High-speed train non-cellular MIMO energy-efficient transmission method, device and storage medium

By constructing a Ricean channel model and a dynamic cooperative clustering mechanism, combined with a multi-agent Markov decision process and a deep deterministic policy gradient algorithm, the real-time and dynamic problems of resource allocation in high-speed train communication are solved, and energy efficiency and spectral efficiency are optimized, making it suitable for green and intelligent transmission in high-speed train communication.

CN123074535APending Publication Date: 2026-11-10BEIJING UNIV OF POSTS & TELECOMM
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611320715.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-28
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing non-cellular massive MIMO resource allocation schemes are insufficient to meet the real-time and dynamic requirements of high-speed train communication scenarios, cannot adapt to the time-varying characteristics of channels brought about by high mobility, and have high computational complexity, making it impossible to optimize energy efficiency and spectral efficiency.

Method used

A Ricean channel model is constructed, employing a dynamic cooperative clustering mechanism and a multi-agent Markov decision process. Power allocation is performed in conjunction with a multi-agent deep deterministic policy gradient algorithm to optimize system energy efficiency, and interference is suppressed through a linear reception algorithm.

Benefits of technology

It achieves an optimal trade-off between energy efficiency and spectral efficiency in high-speed train communication, reduces antenna power variance, adapts to time-varying channel characteristics, and provides a green, intelligent, and reliable transmission solution.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123074535A_ABST
    Figure CN123074535A_ABST
Patent Text Reader

Abstract

A method, apparatus, and storage medium for high-speed train non-cellular MIMO energy-efficient transmission are disclosed. The method includes: constructing a Ricean channel model containing line-of-sight and non-line-of-sight components; determining the service access point subset corresponding to each transmit antenna based on a dynamic cooperative clustering mechanism centered on the transmit antenna, and forming a dynamic cooperative clustering result; modeling the joint optimization problem of the dynamic cooperative clustering result and power allocation as a multi-agent Markov decision process based on the Ricean channel model, with the optimization objective being to maximize system energy efficiency and the constraint being a maximum transmit power limit; solving the multi-agent Markov decision process using a multi-agent deep deterministic policy gradient algorithm to obtain the power allocation strategy for each transmit antenna and outputting the corresponding transmit power; and performing interference processing on the received antenna signals based on a linear receiving algorithm to obtain the processed received signal. This method can adapt to the time-varying channel characteristics brought about by high mobility and large-scale OFDM system deployment.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the field of wireless communication technology, specifically to a method, apparatus, and storage medium for high-speed train non-cellular MIMO energy-efficient transmission. Background Technology

[0002] Cell-free (CF) massive MIMO (Multiple-Input Multiple-Output) technology, as an innovative solution to traditional cellular architecture, deploys a large number of distributed access points (APs) connected to a central processing unit to achieve collaborative services for user equipment. It combines the macro-diversity gain of a distributed architecture with the antenna array gain of massive MIMO, offering advantages such as high throughput, low latency, and high reliability. Since its introduction in 2017, it has become one of the key technologies for 5G and Beyond-5G (B5G) communication.

[0003] High-speed train communication (HST), as a core scenario of railway communication, faces unique challenges brought about by high mobility, such as Doppler Frequency Offset (DFO), Inter-Carrier Interference (ICI), and channel non-stationarity. Existing research has seen some schemes simplify channel representation in HST scenarios using line-of-sight (LoS) channel models, and others designing algorithms such as fractional power control and maximum and spectral efficiency power control to optimize resource allocation. However, most of these schemes assume static scenarios or ignore non-line-of-sight (NLoS) components, making them difficult to adapt to the complex and variable propagation environments of HST (such as tunnel and plain composite scenarios).

[0004] Meanwhile, with the exponential growth in the number of communication devices and the surge in wireless traffic, energy efficiency has become a performance indicator as important as spectrum efficiency. Existing resource allocation schemes for non-cellular massive MIMO mostly focus on spectrum efficiency optimization or use traditional optimization methods to solve the energy efficiency maximization problem. These schemes suffer from high computational complexity and difficulty in handling some observable and continuous action spaces, thus failing to meet the real-time and dynamic requirements of HST scenarios. Summary of the Invention

[0005] At least one embodiment of this application provides a method, apparatus, and storage medium for high-speed train non-cellular MIMO energy-efficient transmission, which solves the problem that existing resource allocation and energy-efficient transmission cannot meet the real-time and dynamic requirements of HST scenarios.

[0006] To solve the above-mentioned technical problems, this application is implemented as follows:

[0007] One embodiment of this application provides a method for energy-efficient, non-cellular MIMO transmission on high-speed trains, including:

[0008] A Ricean channel model is constructed, which includes line-of-sight and non-line-of-sight components, to characterize the time-varying channel characteristics in the high-speed train scenario. The time-varying channel characteristics include Doppler frequency offset and inter-carrier interference.

[0009] Based on the dynamic collaborative clustering mechanism of the transmit antenna center, the service access point subset corresponding to each transmit antenna is determined, and the dynamic collaborative clustering result is formed.

[0010] Based on the Rice channel model, the joint optimization problem of the dynamic cooperative clustering results and power allocation is modeled as a multi-agent Markov decision process, wherein the optimization objective is to maximize the system energy efficiency, and the constraint is the maximum transmit power limit of each transmit antenna.

[0011] The multi-agent Markov decision process is solved using the Multi-Agent Deep Deterministic Policy Gradient (MADDPG) algorithm to obtain the power allocation strategy for each transmit antenna, and the corresponding transmit power is output to each transmit antenna so that each transmit antenna transmits a signal according to the corresponding transmit power.

[0012] The received antenna signals transmitted by each of the transmitting antennas are subjected to interference processing based on a linear receiving algorithm to obtain the processed received signals.

[0013] Preferably, in the high-speed train cellular-free MIMO energy-efficient transmission method described above, the construction of the Ricean channel model includes:

[0014] The line-of-sight component and the non-line-of-sight component are determined based on scene parameters;

[0015] The channel gain expression is determined based on the line-of-sight component and the non-line-of-sight component;

[0016] The Doppler frequency offset effect is determined based on the carrier frequency, train speed, Orthogonal Frequency Division Multiplexing (OFDM) symbol duration, and the angle of arrival of each transmitting antenna and the target receiving point.

