Semantic Relay System, Resource Allocation Method, Electronic Device and Storage Medium

Abstract

An embodiment of the present invention provides a semantic relay system, a resource allocation method, an electronic device, and a computer-readable storage medium. The semantic relay system includes a base station, and the base station includes a semantic communication encoder; a semantic relay device, the semantic relay device includes a semantic receiver, the semantic receiver is data-connected to the semantic communication encoder, and the semantic receiver is configured to perform decoding processing on the semantic encoded information to obtain semantic decoded information and perform source coding processing on the semantic decoded information to obtain bit-encoded information; a user device, the user device includes a bit decoder, and the source decoder is data-connected to the semantic receiver. According to the solution of the embodiment of the present invention, not only the efficiency of text transmission is improved, but also since the semantic decoding process is performed on the semantic relay device, the computing and storage burdens of the user device are well reduced.

Description

Technical Field

[0001] The present invention relates to the field of communication technologies, and particularly to a semantic relay system, a resource allocation method, an electronic device, and a computer-readable storage medium. Background Art

[0002] Semantic communication (SemCom) has great potential to break through the Shannon capacity limit in the future sixth-generation wireless system by leveraging advanced artificial intelligence technologies to improve data transmission efficiency. Specifically, the information transmitted by SemCom is not a bit sequence in traditional communication, but the semantic information extracted from the source data, so as to reduce communication overhead and improve resource utilization efficiency. In recent years, the superior performance of semantic communication has stimulated people's research interest, and the SemCom technology has been applied to various applications, such as education and medical care through virtual reality, human-computer interaction with multiple intelligent devices, and cooperative communication between vehicles. These applications often have strict requirements for end-to-end communication in terms of latency and data rate.

[0003] With the latest research progress of deep learning technologies and the improvement of device computing and storage capabilities, deep neural networks have been widely used in the research of SemCom. Existing research work mainly focuses on three key aspects, namely the SemCom system architecture, semantic representation, and resource management. Although SemCom has great potential for better information transmission efficiency than traditional bit transmission, especially in low signal-to-noise ratio and small bandwidth scenarios. However, most previous work on SemCom simply assumes the deployment of deep learning neural networks on mobile devices, and such an assumption ignores that ordinary mobile devices only have limited computing and storage resources and cannot implement SemCom technology based on deep learning. Summary of the Invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art.

[0005] To this end, the present invention proposes a semantic relay system, which adds a semantic relay device between the base station and the user equipment. By performing semantic communication between the base station and the semantic relay device, not only the efficiency of text transmission is improved, but also the computing and storage burdens of the user equipment are well alleviated because the semantic decoding process is carried out on the semantic relay device.

[0006] The present invention also proposes a resource allocation method implemented based on the above semantic relay system.

[0007] The present invention also proposes an electronic device applying the above resource allocation method.

[0008] The present invention also proposes a computer-readable storage medium applying the above resource allocation method.

[0009] A semantic relay system according to an embodiment of the first aspect of the present invention, the system comprising:

[0010] A base station, the base station comprising a semantic communication encoder for encoding source text information in the base station to generate semantic encoded information;

[0011] A semantic relay device, the semantic relay device comprising a semantic receiver, the semantic receiver being data-connected to the semantic communication encoder, the semantic receiver being configured to decode the semantic encoded information to obtain semantic decoded information and perform source encoding processing on the semantic decoded information to obtain bit-encoded information;

[0012] A user equipment, the user equipment comprising a bit decoder, the bit decoder being data-connected to the semantic receiver, the bit decoder being configured to decode the bit-encoded information to obtain the source text information.

[0013] According to some embodiments of the present invention, the semantic communication encoder comprises a semantic encoder and a first channel encoder, the semantic encoder being connected to the first channel encoder, wherein the semantic encoder is configured to perform a first encoding process on the source text information in the base station to obtain first encoded information, and the first channel encoder is configured to perform a first channel encoding process on the first encoded information to generate the semantic encoded information.

[0014] According to some embodiments of the present invention, the semantic receiver comprises a first channel decoder, a semantic decoder, a source encoder, and a second channel encoder, the first channel decoder, the semantic decoder, the source encoder, and the second channel encoder being connected in sequence, the first channel decoder being data-connected to the first channel encoder, the first channel decoder being configured to perform a first channel decoding process on the semantic encoded information to obtain first decoded information, the semantic decoder being configured to perform a first decoding process on the first decoded information to obtain the semantic decoded information, the source encoder being configured to perform a second encoding process on the semantic decoded information to obtain second encoded information, and the second channel encoder being configured to perform a second channel encoding process on the second encoded information to obtain the bit-encoded information.

[0015] According to some embodiments of the present invention, the bit decoder comprises a second channel decoder and a source decoder, the second channel decoder being connected to the source decoder, the second channel decoder being data-connected to the second channel encoder, the second channel decoder being configured to perform a second channel decoding process on the bit-encoded information to obtain second decoded information, and the source decoder being configured to perform a second decoding on the second decoded information to obtain the source text information.

[0016] The resource allocation method according to the second aspect embodiment of the present invention is applied to the semantic relay system of the above embodiment, and the method includes:

[0017] Obtain the first initial distance and the first initial bandwidth ratio from the base station to the semantic relay device, the second initial distance and the second initial bandwidth ratio from the semantic relay device to the user equipment, and the initial transmission power of the semantic relay device;

[0018] Determine the first channel gain according to the first initial distance and determine the second channel gain according to the second initial distance;

[0019] Calculate the first signal-to-noise ratio according to the first channel gain and the first initial bandwidth ratio, and calculate the second signal-to-noise ratio according to the second channel gain and the second initial bandwidth ratio;

[0020] Determine the semantic similarity according to the first signal-to-noise ratio; and calculate the initial bit transmission rate according to the semantic similarity and the first initial bandwidth ratio, and calculate the initial user achievable bit rate according to the second initial bandwidth ratio and the second signal-to-noise ratio;

[0021] Determine an optimization problem based on the first channel gain, the second channel gain, the initial bit transmission rate, the initial user achievable bit rate, and the initial transmission power;

[0022] Perform a solution process on the optimization problem based on a preset optimization algorithm to obtain resource allocation information.

[0023] According to some embodiments of the present invention, the optimization problem includes a single-user resource optimization problem, the optimization algorithm includes a block coordinate descent optimization algorithm, and performing a solution process on the optimization problem based on a preset optimization algorithm to obtain resource allocation information includes:

[0024] Perform a solution on the single-user resource optimization problem based on the block coordinate descent optimization algorithm to obtain a first sub-optimal solution;

[0025] Update the penalty coefficient in the block coordinate descent optimization algorithm according to the first sub-optimal solution, and re-solve the single-user resource optimization problem using the updated block coordinate descent optimization algorithm until the first optimal solution of the single-user resource optimization problem is obtained;

[0026] Determine a first optimal distance from the base station to the semantic relay device, a first optimal bandwidth ratio, a second optimal distance from the semantic relay device to the user equipment, and a second optimal bandwidth ratio according to the first optimal solution, where the resource allocation information includes the first optimal distance, the second optimal distance, the first optimal bandwidth ratio, and the second optimal bandwidth ratio.

[0027] According to some embodiments of the present invention, the optimization problem includes a multi-user resource optimization problem, the optimization algorithm includes a block coordinate descent resource allocation algorithm, and solving the optimization problem based on a preset optimization algorithm to obtain resource allocation information includes:

[0028] Solving the multi-user resource optimization problem based on the block coordinate descent resource allocation algorithm to obtain a second sub-optimal solution;

[0029] Updating the initial transmission power, the first initial bandwidth ratio, and the second initial bandwidth ratio according to the second sub-optimal solution to obtain a second optimal solution to the multi-user resource optimization problem;

[0030] Determine an optimal transmission power of the semantic relay device, a third optimal bandwidth ratio from the base station to the semantic relay device, and a fourth optimal bandwidth ratio from the semantic relay device to the user equipment according to the second optimal solution, where the resource allocation information includes the optimal transmission power, the third optimal bandwidth ratio, and the fourth optimal bandwidth ratio.

[0031] According to some embodiments of the present invention, the updating the initial transmission power, the first initial bandwidth ratio, and the second initial bandwidth ratio according to the second sub-optimal solution includes:

[0032] Determine the first initial bandwidth ratio and the second initial bandwidth ratio as adjusted bandwidth ratios;

[0033] Update the initial transmission power and the adjusted bandwidth ratio alternately according to the second sub-optimal solution.

