KR102664971B1 - Rate-splitting multiple access method for the base station with limited energy supply to maximize the sum rate based on deep reinforcement learning - Google Patents

Abstract

A rate division multiple access method for a base station with limited energy supply to maximize the sum rate according to an embodiment of the present invention includes determining a first transmission power based on information about the base station in a first state; Splitting data to be transmitted to a plurality of terminals into shared and personal messages, and encoding the shared and personal messages into shared and personal streams, respectively; Designing power allocation and beamforming for each of the encoded shared and private streams within the determined transmission power; Transmitting the data to the plurality of terminals and receiving a sum data rate; and determining a second transmission power based on the received sum data rate and the base station information in the second state. Accordingly, in one embodiment of the present invention, the base station determines the transmission power using deep reinforcement learning based on the sum data rate and base station status information received from a plurality of terminals, and uses an optimization technique within the determined transmission power. By designing power allocation and beamforming for each of the encoded shared and private streams, a base station with limited energy supply can efficiently use the supplied energy to suit changing external energy and communication channel conditions.

Description

Rate-splitting multiple access method for the base station with limited energy supply to maximize the sum rate based on deep reinforcement learning}

The present invention relates to a rate division multiple access method for a base station with limited energy supply to maximize the sum rate, and more specifically, to optimize power allocation and beamforming design in the transmitter using deep reinforcement learning and optimization techniques. , This is about a rate division multiple access method for maximizing the sum rate in a communication environment with limited energy supply.

Since the emergence of LTE-based 4G mobile communication, MIMO (Multiple Input Multiple Output) technology using multiple antennas has become an essential core technology in mobile communication. Recently, MIMO technology has developed into Massive MIMO, and even theoretically infinite antennas are being considered. In addition, MIMO technology has spread in wireless LANs, starting with 802.11n, and all recently released wireless LANs adopt MIMO technology as standard, and multi-user MIMO (MU-U-MIMO) combines beamforming and spatial multiplexing technology. MIMO) technology has emerged.

Multi-user MIMO technology can support services to multiple users simultaneously in the same frequency band using multiple antennas in one base station, thereby increasing the efficiency of the wireless band. Despite these advantages, there is a problem of inter-user interference because one base station uses multiple antennas to provide services to multiple users, and due to the nature of wireless communication, the information in the channel, which is the information transmission medium, is Due to rapid changes, it is virtually impossible to obtain accurate channel information for all users, so signal interference between users receiving services is unavoidable.

Rate-Splitting Multiple Access (RSMA) is a non-orthogonal multi-user access technology that has excellent performance and robustness in a multi-user and multi-antenna communication environment and is capable of supporting multiple users in a variety of channel environments. . This rate division multiple access technology has strengths compared to existing multiple access technologies in terms of energy and frequency band efficiency and robustness to inaccurate channel state information. Meanwhile, the rate division multiple access technology to date uses a base station with no limit on transmission power to provide the maximum sum data rate in preparation for changing channel state information in an environment where the base station can provide services to users with constant transmission power. has been developed to provide

In 6G, the next-generation mobile communication technology that uses a high-frequency band (mmWave) with strong linearity, the importance of mobile base stations utilizing drone-type base station platforms, an unmanned aerial vehicle system, is gradually emerging. The high-frequency band utilizes a wider bandwidth than the existing frequency band, so it is suitable for transmitting large amounts of data, but because it has strong linearity, it is relatively vulnerable to radio wave shadowing, etc. To solve this problem, small The importance of small cells is increasing. Mobile base stations receive energy from external energy such as solar energy and RF signals due to limited transmission power, so they must use it efficiently and appropriately to the communication channel conditions to provide reliable services to users. However, it is difficult to efficiently use the supplied energy appropriately for changing external energy and communication channel conditions. Accordingly, in the rate division multiple access method for a base station with limited energy supply to maximize the sum rate, a multiple access method that can efficiently use the energy supplied appropriately to changing external energy and communication channel conditions is needed.

The present invention is intended to solve the problems of the prior art as described above, and allows a base station with a limited energy supply to efficiently use the energy supplied suitably to changing external energy and communication channel conditions to maximize the sum transmission rate. It provides a rate division multiple access method to perform.