[0017] Based on the line-of-sight component and the non-line-of-sight component, an inter-carrier interference characterization is determined, which includes deterministic inter-carrier interference and statistically averaged inter-carrier interference.

[0018] Preferably, in the high-speed train cellular-free MIMO energy-efficient transmission method described above, the dynamic cooperative clustering mechanism based on the center of the transmitting antenna, which determines the service access point subset corresponding to each transmitting antenna and forms the dynamic cooperative clustering result, includes:

[0019] The cooperative relationship between each service access point and each transmit antenna is determined based on the large-scale fading coefficient and a preset threshold.

[0020] Based on the cooperative relationship, the subset of service access points corresponding to each of the transmitting antennas is determined, and a dynamic cooperative clustering result is formed.

[0021] Preferably, in the high-speed train non-cellular MIMO energy-efficient transmission method described above, during the multi-agent Markov decision-making process, the state space is the aggregated large-scale fading coefficient set of each of the transmitting antennas and the corresponding service access point; the action space is the set of the transmit power coefficients of each of the transmitting antennas; and the reward function is the system energy efficiency.

[0022] Preferably, the high-speed train cellular-free MIMO energy-efficient transmission method described above obtains the power allocation strategy for each transmit antenna by solving the multi-agent Markov decision process using a multi-agent deep deterministic policy gradient algorithm, including:

[0023] Initialize the policy-value network architecture parameters corresponding to each of the aforementioned transmit antennas;

[0024] Randomly sample small batches of training samples from the interaction data between the transmitting antenna and the environment already stored in the experience playback pool;

[0025] The policy-value network architecture is trained based on the mini-batch training samples, and the network parameters in the policy-value network architecture are softly updated.

[0026] After training is completed, the policy-value network architecture obtains the power allocation policy corresponding to each transmit antenna based on the large-scale fading coefficient at the current moment.

[0027] Preferably, in the high-speed train non-cellular MIMO energy-efficient transmission method described above, the step of performing interference processing on the antenna signals transmitted by each of the transmitted antennas based on a linear receiving algorithm to obtain the processed received signal includes:

[0028] The target linear receiving algorithm is determined from among the multiple linear receiving algorithms based on complexity requirements and / or spectral efficiency;

[0029] According to the target linear receiving algorithm, the merging vector corresponding to each of the transmitting antennas is obtained;

[0030] The received antenna signals are combined according to the merging vector to obtain the processed received signal.

[0031] Another embodiment of this application provides a control device for high-speed train non-cellular MIMO energy-efficient transmission, including:

[0032] The first processing module is used to construct a Ricean channel model, which includes line-of-sight components and non-line-of-sight components, and is used to characterize the time-varying channel characteristics in the high-speed train scenario. The time-varying channel characteristics include Doppler frequency offset and inter-carrier interference.

[0033] The second processing module is used to determine the service access point subset corresponding to each transmitting antenna based on the dynamic collaborative clustering mechanism of the transmitting antenna center, and to form the dynamic collaborative clustering result.

[0034] The third processing module is used to model the joint optimization problem of the dynamic cooperative clustering results and power allocation as a multi-agent Markov decision process based on the Ricean channel model, wherein the optimization objective is to maximize the system energy efficiency and the constraint is the maximum transmit power limit of each transmit antenna.

[0035] The fourth processing module is used to solve the multi-agent Markov decision process according to the multi-agent deep deterministic policy gradient algorithm, obtain the power allocation strategy of each transmitting antenna, and output the corresponding transmitting power to each transmitting antenna so that each transmitting antenna transmits a signal according to the corresponding transmitting power.

[0036] The fifth processing module is used to perform interference processing on the antenna signals transmitted by each of the transmitting antennas based on a linear receiving algorithm to obtain the processed received signal.

[0037] Another embodiment of this application provides an electronic device, including: a processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of the high-speed train non-cellular MIMO energy-efficient transmission method as described above.

[0038] Another embodiment of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the high-speed train non-cellular MIMO energy-efficient transmission method as described above.

[0039] Another embodiment of this application provides a computer program product including computer instructions that, when executed by a processor, implement the steps of the high-speed train non-cellular MIMO energy-efficient transmission method as described above.

[0040] Compared with existing technologies, the high-speed train non-cellular MIMO energy-efficient transmission method, device and storage medium provided in this application, through accurate channel modeling, dynamic cooperative clustering, intelligent resource allocation and effective interference suppression, can adapt to the time-varying channel characteristics brought about by the high mobility of HST. While reducing antenna power variance, it achieves the optimal trade-off between energy efficiency and spectral efficiency. Moreover, the algorithm running time increases linearly with the number of subcarriers, making it suitable for large-scale OFDM system deployment, and providing a green, intelligent and reliable transmission solution for high-speed train communication. Attached Figure Description

[0041] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0042] Figure 1 This is one of the flowcharts illustrating the high-speed train non-cellular MIMO energy-efficient transmission method of this application;

[0043] Figure 2 This is the second schematic diagram of the high-speed train non-cellular MIMO energy-efficient transmission method of this application;

[0044] Figure 3 This is the third flowchart illustrating the high-speed train non-cellular MIMO energy-efficient transmission method of this application;

[0045] Figure 4 This is the fourth flowchart illustrating the high-speed train non-cellular MIMO energy-efficient transmission method of this application;

[0046] Figure 5 A schematic diagram of the training process for the policy-value network architecture;

[0047] Figure 6 This is the fifth flowchart illustrating the high-speed train non-cellular MIMO energy-efficient transmission method of this application;

[0048] Figure 7This is a comparative diagram showing the performance of different power control algorithms when using minimum mean square error reception in mixed and complex scenarios.

[0049] Figure 8 A diagram illustrating the comparison of energy efficiency under different scenarios;

[0050] Figure 9 This is the sixth flowchart illustrating the high-speed train non-cellular MIMO energy-efficient transmission method of this application;

[0051] Figure 10 This is a schematic diagram of the control device for high-speed train cellular-free MIMO energy-efficient transmission according to this application.

