EP4690951A1

EP4690951A1 - Traffic offloading in communication system

Info

Publication number: EP4690951A1
Application number: EP23931088.1A
Authority: EP
Inventors: Selim ICKIN; Aman RAPARIA; Yak NG MOLINA; Erik SANDERS; Oleg GORBATOV; Nitin Khanna
Original assignee: Telefonaktiebolaget LM Ericsson AB
Current assignee: Telefonaktiebolaget LM Ericsson AB
Priority date: 2023-03-28
Filing date: 2023-03-28
Publication date: 2026-02-11
Also published as: WO2024205457A1

Abstract

Embodiments provide solutions for traffic offloading in a communication system. Specifically, user information is collected for a plurality of terminal devices in a plurality of cells served by a network device in the communication system. An action is determined, according to a machine learning model, based on the user information and state information of the plurality of cells, and the action is to be implemented by the network device for offloading a group of terminal devices in the plurality of terminal devices from a first cell in the plurality of cells to a second cell. Configuration information is obtained for the network device based on the action, and then transmitted to the network device. Accordingly, the configuration information may be automatically obtained based on the machine learning model, and energy consumption of the network device may be reduced once the configuration information loaded in the network device.

Description

TRAFFIC OFFLOADING IN COMMUNICATION SYSTEM

FIELD

[0001] Embodiments of the present disclosure generally relate to the field of telecommunication, and in particular, to traffic offloading in a communication system based on a machine learning (ML) model.

BACKGROUND

[0002] A large number of network devices are deployed in the communication system. Energy consumptions of these network devices may vary based on active cells served by these network devices and the like. With respect to an active cell which serves only a small number of terminal devices, offloading these terminal devices to another cell and then lock this cell may lead to energy saving. Moreover, offloading traffic between cells in dynamically changing cell profiles and load might also yield energy saving.

[0003] When a cell is predicted to have low number of users and traffic, offloading these users and the corresponding low traffic to another cell that has a higher number of users can enable locking the cell where the users are offloaded from. Locking the cell yields a high amount of energy saving. There are other techniques for energy saving such as micro-sleep, deep-sleep that are rather executed at lower layers in the order of milliseconds, while the actions such as the cell lock are executed at higher layers and can be in the order of few minutes.

[0004] Specifically, users may be offloaded from one cell to the other depending on various parameters, such as the received signal strength of the User Equipment (UE), Signal to Interference plus Noise Ratio (SINR), and the distance of the UE to the base station, and the like. In addition, there are energy saving features, where the base stations can be instructed to be locked or switched off if the cell traffic is below a predetermined threshold or by time (e.g., shutting down at 03:00 at midnight). It leads to less operating cost to lock the cell at a specific time interval, however, this is aggressive and rather risky due to the lack of adaptiveness of the dynamically changing environment, in fact might yield to locking of a cell with significant traffic.

SUMMARY

[0005] In general, embodiments of the present disclosure provide a solution for traffic offloading in a communication system.

[0006] In a first aspect of the present disclosure, there is provided a computer-implemented method for traffic offloading in a communication system. The method comprises: collecting user information of a plurality of terminal devices in a plurality of cells served by a network device in the communication system; determining, according to an ML model, an action that is to be implemented by the network device based on the user information and state information of the plurality of cells, the action being used for offloading a group of terminal devices in the plurality of terminal devices from a first cell in the plurality of cells to a second cell in the plurality of cells; obtaining configuration information for the network device based on the action; and transmitting the configuration information to the network device. With these embodiments, the ML model may determine the offloading action, and then the action may be translated into the configuration information automatically. Once the configuration information is deployed in the network device, appropriate terminal device(s) may be offloaded from the source cell to the destination cell without any human involvement.

[0007] In some embodiments of the present application, the method further comprises: determining energy consumption and Quality of Service (QoS) attributes associated with a third and a fourth cell based on state information of the third and fourth cells; generating a first objective function for training the ML model based on the energy consumption and QoS attributes; and training the ML model based on the first objective function. With these embodiments, the objective function may be generated based on various attributes related to the various cells in an accurate way.

[0008] In some embodiments of the present application, the energy consumption and QoS attributes are determined according to a pre-trained ML model deployed at a first network function in the communication system, the pre-trained ML model describing an association relationship between the state information of the third and fourth cells and estimations for the energy consumption and QoS attributes. With these embodiments, the pre-trained ML model may help to determine the energy consumption and QoS attributes for various cells in an automatic way.

[0009] In some embodiments of the present application, generating the first objective function further comprises: determining an energy saving metric based on a change of the energy consumption attributes caused by offloading one or more terminal devices from the third cell to the fourth cell; determining a QoS reduction metric based on a change of the QoS attributes caused by offloading the one or more terminal devices from the third cell to the fourth cell; and creating the first objective function based on the energy saving metric and the QoS reduction metric. With these embodiments, the objective function that leads to the energy saving without affecting normal operations of the users may be determined in an accurate way.

[0010] In some embodiments of the present application, obtaining the configuration information comprises: generating respective thresholds for the first and second cells in the configuration information based on a type of the action. With these embodiments, the thresholds may be determined in an accurate way by considering the type of the action. P107035W001

[0011] In some embodiments of the present application, a threshold, in the respective thresholds, for a cell in the first and second cells comprises at least one of: a threshold for a number of terminal devices in the cell, or a threshold associated with a physical resource block for the cell. With these embodiments, various aspects of the threshold may be adjusted so as to implement the offloading plan.

[0012] In some embodiments of the present application, the type of the action comprises any of: an increment type indicating that the number of terminal devices in a cell of the first and second cells is to be increased by a first step size, a decrement type indicating that the number of terminal devices in a cell of the first and second cells is to be decreased by a second step size, or a nochange type indicating that the number of terminal devices in a cell of the first and second cell is unchanged. With these embodiments, the offloading may be implemented in a more flexible and effective way.

[0013] In some embodiments of the present application, at least one of the first and second step sizes is determined by adapting a default step size based on a second objective function that is generated by the ML model based on the state information of the plurality of cells. With these embodiments, the offloading may be implemented in a more flexible and effective way.