A rate division multiple access method for a base station with limited energy supply to maximize the sum rate according to the first feature of the present invention includes determining a first transmission power based on information about the base station in a first state; Splitting data to be transmitted to a plurality of terminals into shared and personal messages, and encoding the shared and personal messages into shared and personal streams, respectively; Designing power allocation and beamforming for each of the encoded shared and private streams within the determined transmission power; Transmitting the data to the plurality of terminals and receiving a sum data rate; and determining a second transmission power based on the received sum data rate and the base station information in the second state.

A base station using rate division multiple access to maximize the sum rate in a communication environment with limited energy supply according to the second feature of the present invention includes a transceiver unit for transmitting and receiving signals to and from a plurality of terminals; a processor connected to the transceiver and controlling the operation of the base station; and a memory, wherein the processor determines a first transmission power based on the base station information in a first state; dividing data to be transmitted to the plurality of terminals into shared and personal messages, and encoding the shared and personal messages into shared and personal streams, respectively; An operation of performing power allocation and beamforming design for each of the encoded shared and private streams within the determined transmission power; Transmitting the data to the plurality of terminals and receiving a sum data rate; and determining a second transmission power based on the received sum data rate and the base station information in the second state.

According to the third feature of the present invention, a program of instructions that can be executed by a digital processing device is tangibly implemented to provide rate division multiple access for maximizing the sum rate in a communication environment with limited energy supply, and the digital processing device The recording medium that can be read by is characterized in that a program for executing the rate division multiple access method according to the first feature of the present invention on a computer is recorded thereon.

The rate division multiple access method for a base station with limited energy supply to maximize the sum rate according to an embodiment of the present invention provides the following effects.

The base station determines the transmission power using deep reinforcement learning based on the sum data rate and base station status information received from multiple terminals, and uses an optimization technique within the determined transmission power to allocate power to each of the encoded shared and private streams. And by designing beamforming, a base station with limited energy supply can efficiently use the energy supplied to suit changing external energy and communication channel conditions.

The above-described effect becomes an even greater advantage when the base station is a mobile base station whose energy supply is inevitably limited. This mobile base station serves as a connection between cells, making it easier to secure network connectivity. Service support is possible even in areas where communication is limited because there are no existing base stations.

By using rate division multiple access technology, it is possible to provide reliable services to users using the same frequency band even if the channel status information of users is not accurately known in a multi-user MIMO environment.

Figure 1 is a flowchart of a rate division multiple access method for a base station with limited energy supply to maximize the sum rate based on deep reinforcement learning according to an embodiment of the present invention.
Figure 2 is a block diagram showing a base station using rate division multiple access to maximize the sum rate in a communication environment with limited energy supply according to an embodiment of the present invention.
Figure 3 is a diagram schematically showing a series of processes in which a rate division multiple access method is performed for a base station with limited energy supply to maximize the sum rate according to an embodiment of the present invention.
Figure 4 is a graph comparing the performance of multiple access technologies including rate division multiple access technology according to an embodiment of the present invention.
Figure 5 is a graph comparing the performance of multiple access technologies, including rate division multiple access technology using deep reinforcement learning, in a communication environment with limited energy supply according to an embodiment of the present invention.

Hereinafter, the present invention will be described in detail with reference to examples and drawings. However, the following description is not intended to limit the present invention to specific embodiments, and in describing the present invention, if it is determined that a detailed description of related known technology may obscure the gist of the present invention, the detailed description will be omitted. .

In embodiments of the present invention, a base station generally refers to a fixed station that communicates with a wireless device, but may be mobile and can be replaced by terms such as evolved-NodeB (eNB) and base transceiver system (BTS). there is.

Additionally, in embodiments of the present invention, the terminal may be fixed or mobile, and may include a user equipment (UE), a mobile station (MS), a subscriber station (SS), and a mobile subscriber station (MSS). It can be replaced by terms such as Mobile Subscriber Station, Mobile Terminal, or Advanced Mobile Station (AMS).

Figure 1 is a flow chart of a rate division multiple access method for a base station with limited energy supply to perform sum rate maximization according to an embodiment of the present invention, and Figure 2 is a flow chart for a base station with limited energy supply according to an embodiment of the present invention. This is a block diagram showing a base station using rate division multiple access to perform sum rate maximization in a communication environment.