[0052] Figure 11 This is a schematic diagram of the electronic device of this application. Detailed Implementation

[0053] Exemplary embodiments of the present application will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this application will be thorough and complete, and will fully convey the scope of the present application to those skilled in the art.

[0054] The terms “first,” “second,” etc., used in the specification and claims of this application 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 where appropriate so that the embodiments of this application described herein can be implemented, for example, in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus. The terms “and / or” in the specification and claims indicate at least one of the connected objects.

[0055] The technologies described in this document are not limited to NR systems and Long Time Evolution (LTE) / LTE-Advanced (LTE-A) systems, but can also be used in various wireless communication systems such as Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), Orthogonal Frequency Division Multiple Access (OFDMA), Single-carrier Frequency-Division Multiple Access (SC-FDMA), and other systems. The terms "system" and "network" are often used interchangeably. CDMA systems can implement radio technologies such as CDMA2000 and Universal Terrestrial Radio Access (UTRA). UTRA includes Wideband Code Division Multiple Access (WCDMA) and other CDMA variants. TDMA systems can implement radio technologies such as the Global System for Mobile Communication (GSM). OFDMA systems can implement radio technologies such as Ultra Mobile Broadband (UMB), Evolution-UTRA (E-UTRA), IEEE 802.21 (Wi-Fi), IEEE 802.16 (WiMAX), IEEE 802.20, and Flash-OFDM. UTRA and E-UTRA are part of the Universal Mobile Telecommunications System (UMTS). LTE and more advanced LTE (such as LTE-A) are newer versions of UMTS that use E-UTRA. UTRA, E-UTRA, UMTS, LTE, LTE-A, and GSM are described in documents from an organization called the 3rd Generation Partnership Project (3GPP). CDMA2000 and UMB are described in documents from an organization called 3rd Generation Partnership Project 2 (3GPP2).The techniques described herein can be used with the systems and radio technologies mentioned above, as well as with other systems and radio technologies. However, the following description describes NR systems for illustrative purposes, and NR terminology is used in most of the following description, although these techniques can also be applied to applications beyond NR systems.

[0056] Please see Figure 1 This application provides an embodiment of a high-speed train cellular-free MIMO energy-efficient transmission method, comprising:

[0057] Step S101: Construct a Ricean channel model, which includes line-of-sight and non-line-of-sight components, to characterize the time-varying channel characteristics in the high-speed train scenario. The time-varying channel characteristics include Doppler frequency offset and inter-carrier interference.

[0058] Step S102: Based on the dynamic collaborative clustering mechanism of the transmitting antenna center, determine the service access point subset corresponding to each transmitting antenna and form the dynamic collaborative clustering result;

[0059] Step S103: Based on the Rice channel model, the joint optimization problem of the dynamic cooperative clustering results and power allocation is modeled as a multi-agent Markov decision process, wherein the optimization objective is to maximize the system energy efficiency and the constraint is the maximum transmit power limit of each transmit antenna.

[0060] Step S104: Solve the multi-agent Markov decision process according to the multi-agent deep deterministic policy gradient algorithm to obtain the power allocation strategy of each transmitting antenna, and output the corresponding transmitting power to each transmitting antenna so that each transmitting antenna transmits the signal according to the corresponding transmitting power;

[0061] Step S105: Based on the linear receiving algorithm, the received antenna signals transmitted by each of the transmitting antennas are subjected to interference processing to obtain the processed received signals.

[0062] In this embodiment, since the propagation environment of the HST scenario includes plains, tunnels, and composite scenarios, and there are significant non-line-of-sight components in the tunnel scenario, the Ricean channel model constructed based on the propagation characteristics of line-of-sight and non-line-of-sight components is used to characterize the time-varying channel characteristics in the high-speed train scenario. This model can adapt to various HST propagation environments such as plains, tunnels, and composite scenarios, and improve the accuracy of channel estimation and system performance evaluation. The time-varying channel characteristics include Doppler frequency offset and inter-carrier interference.

[0063] Then, considering that the number of access points in high-speed train communication scenarios is much greater than the number of transmitting antennas, and that the distance between access points and transmitting antennas changes rapidly due to the high-speed movement of the train, a dynamic collaborative clustering mechanism based on the center of the transmitting antenna is adopted to determine the service access point subset corresponding to each transmitting antenna and form a dynamic collaborative clustering result. Only access points that are close to the transmitting antenna and whose large-scale fading coefficient meets a preset threshold are selected to be included in the service access point subset. Service APs are filtered based on real-time large-scale fading information, reducing collaborative overhead and computational complexity while ensuring the clustering optimization effect and adapting to the high mobility of HST. In a specific embodiment, a diagonal matrix can be defined. This is used to represent the cooperative relationship between the access point and the transmitting antenna, where... , ;

[0064]

[0065] in A subset of service access points that provide services to the i-th transmitting antenna. It is an N-order identity matrix (N is the number of antennas for each AP). It is an N-order zero matrix.

[0066] In one specific embodiment, the selection criterion for cooperative access points is: based on large-scale fading coefficient. (in For reference distances, such as path loss at 1km, For the i-th transmitting antenna and the i-th transmitting antenna Euclidean distance between access points (Path loss exponent), only retain Access points exceeding a preset threshold are included in the subset of serving access points corresponding to the i-th transmitting antenna, ensuring channel quality between the cooperating access point and the transmitting antenna.

[0067] In another embodiment, the cooperative access point screening threshold is set to 30% of the maximum large-scale fading coefficient, i.e., only those access points are retained. ≥0.3×max( The access point of the i-th transmit antenna is included in the service access point subset. For example, when the transmitting antenna is located at the midpoint of the track, its service access point subset contains the 3-4 nearest access points; when the transmitting antenna moves to the end point of the track, the service access point subset is dynamically updated to the access points closer to the end point, ensuring that the channel quality between the service access points and the transmitting antenna always meets the requirements.

