as where T is the total number of time slots, and t represents any time slot. Let the total time slot set of the transmission period be The total time slot set of the reflection period is and The transmission coefficient matrix of the STAR-RIS is denoted as The reflection coefficient matrix of the STAR-RIS is denoted as where, represents the amplitude reflection coefficient of the transmission element, the amplitude reflection coefficient of the reflection element, assuming and are identically equal to 1 to obtain the maximum transmission and reflection rates, represents the phase shift of the transmission element, represents the phase shift of the reflection element,

[0080] In the uplink NOMA system, the BS serves K users simultaneously, and the number of time / frequency / code resource blocks (Resource block, abbreviated as RB) is M, and M < K. To ensure the data transmission requirements of multiple users, the K users are assigned to M groups. Users within the same group perform NOMA transmission within the same RB, and users in different groups perform OMA transmission in different RBs. The signal expression received at the BS in the m-th RB is as follows:

[0081]

[0082] where, u m,i ∈{0,1} and u m,j ∈{0,1} represent whether user i and user j are assigned to the m-th RB respectively, 0 means not assigned, and 1 means assigned; p O,i and p O,j represent the signal transmission powers of user i and user j respectively; x i and x j represent the signals sent by user i in the reflection area and user j in the transmission area respectively, and Let \(G\) denote the channel gain coefficient from the STAR-RIS to the BS, \(h\) r,i denote the channel gain coefficient from user \(i\) to the STAR-RIS, \(h\) r,j the channel gain coefficient from user \(j\) to the STAR-RIS, \(n\) m be the additive white Gaussian noise (AWGN) of the \(m\)-th RB, with an average power of \(\sigma\) 2 .

[0083] For each RB, the BS should decode the signals of this group of users in a consecutive order using successive interference cancellation (SIC). Let \(\pi\) m,k denote the decoding order of the signal of user \(k\) in the \(m\)-th RB. For example, \(\pi\) m,k = 3 means that user \(k\) is the third user to be decoded in the \(m\)-th RB. Therefore, the effective received signal of user \(k\) is expressed as follows:

[0084]

[0085] where \(p\) O,k denotes the transmit power of the signal of user \(k\), and \(x\) k denotes the signal sent by user \(k\), denotes that in the \(m\)-th RB, user \(k\) - is decoded after user \(k\). Let the interference signal be the signals of other users that have not been canceled except user \(k\) in this RB.

[0086] The uplink signal-to-noise ratio of user \(k\) for communication is expressed as follows:

[0087]

[0088] The achievable rate of user \(k\) is expressed as follows:

[0089] \(r\) k = B log2(1 + SINR k )

[0090] where \(B\) represents the channel bandwidth.

[0091] Note that for uplink NOMA, SIC operates at the BS. The BS has sufficient power for signal processing and does not have an additional SNR constraint as in downlink NOMA. Assuming that the BS fully knows the channel state information of all users, SIC can completely eliminate interference. In addition, for power control and user grouping, the BS only needs to transmit the power control coefficient and RB metrics to each user. Then, the achievable sum rate (ASR) of the users in the m-th RB (group) is expressed as follows:

[0092]

[0093] It can be understood that although the ASR of each group is not affected by the decoding order of this formula, different decoding orders result in different achievable rates for each user.

[0094] Specifically, the establishment process of the queue model is as follows:

[0095] For each time slot Use to represent the amount of tasks arriving at user k during time slot t. Assume it is independent and identically distributed over all time intervals. At the end of time slot t, all the arriving tasks are added to the user's task queue buffer and then processed in the next time slot. Additionally, assume that the tasks of the application are fine-grained, which means that the data of each task can be partially executed on the local device while another part is offloaded to the edge server. During time slot t, use and to represent the amount of locally computed and offloaded tasks respectively, use to represent the total amount of tasks removed from the task queue, and will not exceed the initial queue length of the current time slot At the beginning of time slot t + 1, the change in the length of the task queue is expressed as follows:

[0096]

[0097] where, when in the 0-th time slot, the queue is empty, expressed as

[0098] Specifically, the establishment process of the computing model is as follows:

[0099] Assume the independence of task inputs, which means they can be divided into subsets of arbitrary size. This is the basic premise for task scheduling and offloading. In this case, a partial offloading mode can be achieved, that is, tasks can be executed on both the user device and the MEC server simultaneously.

