Design method of heterogeneous node coexistence system under multicarrier modulation

Abstract

The invention provides a method for designing a heterogeneous node coexistence system under multicarrier modulation, which belongs to the technical field of semantic communication, and comprises the following steps that: based on a heterogeneous semantic communication and bit communication hybrid uplink communication system, users transmit signals to a base station, the communication system comprises the base station and the users, the users comprise a semantic user and K bit users, the base station is provided with a fixed antenna, each user is provided with a single fluid antenna with T ports, the bit users communicate with the base station by adopting orthogonal multiple access, and the semantic users and all the bit users form power domain non-orthogonal multiple access; the base station decodes the data of the bit users and the data of the semantic users in the received signals by adopting continuous interference cancellation; a joint optimization problem related to power distribution and carrier selection is constructed, and the joint optimization problem is solved based on deep reinforcement learning to maximize the sum rate of semantic users and bit users. According to the method, the semantic transmission rate can be maximized, and meanwhile, the communication quality of BitCom users is ensured.

Description

Technical Field

[0001] The present invention belongs to the technical field of semantic communication, and particularly relates to a design method for a heterogeneous node coexistence system under multi-carrier modulation. Background Art

[0002] With the rapid increase in the number of users and the amount of information in wireless communication networks, the traditional data transmission mode is difficult to meet the growing communication needs. As an emerging technology, semantic communication (SemCom) has the advantage of being able to optimize transmission for specific tasks (such as semantic similarity or image quality) compared to traditional bit communication (BitCom). However, the high computational and storage requirements of SemCom systems make it difficult to completely replace existing BitCom systems. Therefore, how to achieve the efficient coexistence of SemCom and BitCom systems has become the focus of current research.

[0003] Non-orthogonal multiple access (NOMA) technology is considered an effective means to promote the coexistence of SemCom and BitCom because of its ability to support multiple users to transmit signals on the same spectrum resource. However, existing research mainly focuses on fixed antenna devices and cannot fully utilize the spatial degrees of freedom to improve the spectrum efficiency, which makes the coexistence performance of SemCom and BitCom unable to reach the optimal. Summary of the Invention

[0004] The object of the present invention is to propose a design method for a heterogeneous node coexistence system under multi-carrier modulation, which can maximize the semantic transmission rate while ensuring the communication quality of BitCom users.

[0005] The present invention is realized through the following technical solutions:

[0006] A design method for a heterogeneous node coexistence system under multi-carrier modulation includes the following steps:

[0007] Step S1: Based on a heterogeneous semantic communication and bit communication hybrid uplink communication system, users transmit signals to the base station. Among them, the communication system includes a base station and users. The users include one semantic user and K bit users. The base station is equipped with fixed antennas, and each user is equipped with a single-fluid antenna with T ports. The bit users use orthogonal multiple access to communicate with the base station, and the semantic user and all bit users form a power domain non-orthogonal multiple access;

[0008] Step S2: At the base station, successive interference cancellation is used to decode the data of the bit users and the data of the semantic user in the received signal.

[0009] Step S3: Based on the data decoded in Step S2, construct a joint optimization problem involving power allocation and carrier selection, and solve the joint optimization problem based on deep reinforcement learning to maximize the sum rate of the semantic user and the bit user.

[0010] Further, in Step S1, the received signal of the base station on the n-th subcarrier is 1 ≤ n ≤ N, where k = 0 corresponds to the semantic user, 1 ≤ k ≤ K corresponds to the bit user, A k represents the large-scale fading between the base station and the k-th user, a k [n] indicates whether the k-th bit user occupies the n-th subcarrier, h k [n] represents the channel of the k-th user on the n-th subcarrier, P k [n] represents the transmission power of the k-th user on the n-th subcarrier, s k [n] represents the transmission symbol of the k-th user on the n-th subcarrier, z represents Gaussian white noise with a mean of 0 and a variance of σ 2 , and N represents the number of subcarriers.

[0011] Further, in Step S1, it is assumed that there are M k propagation paths between the k-th user and the base station, then the time-domain channel of the k-th user is expressed as Then the channel at the n-th subcarrier is expressed as where B k,t [m] represents the small-scale fading of the m-th path of the k-th user at the t-th port, τ k [m] respectively represent the time delay of the m-th path of the k-th user at the t-th port, C k,t [m] represents the guiding direction of the t-th port, q k [t] indicates whether the k-th user selects the t-th port. If selected, it is 1, otherwise it is 0, δ(τ - τ k [m]) represents the activation function. When τ = τ k [m], δ(τ - τ k [m]) is 1, otherwise it is 0, and τ represents the time delay.