[0034] An electronic device according to an embodiment of the third aspect of the present invention includes: a memory, a processor, and a computer program stored on the memory and executable on the processor, and when the processor executes the computer program, the resource allocation method described above is implemented.

[0035] A computer-readable storage medium according to an embodiment of the fourth aspect of the present invention stores computer-executable instructions, and when the computer-executable instructions are executed by a control processor, the resource allocation method described above is implemented.

[0036] The semantic relay system according to the embodiments of the present invention has at least the following beneficial effects: During the process of semantic communication, the semantic communication encoder of the base station is used to encode the source text information in the base station to generate semantic encoded information; after the semantic receiver of the semantic relay device receives the semantic encoded information sent by the base station, it decodes the semantic encoded information to obtain semantic decoded information, and then performs source encoding on the semantic decoded information to obtain bit encoded information; then the semantic relay device sends the bit encoded information to the user equipment, and the bit decoder in the user equipment can decode the received bit encoded information, and finally the source text information can be obtained on the user equipment; during the entire semantic communication process, the parts of semantic encoding and semantic decoding are set between the base station and the semantic relay device, which not only improves the efficiency of text transmission, but also, since the semantic decoding process is carried out on the semantic relay device, it well reduces the computing and storage burdens of the user equipment.

[0037] Other features and advantages of the present invention will be described in the following specification, and in part will be obvious from the specification, or can be understood by implementing the present invention. The objectives and other advantages of the present invention can be realized and obtained by the structures specifically pointed out in the specification, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The drawings are used to provide a further understanding of the technical solutions of the present disclosure, and constitute a part of the specification. Together with the embodiments of the present disclosure, they are used to explain the technical solutions of the present disclosure, and do not constitute a limitation to the technical solutions of the present disclosure.

[0039] Figure 1 is a schematic structural diagram of a semantic relay system provided by an embodiment of the present invention;

[0040] Figure 2 is a flowchart of a resource allocation method provided by an embodiment of the present invention;

[0041] Figure 3 is a specific flowchart of solving an optimization problem based on a preset optimization algorithm provided by an embodiment of the present invention;

[0042] Figure 4 is a specific flowchart of solving an optimization problem based on a preset optimization algorithm provided by another embodiment of the present invention;

[0043] Figure 5 is a specific flowchart of updating the initial transmission power, the first initial bandwidth ratio, and the second initial bandwidth ratio provided by an embodiment of the present invention;

[0044] Figure 6It is a schematic diagram showing the relationship between the achievable bit rate and the total system bandwidth provided by an embodiment of the present invention;

[0045] Figure 7 It is a schematic diagram showing the relationship between the optimized bandwidth ratio and the total system bandwidth provided by an embodiment of the present invention;

[0046] Figure 8 It is a schematic diagram showing the relationship between the optimized position and the system bandwidth provided by an embodiment of the present invention;

[0047] Figure 9 It is a schematic diagram showing the relationship between the achievable weighted sum rate and the system bandwidth provided by an embodiment of the present invention;

[0048] Figure 10 It is a schematic diagram showing the relationship between the achievable weighted sum rate and the transmit power of the semantic relay device provided by an embodiment of the present invention;

[0049] Figure 11 It is a schematic diagram showing the structure of an electronic device provided by an embodiment of the present invention.

[0050] Reference numerals:

[0051] Base station 100, semantic communication encoder 110, semantic encoder 111, first channel encoder 112, semantic relay device 200, semantic receiver 210, first channel decoder 211, semantic decoder 212, source encoder 213, second channel encoder 214, user equipment 300, bit decoder 310, second channel decoder 311, source decoder 312. Detailed implementation manners

[0052] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0053] In the description of the present invention, the meaning of several is one or more, the meaning of multiple is two or more, greater than, less than, exceeding, etc. are understood as not including the number itself, above, below, within, etc. are understood as including the number itself. If the first and second are described only for the purpose of distinguishing technical features, they cannot be understood as indicating or implying relative importance or implicitly indicating the quantity of the indicated technical features or implicitly indicating the sequence of the indicated technical features.

[0054] In the description of the present invention, unless otherwise clearly defined, words such as setting, installation, connection, etc. should be understood in a broad sense, and those skilled in the art can reasonably determine the specific meanings of the above words in the present invention in combination with the specific content of the technical solution.

[0055] The present invention provides a semantic relay system, a resource allocation method, an electronic device, and a computer-readable storage medium. The semantic relay system includes: a base station, the base station includes a semantic communication encoder for encoding source text information in the base station to generate semantic encoding information; a semantic relay device, the semantic relay device includes a semantic receiver, the semantic receiver is data-connected to the semantic communication encoder, and the semantic receiver is used for decoding the semantic encoding information to obtain semantic decoding information and performing source encoding processing on the semantic decoding information to obtain bit encoding information; a user device, the user device includes a bit decoder, the source decoder is data-connected to the semantic receiver, and the source decoder is used for decoding the bit encoding information to obtain the source text information; through the above technical solution, in the process of semantic communication, the semantic communication encoder of the base station is used for encoding the source text information in the base station to generate semantic encoding information; after the semantic receiver of the semantic relay device receives the semantic encoding information sent by the base station, it decodes the semantic encoding information to obtain semantic decoding information, and then performs source encoding processing on the semantic decoding information to obtain bit encoding information; then the semantic relay device will send the bit encoding information to the user device, and the bit decoder in the user device can decode the received bit encoding information, and finally the source text information can be obtained on the user device; in the whole process of semantic communication, the parts of semantic encoding and semantic decoding are set between the base station and the semantic relay device, which not only improves the efficiency of text transmission, but also well reduces the computing and storage burdens of the user device because the semantic decoding process is performed on the semantic relay device.

[0056] The embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0057] As Figure 1 shown, Figure 1 FIG. 10 is a semantic relay system provided by an embodiment of the present invention. The semantic relay system includes a base station 100, a semantic relay device 200, and a user device 300; wherein, the base station 100 includes a semantic communication encoder 110, and the semantic communication encoder 110 is used for encoding source text information in the base station 100 to generate semantic encoding information; the semantic relay device 200 includes a semantic receiver 210, the semantic receiver 210 is data-connected to the semantic communication encoder 110, and the semantic receiver 210 is used for decoding the semantic encoding information to obtain semantic decoding information and performing source encoding processing on the semantic decoding information to obtain bit encoding information; the user device 300 includes a bit decoder 310, the bit decoder 310 is data-connected to the semantic receiver 210, and the bit decoder 310 is used for decoding the bit encoding information to obtain the source text information.

[0058] It should be noted that during the semantic communication process, the semantic communication encoder 110 of the base station 100 is used to encode the source text information in the base station 100 to generate semantic encoded information; after the semantic receiver 210 of the semantic relay device 200 receives the semantic encoded information sent by the base station 100, it decodes the semantic encoded information to obtain semantic decoded information, and then performs source encoding on the semantic decoded information to obtain bit encoded information; then the semantic relay device 200 sends the bit encoded information to the user equipment 300, and the bit decoder 310 in the user equipment 300 can decode the received bit encoded information, and finally the source text information can be obtained on the user equipment 300; during the entire semantic communication process, the parts of semantic encoding and semantic decoding are set between the base station 100 and the semantic relay device 200, which not only improves the efficiency of text transmission, but also, since the semantic decoding process is carried out on the semantic relay device 200, well reduces the computational and storage burdens of the user equipment 300.

[0059] It is worth noting that by setting the semantic receiver 210 on the semantic relay device 200, semantic communication can be carried out between the base station 100 and the semantic relay device 200. Since semantic communication requires a lot of resources, therefore, for the base station 100 and the semantic relay device 200, it will not cause a large resource occupancy rate for both. However, for the user equipment 300, if the storage and decoding processing of semantic encoded information are carried out on the user equipment 300, it will cause a large resource occupancy rate for the user equipment 300 and affect the normal operation of the user equipment 300. Therefore, by establishing the semantic relay device 200 between the base station 100 and the user equipment 300 and setting the reception and decoding processes of semantic encoded information in the semantic relay device 200, it will not occupy the resources of the user equipment 300, and traditional bit information transmission is still carried out between the semantic relay device 200 and the user equipment 300; through the above technical solution, not only the efficiency of text transmission is improved, but also, since the semantic decoding process is carried out on the semantic relay device 200, well reduces the computational and storage burdens of the user equipment 300.

[0060] At the base station 100 end where computational and storage resources are relatively rich, the semantic communication encoder 110 includes a semantic encoder 111 and a first channel encoder 112. The semantic encoder 111 is connected to the first channel encoder 112. Among them, both the semantic encoder 111 and the first channel encoder 112 are used to encode the source text information in the base station 100 to generate semantic encoded information.