[0100] Local computing can adjust the chip voltage as needed to control the user's CPU frequency f by leveraging the dynamic voltage and frequency scaling (DVFS) technology. k , which is expressed as follows:

[0101]

[0102] where p L,k is the local computing power and κ is the effective capacitance coefficient.

[0103] The amount of tasks computed during a unit time slot is expressed as follows:

[0104]

[0105] where d L,k represents the amount of tasks computed during a unit time slot; τ0 is the length of the time slot in seconds; L k represents the number of CPU cycles required for user k to process a unit amount of task data.

[0106] It can be understood that edge servers usually have sufficient computing resources to process multiple tasks in parallel. In addition, considering that the amount of data unloaded from all user devices to the server within a time slot is also limited and the scale of the computing results is relatively small, the time required for the MEC server in computing and result downloading can be ignored. The edge server can complete all the unloaded tasks within each unit time slot, and the amount of tasks from each user is expressed as follows:

[0107] d O,k = τ0r k

[0108] Furthermore, based on the queue length of the user tasks in the current time slot, the local computing power and signal transmission power of the user in the current time slot, a user computing total cost model is constructed, and the user computing total cost model satisfies the following formula:

[0109]

[0110] where C t represents the computing cost of user k in time slot t, represents the local computing power of user k in time slot t, represents the signal transmission power of user k in time slot t, represents the queue length of user k in time slot t, ω1 is the power weight coefficient, ω2 is the queue length weight coefficient, and ω1 ∈ [0, 1], ω2 ∈ [0, 1], ω1 + ω2 = 1.

[0111] S3: Establish an optimization problem aiming to minimize the total computing cost of users, and split the optimization problem aiming to minimize the total computing cost of users into a first sub-problem and a second sub-problem. The first sub-problem includes: determining an optimal or sub-optimal user grouping scheme. The second sub-problem includes: determining an optimal offloading power allocation scheme aiming to minimize the total cost of users.

[0112] By jointly optimizing the local computing power, task offloading power, and phase shift of the passive phase shifter to minimize the total computing cost of all users in the edge computing network;

[0113] Under the constraint of the maximum average queue length, the optimization problem of the total computing cost of all users satisfies the following formula:

[0114]

[0115] s.t.

[0116]

[0117] where s.t. represents the constraint condition, {u m,k} represents the grouping scheme of all users; {π m,k} represents the decoding order set of all users; represents the set of local computing powers allocated by the system to all users, p L,k represents the local computing power of user k; represents the set of signal transmission powers allocated by the system to all users, p O,k represents the signal transmission power of user k; is the set of phase shifts allocated by the system to all components, represents the phase shift of component n, n represents any component, in the reflection state or transmission state; T is the total number of time slots, and t represents any time slot; represents the average queue length of all users, is the preset queue length value; (b) and (C) indicate that each user can only choose to join one resource group; (d), (e), (f), and (g) respectively limit the local computing, signal transmission power, phase shift range of the RIS component, and the minimum transmission rate of the user; represents the maximum local computing power; represents the maximum signal transmission power, r k represents the achievable rate of user k, is the minimum rate of user k.