[0012] Further, Step S2 specifically includes the following steps:

[0013] Step S21: Regard the data of the semantic user as interference and decode the data of the bit user. Among them, for the k-th bit user, its SINR on the n-th subcarrier is expressed as If k ≥ 1, the sum rate of the k-th bit user is expressed as k ≥ 1;

[0014] Step S22: Eliminate the data of the bit user from the received signal and perform decoding of the semantic user data. Then, the SINR of the semantic user on the nth subcarrier is expressed as If the data transmission of the semantic user adopts the DeepSC network, the communication rate of the semantic user is expressed as where β n is the residual interference coefficient. The residual interference refers to the data of the bit user that has not been correctly decoded after step S21. S is the average number of semantic symbols of the words of the data of the semantic user, I represents the average semantic information of the words of the data of the semantic user, L represents the average number of words in the sentences of the data of the semantic user, and ζ(γ0[n]) is the semantic similarity function.

[0015] Furthermore, in step S3, the joint optimization problem is expressed as where is defined as the user power allocation coefficient, is defined as the bit user carrier selection variable, is the rate of the bit user evaluated by the semantic rate, μ is the average number of bits of the words of the data of the semantic user, R min is the minimum rate of the bit user, is the minimum rate of the semantic user, P max is the maximum power of the user.

[0016] Furthermore, the deep reinforcement learning in step S3 is the dual PPO algorithm. The dual PPO algorithm involves the CP network and the Power network. Among them, the CP network is used to optimize carrier selection, and the Power network is used to optimize power allocation.

[0017] Furthermore, the state space, action space, and reward function of the dual PPO algorithm in step S3 are respectively:[[]]

[0018] State space: At the i-th step, the shared state space is The state of the CP network is defined as The state of the Power network is defined as where β represents the set of residual interference coefficients, γ represents the SINR of the user on the subcarriers it occupies, is the output action of the Power network at the (i - 1)-th step, is the output action of the CP network at the i-th step;

[0019] Action space: At the i-th step, the output action of the CP network is the result of the bit user selecting subcarriers, expressed as Denote the user number selected for the nth subcarrier, and the output action of the Power network is the normalized power allocation result, expressed as where k = 0, 1,.., K;

[0020] Reward function: The reward function is designed as where and k > 0 is the compensation factor, is an approximation of ζ(γ0[n]), and the coefficients D1 - D4 are obtained by fitting the performance of the DeepSC network through data regression.

[0021] The present invention has the following beneficial effects:

[0022] 1. The users of the present invention transmit signals to the base station based on a hybrid uplink communication system of heterogeneous semantic communication and bit communication. At the base station, successive interference cancellation is used to decode the data of bit users and the data of semantic users in the received signal. Based on the decoded data, a joint optimization problem involving power allocation and subcarrier selection is constructed, and the joint optimization problem is solved based on deep reinforcement learning to maximize the sum rate of semantic users and bit users. By deploying a single-fluid antenna on the user side and dynamically adjusting the antenna port position, the signal-to-interference-plus-noise ratio (SINR) of the system is effectively improved, and the interference between non-orthogonal multiple access users is reduced, thereby improving the overall semantic rate and overall communication performance; and by constructing the optimization problem as described above and solving it based on deep reinforcement learning, the present invention can improve the solution efficiency and system throughput, further enhancing the overall performance of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The present invention will be further described in detail below with reference to the accompanying drawings.

[0024] Figure 1 is a flowchart of the present invention.

[0025] Figure 2 is a schematic diagram of the communication system of the present invention.

[0026] Figure 3 is a schematic diagram of the dual PPO algorithm of the present invention.

[0027] Figure 4 is a schematic diagram of the convergence comparison between the present invention and other benchmark algorithms.

[0028] Figure 5 is a schematic diagram of the relationship between the maximum transmit power of users and the system throughput under different algorithm conditions of the present invention.