[0068] Then, based on the Ricean channel model, the joint optimization problem of the dynamic cooperative clustering results and power allocation is modeled as a multi-agent Markov decision process. Each transmitting antenna is treated as an independent agent, and the optimization objective is to maximize the system energy efficiency, with the constraint being the maximum transmit power limit of each transmitting antenna. Specifically, the system energy efficiency is defined as the ratio of total spectral efficiency to total power consumption, where total power consumption includes the transmit power consumption of the transmitting antennas and the baseband processing power consumption of the central processing unit (CPU). By modeling the joint optimization problem as a multi-agent Markov decision process, distributed reinforcement learning can replace centralized analytical optimization. This solves the real-time solution challenge of high-dimensional coupling problems while maintaining the realism of the physical model, and it can adapt to the dynamically changing wireless environment of high-speed trains.

[0069] Then, the multi-agent Markov decision process is solved using the multi-agent deep deterministic policy gradient algorithm to obtain the optimal power allocation strategy for each transmit antenna through interactive learning with the environment, and outputs the corresponding transmit power to each transmit antenna so that each transmit antenna transmits signals according to the corresponding transmit power. This enables adaptive optimization in a partially observable, continuous action space, achieving an optimal trade-off between energy efficiency and spectral efficiency while reducing computational complexity, such as reducing antenna power variance. At the same time, the algorithm's running time increases linearly with the number of subcarriers, making it suitable for large-scale OFDM system deployments.

[0070] Furthermore, interference processing is performed on the antenna signals transmitted by each of the transmitting antennas based on a linear receiving algorithm to obtain the processed received signal, which further suppresses inter-carrier interference and helps reduce computational complexity and improve real-time channel tracking capability.

[0071] In summary, this application, through accurate channel modeling, dynamic cooperative clustering, intelligent resource allocation, and effective interference suppression, can adapt to the time-varying channel characteristics brought about by the high mobility of HST. While reducing antenna power variance, it achieves an optimal trade-off between energy efficiency and spectral efficiency. Furthermore, the multi-agent deep deterministic policy gradient algorithm has a runtime that increases linearly with the number of subcarriers, making it suitable for large-scale OFDM system deployments. This provides a green, intelligent, and reliable transmission solution for high-speed train communication.

[0072] See Figure 2 Preferably, in the high-speed train cellular-free MIMO energy-efficient transmission method described above, the construction of the Ricean channel model includes:

[0073] Step S201: Determine the line-of-sight component and the non-line-of-sight component based on scene parameters;

[0074] Step S202: Determine the channel gain expression based on the line-of-sight component and the non-line-of-sight component;

[0075] Step S203: Determine the Doppler frequency offset effect based on the carrier frequency, train speed, orthogonal frequency division multiplexing symbol duration, and angle of arrival of each transmitting antenna and the target receiving point;

[0076] Step S204: Determine the inter-carrier interference characterization based on the line-of-sight component and the non-line-of-sight component. The inter-carrier interference characterization includes deterministic inter-carrier interference and statistically averaged inter-carrier interference.

[0077] In this embodiment, when constructing the Ricean channel model, firstly, the corresponding line-of-sight components and non-line-of-sight components are determined according to the scenario parameters. To facilitate understanding by those skilled in the art, in one embodiment, the simulation scenario can be set as follows: the HST travels at a speed of 300 km / h, the track length is 700 m, access points are deployed at equal intervals along one side of the track, the number is L=10, and each access point is equipped with N antennas (N=8); the antennas are evenly distributed on the HST, the number is K=8, and the distance between adjacent antennas is 25 m; the carrier frequency is f=1.8 GHz, the bandwidth is B=20 MHz, and the total number of subcarriers is... Each coherent block contains M=8 subcarriers; the path loss exponent α=3, and the path loss at a reference distance of 1km. Rice factor is 20dB in plains scenes and 10dB in tunnel scenes; maximum transmit power. noise power .

[0078] In the above simulation scenario, the line-of-sight component can be expressed as:

[0079]

[0080] in, For the first The transmitting antenna and the first Large-scale fading coefficient of line-of-sight components between access points Rice factor, The total large-scale fading coefficient; For the line-of-sight component, random phase, and ; ( (where λ is the carrier wavelength) represents the normalized antenna spacing. Represents the imaginary unit; For the first The transmitting antenna and the first The angle of arrival of each access point, and , .

[0081] No. The transmitting antenna and the first Spatial covariance matrix of non-line-of-sight components among access points Represented as:

[0082]

[0083] in, Spatial covariance matrix The Line 1 Column elements, These are the service access point antenna indices; For large-scale fading coefficients of non-line-of-sight components; This represents the number of scattering clusters. Let AoA be the angle of arrival of the m-th scattering cluster, and ; For angular extension, in one embodiment .

[0084] Then, based on the obtained line-of-sight and non-line-of-sight components, the channel gain expression of the Ricean channel model is determined, i.e. ,in The line-of-sight component has a known amplitude and a uniformly distributed phase. For non-line-of-sight components, This is the spatial correlation covariance matrix of the non-line-of-sight components, used to describe the spatial correlation of the scattering path.

[0085] Furthermore, this can be determined based on the carrier frequency f (e.g., 1.8 GHz) and train speed. Orthogonal Frequency Division Multiplexing Symbol Duration and the angle of arrival of each transmitting antenna and the target receiving point. Determine the Doppler frequency shift effect Specifically, it can be expressed as: ,in, The speed of light;

[0086] Furthermore, since the Doppler frequency offset effect can cause inter-carrier interference (ICI) between adjacent subcarriers, III can be characterized based on line-of-sight (LAS) and non-LAS components. III characterization includes deterministic III and statistically averaged III. Specifically, deterministic III is caused by the LAS component, and its characteristics can be further defined by... Statistically averaged inter-carrier interference is caused by non-line-of-sight components, which can be specifically identified through... Characterization, where M is the number of subcarriers, In one embodiment, the normalized maximum Doppler frequency offset factor is... ,in, For carrier frequency, For train speed, OFDM symbol duration It is the speed of light. At this time, each service access point... The total received signal can be represented in the frequency domain as:

[0087]

[0088] in, For service access point In the Received signals on each subcarrier; For the number of subcarriers, For the first The actual transmission power of each transmitting antenna; This refers to deterministic inter-carrier interference corresponding to the line-of-sight component; For the line-of-sight component; This refers to the statistically averaged inter-carrier interference corresponding to the non-line-of-sight components; This is the non-line-of-sight component; For the first The transmitting antenna is at the 1st Signals on each subcarrier; For service access point In the Additive white Gaussian noise on each subcarrier.