[0118] It can be understood that in order to handle constraint (a), this optimization problem with long-term constraints can be transformed into a queue stability problem. Specifically, Θ tDefined as the queue backlog matrix of the users, that is Let L(Θ t ) be the Lyapunov function, and the following formula is obtained:

[0119]

[0120] where L(Θ t ) represents the queue backlog status of the users. A larger L(Θ t ) indicates that the task queue backlog of at least one user is relatively large. Only when the task queue backlog of each user is small, L(Θ t ) will be small. Therefore, it is desired to reduce the value of L(Θ t ) to maintain a low congestion state of the user task queue. Define the conditional Lyapunov drift as Δ(Θ t ), specifically:

[0121] Δ(Θ t ) = E{L(Θ t+1 ) - L(Θ t )|Θ t}

[0122] According to the Lyapunov optimization theory, combine the user cost and the queue backlog to obtain the following formula:

[0123] Δ(Θ t ) + VE{C t |Θ t}

[0124] where V is the trade-off parameter between the user cost and the queue backlog, and V ≥ 0.

[0125] At time slot t, let and be the maximum upper bounds of a k and r k . Square both sides of the task queue length change formula in step S2 above to obtain the following formula:

[0126]

[0127] represents the actual task processing volume of the user at time slot t, then:

[0128]

[0129] Because and are non-negative, the following formula can be obtained:

[0130]

[0131] Taking the sum over all users of the above equation and taking the conditional expectation gives:

[0132]

[0133] Since and it can be concluded that and So we have:

[0134]

[0135] From and the above equation, the following equation can be obtained:

[0136]

[0137] Let C be equal to Adding the above equation to VE{C t ∣Θ(t)} gives the following equation:

[0138]

[0139] Let Therefore, the original optimization problem can be reformulated as the following:

[0140]

[0141] s.t.

[0142]

[0143] To solve problem (P2), this complex problem can be decomposed into two independent problems, namely, finding the optimal or sub - optimal uplink NOMA user grouping scheme and finding the optimal offloading strategy power allocation scheme aiming to minimize the total user cost.

[0144] S4: Solve the first sub - problem and the second sub - problem, and optimize the resource allocation of edge computing according to the optimal or sub - optimal solution of the first sub - problem and the optimal solution of the second sub - problem.

[0145] The objective of the user grouping scheme can be considered separately as finding the optimal or sub - optimal solution aiming to maximize the ASR of all users, which is specifically expressed as follows:

[0146]

[0147] s.t.

[0148]

[0149] where α k is the power control coefficient of user k; {αk} is the set of all user power control coefficients. For the convenience of theoretical derivation in the problem-solving process, the signal transmission power p of the user O,k is represented by α k P, where P is a constant representing the maximum value of the transmission power.

[0150] It can be understood that since the optimization variables {u m,k}, {π m,k}, and {α k} are intertwined and difficult to solve directly, and the user grouping variable {π m,k} is a combinatorial integer programming variable, the decoding order and power control coefficients can be optimized first, and then the user grouping scheme can be solved.

[0151] It should be noted that for any fixed user grouping scheme {u m,k}, the decoding order and power control problems are independent between different groups. Therefore, for each group of users, the following problem can be solved:

[0152]

[0153] s.t.

[0154]

[0155] where is the set of users in the m-th group.

[0156] Considering any two users with adjacent decoding orders, denoted as user x and user y respectively. There are two decoding orders for these two users, i.e., decoding x first or decoding y first. Then, the feasible regions of power control α x and α y in different cases can be obtained, and the case that can achieve a higher sum rate can be found.

[0157] Case 1: Decode the signal of user x first. The feasible region of power control can be generated by the following constraint conditions:

[0158]

[0159] where h x and h y represent the channel gain coefficients from user x and user y to the BS through the STAR-RIS respectively; D represents the interference from users with decoding orders higher than user x and user y; represents the minimum signal-to-noise ratio requirement of user k; ω is the upper bound of the interference plus noise generated by users with decoding orders lower than user x and user y due to the minimum rate constraint.

[0160] Case 2: First, decode the signal of user y. The feasible region of power control can be generated by the following constraint conditions:

[0161]

[0162] Without loss of generality, assume |h x | 2 (1 + 1 / η x ) ≥ |h y | 2 (1 + 1 / η y ).