[0014] In some embodiments of the present application, generating the respective thresholds in the configuration information further comprises: determining the respective thresholds according to a pre-trained ML model that is deployed at a second network function in the communication system, the pre-trained ML model describing an association relationship between the type of an action and estimations for the respective thresholds. With these embodiments, the pre-trained ML model may accept the translatable action and then translate the action into thresholds to be set in the network device in an automatic way.

[0015] In some embodiments of the present application, the action is determined in response to receiving an offloading request from a third network function in the communication system. With these embodiments, the offloading procedure may be triggered by a desired network function in a flexible and distributed way.

[0016] In some embodiments of the present application, the method further comprises: obtaining from the third network function a rule defining constraints for the offloading, and wherein the action is determined further based on the rule. With these embodiments, the action may be determined based on constraints that are defined in the rule, therefore the offloading procedure may be controlled in a more flexible way.

[0017] In some embodiments of the present application, transmitting the configuration information to the network device comprises: sending the configuration information to the network device via the third network function. With these embodiments, the configuration information P107035W001 may be obtained in a flexible and distributed way among multiple network functions in the communication system.

[0018] In some embodiments of the present application, the state information of the plurality of cells is obtained via an address of a radio resource control agent provided by a fourth network function in the communication system. With these embodiments, the state information may be obtained in a flexible and distributed way among multiple network functions in the communication system.

[0019] In some embodiments of the present application, the method further comprises: generating, according to the ML model, a third objective function based on the state information of the plurality of cells; and registering, at one or more network functions in the communication system, the configuration information along with the third objective function. With these embodiments, details related to the offloading procedure may be registered at one or more desired network functions for records.

[0020] In some embodiments of the present application, the network device is a base station in the communication system, and the first and second cells are within any of: the same sector of the base station; or different neighboring sectors of the base station. With these embodiments, the terminal device(s) within the same or neighboring sectors may be managed in a flexible and effective way.

[0021] In some embodiments of the present application, the user information comprises any of: mission criticality, tasks of user, user contexts, user locations regarding applications running on the plurality of terminal devices, and terminal device information. With these embodiments, the to-be-offloaded user(s) may be maintained at relatively stable quality levels.

[0022] In some embodiments of the present application, the ML model is pre-trained and serves as a digital twin in the communication system. With these embodiments, the pre-trained ML model may be directly obtained and used for determining the offloading in the communication system.

[0023] In a second aspect, there is provided a computing device. The computing device comprises at least one processor coupled to a computer-readable memory unit, the memory unit comprising instructions that when executed by the at least one processor implement a method for traffic offloading in a communication system. The method comprises: collecting user information of a plurality of terminal devices in a plurality of cells served by a network device in the communication system; determining, according to an ML model, an action that is to be implemented by the network device based on the user information and state information of the plurality of cells, the action being used for offloading a group of terminal devices in the plurality of terminal devices from a first cell in the plurality of cells to a second cell in the plurality of cells; obtaining configuration information for the network device based on the action; and transmitting the configuration information to the network device.

[0024] In some embodiments of the present application, the method further comprises: determining energy consumption and Quality of Service (QoS) attributes associated with a third and a fourth cell based on state information of the third and fourth cells; generating a first objective function for training the ML model based on the energy consumption and QoS attributes; and training the ML model based on the first objective function.

[0025] In some embodiments of the present application, the energy consumption and QoS attributes are determined according to a pre-trained ML model deployed at a first network function in the communication system, the pre-trained ML model describing an association relationship between the state information of the third and fourth cells and estimations for the energy consumption and QoS attributes.

[0026] In some embodiments of the present application, generating the first objective function further comprises: determining an energy saving metric based on a change of the energy consumption attributes caused by offloading one or more terminal devices from the third cell to the fourth cell; determining a QoS reduction metric based on a change of the QoS attributes caused by offloading the one or more terminal devices from the third cell to the fourth cell; and creating the first objective function based on the energy saving metric and the QoS reduction metric.

[0027] In some embodiments of the present application, obtaining the configuration information comprises: generating respective thresholds for the first and second cells in the configuration information based on a type of the action.

[0028] In some embodiments of the present application, a threshold, in the respective thresholds, for a cell in the first and second cells comprises at least one of: a threshold for a number of terminal devices in the cell, or a threshold associated with a physical resource block for the cell. [0029] In some embodiments of the present application, the type of the action comprises any of: an increment type indicating that the number of terminal devices in a cell of the first and second cells is to be increased by a first step size, a decrement type indicating that the number of terminal devices in a cell of the first and second cells is to be decreased by a second step size, or a nochange type indicating that the number of terminal devices in a cell of the first and second cell is unchanged.

[0030] In some embodiments of the present application, at least one of the first and second step sizes is determined by adapting a default step size based on a second objective function that is generated by the ML model based on the state information of the plurality of cells.

[0031] In some embodiments of the present application, generating the respective thresholds in the configuration information further comprises: determining the respective thresholds according to a pre-trained ML model that is deployed at a second network function in the communication system, the pre-trained ML model describing an association relationship between the type of an action and estimations for the respective thresholds.

[0032] In some embodiments of the present application, the action is determined in response to receiving an offloading request from a third network function in the communication system.

[0033] In some embodiments of the present application, the method further comprises: obtaining from the third network function a rule defining constraints for the offloading, and wherein the action is determined further based on the rule.

[0034] In some embodiments of the present application, transmitting the configuration information to the network device comprises: sending the configuration information to the network device via the third network function.

[0035] In some embodiments of the present application, the state information of the plurality of cells is obtained via an address of a radio resource control agent provided by a fourth network function in the communication system.

[0036] In some embodiments of the present application, the method further comprises: generating, according to the ML model, a third objective function based on the state information of the plurality of cells; and registering, at one or more network functions in the communication system, the configuration information along with the third objective function.

[0037] In some embodiments of the present application, the network device is a base station in the communication system, and the first and second cells are within any of: the same sector of the base station; or different neighboring sectors of the base station.

[0038] In some embodiments of the present application, the user information comprises any of: mission criticality, tasks of user, user contexts, user locations regarding applications running on the plurality of terminal devices, and terminal device information.

[0039] In some embodiments of the present application, the ML model is pre-trained and serves as a digital twin in the communication system.

[0040] In a third aspect, a non-transitory computer readable medium is provided. The non- transitory computer readable medium has instructions stored thereon, the instructions when executed by at least one processor cause the at least one processor to perform a method according to any of the embodiments of the first aspect.