[0089] See Figure 3 Preferably, in the high-speed train non-cellular MIMO energy-efficient transmission method described above, the dynamic cooperative clustering mechanism based on the center of the transmitting antenna, which determines the service access point subset corresponding to each transmitting antenna and forms the dynamic cooperative clustering result, includes:

[0090] Step S301: Determine the cooperative relationship between each of the service access points and each of the transmit antennas based on the large-scale fading coefficient and a preset threshold.

[0091] Step S302: Determine the service access point subset corresponding to each of the transmitting antennas based on the cooperation relationship, and form a dynamic cooperative clustering result.

[0092] In this embodiment, when performing dynamic collaborative clustering, considering that the number of service access points in the HST scenario is much greater than the number of transmitting antennas, and that the high-speed movement of the train causes rapid changes in the distance between the service access points and the transmitting antennas, a dynamic collaborative clustering mechanism centered on the transmitting antennas is adopted. Only access points that are close to the transmitting antennas and whose large-scale fading coefficients meet the preset threshold are selected as the service access point subset, which can effectively reduce the collaborative overhead and computational complexity. Specifically:

[0093] Define a diagonal selection matrix Used to identify the cooperative relationship between the access point and the transmitting antenna, where, , ;

[0094]

[0095] in For the first A subset of service access points provided by each transmitting antenna It is an N-order identity matrix (N is the number of antennas for each service access point). It is an N-order zero matrix.

[0096] The service access point selection criteria are as follows: based on large-scale fading coefficient. (in For reference, path loss at a distance of 1km, Let AP and TA be the Euclidean distance. The system uses the path loss index and a preset threshold to determine which values ​​are retained. Service access points exceeding the preset threshold enter This ensures channel quality between the service access point and the transmitting antenna.

[0097] In one specific embodiment, the service access point filtering threshold is set to 30% of the maximum large-scale fading coefficient, that is, only those service access points are retained. Service access points included in the first A subset of service access points for each transmitting antenna For example, when the transmitting antenna is located at the midpoint of the orbit, its serving access point subset... Includes the 3-4 nearest service access points; a subset of service access points is included when the transmitting antenna moves to the orbital endpoint. The service access point is dynamically updated to be closer to the endpoint, ensuring that the channel quality between the service access point and the transmitting antenna always meets the requirements.

[0098] Preferably, in the high-speed train non-cellular MIMO energy-efficient transmission method described above, during the multi-agent Markov decision-making process, the state space is the aggregated large-scale fading coefficient set of each of the transmitting antennas and the corresponding service access point; the action space is the set of the transmit power coefficients of each of the transmitting antennas; and the reward function is the system energy efficiency.

[0099] In this embodiment, to facilitate understanding by those skilled in the art, some parameters in the multi-agent Markov decision-making process are illustrated, wherein the state space S: , defined as the aggregated large-scale fading coefficient set between each transmit antenna and the serving access point, where , For the first The transmitting antenna and the first The large-scale fading coefficient between service access points is used to reflect the long-term statistical characteristics of the channel.

[0100] Action space A: defined as the set of transmit power coefficients for each transmit antenna, i.e. ,in For the first The power coefficient of each transmitting antenna at time t, with a value ranging from [0,1], corresponds to the actual transmitting power. ( (Maximum transmit power);

[0101] The reward function R is defined as the system's energy efficiency (EE), i.e. ,in: For the first The spectral efficiency of a transmit antenna on the s-th subcarrier is determined by the type of receive algorithm (e.g., maximum ratio combining (MR) or minimum mean square error (LP-MMSE)). For the first The transmit power consumption of each transmitting antenna, including antenna circuit power and transmit power; The baseband processing power consumption of the Central Processing Unit (CPU) is expressed as follows: ,in, This refers to the CPU's static power. and These are the encoding and decoding power densities, respectively. B represents the baseband processing complexity, and B represents the bandwidth. For system load;

[0102] The constraints are: , ,in, Maximum transmit power limit for each transmit antenna. This is the transmitted signal vector; It is the conjugate transpose; This refers to the transmission power.

[0103] The optimization objective is to maximize the system's energy efficiency. .

[0104] See Figure 4 Preferably, the high-speed train cellular-free MIMO energy-efficient transmission method described above, wherein the power allocation strategy for each transmit antenna is obtained by solving the multi-agent Markov decision process using a multi-agent deep deterministic policy gradient algorithm, including:

[0105] Step S401: Initialize the policy-value network architecture parameters corresponding to each of the transmitting antennas;

[0106] Step S402: Randomly sample a small batch of training samples from the interaction data between the transmitting antenna and the environment stored in the experience playback pool;

[0107] Step S403: Train the policy-value network architecture based on the mini-batch training samples, and perform soft updates on the network parameters in the policy-value network architecture.

[0108] Step S404: After training is completed, the policy-value network architecture obtains the power allocation policy corresponding to each transmit antenna based on the large-scale fading coefficient at the current moment.

[0109] When solving multi-agent Markov decision processes using the deep deterministic policy gradient algorithm, each transmitting antenna is abstracted as an agent, and a policy-value (Actor-Critic) network architecture is constructed. Each agent contains an Actor network and a Critic network. The Actor network takes the local observation state (e.g., aggregating large-scale fading coefficients) as input and outputs a deterministic power allocation policy. The Critic network takes the global state and the actions of all agents as input and outputs an action-value function to evaluate the decision quality of the Actor network. An experience replay pool is also included, containing samples generated from interactions between agents and the environment. These samples will be stored in the experience replay pool to facilitate random sampling of small batches of training samples during training, reducing sample correlation and improving training stability. The aggregated large-scale fading coefficient set between each transmitting antenna and the serving access point at time t; Let be the set of transmit power coefficients for each transmitting antenna at time t; Let be the system energy efficiency at time t; It is the aggregated large-scale fading coefficient set between each transmitting antenna and the serving access point at time t+1.