[0061] It should be noted that the semantic communication encoder 110 in the base station 100 includes a semantic encoder 111 and a first channel encoder 112. Both the semantic encoder 111 and the first channel encoder 112 are used to encode the source text information in the base station 100 to generate semantic encoded information, so as to make preparations for semantic communication with the semantic relay device 200.

[0062] Among them, the semantic encoder 111 is used to perform a first encoding process on the source text information in the base station 100 to obtain first encoded information, and the first channel encoder 112 is used to perform a first channel encoding process on the first encoded information to generate semantic encoded information.

[0063] In the innovative semantic relay structure, the semantic receiver 210 includes a first channel decoder 211, a semantic decoder 212, a source encoder 213, and a second channel encoder 214. The first channel decoder 211, the semantic decoder 212, the source encoder 213, and the second channel encoder 214 are connected in sequence. The first channel decoder 211 is data-connected to the first channel encoder 112. The first channel decoder 211 and the semantic decoder 212 are used to decode the semantic encoded information to obtain semantic decoded information. The source encoder 213 and the second channel encoder 214 are used to perform source encoding on the semantic decoded information to obtain bit-encoded information.

[0064] It should be noted that the semantic receiver 210 in the semantic relay device 200 includes a first channel decoder 211, a semantic decoder 212, a source encoder 213, and a second channel encoder 214. Among them, the first channel decoder 211 and the semantic decoder 212 are used to decode the semantic encoded information sent from the base station 100 to obtain semantic decoded information. Then, the source encoder 213 and the second channel encoder 214 can be used to perform source encoding on the semantic decoded information to obtain bit-encoded information, so that subsequent traditional bit information transmission with the user equipment 300 will not occupy too much network resources of the user equipment 300, and well relieve the computing and storage burden of the user equipment 300.

[0065] Among them, the first channel decoder 211 is used to perform a first channel decoding process on the semantic encoded information to obtain first decoded information, the semantic decoder 212 is used to perform a first decoding process on the first decoded information to obtain semantic decoded information, the source encoder 213 is used to perform a second encoding process on the semantic decoded information to obtain second encoded information, and the second channel encoder 214 is used to perform a second channel encoding process on the second encoded information to obtain bit-encoded information.

[0066] On the resource - limited mobile user side, the bit decoder 310 includes a second channel decoder 311 and a source decoder 312. The second channel decoder 311 is connected to the source decoder 312. The second channel decoder 311 is data - connected to the second channel encoder 214. The second channel decoder 311 and the source decoder 312 are used to decode the bit - encoded information to obtain the source text information.

[0067] It should be noted that the bit decoder 310 in the user equipment 300 includes a second channel decoder 311 and a source decoder 312. The second channel decoder 311 and the source decoder 312 are used to decode the bit - encoded information sent by the semantic relay device 200, so as to obtain the source text information. Through the above - mentioned technical solution, by means of the semantic relay device 200 for semantic information decoding, it does not occupy too much network resources of the user equipment 300, and there is no need to layout relevant semantic communication modules on the traditional user equipment 300. Among them, the user equipment 300 in the embodiments of the present application can be a mobile phone or other network devices with mobile communication functions.

[0068] Among them, the second channel decoder 311 is used to perform second - channel decoding on the bit - encoded information to obtain second - decoded information, and the source decoder 312 is used to perform second - decoding on the second - decoded information to obtain the source text information.

[0069] In order to enable the semantic relay system to achieve the maximum sum - rate, the present invention proposes a resource allocation method as shown in Figure 2 and is applied to the semantic relay system described in the first - aspect embodiments above. The method includes but is not limited to steps S100 to S600:

[0070] Step S100: Obtain the first initial distance and the first initial bandwidth ratio from the base station to the semantic relay device, the second initial distance and the second initial bandwidth ratio from the semantic relay device to the user equipment, and the initial transmission power of the semantic relay device.

[0071] Step S200: Determine the first channel gain according to the first initial distance and determine the second channel gain according to the second initial distance.

[0072] Step S300: Calculate the first signal - to - noise ratio according to the first channel gain and the first initial bandwidth ratio, and calculate the second signal - to - noise ratio according to the second channel gain and the second initial bandwidth ratio.

[0073] Step S400: Determine the semantic similarity according to the first signal - to - noise ratio; and calculate the initial bit transmission rate according to the semantic similarity and the first initial bandwidth ratio, and calculate the initial user achievable bit rate according to the second initial bandwidth ratio and the second signal - to - noise ratio.

[0074] Step S500: Determine an optimization problem based on the first channel gain, the second channel gain, the initial bit transmission rate, the initial user achievable bit rate, and the initial transmit power.

[0075] Step S600: Solve the optimization problem based on a preset optimization algorithm to obtain resource allocation information.

[0076] It should be noted that during the resource allocation process, first obtain the first initial distance and the first initial bandwidth ratio from the base station to the semantic relay device, the second initial distance and the second initial bandwidth ratio from the semantic relay device to the user equipment, and the initial transmit power of the semantic relay device; then determine the first channel gain according to the first initial distance and the second channel gain according to the second initial distance; then calculate the first signal-to-noise ratio based on the first channel gain and the first initial bandwidth ratio and calculate the second signal-to-noise ratio based on the second channel gain and the second initial bandwidth ratio; then determine the semantic similarity according to the first signal-to-noise ratio; and calculate the initial bit transmission rate according to the semantic similarity and the first initial bandwidth ratio, and calculate the initial user achievable bit rate according to the second initial bandwidth ratio and the second signal-to-noise ratio; then determine an optimization problem based on the first channel gain, the second channel gain, the initial bit transmission rate, the initial user achievable bit rate, and the initial transmit power; finally, solve the optimization problem based on a preset optimization algorithm to obtain resource allocation information. In the embodiments of the present application, by continuously optimizing the optimization problem, update and adjust the first initial distance, the first initial bandwidth ratio, the second initial distance, the second initial bandwidth ratio, and the initial transmit power, so that the semantic relay system can achieve faster and more stable semantic communication processing.

[0077] To illustrate how to allocate communication resources for different transmission links and how to determine the location of the semantic relay in the case of single-user text transmission Figure 3 The proposed optimization problem includes a single-user resource optimization problem, and the optimization algorithm includes a block coordinate descent optimization algorithm. The above step S600 may include but is not limited to steps S610 to S630:

[0078] Step S610: Solve the single-user resource optimization problem based on the block coordinate descent optimization algorithm to obtain a first sub-optimal solution.

[0079] Step S620: Update the penalty coefficient in the block coordinate descent optimization algorithm according to the first sub-optimal solution, and re-solve the single-user resource optimization problem using the updated block coordinate descent optimization algorithm until the first optimal solution of the single-user resource optimization problem is obtained.

[0080] Step S630: Determine the first optimal distance from the base station to the semantic relay device, the first optimal bandwidth ratio, the second optimal distance from the semantic relay device to the user equipment, and the second optimal bandwidth ratio according to the first optimal solution. The resource allocation information includes the first optimal distance, the second optimal distance, the first optimal bandwidth ratio, and the second optimal bandwidth ratio.

[0081] It should be noted that in the process of solving the optimization problem based on the preset optimization algorithm, first, the block coordinate descent optimization algorithm is used to solve the single-user resource optimization problem to obtain the first sub-optimal solution. Then, the penalty coefficient in the block coordinate descent optimization algorithm is updated according to the first sub-optimal solution, and the updated block coordinate descent optimization algorithm is used to re-solve the single-user resource optimization problem until the first optimal solution of the single-user resource optimization problem is obtained. Finally, the first optimal distance from the base station to the semantic relay device, the first optimal bandwidth ratio, the second optimal distance from the semantic relay device to the user equipment, and the second optimal bandwidth ratio are determined according to the first optimal solution. The resource allocation information includes the first optimal distance, the second optimal distance, the first optimal bandwidth ratio, and the second optimal bandwidth ratio. In the embodiment of the present application, the block coordinate descent optimization algorithm is used to re-solve the single-user resource optimization problem to obtain the first optimal distance, the second optimal distance, the first optimal bandwidth ratio, and the second optimal bandwidth ratio. In the subsequent semantic communication process, the network resource setting is adjusted based on the above parameters to improve the efficiency of subsequent text transmission.

[0082] It should be noted that steps S610 to S630 mainly target the single-user resource optimization problem, that is, the semantic relay device performs information transfer processing on a user equipment.

