Destination-initiated semantic communications method

Abstract

This patent is based on a networking approach that realizes a semantic communications framework initiated from the receiver. The goal of this patent is to increase efficiency in terms of data and energy consumption as well as network availability, while also ensuring the completion of the task at hand. By initiating the communication from the receiver side, this patent substitute static approaches with autonomous ones that adapt in real-time to the changing topologies and dynamic condition of next generation network. Once the receiver's request is conveyed to the source, the latter extracts the relevant semantic information from the collected data and transmits it to the destination, where the goal of the communication is achieved in an efficient manner.

Description

[0001] DESCRIPTION

[0002] Destination-initiated semantic communications method Technical field

[0003] This patent introduces a semantics- based networking method, in which the communication is initiated at the destination. The goal of this patent is to transmit the least possible data through the channel in the shortest amount of time - thus, significantly reducing network traffic - while at the same time ensuring a high probability of fulfilling the goal of the communication.

[0004] State of the art

[0005] Even in state-of-the-art beyond 5G networks, multiple operations still depend upon human input and supervision, which results in increased likelihood of errors. Additionally, the limited degree of network automation cannot alleviate the need for customizing each technological solution, device, protocol, etc. To remedy this situation, research activities have been intensified towards designing methods for orchestrating the communications though abstraction techniques applied in the higher levels of the network stack. Any user requests are automatically translated in policies that are distributed throughout all levels. This paradigm shift is intensified in 6G by technologies like network function virtualization and software defined networking that are increasingly attracting more attention.

[0006] However, various limitations still exist, such as scalability and flexibility challenges that accompany the rapid increase in network complexity and number of devices. This phenomenon highlights the importance of resource management and orchestration that can be intelligently optimized by enlisting intelligent approaches. Moreover, another limitation is identified in the high correlation between the envisioned 6G applications and services, which can be mitigated through enlisting a variety of intelligent models that can be selected to perform specific tasks at different levels with extremely high efficiency; thus, alleviating the need for human intervention that is linked to increased workforce costs, delays, and errors. Validation of the lifecycle of operations is necessary to improve the stability and improve on the applied policies; thus, creating a resilient networking ecosystem with improved recovery times.

[0007] This patent provides an alternative to conventional communications and networking paradigms and ensures that services are provided with reduced requirements; thus, ensuring quality of service while alleviating network resources as well as reducing the need for specialized personnel. Short presentation of the invention

[0008] The aforementioned limitations can be overcome by creating a novel communication paradigm that is initiated by the destination and surpasses conventional static approaches by dynamically adapting to the continuously changing nature of next generation networks. This patent presents a framework, in which network nodes can request the relevant information from the respective sources (3) and the system intelligently optimizes communication by taking advantage of novel semantic communications approaches. The various requests are collected by the source (3) that extracts the requested semantic information, encodes it, and transmits it though the communication channel. In turn, the semantic information is received by the destination (2), decoded, and utilized to fulfill the goal of the communication.

[0009] Advantages of the invention

[0010] The key advantages of this patent include:

[0011] A) Improved data and energy efficiency by transmitting only the necessary semantic information from the source (3) to the destination (2).

[0012] B) Improved latency by transmitting less data from the source (3) to the destination (2).

[0013] C) A novel semantic communication framework that overcomes common limitations in semantic information extraction.

[0014] D) Optimized encoding and decoding processes in the source (3) and the destination (2).

[0015] Disclosure of the invention

[0016] Figure 1 illustrates the considered paradigm, in which the destination (2) initiates the communication by announcing its existence and sending a request for the information that is required. This paradigm reduces the network load due to constant exchange of messages and increases data and energy efficiency. This patent realizes destination-initiated communications by transmitting a request that includes the data that are necessary for achieving the goal of the communication. In response, the source (3) transmits only the requested data to the destination (2).

