Power allocation method based on cell-free massive MIMO short-packet communication system

Abstract

Disclosed in the present invention is a power allocation method based on a cell-free massive MIMO short-packet communication system. The method comprises: on the basis of a Rayleigh fading channel model under channel aging, deriving an expression for a downlink achievable rate of a cell-free massive MIMO short-packet communication system under user mobility; with the goal of maximizing a minimum user rate, proposing a downlink power allocation method; and on the basis of a successive convex approximation method and geometric programming, performing iterative solving, wherein an obtained solution is used as a final power allocation result for calculating the rate of each user. The present invention maximizes a minimum user achievable rate of users in a system, effectively solves the problem of uneven quality of service for the users, enhances system fairness, and also has a relatively high convergence rate.

Description

A power allocation method based on decellularized massive MIMO short packet communication system Technical Field

[0001] The present invention belongs to wireless communication technology, and in particular relates to a power allocation method based on a decellularized large-scale MIMO short packet communication system. Background Art

[0002] With the advent of the Industrial Internet of Things (IIoT), communications are gradually expanding from person-to-person to person-to-things and thing-to-things, ultimately enabling the interconnection of all things. To meet the demands of massive connectivity and industrial automation applications, decellularized massive MIMO network architectures and ultra-reliable, low-latency communication transmission have become key technologies. Decellularized massive MIMO network architectures abandon the traditional cell concept, effectively overcoming the severe interference and handover issues inherent in cellular networks, particularly those affecting edge users. Furthermore, a large number of access points (APs) equipped with one or more antennas are randomly distributed around users, significantly shortening communication distances and reducing path loss during signal propagation. This also provides more spatial degrees of freedom and diversity gain, paving the way for massive connectivity. Furthermore, in industrial automation, the number of transmitted instructions is typically small, and synchronous applications have high latency requirements. Therefore, the traditional theory of infinite-length transmission is no longer applicable. Ultra-reliable, low-latency communication, also known as short-packet communication, is a transmission scheme designed to transmit finite-length data packets. Combining this with decellularized massive MIMO short-packet communication systems has become a research hotspot for the realization of the Industrial Internet of Things.

[0003] As mentioned above, in decellularized massive MIMO short packet communication systems, the limited packet length precludes the direct application of theories and techniques analyzed and designed based on infinite-length packets, such as system achievable rate analysis and power optimization allocation schemes. Furthermore, since users are mobile, the channel constantly changes over time, severely impacting communication quality. Against this backdrop, the present invention provides a power allocation method suitable for decellularized massive MIMO short packet communication systems. Summary of the Invention

[0004] Purpose of the invention: The purpose of the present invention is to provide a power allocation method based on a de-cellularized massive MIMO short packet communication system to solve the problem of unfair user services caused by uneven power resource allocation.

[0005] Technical solution: A power allocation method based on a decellularized massive MIMO short packet communication system of the present invention comprises the following steps:

[0006] Step 1: Based on a decellularized massive MIMO short packet communication system in a user mobility scenario, a Rayleigh fading channel model under channel aging is established;

[0007] Step 2: Based on the Rayleigh fading channel model, the user's received data expression during the downlink data transmission phase is given, and the downlink achievable rate is derived from it;

[0008] Step 3: Based on the downlink achievable rate, with the minimum user rate as the objective function, a maximization optimization problem with power constraints is proposed;

[0009] Step 4: For the maximization optimization problem, the objective function is converted into an exponential function using the continuous convex approximation method. The constraints are then processed again using the continuous convex approximation method to meet the requirements of the geometric programming algorithm and obtain the optimal power allocation solution.

[0010] Furthermore, in step 1, the Rayleigh fading channel model is expressed as:

[0011] Among them, h mk (t) represents the channel sequence between APm and user k at the tth moment, h mk (λ) represents the channel sequence between APm and user k at the λth moment, ξ mk (t) represents the random channel sequence between APm and user k at time t and time λ, ρ k (t-λ) represents the channel correlation coefficient between user k at time t and time λ.