[0083] To illustrate how to optimize the system bandwidth and the transmission power of the semantic relay in the case of multi-user text transmission to maximize the sum rate of the multi-user text transmission system, Figure 4 The multi-user resource optimization problem is proposed. The optimization algorithm includes the block coordinate descent resource allocation algorithm. The above step S600 may include but is not limited to steps S640 to S660:

[0084] Step S640: Solve the multi-user resource optimization problem based on the block coordinate descent resource allocation algorithm to obtain the second sub-optimal solution.

[0085] Step S650: Update the initial transmission power, the first initial bandwidth ratio, and the second initial bandwidth ratio according to the second sub-optimal solution to obtain the second optimal solution of the multi-user resource optimization problem.

[0086] Step S660: Determine the optimal transmission power of the semantic relay device, the third optimal bandwidth ratio from the base station to the semantic relay device, and the fourth optimal bandwidth ratio from the semantic relay device to the user equipment according to the second optimal solution, where the resource allocation information includes the optimal transmission power, the third optimal bandwidth ratio, and the fourth optimal bandwidth ratio.

[0087] It should be noted that in the process of solving the optimization problem based on a preset optimization algorithm, first, the block coordinate descent resource allocation algorithm is used to solve the multi-user resource optimization problem to obtain the second-best solution; then, according to the second-best solution, the initial transmission power, the first initial bandwidth ratio, and the second initial bandwidth ratio are updated to obtain the second optimal solution of the multi-user resource optimization problem; finally, according to the second optimal solution, the optimal transmission power of the semantic relay device, the third optimal bandwidth ratio from the base station to the semantic relay device, and the fourth optimal bandwidth ratio from the semantic relay device to the user equipment are determined, where the resource allocation information includes the optimal transmission power, the third optimal bandwidth ratio, and the fourth optimal bandwidth ratio. In the embodiments of the present application, the block coordinate descent resource allocation algorithm is used to solve the multi-user resource optimization problem to obtain the optimal transmission power, the third optimal bandwidth ratio, and the fourth optimal bandwidth ratio, and in the subsequent semantic communication process, the network resource setting is adjusted based on the above parameters to improve the efficiency of subsequent text transmission.

[0088] It is worth noting that steps S640 to S660 mainly target the multi-user resource optimization problem, that is, the semantic relay device performs information transfer processing on multiple user equipment.

[0089] As Figure 5 shown, the above step S650 may include, but is not limited to, steps S651 to S652:

[0090] Step S651: Determine the first initial bandwidth ratio and the second initial bandwidth ratio as the adjusted bandwidth ratio.

[0091] Step S652: Update the initial transmission power and the adjusted bandwidth ratio alternately according to the second-best solution.

[0092] It should be noted that in the process of updating the initial transmission power, the first initial bandwidth ratio, and the second initial bandwidth ratio according to the second optimal solution to obtain the second optimal solution of the multi-user resource optimization problem, first, the first initial bandwidth ratio and the second initial bandwidth ratio are determined as the adjusted bandwidth ratios. Then, according to the second optimal solution, the initial transmission power and the adjusted bandwidth ratios are alternately updated. That is, in one stage of adjustment, during the process of updating the adjusted bandwidth ratios, the initial transmission power is maintained at the adjusted value of the previous adjustment stage; in the subsequent stage, during the process of updating the initial transmission power, the adjusted bandwidth ratios are maintained at the adjusted values of the previous adjustment stage, and so on for repeated adjustment.

[0093] To more clearly illustrate the specific process of the resource allocation method provided by the embodiments of the present invention, the following uses a specific example for illustration.

[0094] With the latest research progress of deep learning technology and the improvement of device computing and storage capabilities, deep neural networks have been widely used in the research of SemCom. Existing research work mainly focuses on three key aspects, namely the SemCom system architecture, semantic representation (or recovery), and resource management. Specifically, for the design of the SemCom system architecture, a new deep learning-based SemCom system (DeepSC) has been proposed in the prior art. It uses a Transformer network to build the SemCom system and designs a new method for joint semantic channel coding. To address the problem that DeepSC only transmits single-modal information, a new task-oriented multi-modal SemCom system (MU-DeepSC) and a structurally integrated multi-modal SemCom (U-DeepSC) system have been further proposed respectively to achieve efficient multi-mode data transmission. For the relevance problem of semantic representation, a cognitive SemCom system has been proposed. This framework uses a knowledge graph shared between the transmitter and the receiver. Experiments have shown that this system is superior to other benchmark systems in terms of data compression ratio and communication reliability. In addition, to utilize the deeper meaning of semantic information that is more useful, explicit and implicit semantics have been combined to propose a novel multi-layer semantic representation method. At the same time, the knowledge graph is also a promising method for representing and recovering semantic information and has received wide attention in text semantic transmission. For the resource management problem of SemCom, several representative performance evaluation metrics have also been proposed in recent research and used to guide the resource allocation of the SemCom system, such as semantic rate (S-R), semantic spectral efficiency (S-SE), and semantic similarity. For example, semantic similarity is used to evaluate the accuracy of semantic transmission and determine whether the semantic transmission can meet the transmission requirements.

[0095] In the embodiments of this application, two scenarios of a single - user and a multi - user system assisted by a semantic relay device are considered. And the achievable weighted sum (bit) rate is maximized, and a joint optimization algorithm for resource allocation is designed for the two scenarios respectively. Numerical results prove the effectiveness of the proposed algorithm, and compared with the traditional decode - and - forward relay, the proposed semantic relay device achieves significant communication performance gains.

[0096] In a communication system assisted by a semantic relay device, the base station transmits text information to N users, and both the BS and the users are equipped with single antennas (this work can be easily extended to the case of multi - antenna base stations and users). Since the BS has rich storage and computing resources, it can perform SemCom by deploying a well - trained DeepSC encoder, while the storage and computing capabilities of users are limited, so they cannot decode the semantic information transmitted by the base station. In addition, it is assumed that the direct link between the base station and the users is blocked due to long distances and obstacles.

[0097] To achieve efficient text transmission from the base station to user equipment, a semantic relay device with a DeepSC decoder and rich computing resources is appropriately deployed to assist text transmission in two stages. Specifically, in the first stage, the base station extracts and transmits the semantic information of the source text through its DeepSC transmitter and sends it to the semantic relay device. In the second stage, the semantic relay device first decodes the text information from the received signal using its channel and semantic decoders, and then forwards the decoded text to the corresponding user equipment through traditional bit - based information transmission. It is assumed that the semantic decoding delay of the semantic relay device can be ignored because it has rich computing resources. In addition, to avoid communication interference, a frequency - division multiple - access scheme is adopted for text transmission assisted by the semantic relay device; among them, semantic transmission is achieved on the link from the base station to the semantic relay device; bit transmission is achieved on the link from the semantic relay device to the user equipment, and both can be carried out on orthogonal frequency bands.

[0098] In the embodiments of this application, a new SemRelay - based text - transmission communication system is proposed for the first time, and the channel modeling of each communication link of the system is given below. Among them, SemRelay is the semantic relay device referred to in the above - mentioned embodiments, and BS is the base station referred to in the above - mentioned embodiments.

[0099] For BS - SemRelay semantic transmission: Let s denote the source text information transmitted by the BS. To extract semantic information from s and accurately transmit it through the wireless fading channel, s is encoded using semantic and channel encoders respectively. Specifically, the encoded information is expressed as

[0100]

[0101] where denotes the semantic-channel joint encoder, and M is the average number of semantic symbols obtained by encoding the text information s. In addition, under the line-of-sight dominant wireless channel, let h br denote the channel gain from the BS to the SemRelay, B denote the total bandwidth of the system, and α br be the bandwidth ratio allocated to the BS→SemRelay link. Then, the received signal-to-noise ratio γ br (in dB) at the SemRelay can be expressed as

[0102]

[0103] where P b denotes the transmit power of the BS, and N0 denotes the power spectral density of the AWGN. Different from the traditional bit-based transmission scheme, semantic transmission inputs each sentence into the DeepSC transceiver to generate a semantic symbol vector where L denotes the average number of words per sentence, and K denotes the average number of semantic symbols per word in the source sentence. In addition, I denotes the semantic unit (unit: suts) that measures the average semantic information of each sentence. Therefore, I / L (unit: suts / word) represents the average amount of semantic information contained in each word of each sentence. To evaluate the performance of the DeepSC model, a new performance metric for measuring the distance of semantic information between two sentences, called semantic similarity, is proposed, and it is pointed out that the semantic similarity is related to the received SNR γb r and K. For any given K, the semantic similarity function ε K (γ br ) ∈ [0,1] is generally "S"-shaped with respect to the signal-to-noise ratio. Therefore, it can be approximated by the sigmoid function as shown below using the generalized logistic regression method

[0104]

[0105] where a1, a2, c1, and c2 are all constant coefficients determined by K. To ensure the accuracy of the recovered data, ε K (γ br ) should satisfy the minimum semantic similarity requirement, that is Based on the above-obtained semantic similarity model, the achievable semantic rate (unit: suts / s) of the BS→SemRelay link can be expressed as

[0106]

[0107] In addition, in traditional text transmission, μ (unit: bits / word) represents the average number of bits per word. Then, the semantic rate of the BS→SemRelay link can be converted into the following effective rate expression based on bit transmission

[0108]

[0109] For SemRelay-user bit transmission: After receiving the semantic coding signal, the SemRelay first decodes the text semantic information by the DeepSC receiver, and then forwards it to the user in the way of traditional bit transmission. Let represent the channel gain from the SemRelay to the user Then, the achievable bit rate (unit: bps) of user n is

[0110]

[0111] where P r (n) represents the SemRelay transmission power allocated to user n, represents the bandwidth ratio allocated to the SemRelay→user link n.