[0041] In a fourth aspect, a computer program product is provided. The computer program product comprises instructions, the instructions when executed by at least one processor cause the at least one processor to perform a method according to any of the embodiments of the first aspect. [0042] It is to be understood that the Summary section is not intended to identify key or essential features of embodiments of the present disclosure, nor is it intended to be used to limit the scope of the present disclosure. Other features of the present disclosure will become easily comprehensible through the following description.

BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Through the more detailed description of some example embodiments of the present disclosure in the accompanying drawings, the above and other objects, features and advantages of the present disclosure will become more apparent, where:

[0044] Fig. 1 illustrates a schematic diagram for a communication system where embodiments of the present disclosure may be applied;

[0045] Fig. 2 illustrates a schematic diagram for generating configuration information for traffic offloading in a communication system based on ML in accordance with embodiments of the present disclosure;

[0046] Fig. 3 illustrates a schematic diagram for interactions between an ML agent and an ML environment in accordance with embodiments of the present disclosure;

[0047] Fig. 4 illustrates a schematic diagram for determining a traffic offloading action based on ML in accordance with embodiments of the present disclosure;

[0048] Fig. 5 illustrates a schematic diagram for estimations from multiple pre-trained ML models in accordance with embodiments of the present disclosure;

[0049] Fig. 6 illustrates a schematic diagram for state information that may be used for determining a traffic offloading action in accordance with embodiments of the present disclosure;

[0050] Fig. 7 illustrates a schematic diagram for details of a traffic offloading action in accordance with embodiments of the present disclosure;

[0051] Fig. 8 illustrates a schematic flowchart for implementing traffic offloading via a configuration file in a communication system in accordance with embodiments of the present disclosure;

[0052] Fig. 9 illustrates a schematic signaling chart for generating a configuration file based on a partially trained (or an untrained) ML model in accordance with embodiments of the present disclosure;

[0053] Fig. 10 illustrates a schematic signaling chart for generating a configuration file based on a trained ML model in accordance with embodiments of the present disclosure;

[0054] Fig. 11 illustrates a schematic flowchart of a method for generating configuration information for traffic offloading in a communication system based on ML in accordance with embodiments of the present disclosure; P107035W001

[0055] Fig. 12 illustrates an example of a communication system 1200 in accordance with some embodiments; and

[0056] Fig. 13 illustrates a network node in accordance with some embodiments.

[0057] Throughout the drawings, the same or similar reference numerals represent the same or similar element.

DETAILED DESCRIPTION

[0058] Some of the embodiments contemplated herein will now be described more fully with reference to the accompanying drawings. The embodiments set forth below represent information to enable those skilled in the art to practice the embodiments and illustrate the best mode of practicing the embodiments. Upon reading the following description in light of the accompanying figures, those skilled in the art will understand the concepts of the disclosure and will recognize applications of these concepts not particularly addressed herein. It should be understood that these concepts and applications fall within the scope of the disclosure.

[0059] Generally, all terms used herein are to be interpreted according to their ordinary meaning in the relevant technical field, unless a different meaning is clearly given and/or is implied from the context in which it is used. All references to a/an/the element, apparatus, component, means, step, etc. are to be interpreted openly as referring to at least one instance of the element, apparatus, component, means, step, etc., unless explicitly stated otherwise. The steps of any methods disclosed herein do not have to be performed in the exact order disclosed, unless a step is explicitly described as following or preceding another step and/or where it is implicit that a step must follow or precede another step. Any feature of any of the embodiments disclosed herein may be applied to any other embodiment, wherever appropriate. Likewise, any advantage of any of the embodiments may apply to any other embodiments, and vice versa. Other objectives, features, and advantages of the enclosed embodiments will be apparent from the following description.

[0060] As used herein, the term “includes” and its variants are to be read as open terms that mean “includes, but is not limited to.” The term “based on” is to be read as “at least in part based on.” The term “one embodiment” and “an embodiment” are to be read as “at least one embodiment.” The term “another embodiment” is to be read as “at least one other embodiment.” The terms “first,” “second,” and the like may refer to different or same objects. Other definitions, explicit and implicit, may be included below.

[0061] As used herein, the term “terminal device” refers to any device having wireless or wired communication capabilities. Examples of the terminal device include, but not limited to, user equipment (UE), personal computers, desktops, mobile phones, cellular phones, smart phones, personal digital assistants (PDAs), portable computers, tablets, wearable devices, internet of things (loT) devices, Ultra-reliable and Low Latency Communications (URLLC) devices, Internet of Everything (loE) devices, machine type communication (MTC) devices, devices on vehicle for V2X communication where X means pedestrian, vehicle, or infrastructure/network, devices for Integrated Access and Backhaul (IAB), Space borne vehicles or Air borne vehicles in Non-terrestrial networks (NTN) including Satellites and High Altitude Platforms (HAPs) encompassing Unmanned Aircraft Systems (UAS), extended Reality (XR) devices including different types of realities such as Augmented Reality (AR), Mixed Reality (MR) and Virtual Reality (VR), the unmanned aerial vehicle (UAV) commonly known as a drone which is an aircraft without any human pilot, devices on high speed train (HST), or image capture devices such as digital cameras, sensors, gaming devices, music storage and playback appliances, or Internet appliances enabling wireless or wired Internet access and browsing and the like. The “terminal device” can further have “multicast/broadcast” feature, to support public safety and mission critical, V2X applications, transparent IPv4/IPv6 multicast delivery, IPTV, smart TV, radio services, software delivery over wireless, group communications and loT applications. It may also incorporate one or multiple Subscriber Identity Module (SIM) as known as Multi-SIM. The term “terminal device” can be used interchangeably with a UE, a mobile station, a subscriber station, a mobile terminal, a user terminal or a wireless device.