[0012] Furthermore, in step 2, the received data of the user in the downlink data transmission phase is expressed as follows: the data received by user k at time t:

[0013] Where M represents the number of APs, ρ d represents the normalized signal-to-noise ratio of each data symbol in downlink transmission, η mk represents the power coefficient allocated by APm to user k, η mk′ represents the power coefficient allocated by APm to user k′, ρ k (t-λ) represents the channel correlation coefficient between user k at time t and time λ, f mk (λ) represents the precoding sequence of APm and user k at the λth moment, f mk′ (λ) represents the precoding sequence of APm and user k′ at the λth moment, q k (t) represents the data symbol sent by AP to user k at time t, q k′ (t) represents the data symbol sent by AP to user k′ at time t, which satisfies E{|q k (t)| 2}=1 and ω k (t) represents the additive Gaussian noise received by user k at the tth moment, DS k represents the expected signal strength of user k, BU k represents the uncertainty of beam gain of user k, CA k Indicates the impact of channel aging on user k, UI kk′ represents the inter-user interference suffered by user k, E{·} and (·) H They represent the expectation operation and matrix conjugate transpose operation respectively.

[0014] Furthermore, in step 2, the derived downlink achievable rate is specifically: the downlink achievable rate R of user k at time t k (t) is expressed as:

[0015] Among them, L represents the packet length, SINR k (t) represents the signal to interference and noise ratio of user k at the tth moment, ε represents the decoding error probability, Q -1 (·) represents the inverse function of the Q function, which is the right tail function of the standard normal distribution. Ε{·} and |·| represent expectation operation and modulo operation respectively.

[0016] Furthermore, step 3 is specifically as follows: the power optimization problem in the downlink data transmission phase is expressed as:

[0017] Among them, R k (t) represents the downlink rate achievable by user k at time t, η mk represents the power coefficient allocated by APm to user k, E{·} and (·) H They represent the expectation operation and matrix conjugate transpose operation respectively, f mk (λ) represents the precoding sequence of APm and user k at the λth time.

[0018] Furthermore, the specific steps of step 4 are:

[0019] Step 1: Introduce auxiliary variable γ k (t) = SINR k (t), we can get

[0020] The original problem is transformed into

[0021] Step 2: Given an initial And calculate the initial value

[0022] Step 3: Use the continuous convex approximation method to approximate ln(1+γ k (t)) and Available,

[0023] From this we can get,

[0024] Step 4: Use continuous convex approximation again to convert SINR k The numerator of (t) is approximately expressed as

[0025] in Therefore, the optimization problem is transformed into

[0026] Step 5: Introduce the auxiliary variable u to convert the original problem into the standard form of geometric programming, as follows

[0027] Then use the CVX tool to solve the problem and get the initial solution for the next iteration

[0028] Step 6: Repeat steps 2 to 5 until convergence and obtain the optimal allocation solution

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] (1) The power allocation method proposed in this invention for a decellularized massive MIMO short packet communication system takes into account the channel aging effect that is not considered in most existing technologies. That is, the channel between the base station and the user changes over time due to the user's movement, making the invention more practical.

[0031] (2) The present invention establishes a Rayleigh fading channel model under the influence of channel aging, derives the downlink achievable rate expression of the decellularized massive MIMO short packet communication system, and then proposes a minimum user rate maximization optimization problem based on this expression, thereby improving the service quality of users with poor channels as much as possible, thereby improving the fairness of the system. At the same time, since the derived rate expression includes the influence of channel aging, the proposed rate optimization problem also alleviates the deterioration of system performance caused by channel aging to a certain extent, solving most of the problems that have not been solved in the existing technology.

[0032] (3) Due to the complexity and non-convexity of the rate optimization problem, this paper applies continuous convex approximation and geometric programming to solve the maximum and minimum optimization problem in such a system for the first time. Simulations also verify the effectiveness and fast convergence of this method.

[0033] (4) Compared with the prior art, the present invention only utilizes channel statistical information that changes slowly over time, avoiding the huge overhead incurred by the system to obtain instantaneous channel information. The optimal power solution obtained by solving the proposed optimization problem once is also applicable to data transmission within multiple consecutive coherent intervals, avoiding frequent data processing and updating. Therefore, the present invention has good feasibility. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 shows the relationship between the average downlink rate and the number of users;

[0035] FIG2 is a diagram showing the number of iterations of the power allocation method. DETAILED DESCRIPTION

[0036] The technical means and effects of the present invention are further described below in conjunction with the accompanying drawings to make the present invention easier to understand. The following examples are only used to more clearly illustrate the technical solutions of the present invention and are not intended to limit the scope of protection of the present invention.

[0037] Example 1

[0038] This embodiment provides a power allocation method based on a decellularized massive MIMO short packet communication system, the purpose of which is to allocate appropriate power to each user during the downlink data transmission phase to maximize the minimum user rate. The method mainly includes the following steps:

[0039] (1) Based on the user mobility-based decellularized massive MIMO short packet communication system, a Rayleigh fading channel model under channel aging is established;

[0040] (2) Based on the channel model, an expression for the user's received data during the downlink data transmission phase is given, and the downlink achievable rate is derived from it;

[0041] (3) Taking the minimum user rate as the objective function, a maximization optimization problem with power constraints is proposed, and then the objective function is transformed into an exponential function form using a continuous convex approximation method;

[0042] (4) The constraints are processed again using continuous convex approximation to meet the requirements of the geometric programming algorithm.