[0163] It can be understood that the above two constraint conditions restrict the lower bounds of α x and α y . That is, for Case 1, there is the following relationship:

[0164]

[0165] For Case 2, there is the following relationship:

[0166]

[0167] If the feasible region of Case 2 is not empty, then and The lower bounds of always lie in the feasible region shown by the above formula.

[0168] It can be understood that the ASR of α x and α y gradually increases. Therefore, the optimal power control solutions in both cases are the upper bounds. For Case 1, the upper bound of α x is 1, and the upper bound of α y is the following formula:

[0169]

[0170] Substitute it into |h x | 2 α x P + |h y | 2 α y P, and the maximum signal power of Case 1 can be obtained as the following formula:

[0171] min{c1, |h x | 2 P + |h y | 2 P, ω - D - σ 2}

[0172] where c1 = |h x |2 P(1 + 1 / η x ) - D - σ 2 。

[0173] Similarly, the maximum signal power of Case 2 is given by the following equation:

[0174] min{c2, |h x | 2 P + |h y | 2 P, ω - D - σ 2}

[0175] where c2 = |h y | 2 P(1 + 1 / η y ) - D - σ 2 。

[0176] It can be understood that since |h x | 2 (1 + 1 / η x ) ≥ |h y | 2 (1 + 1 / η y ), so c1 ≥ c2, that is, the optimal decoding order is to decode the signal of the user with a larger |h k | 2 (1 + 1 / η k ).

[0177] In summary, for multiple users in the same group, the optimal decoding order of any two adjacent users should satisfy this criterion. Therefore, it can be concluded that the best decoding order of uplink NOMA is the decreasing order of |h k | 2 (1 + 1 / η k ).

[0178] In an alternative embodiment, in order to further reduce the computational complexity, an alternative user grouping algorithm is proposed, as shown in Algorithm 1.

[0179]

[0180] Furthermore, the task offloading process is described by a Markov decision process model, and a deep reinforcement learning algorithm is used to find the optimal task offloading power allocation strategy.

[0181] Specifically, the Markov decision process model consists of a quadruple which consists of representing the state space, representing the action space, representing the reward, representing the state transition probability;

[0182] The state space satisfies the following formula:

[0183]

[0184] where, represents the set of all user channel states, represents the set of all user queue task amounts, {u m,k} represents the grouping scheme for all users, {π m,k} represents the set of decoding orders for all users;

[0185] The action space satisfies the following formula:

[0186]

[0187] where, represents the set of local computing powers allocated by the system to all users, represents the set of signal transmission powers allocated by the system to all users, is the set of phase shifts allocated by the system to all components;

[0188] The reward function satisfies the following formula:

[0189] r = -C' k

[0190] where, C' k represents the total computing cost of user k;

[0191] The state transition probability satisfies the following formula:

[0192]

[0193] where, s(t) represents the state at time t, a(t) represents the action at time t, and s(t + 1) represents the state at time t + 1.

[0194] In an alternative embodiment, the DDQN algorithm in DRL is used to find the optimal task offloading power allocation strategy to minimize the long-term total cost of users while satisfying various constraints.

[0195] The specific algorithm flow is shown in Algorithm 2, and the parameter update formula of the neural network in this algorithm is as follows:

[0196]

[0197] where, Q represents the action-value equation of the Q network, r(t) is the reward at time t, and θ(t) is the initial network parameter setting at time t; θ -(t) is the initial target network parameter setting at time t, ρ is the learning rate, ρ ∈ (0, 1), and γ is the attenuation coefficient.