[0112] Next, the SemRelay-assisted single-user and multi-user systems are studied separately. For the single-user system, the achievable bit rate is maximized by jointly optimizing the SemRelay location deployment and system resource allocation. For this problem, a penalty-based two-layer algorithm is proposed. Given the penalty coefficient, the inner-layer alternately iteratively optimizes the variables using the Block Coordinate Descent (BCD) and Successive Convex Approximation (SCA) algorithms; the outer-layer updates the penalty coefficient. For the multi-user system, the weighted bit rate of all users is maximized by jointly optimizing the SemRelay transmission power and system bandwidth resource allocation, and an algorithm based on BCD and SCA is used to solve this problem. Among them, the block coordinate descent method in the embodiments of this application is the above-mentioned block coordinate descent optimization algorithm.

[0113] Solution 1: Optimization design of location deployment and resource allocation for the SemRelay-assisted single-user text transmission system

[0114] For the considered SemRelay-assisted single-user text transmission system, i.e., N = 1. Therefore, for the sake of convenient expression, the superscripts of all variables in the above achievable bit rate formula are omitted hereinafter, that is, the single-user achievable rate expression is Considering the line-of-sight dominant channel model, the channel gain expressions for the BS→SemRelay link and the SemRelay→user link are as follows:

[0115]

[0116] where d br and d ru represent the horizontal distances from the SemRelay to the BS and the user respectively, ρ0 represents the channel power gain at the reference distance d0 = 1 m, and β represents the path loss exponent.

[0117] The goal is to maximize the end-to-end achievable bit rate by jointly optimizing the location deployment of the SemRelay and the system bandwidth allocation Substituting the above gain formulas into the rate expression respectively, this optimization problem can be formulated as

[0118] (P1-1)

[0119]

[0120] α br ≥0, α ru ≥0, (9e)

[0121] α br + α ru =1, (9f)

[0122] d br ≥0, d ru ≥0, (9g)

[0123] d br + d ru =D, (9h) (7)

[0124] where (9a) and (9b) are the rate constraints based on bit and semantic transmissions respectively, (9f) is the total bandwidth constraint, and (9c) is the minimum semantic similarity constraint to be satisfied. Among them, (9d) can be equivalently transformed into the following form:

[0125]

[0126] Combining the received signal-to-noise ratio γ br formula at the SemRelay side, the channel gain formula from the BS to the SemRelay, and the constraint formula that (9d) can be equivalently transformed, the upper bound of the bandwidth required for semantic transmission can be obtained: This formula indicates that semantic communication is more advantageous in the low-bandwidth region.​

[0127] Since the constraints (9a)-(9b) are non-convex optimization problems with variable coupling, problem (P1-1) is a non-convex optimization problem. Although the optimal solution of problem (P1-1) can be obtained by two-dimensional exhaustive search, it brings extremely high computational complexity, which is unbearable in practice. In addition, numerical results show that the alternating optimization (AO) method cannot be directly used to solve this problem because the optimization variables are strongly coupled in (9a), (9b), and (9d), which makes the AO method fall into a low-quality local solution.

[0128] To solve the above problems, a penalty-based two-layer algorithm is proposed to sub-optimally solve problem (P1-1). Specifically, given a fixed penalty coefficient, the inner layer applies the block coordinate descent (BCD) method to solve the penalty optimization problem, and the outer layer updates the penalty coefficient until the algorithm converges, and finally a high-quality sub-optimal solution of (P1) is obtained. The specific algorithm is given in the following embodiments.

[0129] Reconstruction of the original problem:

[0130] First, to solve the variable coupling problem in constraints (9a), (9b), and (9d), a set of auxiliary variables is introduced for constraints (9f) and (9h). And define

[0131]

[0132] Therefore, constraints (9f) and (9h) can be expressed as

[0133]

[0134] Replacing constraints (9f) and (9h) with equations (10) and (11) respectively, problem (P1-1) is equivalently transformed into

[0135] (P1-2)

[0136] s.t. (9a), (9b), (9d), (9e), (9g), (10)-(14)(13)

[0137] Next, the equality constraints (11) and (12) are added as penalty terms to the objective function of problem (P1-2), thus obtaining the following problem

[0138] (P1-3)

[0139] s.t. (9a), (9b), (9d), (9e), (9g), (10), (13), (14), (14)

[0140] where λ represents the penalty coefficient for penalizing the equality constraints in (P1-2), and ν is the weight for balancing the influence magnitudes of the bandwidth and distance penalty terms on the objective function. By gradually decreasing the value of λ, a solution that satisfies all the equality constraints in (P1-2) within a predefined accuracy range for (P1-3) can be obtained.

[0141] However, given λ > 0, due to the non-convex constraints of (9a), (9b), and (9d), problem (P1-3) remains a non-convex problem. To solve this problem, the BCD method is applied to divide problem (P1-3) into three sub-problems: 1) SemRelay deployment sub-problem, 2) bandwidth allocation sub-problem, and 3) auxiliary variable optimization sub-problem. The above three sub-problems are alternately iteratively optimized and solved until convergence.

[0142] Inner Layer: Solving Problem (P1-3) by the BCD Method

[0143] 1) Location Deployment Optimization:

[0144] For any given bandwidth allocation α and auxiliary variable z, problem (P1-3) can be simplified to the following optimization sub-problem regarding the SemRelay location deployment

[0145] (P1-4)

[0146] s.t. (9a), (9b), (9d), (9g), (10), (15)

[0147] Due to the non-convex constraints of (9a), (9b), and (9d), problem (P1-4) is a non-convex problem. The successive convex approximation (SCA) method can be used to effectively solve this problem, and the specific solution method is as follows.

[0148] Lemma 1. For constraint (9a), is a convex function with respect to Given any local point The lower bound is given by the following expression

[0149]

[0150] where the expressions for the coefficients E1 and E2 are and

[0151] Lemma 2. For constraint (9b), is about a convex function of. Given any local point, the lower bound of ψ is the expression shown below

[0152]

[0153] where χ = c1γ br + c2, and the coefficients The expressions of E3 and E4 are respectively and

[0154] Next, for the non - affine constraint (9d), first relax it to the expression shown below

[0155]

[0156] It can be proved by contradiction that the equality in constraint (20) always holds in the optimal solution of the relaxation problem. In addition, there are the following results.

[0157] Lemma 3. For constraint (20), is a concave function of Given any local point The lower bound of φ is the expression as follows

[0158]

[0159] where the expressions of coefficients E5 and E6 are respectively and

[0160] Based on Lemmas 1 - 3, replace the in constraint (9a), ψ in constraint (9b), and φ in constraint (17) with the corresponding lower bounds or upper bounds shown in (18), (19), and (21) respectively, then problem (P1 - 4) can be approximated as the following optimization problem

[0161] (P1 - 5)

[0162]

[0163] (9g),(10).(20)

[0164] Problem (P1 - 5) is a convex optimization problem and can be solved using a CVX solver.