[0043] Furthermore, the present invention considers a decellularized massive MIMO short packet communication system with M access points (APs) and K mobile users, where each AP has N antennas. In this system, affected by channel aging, the channel model is

[0044] Among them, h mk (t) represents the channel sequence between APm and user k at the tth moment, h mk(λ) represents the channel sequence between APm and user k at the λth moment, ξ mk (t) represents the random channel sequence between APm and user k at time t and time λ, ρ k (t-λ)=J0(2πf Dk T s |t-λ|) represents the channel correlation coefficient between user k at time t and time λ, J0(·) represents the first kind of zero-order Bessel function, T s represents the sampling time, f Dk =f c v / c represents the Doppler shift of user k at speed v, f c represents the channel carrier frequency, and c represents the speed of light.

[0045] Furthermore, under this channel model, the data received by user k from all APs at time t is

[0046] Where M represents the number of APs, ρ d represents the normalized signal-to-noise ratio of each data symbol in downlink transmission, η mk represents the power coefficient allocated by APm to user k, η mk′ represents the power coefficient allocated by APm to user k′, ρ k (t-λ) represents the channel correlation coefficient between user k at time t and time λ, f mk (λ) represents the precoding sequence of APm and user k at the λth moment, f mk′ (λ) represents the precoding sequence of APm and user k′ at the λth moment. k (t) represents the data symbol sent by AP to user k at time t, q k′ (t) represents the data symbol sent by AP to user k′ at time t, which satisfies E{|q k (t)| 2}=1 and ω k (t) represents the additive Gaussian noise received by user k at time t. (·) H Represents the matrix conjugate transpose operation.

[0047] Furthermore, using the use-and-then-forget technique, k (t) can be rewritten as follows

[0048] Among them, DS k represents the expected signal strength of user k, BU krepresents the uncertainty of beam gain of user k, CA k Indicates the impact of channel aging on user k, UI kk′ represents the inter-user interference suffered by user k. E(·) and (·) H They represent the expected value operation and the matrix conjugate transpose operation respectively. At this time, the downlink rate achievable by user k at time t is R k (t)

[0049] Among them, L represents the packet length, SINR k (t) represents the signal to interference and noise ratio of user k at the tth moment, ε represents the decoding error probability, Q -1 (·) represents the inverse function of the Q function, which is the right tail function of the standard normal distribution. Ε{·} and |·| represent expectation operation and modulo operation respectively.

[0050] Furthermore, in order to ensure fairness, with the minimum user rate as the objective function, the power optimization problem can be expressed as

[0051] Furthermore, in order to solve this problem, we first introduce the auxiliary variable γ k (t) = SINR k (t), we can get

[0052] The original problem is now transformed into

[0053] Furthermore, for a given initial value and Using the continuous convex approximation method, ln(1+γ k (t)) and It can be approximated as,

[0054] From this we can get, Equivalent to in

[0055] Furthermore, for constrained SINR k (t)≥γ k (t), firstly, the SINR is calculated using the continuous convex approximation method. k (t) The molecule is processed as follows

[0056] in, At this time, for f(x1,x2,…,x M ) is expanded by the first order Taylor, and we can get

[0057] Further, let and substitute and Where m=1,2,…,M, we can get

[0058] Removing the logarithmic function now gives

[0059] in

[0060] Furthermore, the optimization problem is transformed into

[0061] Furthermore, by introducing the auxiliary variable u, the original problem can be converted into the standard form of geometric programming, as follows

[0062] Then use the CVX tool to solve the problem and get the initial solution for the next iteration and Repeat the above steps until convergence to obtain the optimal allocation solution

[0063] Furthermore, the optimal allocation scheme is substituted into the rate formula to calculate the downlink achievable rate for each user. Referring to Figure 1, it can be found that channel aging significantly reduces the average downlink achievable rate. When there is no power optimization, as shown by the solid line, when the number of users K = 10, the average downlink achievable rate is 1.69 bit / s / Hz without channel aging, and 1.23 bit / s / Hz with channel aging, which is a decrease of about 27.22%. Therefore, the impact of channel aging on the system needs to be considered in the actual system design; in addition, as the number of users increases, the competition among users for fixed resources increases, and the average downlink achievable rate gradually decreases. However, the power allocation method proposed in the present invention can significantly improve the average downlink achievable rate, as shown by the dotted line. Similarly, when the number of users K = 10, the present invention improves by about 14.20% compared to the non-optimization case without channel aging, and by 27.64% with channel aging, which fully demonstrates the effectiveness of the present invention in improving system performance and can reduce the impact of channel aging.