[0198]

[0199]

[0200] The resource allocation optimization method based on STAR-RIS assisted NOMA edge computing provided by the embodiments of this application constructs an edge computing network architecture assisted by STAR-RIS in a NOMA environment, establishes a communication model, a queue model, and a computing model for the edge computing network architecture, constructs a user computing total cost model according to the established communication model, queue model, and computing model, establishes an optimization problem with the goal of minimizing the user computing total cost, and converts the optimization problem into solving an optimal or sub-optimal user grouping scheme and solving an optimal offloading power allocation scheme with the goal of minimizing the user total cost, and optimizes the resource allocation of edge computing according to the solution results. This method can guide users to use the optimal power to transmit tasks at the beginning of any time slot, so as to minimize the long-term total computing cost of all users.

[0201] Figure 3 It is a schematic structural diagram of a resource allocation optimization device based on STAR-RIS assisted NOMA edge computing provided by the embodiments of this application. As Figure 3 shown, the resource allocation optimization device 300 based on STAR-RIS assisted NOMA edge computing provided by this embodiment includes:

[0202] A first processing module 301, configured to construct an edge computing network architecture assisted by STAR-RIS in a NOMA environment, where the edge computing network architecture includes: a base station, a STAR-RIS, and multiple users, and define the communication methods between the base station, the STAR-RIS, and the users;

[0203] A second processing module 302, configured to establish a communication model, a queue model, and a computing model for the edge computing network architecture, and construct a user computing total cost model according to the established communication model, queue model, and computing model;

[0204] A third processing module 303, configured to establish an optimization problem with the goal of minimizing the user computing total cost, and split the optimization problem with the goal of minimizing the user computing total cost into a first sub-problem and a second sub-problem, where the first sub-problem includes: determining an optimal or sub-optimal user grouping scheme, and the second sub-problem includes: determining an optimal offloading power allocation scheme with the goal of minimizing the user total cost;

[0205] The fourth processing module 304 is configured to solve the first sub-problem and the second sub-problem, and optimize the resource allocation for edge computing according to the optimal or sub-optimal solution of the first sub-problem and the optimal solution of the second sub-problem.

[0206] The resource allocation optimization device for STAR-RIS-assisted NOMA edge computing provided in this embodiment can execute the resource allocation optimization method for STAR-RIS-assisted NOMA edge computing provided in the above method embodiment. The implementation principle and technical effects are similar, and will not be elaborated here.

[0207] Figure 4 It is a schematic structural diagram of the resource allocation optimization device for STAR-RIS-assisted NOMA edge computing provided in an embodiment of the present application. As Figure 4 shown, the resource allocation optimization device for STAR-RIS-assisted NOMA edge computing provided in an embodiment of the present application, the resource allocation optimization device 400 for STAR-RIS-assisted NOMA edge computing includes: a receiver 401, a transmitter 402, a processor 403, and a memory 404.

[0208] The receiver 401 is configured to receive instructions and data;

[0209] The transmitter 402 is configured to send instructions and data;

[0210] The memory 404 is configured to store computer-executable instructions;

[0211] The processor 403 is configured to execute the computer-executable instructions stored in the memory 404 to implement each step performed by the resource allocation optimization method for STAR-RIS-assisted NOMA edge computing in the above embodiment. Specifically, reference can be made to the relevant descriptions in the foregoing resource allocation optimization method embodiment for STAR-RIS-assisted NOMA edge computing.

[0212] Optionally, the above memory 404 can be either independent or integrated with the processor 403.

[0213] When the memory 404 is independently provided, the electronic device further includes a bus for connecting the memory 404 and the processor 403.

[0214] An embodiment of the present application also provides a computer storage medium, in which computer-executable instructions are stored. When the processor executes the computer-executable instructions, the resource allocation optimization method for STAR-RIS-assisted NOMA edge computing executed by the resource allocation optimization device for STAR-RIS-assisted NOMA edge computing as described above is implemented.