Referring to FIG. 1, the rate division multiple access method 100 for the base station 200 with limited energy supply to maximize the sum rate according to an embodiment of the present invention is based on the base station information in the first state. Determining the first transmission power (110); Splitting data to be transmitted to a plurality of terminals 300 into shared and personal messages, and encoding the shared and personal messages into shared and personal streams, respectively (120); Step 130 of designing power allocation and beamforming for each of the encoded shared and private streams within the determined transmission power; Transmitting the data to the plurality of terminals 300 and receiving a sum data rate (140); and determining a second transmission power based on the received sum data rate and the base station information in the second state (150).

Referring to FIG. 2, a base station 200 using rate division multiple access to perform sum rate maximization based on deep reinforcement learning in a communication environment with limited energy supply according to an embodiment of the present invention includes a plurality of terminals. A transmitting and receiving unit 210 that transmits and receives signals with the fields 300; A processor 220 connected to the transceiver 210 and controlling the operation of the base station 200; and a memory 230, wherein the processor 220 performs an operation 110 of determining a first transmission power based on the base station information in a first state; An operation 120 of dividing data to be transmitted to the plurality of terminals 300 into shared and personal messages and encoding the shared and personal messages into shared and personal streams, respectively; An operation 130 of performing power allocation and beamforming design for each of the encoded shared and private streams within the determined transmission power; An operation 140 of transmitting the data to the plurality of terminals 300 and receiving a sum data rate; and an operation 150 of determining a second transmission power based on the received sum data rate and the base station information in the second state.

The base station 200 determines transmission power based on base station status information. In one embodiment of the present invention, the status information of the base station may include the amount of energy harvested by the base station, the amount of energy stored in the base station, and channel information fed back to the plurality of communicating terminals 300. .

This transmission power is determined by the processor 220 of the base station 200 receiving the status information of the base station through the transceiver 210 and the memory 230 and performing a deep reinforcement learning algorithm.

In one embodiment of the present invention, the deep reinforcement learning algorithm is a machine learning-based model and may be a reinforcement learning-based model. Reinforcement learning-based models can be continuously or periodically updated as they learn without administrator intervention.

In a reinforcement learning-based model, an agent takes an action from the environment, and accordingly the agent is given a reward and a state, and the agent is selected so that the reward is the largest. It can be a model to train. Reinforcement learning-based models can be divided into models using a model-based algorithm or models using a model-free algorithm, depending on whether the agent has prior knowledge.

Deep reinforcement learning is an algorithm that uses a deep neural network to solve problems in continuous and complex state spaces and action spaces such as real environments using reinforcement learning as described above. In one embodiment of the present invention, the agent is a base station, the environment is a rate division multiple access communication environment with limited energy supply, the action is transmission from the base station to the user, the compensation is the sum transmission rate, the state is the amount of energy harvested, and the base station It can be set based on the amount of stored energy and channel information fed back from multiple communicating terminals.

In one embodiment of the present invention, the deep reinforcement learning algorithm used to determine transmission power may be a soft actor critic (SAC) algorithm.

The base station 200 uses rate division multiple access technology to divide data to be transmitted to a plurality of terminals 300 into two types: a common message and a private message. Shared messages are encoded into one shared stream using a codebook shared by all users, which can be decoded by all users. Personal messages are encoded into personal streams through a codebook that only the specific user corresponding to each message has, and can be decoded only by the specific user.

Users decrypt the shared stream and obtain the part of the shared message that corresponds to their own message. And the shared stream is removed from the received signal through sequential interference cancellation (SIC). Sequential interference cancellation technology is used to simultaneously process multiple received signals on the receiver side. This technology uses signal strength differences among target signals to process signals at the receiver. That is, after decoding the stronger signal, the receiver first extracts the strong signal from the overlapping signal and then removes the weak signal from the remaining signals.

After removing the shared stream, the user only has interference from the personal stream corresponding to other users, and this is treated as noise to decode the desired personal stream. By decoding the shared stream and private stream using this sequential interference cancellation technique, each user can obtain the original message corresponding to him/her.