[0029] Figure 6 Schematic diagram of the relationship between the number T of fluid antenna ports and the system throughput of the present invention under different algorithm conditions. Detailed implementation manners

[0030] As Figure 1 shown, the design method for a heterogeneous node coexistence system under multi-carrier modulation includes the following steps:

[0031] Step S1, a user transmits a signal to a base station based on a heterogeneous semantic communication and bit communication hybrid uplink communication system. Among them, as Figure 2 shown, the communication system includes a base station and a user. The user includes a semantic user and K bit users. The base station is equipped with a fixed antenna, and each user is equipped with a single-fluid antenna with T ports. The bit users communicate with the base station using orthogonal multiple access, that is, one subcarrier is only assigned to one bit user. The semantic user and all bit users form a power-domain non-orthogonal multiple access, that is, the semantic user multiplexes all subcarriers to communicate with the base station. In addition, to be compatible with existing communication protocols, the present invention uses the CP-OFDM waveform as the modulation scheme.

[0032] Specifically, the received signal of the base station on the nth subcarrier is 1 ≤ n ≤ N, where k = 0 corresponds to the semantic user, and 1 ≤ k ≤ K corresponds to the bit users. A k represents the large-scale fading between the base station and the kth user, expressed as A k = -128 - 37.6logd k , d k is the distance between the control center of the base station and the user k, a k [n] represents whether the kth bit user occupies the nth subcarrier. a k [n] = 1 means that the kth bit user occupies the nth subcarrier, otherwise a k [n] = 0, h k [n] represents the channel of the kth user on the nth subcarrier, P k [n] represents the transmission power of the kth user on the nth subcarrier, s k [n] represents the transmission symbol of the kth user on the nth subcarrier, z represents Gaussian white noise with a mean of 0 and a variance of σ 2 , and N represents the number of subcarriers.

[0033] Assume that there are M k propagation paths between the kth user and the base station. Then the time-domain channel of the kth user is expressed as Then the channel at the nth subcarrier is expressed as Among them, Bk,t [m] represents the small-scale fading of the m-th path at the t-th port for the k-th user, τ k [m] represents the delay of the m-th path at the t-th port for the k-th user respectively. To model the correlation between ports, the small-scale fading B k,t [m] is parameterized as and are independent and identically distributed Gaussian variables with zero mean and unit variance. The parameter controls the correlation between two ports, represents the steering direction of the t-th port, θ k,t [m] represents the AoD (Angel of Departure) from the k-th user to the base station, q k [t] represents whether the k-th user selects the t-th port. If selected, it is 1; otherwise, it is 0. δ(τ - τ k [m]) represents the activation function. When τ = τ k [m], δ(τ - τ k [m]) is 1; otherwise, it is 0. τ represents the delay.

[0034] Step S2: At the base station, successive interference cancellation is used to decode the data of the bit user and the data of the semantic user in the received signal;

[0035] Specifically, it includes the following steps:

[0036] Step S21: Treat the data of the semantic user as interference and decode the data of the bit user. Among them, for the k-th bit user, its SINR on the n-th subcarrier is expressed as If k ≥ 1, the sum rate of the k-th bit user is expressed as k ≥ 1;

[0037] To have a unified performance metric, the present invention uses the semantic rate to evaluate the performance of the bit user, that is where I represents the average semantic information of the words of the data of the semantic user, μ represents the average number of bits of the sentences of the data of the semantic user, and L represents the average number of words of the sentences of the data of the semantic user;

[0038] Step S22: Eliminate the data of the bit user from the received signal and perform decoding of the data of the semantic user. Then the SINR of the semantic user on the n-th subcarrier is expressed as The data transmission of the semantic user uses the DeepSC network, and the communication rate of the semantic user is expressed as where β nis the residual interference coefficient. Residual interference refers to the bit user data that has not been correctly decoded after step S21 (in this embodiment, considering the case where residual interference definitely exists and based on the statistical assumption of the proportion of residual interference), S is the average number of semantic symbols of the words in the semantic user's data, I represents the average semantic information of the words in the semantic user's data, L represents the average number of words in the sentences of the semantic user's data, and ζ(γ0[n]) is the semantic similarity function, which can be approximated as The coefficients D1 - D4 are obtained by fitting the performance of the DeepSC network through data regression.