[0215] Those of ordinary skill in the art will understand that all or some of the steps in the methods disclosed above, and the functional modules / units in systems and devices, can be implemented as software, firmware, hardware, and appropriate combinations thereof. In the hardware implementation, the division of functional modules / units mentioned above does not necessarily correspond to the division of physical components; for example, one physical component may have multiple functions, or one function or step may be executed by several physical components in cooperation. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or may be implemented as hardware, or may be implemented as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include a computer storage medium (or non-transitory medium) and a communication medium (or transitory medium). As is well known to those of ordinary skill in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules, or other data. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cassette, tape, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, it is well known to those of ordinary skill in the art that a communication medium typically contains computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transmission mechanism, and can include any information delivery medium.

[0216] After considering the specification and practicing the invention disclosed herein, those skilled in the art will readily conceive of other embodiments of the present application. The present application is intended to cover any variations, uses, or adaptations of the present application, which follow the general principles of the present application and include known common knowledge or conventional technical means in the technical field not disclosed in the present application. The specification and examples are only illustrative, and the true scope and spirit of the present application are pointed out by the following claims.

[0217] It should be understood that the present application is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present application is only limited by the appended claims.

Claims

1. A resource allocation optimization method based on STAR-RIS assisted NOMA edge computing, characterized in that The method includes: Construct a STAR-RIS assisted edge computing network architecture in the NOMA environment, where the edge computing network architecture includes: a base station, a STAR-RIS, and multiple users, and define the communication methods between the base station, the STAR-RIS, and the users; Establish a communication model, a queue model, and a computing model for the edge computing network architecture, and construct a user computing total cost model according to the established communication model, queue model, and computing model; Establish an optimization problem with the goal of minimizing the total user computing cost, and split the optimization problem with the goal of minimizing the total user computing cost into a first sub-problem and a second sub-problem. The first sub-problem includes: determining the optimal or sub-optimal user grouping scheme, and the second sub-problem includes: determining the optimal offloading power allocation scheme with the goal of minimizing the total user cost; Solve the first sub-problem and the second sub-problem, and optimize the resource allocation of edge computing according to the optimal or sub-optimal solution of the first sub-problem and the optimal solution of the second sub-problem.

2. The method according to claim 1, wherein The constructing a user computing total cost model according to the established communication model, queue model, and computing model includes: Determine the queue length of the user task in the current time slot through the queue model; Determine the local computing power and signal transmission power of the user in the current time slot through the computing model, and construct a user computing total cost model according to the queue length of the user task in the current time slot, the local computing power and signal transmission power of the user in the current time slot; The user computing total cost model satisfies the following formula: Among them, C t represents the computing cost of user k at time slot t, denotes the local computing power of user k at time slot t, represents the signal transmission power of user k at time slot t, represents the queue length of user k at time slot t, ω1 is the power weight coefficient, ω2 is the queue length weight coefficient, and ω1 ∈ [0, 1], ω2 ∈ [0, 1], ω1 + ω2 = 1.

3. The method according to claim 2, wherein After establishing the optimization problem with the goal of minimizing the total user computing cost, the method further includes: Based on the queue model, use the Lyapunov optimization theory to transform the optimization problem with the goal of minimizing the total user computing cost into a queue stability problem in different time slots.

4. The method according to claim 2, wherein The establishing an optimization problem with the goal of minimizing the total user computing cost includes: Jointly optimize the local computing power, signal transmission power, and phase shift of the passive phase shifter to minimize the total computing cost of all users in the edge computing network; Under the constraint of the maximum average queue length, the optimization problem of the total computing cost of all users satisfies the following formula: Among them, s.t. represents the constraint condition, and {u m,k} represents the grouping scheme of all users; {π m,k} represents the decoding order set of all users; represents the set of local computing powers allocated by the system to all users, and p L,k represents the local computing power of user k; represents the set of signal transmission powers allocated by the system to all users, and p O,k represents the signal transmission power of user k; is the set of phase shifts allocated by the system to all components, represents the phase shift of component n; T is the total number of time slots, and t represents any time slot; represents the average queue length of all users, is the preset queue length value; represents the maximum local computing power; represents the maximum signal transmission power; r k represents the achievable rate of user k, is the minimum rate of user k.