Therefore, in situations where the base station does not accurately know the channel state information of the users who will be provided with the service (for example, due to quantization error, channel state change, etc.), a shared message is used to multicast part of the message to be transmitted to the users. By transmitting via (multicast), signal interference between adjacent users can be minimized.

In order to transmit divided messages to multiple users, the base station uses a precoder to design power allocation and beamforming for one shared stream, and also design power allocation and beamforming for each private stream. Perform forming design. In this case, the power allocation and beamforming design of the shared stream and the power allocation of each private stream are non-convex problems.

Convex problems have one minimum value (global minimum) or one maximum value (global maximum), so it is relatively easy to calculate them. However, in the case of non-convex problems such as cubic equations, there is an additional minimum or maximum value. Since it has a local minimum and a local maximum, calculating the minimum or maximum value is relatively difficult. For example, in the case of a non-convex problem, if the problem is solved using methods such as gradient descent, which is a representative method for solving convex problems, the problem will not converge to the minimum or maximum value and will fall into a minimum or maximum value. There may be cases where the intended minimum or maximum value cannot be calculated.

Accordingly, in one embodiment of the present invention, the power allocation and beamforming design for each of the encoded shared and private streams is preferentially performed using the zero-forcing technique. And the power allocation of the personal stream, the power allocation of the shared stream, and the beamforming design of the shared stream can be performed using an optimization algorithm.

In one embodiment of the present invention, the optimization algorithm used for power allocation of a private stream, power allocation of a shared stream, and beamforming design of the shared stream may be a sequential least squares programming (SLSQP) algorithm.

In one embodiment of the present invention, the SLSQP algorithm simplifies the power allocation of the private stream, the power allocation of the shared stream, and the beamforming design of the shared stream, calculates the objective function by approximating it to a convex quadratic equation, predicts the next point, and repeats it again. Non-convex/non-linear problems can be optimally solved by repeating a series of processes using the same method to solve the problem.

In one embodiment of the present invention, the SLSQP algorithm has the constraint that the base station cannot set the power allocation larger than the transmission power determined as a result of deep reinforcement learning with respect to the power allocation of private streams and shared streams, and all the data to be transmitted. It operates under the upper and lower limit conditions that the sum of the power allocation ratios of the messages must be less than or equal to 1, and with respect to the beamforming design of the shared stream, the square of the L2 norm of the normalized beamforming vector of the shared stream must be less than or equal to 1. Operates under conditions.

Figure 3 is a diagram schematically showing a series of processes in which a rate division multiple access method is performed for a base station with limited energy supply to maximize the sum rate according to an embodiment of the present invention.

Referring to FIG. 2 again along with FIG. 3 , the base station 200 includes a transceiver 210, a processor 220, and a memory 230. The processor 220 may be configured to implement the procedures and/or methods proposed in the present invention. The memory 230 is connected to the processor 220 and stores various information related to the operation of the processor 220. The transceiver 210 is connected to the processor 220 and communicates with a plurality of terminals 300 to transmit and/or receive wireless signals.

In Figure 3, B _i is the amount of energy stored in the base station in the ith step, E _i is the amount of energy harvested in the ith step, and h _i is the channel status that a plurality of terminals will feed back to the base station in the ith step. Information, H _i represents channel state information fed back by the base station from a plurality of terminals in the ith step, and e _i represents channel state information loss that occurs in the ith step.

In one embodiment of the present invention, the base station 200 may transmit data to a plurality of terminals 300 and receive the sum data rate 10. Additionally, the transmission power can be re-determined based on the received sum data rate (10) and the next state base station information (20).

Referring to FIG. 3, the base station information 20 in the next state includes the amount of energy stored in the base station in the ith step (B _i ), the amount of energy harvested in the ith step (E i ), and the amount of energy harvested in the ith step (E _i ). This is information including the feedback channel information status (H _i ).

A deep reinforcement learning algorithm may be used to determine transmission power, and in one embodiment of the present invention, the deep reinforcement learning algorithm may be a soft actor critic (SAC) algorithm 30.