[0165] 2) Bandwidth allocation optimization problem:

[0166] For any given deployment \(d\) of the SemRelay and the auxiliary variable \(z\), problem (P1-3) can be simplified to the following optimization sub-problem regarding bandwidth allocation

[0167] (P1-6)

[0168] s.t. (9a), (9b), (9d), (9e), (10), (21)

[0169] Due to the non-convexity of constraints (9b) and (9d), problem (P1-6) is a non-convex optimization problem. To solve this difficult problem, first, the constraints of (P1-6) are equivalently rewritten as the following expressions

[0170]

[0171] Although (24) is still a non-convex optimization constraint, it can be processed using the SCA method

[0172] Lemma 4. For constraint (24), is a convex function with respect to \(\alpha\) br and \(S\). Given any local point and The lower bound of \(\delta\) is the following expression

[0173]

[0174] Based on Lemma 4, constraint (24) can be approximated as the following constraint

[0175]

[0176] Next, for the non-affine constraint (25), first, it is relaxed to the following expression

[0177]

[0178] It can be proven by contradiction that the equality in constraint (28) always holds in the optimal solution of the relaxed problem, so it does not affect the optimality of the optimization result

[0179] In addition, it can be observed that, given \(d\) and \(z\), the non-convex constraints (28) and (9d) in problem (P1-6) have similar forms to the constraints (9b) and (9d) in problem (P1-4) respectively. Following similar steps based on the SCA method, the constraints (28) and (9d) in problem (P1-6) can be transformed into

[0180]

[0181] where \(\tau = c1\gamma\) br+c2. Given a local point and The coefficients in the above formula are respectively and

[0182] By substituting the constraints (9b) and (9d) in problem (P1-6) with the constraints shown in (27), (29) and (30), problem (P1-6) can be approximated as the following problem

[0183] (P1-7)

[0184] s.t. (9a), (9e), (10), (27), (29), (30). (29)

[0185] Since the known problem (P1-7) is a convex optimization problem, the CVX solver can be used to solve it.

[0186] 3) Auxiliary variable optimization:

[0187] For any given position deployment d and bandwidth allocation of SemRelay, problem (P1-3) can be simplified to the following optimization sub-problem regarding the auxiliary variable

[0188] (P1-8)

[0189] s.t. (13), (14). (30)

[0190] Since the known problem (P1-8) is a convex problem, the CVX solver can be used to solve it.

[0191] In summary, the sub-problems (P1-5), (P1-7) and (P1-8) can be solved by alternating iteration until the algorithm converges, and finally a sub-optimal solution to problem (P1-3) can be obtained.

[0192] Phase 2: Outter Layer: Update the loss coefficient

[0193] The objective of the outer-layer is to update the penalty coefficient λ with λ = cλ to ensure that the convergence result can satisfy the equality constraint in (P1-2), where c is a constant scaling factor.

[0194] Scheme 2: Joint allocation design of the transmission power and system bandwidth of the SemRelay-assisted multi-user text transmission system:

[0195] For the considered SemRelay-assisted multi-user text transmission system, where the SemRellay is deployed near the users. The channel model is considered to be a Rice channel. The goal is to maximize the multi-user weighted total rate (i.e., the effective bit rate) by jointly optimizing the SemRelay transmit power allocation and the system bandwidth allocation The optimization problem can be formulated in the following form:

[0196] (P2-1)

[0197]

[0198] where, ω (n) is the weight of the nth user, represents the total transmit power of the SemRelay. Among them, (33a) is the information causality constraint, that is, the SemRelay can only forward the information it has received to the users. In addition, the constraint condition (33c) is the same as the constraint (9d) in the problem (P1-1), which represents the minimum semantic similarity requirement. Therefore, (33c) can be rewritten as a linear constraint as shown in (10). The constraint conditions (33d)-(33g) are the total bandwidth constraint condition and the transmit power constraint condition of the SemRelay respectively.

[0199] Since (33a) and (33b) are non-convex constraints, (P2-1) is a non-convex optimization problem, and there is a serious variable coupling phenomenon in the constraint (33a) in the problem (P2-1). To address these problems, an efficient algorithm is proposed to obtain a high-quality suboptimal solution.

[0200] Problem reconstruction:

[0201] To solve the problem (P2-1), a set of slack variables is first introduced Therefore, there is

[0202] (P2-2)

[0203]

[0204] (10),(33b),(33d)-(33g).(32)

[0205] The solution of the problem (P2-2) can be obtained by solving a relaxed optimization problem, which is shown as follows.

[0206] Lemma 1. By relaxing the equality constraints in (33b) and (34b), the solution of the problem (P2-2) can be obtained by solving the following problem

[0207] (P2-3)

[0208]

[0209] (10),(33d)-(33g),(34a).(33)

[0210] However, due to the constraints in (34a), (35a), and (35b), problem (P2-3) remains a non-convex problem. To solve this problem, a BCD-based algorithm is proposed to decompose problem (P2-3) into two optimization sub-problems: 1) the SemRelay transmit power allocation P r sub-problem, 2) the system bandwidth allocation α sub-problem. Subsequently, these two sub-problems are solved by alternating iteration until the algorithm converges. Among them, the BCD algorithm in the embodiments of this application is the block coordinate descent resource allocation algorithm mentioned above.

[0211] Phase 1: SemRelay transmit power allocation

[0212] For any given system bandwidth allocation α, problem (P2-3) is simplified to the following SemRelay transmit power allocation optimization problem

[0213] (P2-4)

[0214] s.t.(10),(33f),(33g),(34a),(35a),(35b).(34)

[0215] Due to the non-convex constraint conditions in (34a) and (35a), problem (P2-4) is a non-convex optimization problem. The SCA method can be used to solve them, and the specific method is as follows.

[0216] Lemma 2. For the constraint condition (34a), R br is a convex function with respect to . Given any local point R br its lower bound can be taken as

[0217]

[0218] where the coefficients F1, F2 are

[0219] Lemma 3. For the constraint condition (35a), is a concave function with respect to P r (n) . Given any local point its upper bound can be taken as

[0220]

[0221] where the coefficients F3 and F4 are

[0222] Based on Theorems 2 and 3, replace the RHS of (34a) and (35a) with the corresponding lower or upper bounds, i.e., R in (37) br(lb) and in (38) Problem (P2-4) can be transformed into the following approximate form

[0223] (P2-5)

[0224]

[0225] (10), (33f), (33g), (35b), (37)

[0226] At this time, problem (P2-5) is a convex optimization problem and can be efficiently solved using a CVX solver.

[0227] Phase II: System bandwidth allocation

[0228] For any given SemRelay transmit power P r , problem (P2-3) can be simplified to the following optimization sub-problem of bandwidth allocation

[0229] (P2-6)

[0230] s.t. (10), (33d), (33e), (34a), (35a), (35b), (38)

[0231] Since (34a), (35a), and (35b) are non-convex constraints, problem (P2-6) is a non-convex optimization problem. In addition, constraint (34a) shows strong coupling between the optimization variables α br and γ br , which poses a great challenge to solving the problem. To solve this problem, first introduce an auxiliary variable

[0232]

[0233] Then the constraint condition (34a) can be equivalently rewritten as

[0234]

[0235] It can be seen that (42) is still a non-convex constraint. Given that constraints (41) and (42) have the same form as constraints (24) and (25) in problem (P1-6), the same steps as the SCA-based method can be followed to handle constraints (41) and (42), and the following constraints are obtained

[0236]

[0237] where is the function at any given local point and is the lower bound function obtained by the first-order Taylor series expansion of the function. The coefficients E5 and E6 are expressed as

[0238] In addition, for the non-convex constraint conditions (35a) and (35b), the SCA method can also be effectively used to solve them.

[0239] Lemma 5. For the constraint condition (35a), the right side of the inequality is a concave function related to and its upper bound is

[0240]

[0241] where the coefficients E7 and E8 are expressed as

[0242] Lemma 6. For the constraint condition (35b), is a convex function of α br . For any local point ψ can take its lower bound as

[0243]

[0244] where the coefficients E9 and E 10 are expressed as

[0245] According to Lemmas 4-6 and (43), (44), problem (P2-7) can be transformed into the following approximate problem

[0246] (P2-7)

[0247]

[0248] γ br ≤ψ (lb) ,(47b)

[0249] (10),(33d),(33e),(43),(44),(45)

[0250] At this time, problem (P2-7) is a convex optimization problem and can be efficiently solved using a CVX solver.

[0251] Simulation analysis:

[0252] In the embodiments of the present application, the simulation results of single-user and multi-user communication systems are respectively subjected to simulation analysis to evaluate the effectiveness of the proposed algorithm. To ensure reliable transmission, the required semantic similarity is set to Consider using the ASCII encoding method to encode each letter. This method uses 8-bit binary numbers to represent a character and can represent 256 different characters. Assuming that the average number of letters per word is 5, the average number of bits per word is μ = 40 bits / word.