[0017] This patent takes advantage of semantic communications in order to increase even more the efficiency of the network by reducing the amount of data transmitted through semantic extraction. The combination of semantic communications with destination (2) -initiated communications can lead to a significant improvement compared to singular applications of each technique. Specifically, the destination (2) requests only the indexes of the data that are necessary for completing the communication goal. These indexes are found in the common knowledge base between the source (3) and the destination (2) and point to specific semantic ontologies. The source (3) capitalizes on the existence of common knowledge to extract semantic information from the collected data and transmit only them to the destination (2). As depicted in Figure 1, the destination (2) requests only the data pertinent to the existence of people and the source (3) responded by sending only the relative semantic information. If higher energy efficiency was required, the source (3) would transmit only the location of the people extracted in the image and transmit them in text format; thus, achieving the goal of communication with the least consumed resources.

[0018] Another advantage of this patent is the alleviation of network traffic and therefore the improvement in scheduling. This improvement results in reduced collisions between messages and consequently reduced number of retransmissions that can lead to more dense networks without sacrificing performance. This patent contributes towards this direction by ensuring that data are transmitted through the network only when it is necessary.

[0019] From a technical perspective, Figure 2 presents the architecture of this patent. The destination (2) creates a request, r, based on the common knowledge base and the data that it requires from the source (3). Next, this request is transmitted to the source (3) through the channel, h. The received message at the source (3) can be expressed as

[0020] r' = hr + n.

[0021] Based on the received request, the source (3) collects that necessary data, d, and forwards it to the semantic encoder, S. The role of the semantic encoder is to extract only the semantic information, which are given by

[0022] m = S(d) .

[0023] The semantic encoder is an intelligent agent that can support a broad spectrum of functionalities in image analysis and object detection. Specifically, it is capable of identifying a plethora of objects and extracting their characteristics based on the common knowledge base. This is achieved through deep learning, neural networks, and pattern recognition techniques that enable adaptation to new object categories. As a result, it enables the extraction of the semantic information and filters out any unnecessary data from the communication. The output of the semantic encoder is forwarded to the source encoder, channel encoder, and modulator that convert the signal to the appropriate format, %, in order to be transmitted through the channel.

[0024] The response that is received by the destination (2) can be expressed as

[0025] y = hx + n ,

[0026] and, once received, it is propagated to the demodulator, channel decoder, and source decoder, which reconstruct the information of the message in their original state with high accuracy.

[0027] Based on the needs of the communication, the received message is conveyed to the appropriate semantic decoder, which with the help of the common knowledge base converts the received semantic information into the appropriate form, m', that can be used to fulfill the goal of the communication.

[0028] Figure descriptions

[0029] Figure 1 is a functional illustration of the information flow between the source node (1), in this case a drone capturing a picture, and the semantic information as it is received by the destination (2). The destination (2) requests only the relevant semantic indices (A) and receives only the corresponding semantic information (B). An example of the semantic indices (A) is shown in the following table.

[0030] Common knowledge base

[0031] Index Feature

[0032] 1 People

[0033] 2 Cars

[0034] 3 Road

[0035] 4 Trees

[0036]

[0037] Figure 2 depicts the architecture of the patent.

Claims

CLAIMSClaim 1: Communications method based on selective extraction and transmission of semantic information from images (1). The method is based on the network nodes requesting only the information that is required. The workflow is as follows: A) The destination requests the information that it needs from the source (3) to complete a task.B) The destination (2) initiates the communication by conveying the request to the source (3) over the channel.C) The source (3) collects the data pertinent to the request and employs its semantic encoder to extract the necessary semantic information.D) The extracted information is forwarded to the channel encoder and afterwards to the modulator, which transform the message that will be transmitted to the destination (2) into the appropriate form.E) The destination (2) forwards the received message to the demodulator, to the channel decoder, and finally to the semantic decoder.F) The semantic decoder utilizes the common knowledge base to transform the received message into the appropriate form and fulfill the goal of the communication.Claim 2: Semantic communication method for images (1), based on axiom 1, that is based on initiating the communication at the destination (2) that only requests the necessary semantic information.Claim 3: Semantic communications method for images (1), based on the previous axioms, that is characterized by extracting the relevant semantic information from the images (1) and transmitting only them through the channel.Claim 4: Semantic communications method for images (1), based on the previous axioms, in which the destination (2) and the source (3) employ artificial intelligence both for extracting semantic information from images (1) and recreating it at the destination (2).Claim 5: Semantic communications method for images (1), based on the previous axioms, in which the destination (2) and the source (3) utilize a common knowledge base for intelligently encoding and decoding the semantic information.

Images