The base station 200 determines the transmission power to be used based on the amount of energy stored in the base station, the amount of harvested energy, and the status of the channel information received. Afterwards, the base station 200 receives the sum data rate 10 as a reward from the communication environment (i.e., a plurality of terminals 300) and is evaluated for the action performed, and at the same time receives the base station information 20 of the next state. do. The agent can refer to this and determine the optimal transmission power by repeating a series of learning processes in the direction of receiving greater rewards.

In this case, the amount of energy stored in the base station must satisfy certain conditions, and the amount of energy stored in the base station is b1 and b2 in each step, the maximum amount of energy that can be stored in the base station is max b, and the transmission power is p. If time is T and the amount of harvested energy is E, the conditions are as follows.

b2 = m(b1, p, E) = min(max b, b1-pT+E)

Therefore, in one embodiment of the present invention, the amount of energy stored in the base station in the second state is (a) the maximum amount of energy that the base station can hold, (b) the amount of energy stored in the base station in the first state The amount of energy obtained by subtracting the product of the determined transmission power in the first state and the time at which data was transmitted and adding the amount of energy harvested to the base station in the second state, in which case the smaller energy of (a) and (b) above. It can be determined by the amount of .

The transmission power (40) determined again through the soft actor critic (SAC) algorithm (30), a deep reinforcement learning algorithm, is used to design power allocation and beamforming for each of the encoded shared and private streams in the i-th step, and these operations In performing, the SLSQP (sequential least squares programming) algorithm 50, which is an optimization technique, is used, and in the i-th step, data is transmitted to a plurality of terminals 300, which are a large number of users, using the optimal precoder. .

Afterwards, the base station 200 receives a new sum rate as compensation from the plurality of terminals 300 and receives an evaluation of the action performed. As described above, the rate division multiple access method for maximizing the sum rate is repeated. It can be done.

In one embodiment of the present invention, a base station in a communication environment with limited energy supply may be a mobile base station. The drone-type base station platform serves as a link between cells, making it easier to secure network connectivity and supporting services even in areas where communication is limited due to the absence of a base station. In addition, by increasing the altitude of the base station, line of sight can be secured as much as possible to compensate for vulnerabilities such as shadowing in high frequency band communication, and continuous service can be provided even if the existing base station is destroyed due to natural disasters or other reasons. This is possible.

The mobile base station receives energy through external energy such as solar energy and RF signals due to limited transmission power, so it must be used efficiently in accordance with the communication channel status to provide reliable services to users. The sum data rate according to the present invention is Through the rate division multiple access method to perform maximization, it is possible to efficiently use the energy supplied appropriately to changing external energy and communication channel conditions.

The following describes 1) performance comparison of rate division multiple access technology and other multiple access technologies, and 2) performance comparison of rate division multiple access technology using deep reinforcement learning and other multiple access technologies in a communication environment with limited energy supply. The experimental process and results are explained in detail.

Figure 4 is a graph comparing the performance of multiple access technologies including rate division multiple access technology according to an embodiment of the present invention, and Figure 5 is a graph comparing deep reinforcement learning in a communication environment with limited energy supply according to an embodiment of the present invention. This is a graph comparing the performance of multiple access technologies including the rate division multiple access technology used.

The simulation environment for the experiment is a communication environment in which a mobile base station receives external energy equal to the battery's rechargeable capacity with a 50% probability through energy harvesting and then provides services to multiple users. In this case, the user is completely aware of the channel state information, and in the process of feeding it back to the base station, channel state information loss occurs due to quantization error and sudden changes in channel state. Accordingly, the base station receives incomplete channel state information as feedback.

In the graphs of Figures 4 and 5, the x-axis represents the number of learning iterations, and in the simulation process, 100 steps were set to 1 iteration. The y-axis is a relative measure of the performance evaluated by the learned agent against the designated verification environment every time 1000 steps are performed. Here, step refers to the interaction between the agent and the environment.

The verification environment was set up the same as the learning environment. The mobile base station receives external energy equal to the battery's chargeable capacity with a 50% probability through energy harvesting, and then operates optimally under conditions identically configured with the amount of energy harvested, the amount of energy remaining in the battery, and the channel information received. Provides service to multiple users by determining the transmission power. Afterwards, compensation, that is, the sum data rate, is received from the environment. For clearer visualization when comparing performance, the reward received from the environment was multiplied by 1/10000 to calculate the value and set it to be shown on the y-axis of the graph. To ensure consistency in the evaluation process, the verification model was set constant.