[0253] SemRelay-assisted single-user system:

[0254] The embodiments of the present application study the performance of a SemRelay-assisted single-user system. The horizontal distance between the BS and the user is 100 m, and the relay is deployed at a height of 10 m. Considering that the system channel is a LoS channel model, the path loss is ρ0 = -60 dB, and the path loss exponent is set to β = 3. The average number of semantic symbols per word is set to K = 4, and the corresponding parameters of the semantic similarity function in formula (4) can be numerically calculated as a1 = 0.3980, a2 = 0.5385, c1 = 0.2815, and c2 = -1.3135. Other parameter settings are respectively λ = 1000, c = 0.9, ν = 10 -4 , P b = P r = 0.1 W, N0 = -169 dBm / Hz, and ∈1 = 10 -8 . The following four benchmark schemes are considered for performance comparison: 1) Design of a SemRelay communication system based on the exhaustive search method; 2) Optimize the SemRelay location deployment strategy alone when the system bandwidth allocation is equal; 3) Fix the SemRelay location in the middle of the BS and the User and optimize the system bandwidth allocation strategy alone; 4) An optimization algorithm for joint relay location deployment and system bandwidth allocation for a traditional DF relay-assisted communication system.

[0255] As Figure 6 shown, Figure 6 represents the achievable bit rate obtained by different schemes as the total bandwidth B changes. First, it can be observed that for the SemRelay-assisted text transmission system, the proposed loss-based algorithm can achieve nearly optimal performance comparable to the exhaustive search algorithm.

[0256] Second, when B is small, the proposed SemRelay-assisted communication system can achieve a higher bit rate than the traditional decode-and-forward relay because the BS→User semantic transmission link only transmits the semantic information extracted from the original text, thus achieving a higher spectral efficiency. Therefore, in the case of limited bandwidth, the bandwidth resource allocation design can bring more degrees of freedom to further improve the performance of the SemRelay-assisted communication system. On the other hand, when B is large enough, the traditional decode-and-forward relay has a tendency to achieve a larger effective bit rate than SemRelay because the minimum semantic similarity requirement limits the upper bound of the bandwidth required for semantic transmission, thus resulting in a slower growth rate of the rate with the increase of the total bandwidth. In addition, the proposed penalty-based algorithm significantly outperforms the benchmark schemes that separately optimize the bandwidth allocation or the SemRelay location deployment, while the scheme that only optimizes the bandwidth allocation separately achieves better performance than the scheme that only optimizes the location deployment. Moreover, in the scheme of independently optimizing the bandwidth allocation, when the total bandwidth is large, the achievable rate obtained by SemRelay under the optimal bandwidth allocation does not increase because, given a fixed SemRelay location deployment, the minimum semantic similarity requirement limits the upper bound of the bandwidth allocated to the BS→SemRelay link (see Figure 7 ).

[0257] Figure 7 And Figure 8 respectively represent the relationships between the optimal bandwidth allocation and the relay location deployment obtained by different schemes and the total bandwidth B.

[0258] As can be seen from Figure 7 , different from the traditional decode-and-forward relay system with equal bandwidth allocation, in the SemRelay-assisted system, the BS→SemRelay link is allocated less bandwidth than the SemRelay→User link because semantic communication only needs to transmit compressed semantic information and can achieve a higher spectral efficiency than traditional bit transmission.

[0259] As can be seen from Figure 8 , to facilitate downlink communication, when the total bandwidth decreases, SemRelay should be deployed closer to the user. This is because the spectral efficiency of the SemRelay→User bit transmission link is lower than that of the BS→SemRelay link based on semantic transmission. Therefore, in the case of less bandwidth resources, the performance of the SemRelay→User link needs to be improved by reducing the path loss, and the spectral efficiency of this link is lower than that of the BS→SemRelay link.

[0260] SemRelay-assisted multi-user system:

[0261] In the embodiments of this application, the performance of a multi-user system assisted by SemRelay is studied. It is assumed that there are N = 10 users in the system, and these users are randomly and uniformly distributed within a maximum distance of 40 m from SemRelay. Considering that all communication channels in the system are Rice channel models with a Rice factor of 20 dB and the BS-SemRelay distance is d br = 60 m. The path loss is ρ0 = -60 dB, and the path loss exponent is set to β = 3.5. In addition, it is assumed that the weights of all users are the same, i.e., ω (n) = 1, The average number of semantic symbols per word is set to K = 5, and the corresponding parameters of the semantic similarity function in formula (4) can be numerically calculated as a1 = 0.3760, a2 = 0.5970, c1 = 0.2634, and c2 = -0.8151

[11] . Other parameters are set to N0 = -169 dBm / Hz, τ = 10 -6 and P b = 2 W. To verify the performance advantages of the proposed algorithm, the following three benchmark schemes are considered for performance comparison: 1) Separate optimization of the SemRelay transmit power is adopted when the system bandwidth allocation is equal; 2) Separate optimization of the system bandwidth allocation is adopted when the SemRelay transmit power allocation is equal; 3) The traditional decode-and-forward relay with joint allocation of the DF relay transmit power and the system bandwidth is adopted.

[0262] As Figure 9 shown, Figure 9 represents the relationship between the (effective) weighted sum rate achieved by various schemes and the total bandwidth B when the total transmit power of SemRelay is . First, the weighted sum rate achieved by the proposed BCD-based algorithm is much higher than the scheme that only adopts power allocation. Compared with the baseline scheme that adopts bandwidth optimization, the improvement in its performance decreases. This indicates that bandwidth allocation has a greater impact on improving the weighted sum rate. Second, the performance of the SemRelay-assisted communication system is significantly better than that of the traditional decode-and-forward relay, especially in the small bandwidth region. This is because the BS-SemRelay transmission link only transmits the semantic information extracted from the original text, thus improving the spectral efficiency. It should be noted that the semantic similarity of data transmission is related to the total bandwidth. When the total bandwidth increases, the noise that appears in data transmission will have a negative impact on the transmitted semantic symbols, resulting in a decrease in semantic similarity.

[0263] In addition, for strict SemCom, i.e., SemCom with high semantic similarity requirements, there is a certain upper limit on the system bandwidth to ensure the minimum semantic similarity required for high-quality semantic transmission. When the bandwidth is large enough (e.g., B > 17 MHz), the power optimization scheme with a fixed bandwidth cannot achieve the required semantic similarity, so the required data transmission cannot be realized, and the achievable bit rate is 0.

[0264] As Figure 10 shown, Figure 10 it shows the relationship between the weighted sum rate achieved by different schemes and the SemRelay transmit power when the total bandwidth is B = 10 MHz. It can be seen that when is greater than 1 W, the growth rate of the weighted sum rate of all schemes gradually slows down with the increase of .

[0265] In addition, by comparing Figure 9 and Figure 10 , it can be concluded that in terms of the system weighted sum rate, the impact of system bandwidth allocation is more significant than that of SemRelay transmit power allocation.

[0266] Through the above technical solutions, for the SemRelay-assisted single-user communication system, an optimization problem is constructed to maximize the end-to-end achievable bit transmission rate by jointly optimizing the location deployment of SemRelay and bandwidth resource allocation. A penalty-based two-layer algorithm that can obtain the optimal solution is proposed. Among them, when a fixed penalty coefficient is given, the inner-layer uses an optimization algorithm based on BCD to jointly design the location deployment of SemRelay and bandwidth resource allocation; the outer-layer updates the penalty coefficient. And the simulation results show that the SemRelay-assisted communication system can always obtain better performance than the traditional relay-assisted communication system when the bandwidth resources are scarce; in addition, different from the traditional DF relay-assisted communication system, SemRelay is not always deployed in the middle of the BS and the user, but gradually approaches the user end with the increase of bandwidth resources.

[0267] In addition, for the SemRelay-assisted multi-user communication system, an optimization problem is constructed to jointly optimize the SemRelay transmission power and system bandwidth allocation to maximize the achievable weighted sum (bit) rate. A BCD-based resource allocation algorithm is proposed to solve the above optimization problem, and a high-quality sub-optimal solution is obtained by alternately optimizing the SemRelay transmission power allocation and system bandwidth allocation (with the other fixed). Simulation results show that the SemRelay-assisted multi-user communication system can always achieve better performance than the traditional relay-assisted communication system when the system bandwidth resources are scarce and the SemRelay transmission power is insufficient; in addition, the impact of system bandwidth allocation on the system weighted sum rate is more significant than that of SemRelay transmission power allocation.