[0039] Step S3: Based on the data decoded in step S2, construct a joint optimization problem involving power allocation and carrier selection, and solve the joint optimization problem based on deep reinforcement learning to maximize the sum rate of the semantic user and the bit user;

[0040] Specifically, the joint optimization problem is expressed as where is defined as the user power allocation coefficient, is defined as the bit user carrier selection variable, R min is the minimum rate of the bit user, is the minimum rate of the semantic user, P max is the maximum power of the user. That is, the first two constraints represent the minimum rate constraints of the bit user and the semantic user, and the last constraint represents the power constraint of the user.

[0041] In the above joint optimization problem, the variables are highly coupled, and it is difficult to apply traditional convex optimization methods. Moreover, traditional reinforcement learning methods are difficult to solve the optimization problem of the mixed action space. Therefore, the present invention uses the dual PPO algorithm to solve the joint optimization problem. The dual PPO algorithm involves a CP network and a Power network. Among them, the CP network is used to optimize carrier selection, and the Power network is used to optimize power allocation. The state space, action space, and reward function of the dual PPO algorithm are as follows:

[0042] State space: At the i - th step, the shared state space is The state of the CP network is defined as The state of the Power network is defined as where β = {β n} represents the set of residual interference coefficients, represents the SINR of the user on the subcarriers it occupies, is the output action of the Power network at the (i - 1) - th step, is the output action of the CP network at the i - th step;

[0043] Action space: At the i-th step, the output action of the CP network is the carrier selection variable, which is the result of subcarrier selection for bit users. To ensure that each subcarrier is only assigned to one bit user, the present invention transforms the carrier selection problem into a user selection problem, that is, the CP network outputs the result of bit user selection of subcarriers, that is Let denote the user number selected for the n-th subcarrier, then the output of the CP network is expressed as For the Power network, its corresponding output is the power allocation result. To meet the power constraint, the output action of the Power network should be normalized, that is where, k = 0, 1,.., K, {.} represents a set;

[0044] Reward function: Due to the existence of the minimum rate constraint, the reward function is designed as where, and k > 0 is a compensation factor, is an approximation of ζ(γ0[n]), and the coefficients D1 - D4 are obtained by fitting the performance of the DeepSC network through data regression. The schematic diagram of the double PPO algorithm is as Figure 3 shown.

[0045] Figure 4 shows the convergence trends of DPPO (the present invention), PPO, and the random algorithm during training. The abscissa is the number of training steps, and the ordinate is the cumulative reward value. It can be seen that the present invention converges within 10,000 steps, and its final cumulative reward value is significantly higher than that of PPO and the random algorithm, indicating its superiority in optimizing resource allocation.

[0046] Figure 5 shows the schematic diagram of the relationship between the maximum transmit power of users and the system throughput under different algorithm conditions. This figure compares the total throughput that can be achieved by various optimization schemes (DPPO, PPO, orthogonal transmission algorithm, and random algorithm) at different transmission powers. The abscissa is the maximum transmit power (dBm), and the ordinate is the system total throughput (suts / s / Hz). The results show that the NOMA strategy can significantly improve the system throughput, and the NOMA scheme optimized by the DPPO algorithm performs best at all power levels.

[0047] Figure 6The schematic diagram showing the relationship between the number of fluid antenna ports T and the system throughput under different algorithms (DPPO, PPO, orthogonal transmission algorithm, and random algorithm) is presented. This figure shows the influence of different numbers of fluid antenna ports (T) on the system throughput. The abscissa represents the number of antenna ports, and the ordinate represents the total system throughput (suts / s / Hz). The results show that as the number of ports increases, the throughput gradually improves. However, after the number of ports exceeds a certain threshold, the system throughput tends to saturate, indicating that too many antenna ports may lead to diminishing marginal returns.

[0048] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above-described exemplary embodiments, and the present invention can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-restrictive. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be embraced within the present invention. Any reference signs in the claims should not be construed as limiting the claims involved.

Claims

1. A design method for a heterogeneous node coexistence system under multi-carrier modulation, characterized in that: It includes the following steps: Step S1: Based on a hybrid uplink communication system of heterogeneous semantic communication and bit communication, a user transmits a signal to a base station. The communication system includes a base station and a user. The user includes a semantic user and K bit users. The base station is equipped with fixed antennas, and each user is equipped with a single-fluid antenna with T ports. The bit users communicate with the base station using orthogonal multiple access, and the semantic user and all bit users form a power-domain non-orthogonal multiple access; Step S2: At the base station, successive interference cancellation is used to decode the data of the bit users and the data of the semantic user in the received signal; Step S3: Based on the data decoded in Step S2, a joint optimization problem involving power allocation and carrier selection is constructed, and the joint optimization problem is solved based on deep reinforcement learning to maximize the sum rate of the semantic user and the bit users.