[0062] As used herein, the term “network node” may also be referred to as a network function (NF), a network entity, or a network device, and refers to a physical, virtual or hybrid function or entity which is deployed at a network side and provides one or more services to clients/consumers. For example, an NF may be arranged at a device in a radio access network (RAN) or a core network (CN) of a communication system. The network node may be implemented in hardware, software, firmware, or some combination thereof. Examples of a network node in a RAN include, but not limited to, a Node B (NodeB or NB), an evolved NodeB (eNodeB or eNB), a next generation NodeB (gNB), a transmission reception point (TRP), a remote radio unit (RRU), a radio head (RH), a remote radio head (RRH), an IAB node, a low power node such as a femto node, a pico node, a reconfigurable intelligent surface (RIS), and the like. Examples of a network node in a CN include, but not limited to, a Mobility Management Entity (MME), a Packet Data Network Gateway (P-GW), a Service Capability Exposure Function (SCEF), a Home Subscriber Server (HSS), or the like. Some other examples of a core network node include a node implementing a Access and Mobility Management Function (AMF), a User Plane Function (UPF), a Session Management Function (SMF), an Authentication Server Function (AUSF), a Network Slice Selection Function (NSSF), a Network Exposure Function (NEF), a Network Function (NF) Repository Function (NRF), a Policy Control Function (PCF), a Unified Data Management (UDM), or the like. [0063] Communications in a communication environment may conform to any suitable standards including, but not limited to, Global System for Mobile Communications (GSM), Long Term Evolution (LTE), LTE-Evolution, LTE- Advanced (LTE-A), New Radio (NR), Wideband Code Division Multiple Access (WCDMA), Code Division Multiple Access (CDMA), GSM EDGE Radio Access Network (GERAN), Machine Type Communication (MTC) and the like. The embodiments of the present disclosure may be performed according to any generation communication protocols either currently known or to be developed in the future. Examples of the communication protocols include, but not limited to, the first generation (1G), the second generation (2G), 2.5G, 2.75G, the third generation (3G), the fourth generation (4G), 4.5G, the fifth generation (5G) communication protocols, 5.5G, 5G-Advanced networks, or the sixth generation (6G) networks and beyond and the like, wireless local network communication protocols such as Institute for Electrical and Electronics Engineers (IEEE) 802.11 and the like, and/or any other protocols currently known or to be developed in the future. Moreover, the communication may utilize any proper wireless communication technology, comprising but not limited to: Code Division Multiple Access (CDMA), Frequency Division Multiple Access (FDMA), Time Division Multiple Access (TDMA), Frequency Division Duplex (FDD), Time Division Duplex (TDD), Multiple-Input Multiple-Output (MIMO), Orthogonal Frequency Division Multiple (OFDM), Discrete Fourier Transform spread OFDM (DFT-s-OFDM) and/or any other technologies currently known or to be developed in the future.

[0064] As used herein, the term “model” is referred to as an association between an input and an output learned from training data, and thus a corresponding output may be generated for a given input after the training. The generation of the model may be based on ML techniques. The ML techniques may also be referred to as artificial intelligence (Al) techniques. In general, an ML model can be built, which receives input information and makes predictions based on the input information. For example, a classification model may predict a class of the input information among a predetermined set of classes. As used herein, “model” may also be referred to as “ML model”, “learning model”, “ML network”, or “learning network,” which are used interchangeably herein.

[0065] Generally, ML may usually involve three stages, i.e., a training stage, a validation stage, and an application stage (also referred to as an inference stage). At the training stage, a given ML model may be trained (or optimized) iteratively using a great amount of training data until the model can obtain, from the training data, consistent inference similar to those that human intelligence can make. During the training, a set of parameter values of the model is iteratively updated until a training objective is reached. Through the training process, the ML model may be regarded as being capable of learning the association between the input and the output (also referred to an input-output mapping) from the training data. At the validation stage, a validation input is applied to the trained ML model to test whether the model can provide a correct output, so as to determine the performance of the model. Generally, the validation stage may be considered as a step in a training process, or sometimes may be omitted. At the inference stage, the trained ML model may be used to process a real-world model input based on the set of parameter values obtained from the training process and to determine the corresponding model output.

[0066] Fig. 1 illustrates a schematic diagram 100 for a communication system where embodiments of the present disclosure may be applied. In Fig. 1, a network device 130 (such as a base station) is deployed in the communication system, and the network device 130 may provide a plurality of cells in one or more sectors. For example, a first cell 110 and a second cell 120 are within a sector 140, and these cells 110 and 120 may serve multiple users (i.e., terminal devices such as a terminal device 150 and the like). Although Fig. 1 illustrates only one network device 130 with only one sector 140 and two cells 110 and 120, there may be multiple network devices that provide multiple sectors and cells in the communication system. Usually, when a cell (for example, the first cell 110) is predicted to have a low number of terminal devices with low traffic, offloading those terminal devices and corresponding traffic to another cell (for example, the second cell 120) that has a higher number of terminal devices may enable locking the first cell 110, and thus energy consumption of the network device 130 may be reduced.

[0067] Solutions have been proposed for determining offloading plans so as to reduce the energy consumption in the communication system. For example, terminal devices may be offloaded from one cell to another depending on multiple parameters (such as the received signal strength of UE, SINR, and the distance of the UE to the base station). In addition, the network device may be instructed to be locked or switched off if the cell traffic reaches certain thresholds, or by time (e.g., shut down at 03:00 at midnight). Although it may reduce the energy consumption, it is aggressive and risky due to the lack of adaptiveness in the dynamical or real time communication environment. According to some energy saving solutions, the network device may be locked when predefined criteria are met, however, considering the states of the cells and users vary all the time, it is difficult to define specific criteria in advance. Sometimes, the energy saving plan may strongly affect the normal operations of the users. By now, ML models have been employed in communication systems to improve the performance of communications systems. At this point, it is desired that the ML models may facilitate the communication system to work out an energy saving plan. Specifically, it is desired that a traffic offloading plan may be determined in a more effective and accurate way, so as to reduce the energy consumption of the communication system without affecting the normal operations of the users. [0068] In order to remove at least some of the above drawbacks, embodiments of the present disclosure provide a computer-implemented method for traffic offloading in a communication system. Referring to Fig. 2 for a brief description of the proposed solution, here Fig. 2 illustrates a schematic diagram 200 for generating configuration information for traffic offloading in a communication system based on ML in accordance with embodiments of the present disclosure. In Fig. 2, cell state information 210 may be collected for a plurality of cells (such as the first and second cells 110 and 120, and the like) served by the network device 130. Here, the cell state information 210 may comprise various aspects of the cells such as number of users in each cell, maximum power, and the like. User information 212 may be collected from the plurality of terminal devices (such as the terminal device 150, and the like) in the plurality of cells, and the user information 212 may comprise various aspects of the users of the terminal devices, such as the mission criticality, tasks, and the like.