[0268] To solve the problem that the user side cannot support semantic communication due to insufficient computing resources and to achieve better communication efficiency, the present invention proposes a novel communication relay (SemRelay)-assisted text transmission communication system. Specifically, the proposed SemRelay is equipped with a DeepSC receiver, allowing semantic transmission on the BS→SemRelay link and bit transmission on the SemRelay→user link. At the same time, for the SemRelay-assisted single-user communication system, an optimization problem is constructed to maximize the end-to-end achievable bit transmission rate by jointly optimizing the location deployment of SemRelay and bandwidth resource allocation. And a penalty-based two-layer algorithm that can obtain the optimal solution is proposed for this problem. Simulation results show that the SemRelay-assisted communication system can always achieve better performance than the traditional relay-assisted communication system when the bandwidth resources are scarce; in addition, different from the traditional decode-and-forward relay-assisted communication system, SemRelay is not always deployed in the middle of the BS and the user, but gradually approaches the user side as the bandwidth resources increase. In addition, for the considered multi-user system, an optimization problem is proposed, that is, to maximize the weighted sum (bit) rate by jointly designing the SemRelay transmission power allocation and system bandwidth allocation. Although this problem is a non-convex problem that is difficult to solve, Solution 2 proposes an efficient algorithm to obtain the corresponding sub-optimal solution using the BCD and SCA methods. Numerical results show that the proposed SemRelay has significant rate performance gains compared with the traditional decode-and-forward relay and other benchmarks. The solution can be extended to resource management research in other SemCom systems, such as semantic-aware image / video transmission and knowledge-graph-based semantic transmission.

[0269] In some embodiments of the present invention, such as Figure 11As shown, an embodiment of the present invention further provides an electronic device 700, including: a memory 720, a processor 710, and a computer program stored on the memory 720 and executable on the processor 710. When the processor 710 executes the computer program, it implements the resource allocation method in the above embodiment. For example, it executes the method steps S100 to S600 described above Figure 2 in the method steps S100 to S600, Figure 3 in the method steps S610 to S630, Figure 4 in the method steps S640 to S660, and Figure 5 in the method steps S651 to S652.

[0270] In some embodiments of the present invention, an embodiment of the present invention further provides a computer-readable storage medium storing computer-executable instructions. When the computer-executable instructions are executed by a processor or a controller, for example, executed by a processor in the above device embodiment, the above processor can be caused to execute the resource allocation method in the above embodiment. For example, it executes the method steps S100 to S600 described above Figure 2 in the method steps S100 to S600, Figure 3 in the method steps S610 to S630, Figure 4 in the method steps S640 to S660, and Figure 5 in the method steps S651 to S652.

[0271] Those of ordinary skill in the art will understand that all or some of the steps and systems disclosed above in the methods can be implemented as software, firmware, hardware, and their appropriate combinations. Some physical components or all physical components can be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or can be implemented as hardware, or can be implemented as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include a computer storage medium (or non-transitory medium) and a communication medium (or transitory medium). As is well known to those of ordinary skill in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules, or other data. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, as is well known to those of ordinary skill in the art, communication media typically contain computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transmission mechanism, and can include any information delivery medium.

[0272] The above has specifically described the preferred embodiments of the present invention, but the present invention is not limited to the above embodiments. Those skilled in the art can make various equivalent deformations or substitutions without departing from the spirit of the present invention, and these equivalent deformations or substitutions are all included within the scope defined by the claims of the present invention.

Claims

1. A resource allocation method, characterized in that, Applied to a semantic relay system, the semantic relay system includes a base station, a semantic relay device, and a user device. The base station includes a semantic communication encoder for encoding source text information in the base station to generate semantic encoded information. The semantic relay device includes a semantic receiver, which is data-connected to the semantic communication encoder. The semantic receiver is used to decode the semantic encoded information to obtain semantic decoded information and perform source encoding on the semantic decoded information to obtain bit-encoded information. The user device includes a bit decoder, which is data-connected to the semantic receiver. The bit decoder is used to decode the bit-encoded information to obtain the source text information. The method includes: Obtaining a first initial distance and a first initial bandwidth ratio from the base station to the semantic relay device, a second initial distance and a second initial bandwidth ratio from the semantic relay device to the user device, and an initial transmission power of the semantic relay device. Determining a first channel gain according to the first initial distance and a second channel gain according to the second initial distance. Calculating a first signal-to-noise ratio according to the first channel gain and the first initial bandwidth ratio, and calculating a second signal-to-noise ratio according to the second channel gain and the second initial bandwidth ratio. Determining a semantic similarity according to the first signal-to-noise ratio; calculating an initial bit transmission rate according to the semantic similarity and the first initial bandwidth ratio, and calculating an initial user achievable bit rate according to the second initial bandwidth ratio and the second signal-to-noise ratio. Determining an optimization problem based on the first channel gain, the second channel gain, the initial bit transmission rate, the initial user achievable bit rate, and the initial transmission power. Solving the optimization problem based on a preset optimization algorithm to obtain resource allocation information.

2. The resource allocation method according to claim 1, characterized in that, The optimization problem includes a single-user resource optimization problem. The optimization algorithm includes a block coordinate descent optimization algorithm. Solving the optimization problem based on the preset optimization algorithm to obtain resource allocation information includes: Solving the single-user resource optimization problem based on the block coordinate descent optimization algorithm to obtain a first sub-optimal solution. Updating a penalty coefficient in the block coordinate descent optimization algorithm according to the first sub-optimal solution, and re-solving the single-user resource optimization problem using the updated block coordinate descent optimization algorithm until a first optimal solution of the single-user resource optimization problem is obtained. Determining a first optimal distance and a first optimal bandwidth ratio from the base station to the semantic relay device and a second optimal distance and a second optimal bandwidth ratio from the semantic relay device to the user device according to the first optimal solution. The resource allocation information includes the first optimal distance, the second optimal distance, the first optimal bandwidth ratio, and the second optimal bandwidth ratio.

3. The resource allocation method according to claim 1, characterized in that, The optimization problem includes a multi-user resource optimization problem, the optimization algorithm includes a block coordinate descent resource allocation algorithm, and solving the optimization problem based on the preset optimization algorithm to obtain resource allocation information includes: Solving the multi-user resource optimization problem based on the block coordinate descent resource allocation algorithm to obtain a sub-optimal solution; Updating the initial transmit power, the first initial bandwidth ratio, and the second initial bandwidth ratio according to the sub-optimal solution to obtain the second optimal solution of the multi-user resource optimization problem; Determining the optimal transmit power of the semantic relay device, the third optimal bandwidth ratio from the base station to the semantic relay device, and the fourth optimal bandwidth ratio from the semantic relay device to the user equipment according to the second optimal solution, where the resource allocation information includes the optimal transmit power, the third optimal bandwidth ratio, and the fourth optimal bandwidth ratio.

4. The resource allocation method according to claim 3, characterized in that, The updating the initial transmit power, the first initial bandwidth ratio, and the second initial bandwidth ratio according to the sub-optimal solution includes: Determining the first initial bandwidth ratio and the second initial bandwidth ratio as the adjusted bandwidth ratios; Updating the initial transmit power and the adjusted bandwidth ratios alternately according to the sub-optimal solution.

5. The resource allocation method according to claim 1, characterized in that, The semantic communication encoder includes a semantic encoder and a first channel encoder, and the semantic encoder is connected to the first channel encoder, where the semantic encoder is used to perform a first encoding process on the source text information in the base station to obtain first encoded information, and the first channel encoder is used to perform a first channel encoding process on the first encoded information to generate the semantic encoded information.

6. The resource allocation method according to claim 5, characterized in that, The semantic receiver includes a first channel decoder, a semantic decoder, a source encoder, and a second channel encoder. The first channel decoder, the semantic decoder, the source encoder, and the second channel encoder are connected in sequence. The first channel decoder is data-connected to the first channel encoder. The first channel decoder is used to perform a first channel decoding process on the semantic encoded information to obtain first decoded information. The semantic decoder is used to perform a first decoding process on the first decoded information to obtain the semantic decoded information. The source encoder is used to perform a second encoding process on the semantic decoded information to obtain second encoded information. The second channel encoder is used to perform a second channel encoding process on the second encoded information to obtain the bit-encoded information.

7. The resource allocation method according to claim 6, characterized in that, The bit decoder includes a second channel decoder and a source decoder. The second channel decoder is connected to the source decoder. The second channel decoder is data-connected to the second channel encoder. The second channel decoder is used to perform a second channel decoding process on the bit-encoded information to obtain second decoded information. The source decoder is used to perform a second decoding on the second decoded information to obtain the source text information.

8. An electronic device, characterized in that, Including: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the resource allocation method according to any one of claims 1 to 7 is implemented.

9. A computer-readable storage medium storing computer-executable instructions, characterized in that, When the computer-executable instructions are executed by a control processor, the resource allocation method according to any one of claims 1 to 7 is implemented.

Images