2. A design method for a heterogeneous node coexistence system under multi-carrier modulation according to claim 1, characterized in that: In the step S1, the received signal of the base station on the n-th subcarrier is where k = 0 corresponds to the semantic user, 1 ≤ k ≤ K corresponds to the bit user, A k represents the large-scale fading between the base station and the k-th user, a k [n] indicates whether the k-th bit user occupies the n-th subcarrier, h k [n] represents the channel of the k-th user on the n-th subcarrier, P k [n] represents the transmit power of the k-th user on the n-th subcarrier, s k [n] represents the transmission symbol of the k-th user on the n-th subcarrier, z represents the Gaussian white noise with a mean of 0 and a variance of σ 2 , and N represents the number of subcarriers.

3. A design method for a heterogeneous node coexistence system under multi-carrier modulation according to claim 2, characterized in that: In the step S1, it is assumed that there are M k propagation paths between the k-th user and the base station, then the time-domain channel of the k-th user is expressed as Then the channel representation at the n-th subcarrier is where B k,t [m] represents the small-scale fading of the m-th path of the k-th user at the t-th port, and τ k [m] respectively represent the time delay of the m-th path of the k-th user at the t-th port, C k,t [m] represents the guiding direction of the t-th port, q k [t] represents whether the k-th user selects the t-th port. If it is selected, it is 1, otherwise it is 0. δ(τ - τ k [m]) represents the activation function. When τ = τ k [m], δ(τ - τ k [m]) is 1, otherwise it is 0, and τ represents the time delay.

4. A design method for a heterogeneous node coexistence system under multi-carrier modulation according to claim 3, characterized in that: The specific steps of Step S2 include the following steps: Step S21: Treat the data of the semantic user as interference and decode the data of the bit user. Among them, for the k-th bit user, its SINR on the n-th subcarrier is expressed as Then the sum rate of the k-th bit user is expressed as Step S22: Eliminate the data of the bit user from the received signal and perform decoding of the semantic user data. Then, the SINR of the semantic user on the n-th subcarrier is expressed as If the data transmission of the semantic user adopts the DeepSC network, the communication rate of the semantic user is expressed as where β n is the residual interference coefficient. The residual interference refers to the data of the bit user that has not been correctly decoded after step S21. S is the average number of semantic symbols of the words in the data of the semantic user, I represents the average semantic information of the words in the data of the semantic user, L represents the average number of words in the sentences of the data of the semantic user, and ζ(γ0[n]) is the semantic similarity function.

5. A design method for a heterogeneous node coexistence system under multi-carrier modulation according to claim 4, characterized in that: In Step S3, the joint optimization problem is expressed as Among them, is defined as the user power allocation coefficient, is defined as the bit user carrier selection variable, is the bit user rate evaluated by the semantic rate, μ is the average number of bits of the words of the semantic user's data, R min is the minimum rate of the bit user, is the minimum rate of the semantic user, P max is the maximum power of the user.

6. A design method for a heterogeneous node coexistence system under multi-carrier modulation, as claimed in claim 5, wherein: The deep reinforcement learning in Step S3 is the dual PPO algorithm. The dual PPO algorithm involves a CP network and a Power network. The CP network is used to optimize carrier selection, and the Power network is used to optimize power allocation.

7. A design method for a heterogeneous node coexistence system under multi-carrier modulation according to claim 6, characterized in that: In Step S3, the state space, action space, and reward function of the dual PPO algorithm are respectively: State space: At the $i$-th step, the shared state space is The state of the CP network is defined as The state of the Power network is defined as where $\beta$ represents the set of residual interference coefficients, $\gamma$ represents the SINR of the user on the subcarriers it occupies, is the output action of the Power network at the $(i - 1)$-th step, is the output action of the CP network at the $i$-th step; Action space: At the i-th step, the output action of the CP network The result of subcarrier selection for bit users, denoted as Denote the user number selected by the n-th subcarrier, the output action of the Power network The power allocation result after normalization, denoted as where Reward function: The reward function is designed as where and are compensation factors, is an approximation of ζ(γ0[n]), and the coefficients D1 - D4 are obtained by fitting the performance of the DeepSC network through data regression.

Images