5. The method according to claim 3, characterized in that, The using the Lyapunov optimization theory based on the queue model to transform the optimization problem with the goal of minimizing the total user computing cost into a queue stability problem in different time slots satisfies the following formula: Among them, V represents the trade-off parameter between user cost and queue backlog, V≥0, represents the total amount of tasks removed by user k from the task queue at time slot t; s.t. represents the constraint condition, {u m,k} represents the grouping scheme of all users; {π m,k} represents the decoding order set of all users; represents the set of local computing powers allocated by the system to all users, p L,k represents the local computing power of user k; represents the set of signal transmission powers allocated by the system to all users, p O,k represents the signal transmission power of user k; is the set of phase shifts allocated by the system to all components, represents the phase shift of component n; T is the total number of time slots, and t represents any time slot; represents the maximum local computing power; represents the maximum signal transmission power; r k represents the achievable rate of user k, is the minimum rate of user k.

6. The method according to claim 1, wherein The solving the first sub-problem and the second sub-problem includes: Find the optimal or sub-optimal solution of the first sub-problem with the goal of maximizing the sum of the data transmission rates of all users; Describe the task offloading process through a Markov decision process model, and use a deep reinforcement learning algorithm to determine the optimal solution of the second sub-problem.

7. The method according to claim 6, wherein The Markov decision process model consists of a quadruple and represents the state space, represents the action space, represents the reward, represents the state transition probability; The state space satisfies the following formula: Among them, represents the set of all user channel states, represents the set of all user queue task amounts, {u m,k} represents the packetization scheme for all users, {π m,k} represents the decoding order set for all users; The action space satisfies the following formula: Among them, represents the set of local computing power allocated by the system to all users, represents the set of signal transmission power allocated by the system to all users, is the set of phase shifts allocated by the system to all components; The reward function satisfies the following formula: r＝-C' k Among them, C' k represents the total computing cost of user k; The state transition probability satisfies the following formula: Where, s(t) represents the state at time t, a(t) represents the action at time t, and s(t + 1) represents the state at time t + 1.

8. A resource allocation optimization device based on STAR-RIS assisted NOMA edge computing, characterized in that, The device includes: The first processing module is used to construct a STAR-RIS assisted edge computing network architecture in the NOMA environment. The edge computing network architecture includes: a base station, a STAR-RIS, and multiple users, and defines the communication methods among the base station, the STAR-RIS, and the users; The second processing module is used to establish a communication model, a queue model, and a computing model for the edge computing network architecture, and construct a user computing total cost model according to the established communication model, queue model, and computing model; The third processing module is used to establish an optimization problem with the goal of minimizing the total user computing cost, and split the optimization problem with the goal of minimizing the total user computing cost into a first sub-problem and a second sub-problem. The first sub-problem includes: determining an optimal or sub-optimal user grouping scheme, and the second sub-problem includes: determining an optimal offloading power allocation scheme with the goal of minimizing the total user cost; The fourth processing module is used to solve the first sub-problem and the second sub-problem, and optimize the resource allocation of edge computing according to the optimal or sub-optimal solution of the first sub-problem and the optimal solution of the second sub-problem.

9. A resource allocation optimization device based on STAR-RIS assisted NOMA edge computing, characterized in that, The device includes: A memory; A processor; Wherein, the memory stores computer execution instructions; The processor executes the computer execution instructions stored in the memory to implement the resource allocation optimization method for STAR-RIS assisted NOMA edge computing according to any one of claims 1-7.

10. A computer storage medium, characterized in that, Computer execution instructions are stored in the computer storage medium, and when the computer execution instructions are executed by the processor, they are used to implement the resource allocation optimization method for STAR-RIS assisted NOMA edge computing according to any one of claims 1-7.

Images