[0069] Further, according to an ML model 220, an action 230 may be determined based on the user information 212 and the cell state information 210 of the plurality of cells. Here, the action 230 may be implemented by the network device 130 for offloading a group of terminal devices in the plurality of terminal devices from the first cell 110 to the second cell 120. Here, the ML model 220 may be trained by previously collected training data and has knowledge about which terminal device(s) should be offloaded from a source cell (also referred to as a capacity cell) to a destination cell (also referred to as a coverage cell) based on the cell state information 210 and the user information 212. Once the ML model 220 is trained, the ML model 220 may automatically output the action 230 for traffic offloading. Then, the action 230 may be translated into configuration information 240 (for example, a configuration file that may be used in setting the network device 130). Further, the configuration information 240 may be transmitted to the network device 130, and then the traffic offloading may be automatically implemented.

[0070] Due to the dynamical environment in the communication system, it is hard to manually select the source cell, the destination cell and the terminal devices that are to be offloaded from the source cell to the destination cell. Further, various operators may have their specific restrictions and requirements, and thus the difficulty level for implementing the traffic offloading is further increased. For example, the restrictions and requirements may relate to both of the mobility strategy and the quality of service that the operators want to offer to the users. In one example, the standard usage of the network is to prioritize the users in the capacity cell having a coverage cell underneath, so that if the utilization of the network is low enough, the users may be offloaded to the coverage cell to save energy in the capacity cell. In another example, if there is a user with a low latency service requirement, maybe the coverage cell is not enough for it because the bandwidth in this frequency layer is limited and only will achieve the quality of service that was promised in the capacity cell. Specifically, it is highly depended on the application type and requirements. In an outdoor gaming scenario with Augmented Reality (AR), it is preferable to keep the user in the capacity cell no matter what to meet the latency requirements. In contrast, if a user is hiking and need a connection just for text of status update (e.g., in emergency cases), a coverage cell probably is a better option.

[0071] With the proposed ML model 220, the action 230 may be automatically determined based on the cell state information 210 and the user information 212 that are collected in real time. Therefore, the action 230 describes an accurate traffic offload plan among the plurality of cells. Specifically, the action 230 may be translated into configuration information 240 (for example, by a generative ML model), once the configuration information 240 is deployed in the network device 130, appropriate terminal device(s) may be offloaded from the source cell to the destination cell without any human involvement. Meanwhile, as the user information 212 carries data related to the criticality of the services being served by the network device 130, QoS for the to-be-offloaded users may be maintained at relatively stable levels. For example, as compared to users who are occasionally sending messages, it would be undesirable to offload users who are heavily streaming video.

[0072] In some embodiments of the present application, the ML model 220 may be pre-trained and serve as a digital twin in the communication system. Here, the ML model 220 may be built according to any ML network structure that is known in the ML field. Alternatively and/or in addition, the ML model 220 may also be built based on any ML network structure that is to be developed in the future. Afterwards, the ML model 220 may be trained by training data during the training procedure. Here, anyone may implement the training procedure as long as reliable training data is obtained. For example, engineers from the operators and/or a third party may train the ML model 220. Further, the well-trained and/or partly-trained ML model 220 may be trained gradually in an iteration way after it is deployed in the communication system for estimating the offloading action. At this point, the ML model 220 may further be optimized in a real working environment, and thus the accuracy level of the ML model 220 may be increased.

[0073] Having provided the brief description of the embodiments of the present application, reference will be made to Fig. 3 for more details about the traffic offloading solution. Fig. 3 illustrates a schematic diagram 300 for interactions between an ML agent (i.e., a learner) and an ML environment in accordance with embodiments of the present disclosure. In some embodiments of the present application, the ML model 220 may be implemented based on Reinforcement Learning (RL). RL relates to decision making and it aims to learn the optimal behavior in an environment to obtain the maximum reward. This optimal behavior is learned through interactions with the environment and observations of how it responds. At this point, the learner may independently discover the sequence of actions that maximize the reward based on a trial-and-error search. Therefore, actions that result in eventual success may be learned in an unseen way without any help from a supervisor.

[0074] As illustrated in Fig. 3, an ML agent 310 is provided for interacting with an environment 320. Here, in the i^th state, with inputs of a state 332 (represented as S and a reward 330 (represented as ), the ML agent 310 aims to output the action 230 (represented as Ai) with the goal to maximize the reward by taking the action. As a result of the ML agent 310 being interacting with the environment 320, the action 230 may be implemented in the environment 320 and then a new state (the (i + l)^th state) is formed (represented as a state S_i+1 440). The new updated state is then sent as input to the estimation models and the output of the estimation models yield a joint reward R_i+1 450. Thus, the ML agent 310 may continuously observe the new state and output action accordingly. By this way, the offloading problem may be formulated based on the RL solution, and then the ML model 220 is trained by maximizing the reward gradually.

[0075] In some embodiments of the present application, in the training procedure, energy consumption and Quality of Service (QoS) attributes associated with cells (for example, a third cell that acts as the source cell and a fourth cell that acts as the destination cell involved in the training procedure) may be determined based on state information of the third and fourth cells. Here, additional estimation ML models may be used for determining these attributes. Referring to Fig. 4 for more details, here Fig. 4 illustrates a schematic diagram 400 for determining a traffic offloading action based on ML in accordance with embodiments of the present disclosure. According to a cell selection policy 412, cells 410 within the same or different neighboring sectors may be selected. Here, states about these cells 410 may be inputted into multiple estimation models to obtain the energy consumption and Quality of Service (QoS) attributes, respectively.