[0110] In the specific calculation, initialization is first performed, that is, the parameters of the Actor and Critic networks are randomly initialized, and the target network parameters are consistent with the current network parameters. The input of the agent is the channel state observed by the antenna, and the output is the output power of the antenna. In each round, a small batch of training samples is randomly sampled from the interaction data between the transmitting antenna and the environment stored in the experience replay pool, and the Actor network and Critic network are trained based on the small batch of training samples. Specifically, it can be as follows:

[0111] Critic network training: Calculating the target Q-value based on mini-batch training samples ,in Given the target Critic network parameters, minimize the loss function. Update the Critic network parameters;

[0112] Actor network training: Based on the output gradient of the Critic network, the Actor network parameters are updated through policy gradient ascent, and the cumulative reward is maximized;

[0113] Target network update: The target Actor and target Critic network parameters are updated using a soft update mechanism, i.e. , ,in, For the policy-value network parameter weights, The target network parameters are weighted; in one embodiment, the target network is updated every 10 rounds.

[0114] After training convergence, the Actor network interface outputs the optimal power allocation strategy for each transmit antenna based on real-time observed large-scale fading coefficients. The convergence determination method includes, but is not limited to, determining training convergence when the average reward (energy efficiency) fluctuation over a consecutive preset number of rounds (e.g., 100 rounds) is less than a preset value (e.g., 1%).

[0115] For ease of understanding by those skilled in the art, please refer to Figure 5 , Figure 5 This is a schematic diagram of the training process for the policy-value network architecture.

[0116] See Figure 6 Preferably, in the high-speed train non-cellular MIMO energy-efficient transmission method described above, the step of performing interference processing on the antenna signals transmitted by each of the transmitted antennas based on a linear receiving algorithm to obtain the processed received signal includes:

[0117] Step S601: Determine the target linear receiving algorithm from the plurality of linear receiving algorithms based on complexity requirements and / or spectral efficiency;

[0118] Step S602: According to the target linear receiving algorithm, obtain the merging vector corresponding to each of the transmitting antennas;

[0119] Step S603: Combine the received antenna signal according to the combining vector to obtain the processed received signal.

[0120] This embodiment illustrates the steps for interference processing based on linear receiving algorithms. First, multiple linear receiving algorithms are pre-configured. Then, during interference processing, the appropriate algorithm is selected based on the system deployment requirements of the specific scenario. In this embodiment, the Maximum Ratio (MR) and Minimum Mean Square Error (MMSE) linear receiving algorithms are used as examples.

[0121] First, a target linear receiving algorithm is determined from among several linear receiving algorithms based on complexity requirements and / or spectral efficiency. Maximum ratio combining (MRB) is suitable for scenarios with lower computational complexity requirements. The transmitting antenna is at the 1st Combining vectors on subcarriers Solving by maximizing the signal-to-interference-plus-noise ratio:

[0122]

[0123] in, The expected power diagonal matrix of the merged vector has its l-th diagonal element as follows: ; For merging vectors The local merge vector component corresponding to the l-th access point; Noise power; For the first The transmission power of each transmitting antenna; To represent the equivalent channel response matrix (with dimensions CL×1, where L is the total number of serving access points) that integrates line-of-sight components, non-line-of-sight components, Doppler frequency offset, and inter-carrier interference, the matrix characterizes the [missing information - likely a specific type of channel response matrix]. The transmitting antenna to the first The signal and interference coupling relationship of each transmitting antenna corresponding to the service access point (where, (for interference links). For one of them The value of the frequency offset index being 0 represents the equivalent channel vector of the desired (useful) signal.

[0124] The corresponding spectral efficiency is:

[0125]

[0126] in, , for the first The transmitting antenna is at the 1st Effective signal-to-interference-plus-noise ratio (SINNR) when using two layers (Large Scale Fading Decoding, LSFD) for reception on each subcarrier. For the desired signal power, Add noise covariance to the interference; For the first The transmission power of each transmitting antenna.

[0127] The minimum mean square error is suitable for scenarios with high requirements for spectral efficiency. The transmitting antenna is at the 1st Combining vectors on subcarriers Solve by minimizing the mean square error:

[0128] ,

[0129] in, ; Let be the channel estimation error covariance matrix, and let be the access point. With transmitting antenna An equivalent channel that integrates deterministic inter-carrier interference and statistically averaged inter-carrier interference; It is an N-order identity matrix.

[0130] The corresponding spectral efficiency is:

[0131]

[0132] in, , for the first Each transmit antenna on the subcarrier Effective signal-to-interference-plus-noise ratio (SINORN) of a fully centralized MMSE receiver.

[0133] For ease of understanding by those skilled in the art, please refer to Figure 7 , Figure 7 This diagram illustrates the performance comparison of different power control algorithms when using minimum mean square error (MMSE) reception in a mixed and complex scenario, as provided in this application embodiment. As can be seen from the diagram, in terms of spectral efficiency (SE) performance, the MADDPG algorithm of this application embodiment is the best, followed by the maximum sum algorithm. In terms of antenna power consumption, the MADDPG algorithm is significantly superior to the other two algorithms. Therefore, overall, the MADDPG algorithm exhibits better SE performance and more stable power control.

[0134] Figure 8 This is a schematic diagram comparing energy efficiency in different scenarios provided by the embodiments of this application. As can be seen from the figure, in the three scenarios of plains, tunnels, and mixed plains and tunnels, the MADDPG scheme of the embodiments of this application has significantly higher energy efficiency than fractional power control and maximum summation algorithms under both receiving algorithms, and can still maintain stable performance advantages in the composite scenario.

[0135] Table 1 compares the algorithm running time under different numbers of subcarriers provided in the embodiments of the present invention.