Referring again to FIG. 4, the following four cases were compared. Equal (Equal Power Allocation Precoder) Greedy, WF (Water Filling Precoder) Greedy, SLSQP Greedy (if the base station recognizes that the incomplete channel state information it received as feedback is incomplete), SLSQP Greedy (no-info σ _e )( When the base station recognizes that the incomplete channel state information it received as feedback is complete).

Greedy means that deep reinforcement learning is not applied. In other words, when a base station harvests energy, it uses up all the energy as it is harvested without considering the harvested energy, the amount of energy remaining in the battery, and the channel information received feedback.

SLSQP Greedy and SLSQP Greedy (no-info σ _e ) are the result of using rate division multiple access technology to divide messages into personal messages and shared messages and transmit them together, while Equal Greedy and WF Greedy do not divide messages into personal messages and shared messages. It is a general multiple access technology that transmits messages only in unicast mode.

Through the experimental results in Figure 4, it can be seen that the rate division multiple access technology, which divides messages into personal messages and shared messages and transmits them, shows superior performance in terms of sum transmission rate compared to other multiple access technologies that do not use shared messages. .

Referring again to FIG. 5, the following eight cases were compared. SAC+SLSQP, SAC+SLSQP(no-info σ _e ), SAC+WF, SAC+Equal, Equal Greedy, WF Greedy, SLSQP Greedy, SLSQP Greedy(no-info σ _e ).

Through the experimental results in Figure 5, it can be confirmed that the results of the four cases using deep reinforcement learning in an energy harvesting environment each show superior performance in terms of sum transfer rate compared to the results of the four cases without deep reinforcement learning. there is. In addition, when comparing the case of using deep reinforcement learning and the case of greedy under the same conditions, it can be confirmed that the rate division multiple access technology shows superior performance compared to other multiple access technologies that do not use shared messages. Therefore, it can be seen that the rate division multiple access method for maximizing the sum data rate based on deep reinforcement learning considering energy harvesting according to the present invention shows superior performance compared to existing multiple access technologies in a communication environment under the same conditions.

As seen so far, the rate division multiple access method for a base station with limited energy supply to maximize the sum rate according to an embodiment of the present invention is based on the sum rate and base station status information received by the base station from a plurality of terminals. By determining the transmission power using deep reinforcement learning and designing power allocation and beamforming for each encoded shared and private stream using an optimization technique within the determined transmission power, a base station with limited energy supply changes external energy and communication It provides the effect of efficiently using the energy supplied to suit the channel condition.

Embodiments according to the present invention may be implemented by various means, for example, hardware, firmware, software, or a combination thereof. In the case of implementation by hardware, an embodiment of the present invention includes one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), and FPGAs ( It can be implemented by field programmable gate arrays, processors, controllers, microcontrollers, microprocessors, etc.

In the case of implementation by firmware or software, an embodiment of the present invention may be implemented in the form of a module, procedure, function, etc. that performs the functions or operations described above. Software code can be stored in a memory unit and run by a processor. The memory unit is located inside or outside the processor and can exchange data with the processor through various known means.

Meanwhile, embodiments of the present invention can be implemented as computer-readable code on a computer-readable recording medium. Computer-readable recording media include all types of recording devices that store data that can be read by a computer system.

Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disk, and optical data storage devices, and can also be implemented in the form of a carrier wave (e.g., transmission via the Internet). Includes. Additionally, the computer-readable recording medium can be distributed across computer systems connected to a network, so that computer-readable code can be stored and executed in a distributed manner. And functional programs, codes, and code segments for implementing the present invention can be easily deduced by programmers in the technical field to which the present invention pertains.

Since various modifications may be made to the configurations and methods described and illustrated herein without departing from the scope of the present invention, all matters contained in the foregoing detailed description or shown in the accompanying drawings are exemplary and are not intended to limit the present invention. It's not. Accordingly, the scope of the present invention should not be limited by the above-described exemplary embodiments, but should be determined only by the following claims and their equivalents.