[0076] Specifically, an energy estimation model 420 may be provided for determining the energy consumption, and a KPI estimation model 430 may be provided for determining the QoS attributes for the cells. Here, both of the estimation models may be built based on any available ML network structure, and may be trained in advance according to solutions that have been proposed and/or to be developed in the future. Here, the KPI estimation model 430 may comprise a plurality of ML models for estimating various QoS attributes, respectively. In some embodiments of the present application, the energy estimation model 420, the KPI estimation model 430, and the ML model 220 may be deployed in the same computing device. Alternatively and/or in addition, they may be deployed in one or more network functions in the communication system. Therefore, the traffic offloading may be implemented in a more flexible and easy way in the communication system. [0077] Fig. 5 illustrates a schematic diagram 500 for estimations from multiple pre-trained ML models in accordance with embodiments of the present disclosure. As illustrated in Fig. 5, a graph 510 shows the Physical Resource Block (PRB) utilization that is estimated by a pre-trained ML utilization estimation model, where different colors in the graph 510 show data associated with different hours of the day 512, respectively. Here, the vertical axis represents the PRB utilization estimation, and the horizontal axis represents the number of the users during a day. Although Fig. 5 illustrates only the PRB utilization estimation, there may be more estimation ML models for estimating other attributes. For example, an interference estimation model may be pretrained for estimating the average PUSCH/PUCCH Signal to Interference plus Noise Ratio (SINR), a downlink throughput estimation model may be pre-trained for estimating the user throughput, and the like. With these pre-trained ML models, important QoS attributes may be directly estimated for the cells, and then be inputted into the ML agent 310 for training the ML model 220 towards a direction that the outputted action 230 may maximize the reward. Similarly, a graph 520 shows the energy consumption that is estimated by the pre-trained energy estimation model 420, where different colors in the graph 510 show data associated with different hours of the day 522, respectively. With the energy estimation model, the energy consumption may be directly estimated for determining the offloading plan.

[0078] Returning back to Fig. 4, the ML agent 310 may be implemented by any suitable RL agent such as a Deep Q-Leaming (DQN) agent, and the like, and the environment 320 here may relate to RL development environment such as a GYM environment, and the like. The DQN agent may interact with the GYM environment, and thus the environment state S_L may be updated to the new state S_i+1 after the interaction. At this point, an objective function (represented as the first objective function) may be generated as the reward for training the ML model 220 based on the energy consumption and QoS attributes, and then the ML model 220 may be trained by maximizing the reward. With these embodiments, the reward may assure that the ML model 220 is optimized gradually without a supervisor.

[0079] In some embodiments of the present application, the ML agent 310 works as a learner that aims to maximize the reward for a corresponding action under the current state. The ML agent 310 may be formulated according to any suitable network structure. In one specific example, the ML agent 310 may be implemented by a 2-layer neural network, that takes in a state vector consisting of multiple fields (such as 2N fields, where N indicates a positive integer), and estimates an action. There may be multiple neurons at the intermediate layer, and the activations functions may be set to “ReLu.” It is to be understood that the above network structure is just an example and other network structures having more layers may be adopted in other embodiments of the present application. [0080] Fig. 6 illustrates a schematic diagram 600 for state information that may be used for determining a traffic offloading action in accordance with embodiments of the present disclosure. As illustrated in Fig. 6, the state information 640 may comprise a cell state 610 for the source cell and a cell state 620 for the destination cell. The two cell states 610 and 620 may have the same structure and the cell state 610 is expanded for more descriptions. In Fig. 6, the cell state 610 may comprises multiple fields. For example, an average user number 612 indicates the average number of users that are served in the cell, a max energy 614 indicates the maximum energy /power consumption of the cell, a radio band 616 indicates band(s) that is supported by the cell, and a revision 618 indicates the hardware revision information related to the cell, and the like. It is to be understood that the above fields are just examples of the cell state. Alternative and/or in addition, the cell state may include more, less, or different fields depending on the specific requirements of the operator.

[0081] Similarly, the cell state 620 may be represented in the same format. Supposing each cell state has N fields, then the state information 640 may have 2 A fields in total. Alternatively and/or in addition , the cell state information 210 may further comprise an interaction 630, for indicating a degree of interaction between the cells, depending on the number of cells around a sector and their relation such as handover events. At this point, the cell state information 210 may include 2N fields, where fields [1 7V] indicate the cell state 610 for the source cell, fields [A + 1,2 A] indicate the cell state 620 for the source cell. The above fields may be concatenated together to form the state information 640 in a vector format.

[0082] In some embodiments of the present application, the state information 640 may be inputted into the energy estimation model 420 and the KPI estimation model 430, and then the objective function may be generated based on various estimations outputted from the above estimation models. Specifically, an energy saving metric may be determined based on a change in the energy consumption attributes caused by offloading one or more terminal devices from the third cell to the fourth cell. Here, the energy saving metric may be determined based on a difference between the current energy consumption and the estimation of the energy consumption that is outputted from the energy estimation model 420. It is desired that the estimation of the energy consumption may be lower than the current energy consumption, and thus the energy may be saved after the offloading.

[0083] Further, a QoS reduction metric may be determined based on a change of the QoS attributes caused by offloading the one or more terminal devices from the third cell to the fourth cell. Here, the QoS reduction metric may be determined based on a difference between the current QoS attributes and the estimation of the QoS attributes that are outputted from the KPI estimation model 430. It is desired that the offloading will not affect the QoS of the to-be-offloaded terminal devices, such that users of the terminal devices may continue to enjoy the stable services after the offloading. Then, the objective function may be created based on the energy saving metric and the QoS reduction metric. At this point, the objective function may be defined in proportional to the energy saving metric but in an inverse proportional to the QoS reduction metric. Therefore, the ML model 220 may be trained towards a direction that may increase the energy saving but does not decrease the QoS.

[0084] As the QoS attributes may relate to multiple attributes, the QoS reduction metric may also relate to multiple aspects such as the user throughput reduction, the average PUCCH/PUSCH SINR reduction, and the like. In some embodiments, the following Formula 1 may be used to define the objective function in the RL: reward = nor m^ener gy Saving) — normfuserThroughputRedu)

— norm(avg PucchSinrRedu) — norm(avgPuschSinrRedu)

Formula 1

[0085] In this Formula, reward represents the objective function, normQ represents a normalization function, energySaving represents the estimation of the energy saving, userThroughputRedu represents the user throughput reduction, avg PucchSinrRedu represents the average PUCCH SINR reduction, and avgPuschSinrRedu represents the average PUSCH SINR reduction. It is to be understood that the above Formula 1 is just an example for determining the reward. Alternatively and/or in addition, the reward may be determined according to another formula as long as the ML model 220 may be optimized towards a direction that may increase the energy saving but does not decrease the QoS. For example, the following Formula 2 may be used for determining the reward, here M represents the number of the QoS attributes. reward

— - norm(avgPuschSinrRedu)

Formula 2

[0086] In some embodiments of the present application, constraints may be set such that the ML agent may be penalized (with reward = — 1) when it takes an action in one or more of the following: 1) If the number of terminal devices in the coverage cell is less than 1, i.e., in this situation coverage cell is not locked; and 2) If one of interference related QoS attributes, i.e., average interference (AvgPucchSinr or AvgPuschSinr) is less than 20 or another predefined value. On the other hand, the ML agent may be rewarded when it offloads all the users from capacity cell to coverage cell assuming that an operator does not prefer to lock the coverage cell, since then the capacity cell can be completely locked and significant energy savings can be achieved. With the above objective function, the ML model may be trained by maximizing the reward and thus the trained ML model may learn knowledges about the offloading action and the state information of the plurality of cells. In this way, the trained ML model may not only select the source cell and the destination cell from the plurality of cells, but also select a certain number of terminal devices that should be offloaded.