[0136] Table 1

[0137] Number of subcarriers Traditional methods MADDPG algorithm 8 16.0279 seconds 44.7238 seconds 16 49.2916 seconds 45.8600 seconds 32 377.8561 seconds 48.2034 seconds 64 1623.1606 seconds 50.8121 seconds

[0138] As can be seen from the table, when the number of subcarriers exceeds 8, the running time of fractional power control and maximum sum spectral efficiency benchmark schemes increases exponentially, while the running time of the MADDPG scheme in this embodiment of the invention only increases slowly. When the number of subcarriers is 64, the running time is only 50.8s, which is much lower than the 1623.1s of the traditional method, verifying its applicability in large-scale OFDM systems.

[0139] To facilitate understanding of the implementation scheme of this application by those skilled in the art, please refer to [link / reference needed]. Figure 9 , Figure 9 This is a flowchart illustrating the energy-efficient transmission method for non-cellular MIMO on high-speed trains in this application, where AP is the access point, TA is the transmitting antenna, Agent is the intelligent agent, Action is the action in the Actor network, and Reward is the reward function.

[0140] The various methods of the embodiments of this application have been described above. Apparatus for implementing the above methods will now be provided.

[0141] See Figure 10 Another embodiment of this application provides a control device for high-speed train cellular-free MIMO energy-efficient transmission, comprising:

[0142] The first processing module 1001 is used to construct a Ricean channel model, which includes line-of-sight components and non-line-of-sight components, and is used to characterize the time-varying channel characteristics in the high-speed train scenario. The time-varying channel characteristics include Doppler frequency offset and inter-carrier interference.

[0143] The second processing module 1002 is used to determine the service access point subset corresponding to each transmitting antenna based on the dynamic collaborative clustering mechanism of the transmitting antenna center, and to form the dynamic collaborative clustering result.

[0144] The third processing module 1003 is used to model the joint optimization problem of the dynamic cooperative clustering results and power allocation as a multi-agent Markov decision process based on the Ricean channel model, wherein the optimization objective is to maximize the system energy efficiency and the constraint is the maximum transmission power limit of each of the transmitting antennas.

[0145] The fourth processing module 1004 is used to solve the multi-agent Markov decision process according to the multi-agent deep deterministic policy gradient algorithm, obtain the power allocation strategy of each transmitting antenna, and output the corresponding transmitting power to each transmitting antenna so that each transmitting antenna transmits a signal according to the corresponding transmitting power.

[0146] The fifth processing module 1005 is used to perform interference processing on the antenna signals transmitted by each of the transmitting antennas based on a linear receiving algorithm to obtain the processed received signal.

[0147] Preferably, in the control device for high-speed train cellular MIMO energy-efficient transmission as described above, the first processing module includes:

[0148] The first processing unit is configured to determine the line-of-sight component and the non-line-of-sight component based on scene parameters;

[0149] The second processing unit is used to determine the channel gain expression based on the line-of-sight component and the non-line-of-sight component;

[0150] The third processing unit is used to determine the Doppler frequency offset effect based on the carrier frequency, train speed, orthogonal frequency division multiplexing symbol duration, and angle of arrival of each transmitting antenna and the target receiving point.

[0151] The fourth processing unit is used to determine the inter-carrier interference characterization based on the line-of-sight component and the non-line-of-sight component, wherein the inter-carrier interference characterization includes deterministic inter-carrier interference and statistically averaged inter-carrier interference.

[0152] Preferably, in the control device for high-speed train cellular MIMO energy-efficient transmission as described above, the second processing module includes:

[0153] The fifth processing unit is used to determine the cooperative relationship between each of the service access points and each of the transmitting antennas based on the large-scale fading coefficient and a preset threshold.

[0154] The sixth processing unit is used to determine the service access point subset corresponding to each of the transmitting antennas based on the cooperative relationship, and to form a dynamic cooperative clustering result.

[0155] Preferably, in the control device for high-speed train non-cellular MIMO energy-efficient transmission as described above, during the multi-agent Markov decision-making process, the state space is the aggregated large-scale fading coefficient set of each of the transmitting antennas and the corresponding service access point; the action space is the set of the transmit power coefficients of each of the transmitting antennas; and the reward function is the system energy efficiency.

[0156] Preferably, the control device for high-speed train cellular MIMO energy-efficient transmission as described above, the fourth processing module includes:

[0157] The seventh processing unit is used to initialize the policy-value network architecture parameters corresponding to each of the transmitting antennas;

[0158] The eighth processing unit is used to randomly sample small batches of training samples from the interaction data between the transmitting antenna and the environment stored in the experience playback pool;

[0159] The ninth processing unit is used to train the policy-value network architecture based on the mini-batch training samples and to perform soft updates on the network parameters in the policy-value network architecture.

[0160] The tenth processing unit is used to obtain, after training, the policy-value network architecture obtains the power allocation strategy corresponding to each of the transmit antennas based on the large-scale fading coefficient at the current moment.

[0161] Preferably, in the high-speed train non-cellular MIMO energy-efficient transmission method described above, the fifth processing module includes:

[0162] The eleventh processing unit is used to determine a target linear receiving algorithm from among the multiple linear receiving algorithms based on complexity requirements and / or spectral efficiency;

[0163] The twelfth processing unit is used to obtain the merging vector corresponding to each of the transmitting antennas according to the target linear receiving algorithm;

[0164] The thirteenth processing unit is used to perform signal merging on the received antenna signal according to the merging vector to obtain the processed received signal.

[0165] The apparatus embodiments of this application are apparatuses corresponding to the embodiments of the methods described above. All implementation means in the method embodiments described above are applicable to the apparatus embodiments and can achieve the same technical effects. The apparatus provided in this application embodiments can implement all the method steps implemented in the method embodiments described above and can achieve the same technical effects. Therefore, the parts and beneficial effects that are the same as those in the method embodiments in this embodiment will not be described in detail here.