[0064] Referring to FIG. 2 , when the number of APs is 100 and the number of users is 30, the present invention only needs about 5 iterations to reach convergence, which effectively illustrates the feasibility of the present invention.

[0065] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A power allocation method based on a decellularized massive MIMO short packet communication system, characterized in that: The following steps are involved: Step 1: Based on a decellularized massive MIMO short packet communication system in a user mobility scenario, a Rayleigh fading channel model under channel aging is established; Step 2: Based on the Rayleigh fading channel model, the user's received data expression during the downlink data transmission phase is given, and the downlink achievable rate is derived from it; The derivation of the downlink achievable rate is specifically: the downlink achievable rate R of user k at time t k (t) is expressed as: Among them, L represents the packet length, SINR k (t) represents the signal to interference and noise ratio of user k at the tth moment, ε represents the decoding error probability, Q -1 (·) represents the inverse function of the Q function, which is the right tail function of the standard normal distribution. E{·} and |·| represent the expectation operation and the modulo operation respectively; Step 3: Based on the downlink achievable rate, with the minimum user rate as the objective function, a maximization optimization problem with power constraints is proposed; The power optimization problem in the downlink data transmission phase is expressed as: Among them, R k (t) represents the downlink rate achievable by user k at time t, η mk represents the power coefficient allocated by APm to user k, E{·} and (·) H They represent the expectation operation and matrix conjugate transpose operation respectively, f mk (λ) represents the precoding sequence of APm and user k at the λth time; Step 4: For the maximization optimization problem, the objective function is converted into an exponential function using a continuous convex approximation method. The constraints are then processed using the continuous convex approximation method again to meet the requirements of the geometric programming algorithm and obtain the optimal power allocation solution. Step 4: Step 4-1: Introduce auxiliary variable γ k (t) = SINR k (t), we can get The original problem is transformed into Step 4-2: Given an initial And calculate the initial value Step 4-3: Use the continuous convex approximation method to approximate ln(1+γ k (t)) and Available, From this we can get, Step 4-4: Use continuous convex approximation again to convert SINR k The numerator of (t) is approximately expressed as in Therefore, the optimization problem is transformed into Step 4-5: Introduce the auxiliary variable u to convert the original problem into the standard form of geometric programming, as follows Then use the CVX tool to solve the problem and get the initial solution for the next iteration Step 4-6: Repeat steps 4-2 to 4-5 until convergence and the optimal allocation solution is obtained.

2. The power allocation method based on the decellularized massive MIMO short packet communication system according to claim 1, characterized in that: In step 1, the Rayleigh fading channel model is expressed as: Among them, h mk (t) represents the channel sequence between APm and user k at the tth moment, h mk (λ) represents the channel sequence between APm and user k at the λth moment, ξ mk (t) represents the random channel sequence between APm and user k at time t and time λ, ρ k (t-λ) represents the channel correlation coefficient between user k at time t and time λ.

3. The power allocation method based on the decellularized massive MIMO short packet communication system according to claim 1, characterized in that: In step 2, the received data of the user in the downlink data transmission phase is expressed as follows: the data received by user k at time t: Where M represents the number of APs, ρ d represents the normalized signal-to-noise ratio of each data symbol in downlink transmission, η mk represents the power coefficient allocated by APm to user k, η mk′ represents the power coefficient allocated by APm to user k′, ρ k (t-λ) represents the channel correlation coefficient between user k at time t and time λ, f mk (λ) represents the precoding sequence of APm and user k at the λth moment, f mk′ (λ) represents the precoding sequence of APm and user k′ at the λth moment, q k (t) represents the data symbol sent by AP to user k at time t, q k′ (t) represents the data symbol sent by AP to user k′ at time t, which satisfies E{|q k (t)| 2 }=1 and ω k (t) represents the additive Gaussian noise received by user k at the tth moment, DS k represents the expected signal strength of user k, BU k represents the uncertainty of beam gain of user k, CA k Indicates the impact of channel aging on user k, UI kk′ represents the inter-user interference suffered by user k, E{·} and (·) H They represent the expectation operation and matrix conjugate transpose operation respectively.

Images