[0087] In some embodiments of the present application, the action may comprise multiple types, referring to Fig. 7 for more details about the action. Fig. 7 illustrates a schematic diagram 700 for details of a traffic offloading action in accordance with embodiments of the present disclosure. As illustrated in Fig. 7, the pre-trained ML model 220 may accepts cell state information 210 about the plurality of cells and then it may output the action 230 for offloading a group of terminal devices 150 and the like (with the number of devices 710) from the first cell 110 to the second cell 120. Here, the action 230 may comprise any of an increment type 720, a decrement type 722 and a no-change type 724.

[0088] The increment type 720 indicates that the number of terminal devices in a cell of the first and second cells is to be increased by a first step size. Regarding a cell, the number of terminal devices in the cell may be increased by the predetermined first step size. For example, the first step size may be defined to 5 (or another integer number). At this point, for an action with the increment type 720, 5 users may be increased for the cell. In another example, the first step size may be set to 10, and thus 10 users may be increased by the increment action. Similarly, the decrement type 722 indicates that the number of terminal devices in a cell of the first and second cells is to be decreased by a second step size, and the second step size may be set in advance for defining the group size for decreasing the user number. The no-change type 724 indicates that the number of terminal devices in a cell of the first and second cell is unchanged.

[0089] In some embodiments of the present application, the first and second step sizes may be set to the same value (for example, a default value). Alternatively and/or in addition, the first and second step sizes may be set to different values, respectively. Further, the first and second step sizes may be adjusted, for example, based on the total number of users in the cell. For example, if there are only 4 users in the source cell, at this point it is improper to set the decrement step size to a value that is greater than 4. Therefore, the decrement step size may be corrected to 4.

[0090] Alternatively and/or in addition , both step sizes may be determined by adapting the default step size based on an objective function (for example, a second objective function) that is generated by the ML model 220 based on the cell state information 210 and the user information 212. For example, if the optimum step size is 10 and the default size at the initial stage is 50, based on the reward value, the step size may be adapted to minimize the number of iterations. At this point, more users may be offloaded in each iteration and thus the number of iterations may be reduced. With these embodiments, the offloading may be implemented in a more flexible and effective way.

[0091] In some embodiments of the present application, the action 230 may be translated into the configuration information for the network device. Specifically, with respect to a cell in the first and second cells 110 and 120, respective thresholds in the configuration information may be generated for the cell based on the type of action. Depending on requirements of the operator, various thresholds may be defined in the configuration information for the network device. Considering a situation where the network device is a base station in the communication system, the configuration information may be a configuration file for the base station, and the threshold may comprise at least one of: a threshold for a number of terminal devices in the cell, or a threshold associated with a physical resource block for the cell, and the like.

[0092] Accordingly, these embodiments may decide which threshold(s) should be defined according to the specific requirements of the base station, and then respective values may be set for the respective thresholds. In some embodiments of the present application, another pre-trained ML model may be provided for generating the thresholds in the configuration information. Here, the ML model may be built based on any suitable ML network structure and be trained by training data. For example, any generative ML model based on a conditional variational, or a variational autoencoder with optimized latent search, or any other ML model may be used to translate the offloading action into configuration file that may be easily installed on the network device. One key step is to generate the threshold values in the configuration file based on the action as well as the requirements and restrictions of the operator. Any generative model (M: X, T) may be trained, where the input ^represents the action and the output T represents the threshold value(s) to be set in the configuration file.

[0093] After the ML model is trained, it may describe an association relationship between the type of an action and estimations for the respective thresholds. With these embodiments, the generative ML model may accept translatable action 230 and then translate the action 230 into thresholds to be set in the network device. In one example, the configuration file that consists of a set of threshold values that are instructed for a base station may be provided as below:

{ 'coverage user threshold' :<float>,

'coverage _prb_threshold': <float>,

'capacity user threshold' : <float>, 'capacity prb threshold': <float>}

[0094] In the above configuration information, line 1 defines the threshold for the number of users in the coverage cell, line 2 defines the threshold for the PRB in the coverage cell, line 3 defines the threshold for the number of users in the capacity cell, line 4 defines the threshold for the PRB in the capacity cell. It is to be understood that the above configuration file is just a schematic example. In other embodiments, the configuration file may have a different format for storing the necessary thresholds.

[0095] Once the configuration information is determined, it may be transmitted to the network device and then be loaded in the network device. With these embodiments, with the configuration information, undesired terminal devices may be automatically offloaded from the first cell 110 into the second cell 120. Therefore, the first cell 110 may be locked or the energy cost may be reduced to a certain level, and then the total energy consumption may be lowered.

[0096] In some embodiments of the present application, users may be offloaded between cells within the same sector of the base station. Alternatively and/or in addition , users may be offloaded between cells within different neighboring sectors of the base station. With these embodiments, users within the same or neighboring areas may be managed in a flexible and effective way, and user(s) in a source cell with a small user number and low traffic may be offloaded to another cell. At this point, the source cell may be locked down so as to reduce the energy consumption.

[0097] In some embodiments of the present application, QoS levels of the to-be-offloaded users may be considered in the offloading. Accordingly, the user information may be collected to ensure that changes in the QoS levels related to the offloading will not affect the users’ normal operation. Here, the user information may comprise but not be limited to mission criticality, tasks of user, user contexts, user locations regarding applications running on the plurality of terminal devices, and terminal device information, and the like. For example, if the mission criticality shows that a user is running a video application demanding a heavy streaming traffic, it is preferred that the user should not be offloaded to another cell. Further, the ML model may consider other user information such as the task of the user, the user context, and the like in determining the to- be-offloaded user(s). With these embodiments, impacts of the offloading on the users’ normal operations may be maintained at a relative lower level.