[0166] See Figure 11 Another embodiment of this application provides an electronic device 110, including: a processor 1101, a memory 1102, and a program stored in the memory 1102 and executable on the processor 1101. When the program is executed by the processor 1101, it implements the steps of the high-speed train non-cellular MIMO energy-efficient transmission method as described above, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0167] Another embodiment of this application provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the steps of the high-speed train non-cellular MIMO energy-efficient transmission method described above, achieving the same technical effect. To avoid repetition, it will not be described again here. The computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0168] Another embodiment of this application provides a computer program product, including computer instructions, which, when executed by a processor, implement the steps of the high-speed train non-cellular MIMO energy-efficient transmission method as described above, and achieve the same technical effect. To avoid repetition, it will not be described again here.

[0169] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0170] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0171] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. A method for energy-efficient transmission without cellular MIMO in high-speed trains, characterized in that, include: A Ricean channel model is constructed, which includes line-of-sight and non-line-of-sight components, to characterize the time-varying channel characteristics in the high-speed train scenario. The time-varying channel characteristics include Doppler frequency offset and inter-carrier interference. Based on the dynamic collaborative clustering mechanism of the transmit antenna center, the service access point subset corresponding to each transmit antenna is determined, and the dynamic collaborative clustering result is formed. Based on the Rice channel model, the joint optimization problem of the dynamic cooperative clustering results and power allocation is modeled as a multi-agent Markov decision process, wherein the optimization objective is to maximize the system energy efficiency, and the constraint is the maximum transmit power limit of each transmit antenna. The multi-agent Markov decision process is solved by the multi-agent deep deterministic policy gradient algorithm to obtain the power allocation strategy of each transmitting antenna, and the corresponding transmitting power is output to each transmitting antenna so that each transmitting antenna transmits a signal according to the corresponding transmitting power. The received antenna signals transmitted by each of the transmitting antennas are subjected to interference processing based on a linear receiving algorithm to obtain the processed received signals.

2. The high-speed train non-cellular MIMO energy-efficient transmission method according to claim 1, characterized in that, The construction of the Ricean channel model includes: The line-of-sight component and the non-line-of-sight component are determined based on scene parameters; The channel gain expression is determined based on the line-of-sight component and the non-line-of-sight component; The Doppler frequency offset effect is determined based on the carrier frequency, train speed, orthogonal frequency division multiplexing symbol duration, and angle of arrival of each transmitting antenna and the target receiving point. Based on the line-of-sight component and the non-line-of-sight component, an inter-carrier interference characterization is determined, which includes deterministic inter-carrier interference and statistically averaged inter-carrier interference.

3. The high-speed train non-cellular MIMO energy-efficient transmission method according to claim 1, characterized in that, The dynamic collaborative clustering mechanism based on the transmit antenna center determines the service access point subset corresponding to each transmit antenna and forms the dynamic collaborative clustering result, including: The cooperative relationship between each service access point and each transmission antenna is determined based on the large-scale fading coefficient and a preset threshold. Based on the cooperative relationship, the subset of service access points corresponding to each of the transmitting antennas is determined, and a dynamic cooperative clustering result is formed.

4. The high-speed train non-cellular MIMO energy-efficient transmission method according to claim 1, characterized in that, In the multi-agent Markov decision-making process, the state space is the aggregated large-scale fading coefficient set of each transmit antenna and its corresponding service access point; the action space is the transmit power coefficient set of each transmit antenna; and the reward function is the system energy efficiency.

5. The high-speed train non-cellular MIMO energy-efficient transmission method according to claim 1, characterized in that, The power allocation strategy for each transmit antenna is obtained by solving the multi-agent Markov decision process using the multi-agent deep deterministic policy gradient algorithm, including: Initialize the policy-value network architecture parameters corresponding to each of the aforementioned transmit antennas; Randomly sample small batches of training samples from the interaction data between the transmitting antenna and the environment already stored in the experience playback pool; The policy-value network architecture is trained based on the mini-batch training samples, and the network parameters in the policy-value network architecture are softly updated. After training is completed, the policy-value network architecture obtains the power allocation policy corresponding to each transmit antenna based on the large-scale fading coefficient at the current moment.

6. The high-speed train non-cellular MIMO energy-efficient transmission method according to claim 1, characterized in that, The method of performing interference processing on the received antenna signals transmitted by each of the transmitting antennas based on a linear receiving algorithm to obtain the processed received signal includes: The target linear receiving algorithm is determined from among the multiple linear receiving algorithms based on complexity requirements and / or spectral efficiency; According to the target linear receiving algorithm, the merging vector corresponding to each of the transmitting antennas is obtained; The received antenna signals are combined according to the merging vector to obtain the processed received signal.

7. A control device for high-speed train MIMO energy-efficient transmission without cellular networks, characterized in that, include: The first processing module is used to construct a Ricean channel model, which includes line-of-sight components and non-line-of-sight components, and is used to characterize the time-varying channel characteristics in the high-speed train scenario. The time-varying channel characteristics include Doppler frequency offset and inter-carrier interference. The second processing module is used to determine the service access point subset corresponding to each transmitting antenna based on the dynamic collaborative clustering mechanism of the transmitting antenna center, and to form the dynamic collaborative clustering result. The third processing module is used to model the joint optimization problem of the dynamic cooperative clustering results and power allocation as a multi-agent Markov decision process based on the Ricean channel model, wherein the optimization objective is to maximize the system energy efficiency and the constraint is the maximum transmit power limit of each transmit antenna. The fourth processing module is used to solve the multi-agent Markov decision process according to the multi-agent deep deterministic policy gradient algorithm, obtain the power allocation strategy of each transmitting antenna, and output the corresponding transmitting power to each transmitting antenna so that each transmitting antenna transmits a signal according to the corresponding transmitting power. The fifth processing module is used to perform interference processing on the antenna signals transmitted by each of the transmitting antennas based on a linear receiving algorithm to obtain the processed received signal.

8. An electronic device, characterized in that, include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of the high-speed train non-cellular MIMO energy-efficient transmission method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the high-speed train non-cellular MIMO energy-efficient transmission method as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, It includes computer instructions that, when executed by a processor, implement the steps of the high-speed train non-cellular MIMO energy-efficient transmission method as described in any one of claims 1 to 6.