10: Sum transmission rate of step i-1 20: Base station status information of step i
30: SAC algorithm 40: determined transmission power of step i
50: SLSQP algorithm 200: base station
210: Transmitter and receiver 220: Processor
230: memory 300: user terminals

Claims (15)

In the rate division multiple access method for a base station with limited energy supply to maximize the sum rate,
Determining a first transmission power based on base station information in the first state;
Splitting data to be transmitted to a plurality of terminals into shared and personal messages, and encoding the shared and personal messages into shared and personal streams, respectively;
Designing power allocation and beamforming for each of the encoded shared and private streams within the determined transmission power;
Transmitting the data to the plurality of terminals and receiving a sum data rate; and
determining a second transmission power based on the received sum data rate and the base station information in a second state;
Including,
The power allocation and beamforming design for each of the encoded shared and private streams is performed preferentially using a zero-forcing technique, and the beamforming design of the private stream is performed preferentially, and the power allocation of the private stream and the shared Power allocation of the stream and beamforming design of the shared stream are performed using a first optimization algorithm,
The base station information in the first and second states includes the amount of energy harvested by the base station, the amount of energy stored in the base station, and channel information fed back to the plurality of terminals,
A rate division multiple access method, characterized in that the determination of the transmission power is performed using a second deep reinforcement learning algorithm. delete According to paragraph 1,
A rate division multiple access method, wherein the first optimization algorithm is a sequential least squares programming (SLSQP) algorithm. delete According to paragraph 1,
The amount of energy stored in the base station in the second state is,
(a) Maximum amount of energy the base station can hold;
(b) subtract the product of the determined transmission power in the first state and the time at which the data was transmitted from the amount of energy stored in the base station in the first state, and add the amount of energy harvested in the base station in the second state. amount of energy;
A rate division multiple access method characterized by determining the smaller energy amount among (a) and (b). According to paragraph 1,
The second deep reinforcement learning algorithm is a rate division multiple access method, characterized in that the SAC (soft actor critic) algorithm. According to paragraph 1,
A rate division multiple access method, wherein the base station is a mobile base station. In a base station using rate division multiple access to maximize the sum rate in a communication environment with limited energy supply,
A transceiver unit that transmits and receives signals to and from a plurality of terminals;
a processor connected to the transceiver and controlling the operation of the base station; and memory,
The processor,
An operation of determining a first transmission power based on base station information in a first state;
dividing data to be transmitted to the plurality of terminals into shared and personal messages, and encoding the shared and personal messages into shared and personal streams, respectively;
An operation of performing power allocation and beamforming design for each of the encoded shared and private streams within the determined transmission power;
Transmitting the data to the plurality of terminals and receiving a sum data rate; and
determining a second transmission power based on the received sum data rate and the base station information in a second state;
Do this,
The power allocation and beamforming design for each of the encoded shared and private streams is performed preferentially using a zero-forcing technique, and the beamforming design of the private stream is performed preferentially, and the power allocation of the private stream and the shared Power allocation of the stream and beamforming design of the shared stream are performed using a first optimization algorithm,
The base station information in the first and second states includes the amount of energy harvested by the base station, the amount of energy stored in the base station, and channel information fed back to the plurality of terminals,
A base station, characterized in that the determination of the transmission power is performed using a second deep reinforcement learning algorithm. delete According to clause 8,
A base station, wherein the first optimization algorithm is a sequential least squares programming (SLSQP) algorithm. delete According to clause 8,
The amount of energy stored in the base station in the second state is,
(a) Maximum amount of energy the base station can hold;
(b) subtract the product of the determined transmission power in the first state and the time at which the data was transmitted from the amount of energy stored in the base station in the first state, and add the amount of energy harvested in the base station in the second state. amount of energy;
A base station characterized in that it is determined by the smaller amount of energy among (a) and (b). According to clause 8,
A base station wherein the second deep reinforcement learning algorithm is a SAC (soft actor critic) algorithm. According to clause 8,
A base station, characterized in that the base station is a mobile base station. A recording medium that has a tangible implementation of a program of instructions that can be executed by a digital processing device to provide rate division multiple access for maximizing the sum rate in a communication environment with limited energy supply and that can be read by the digital processing device. as,
A computer-readable recording medium recording a program for executing the method of any one of claims 1, 3, and 5 to 7 on a computer.

Images