[0098] In some embodiments of the present application, the user information is one of the most crucial elements to ensure that the offloading will not affect the normal operations of the users. Accordingly, the user information may include all the UE metrics related to QoS and/or the user contexts, e.g., metrics obtained via sensors of operating systems of the terminal devices. Specifically, these sensors may include but not be limited to an accelerometer, a Global Position System (GPS), a battery level, a predicted application type, and the like. Further, the user information may include metadata information of the user and/or other data being involved in the decision, such as priorities, energy vs network footprint, and the like. Alternatively and/or in addition, the user information may be represented in any appropriate formats. For example, the user information may be shared not only in raw data means to the base station but also in the compressed and/or anonymized and/or encrypted forms to meet the privacy concerns.

[0099] The above paragraphs have provided details of individual steps in determining and implementing the traffic offloading. Hereinafter, reference will be made to Fig. 8 for a whole traffic offloading procedure, here Fig. 8 illustrates a schematic flowchart 800 for implementing traffic offloading via a configuration file in a communication system in accordance with embodiments of the present disclosure. At a block 810, a request may be received for determining whether traffic offloading may be implemented in the plurality of cells that are served by the network device 130. Here, depending on specific requirements of the operator, the request may be represented in various formats. For example, the request may indicate a rule for the traffic offloading, a network state for the network device 130, metadata related to the cells, and quality index distribution related to users of terminal devices that are served in these cells. Alternatively and/or in addition, the request may comprise more or less data as long as subsequent steps may obtain enough information from the request.

[0100] At a block 820, the traffic offloading policy may be learned by retaining users in the wanted quality levels. This step here relates to the training procedure, and the ML model 220 may be trained by maximizing the reward as described in the above paragraphs. With this step, the ML model 220 may be trained based on the RL technique in an accurate and effective way. At a block 830, the traffic offloading action may be obtained from the ML model 220 based on the above rules and/or other constraints. Specifically, the traffic offloading action here may indicate that a group of terminal devices should be offloaded from the first cell 110 to the second cell 120 by retaining necessary quality levels. In other words, after the traffic offloading, energy consumption of the network device 130 will be reduced and the quality levels of the offloaded terminal devices will not be affected, therefore the purpose of energy saving may be achieved.

[0101] At a block 840, the traffic offloading action may be translated into the configuration file. Here, a pre-trained generative ML model may be used for the translation, and thus the translation procedure does not involve any human workload. Next, at a block 850, the configuration file may be sent to the desired consumer, such as the base station whose energy consumption needs to be controlled. [0102] In some embodiments of the present application, the offloading procedure may be implemented in a centralized way. For example, the above procedure as illustrated in Fig. 8 may be implemented in an individual computing device that may access the necessary data (such as the user information 212 for users in the plurality of cells, the cell state information 210 for the cells, and the ML model 220) for the offloading. Alternatively and/or in addition, the offloading procedure may be implemented in a distributed way. At this point, the offloading procedure may relate to interactions among multiple services (for example, provided by one or more computing devices) in the communication system. With these embodiments, the offloading procedure may be implemented in an easy and flexible way, and thus various network devices may be benefitted from the offloading solution.

[0103] Fig. 9 illustrates a schematic signaling chart 900 for generating a configuration file based on a partially trained (or an untrained) ML model in accordance with embodiments of the present disclosure. As illustrated in Fig. 9, multiple Network Functions (NFs) are provided in the communication system, and the offloading procedure may be implemented among these NFs. In some embodiments of the present application, the offloading procedure may be started by an offloading request from a network function in the communication system. Specifically, at the beginning, a NF 920 ,such as a Network Data Analytics Function (NWDAF) sends 901 a discovery request to a Network Repository Function (NRF) 922 to start the offloading procedure. The discovery request may comprise a unique identification for identifying the offloading procedure.

[0104] In some embodiments of the present application, the cell state information 210 of the plurality of cells may be obtained in various ways. For example, the state information is obtained via an address of a radio resource control agent provided by a network function in the communication system. As illustrated in Fig. 9, the discovery request may be sent to NRF 922 to obtain the address of a Radio Resource Control (RRC) agent. In response to the discovery request, the NRF 922 responds 902 back to the NF 920 with the address of the RRC agent. Once the address is received, the NF 920 may obtain the desired state information via the address and then sends 903 a model request to the NF 924 (with the RRC agent) to obtain a configuration file from the RRC agent for the given network state. Here, the ML model 220 may be deployed in the NF 924.

[0105] In some embodiments of the present application, a rule defining constraints for the offloading is obtained from the NF 920, and the action is determined further based on the rule. It also sends operator specific rules to the NF 924 including the number of users in different QoS levels (for example, 5QI or PQI in the v2v communication, or CQI in 4G). In some embodiments of the present application, the rule may be defined in various formats, for example, a rule may be defined as:

[0106] “if (self.this_quality_index_group_list[quality index] < self, prev this quality index group list [quality index]): reward = -1.”

[0107] In the above rule, “this quality index group list” includes a histogram of different quality indices, i.e., the number of RRC users are given in each quality index. If the decision is given to reduce the number of users in the 5QI value where no users should be offloaded, a reward value of -1 is returned to penalize the action. With these embodiments, the reward may be updated based on the rules, and thus the ML model 220 may be adjusted in a flexible way accordingly.

[0108] In response to the model request from the NF 920, the NF 924 simulates the state in GYM environment. If the ML model 220 does not exist, for example, untrained and/or partly- trained, the NF 924 initiates 904 a training procedure with the received constraints for e.g., rules specific to the operator so that may be used as input to the reward function. In some embodiments of the present application, the energy consumption and QoS attributes used in the training procedure may be determined according to the pre-trained ML model deployed at the NF 928 in the communication system. The pre-trained ML model describes an association relationship between the state information of the related cells and estimations for the energy consumption and QoS attributes. At this point, the training procedure includes interactions with the NF 928 (with the estimation models), and then the NF 928 estimates the consequences of the actions in the form of reward. Further, the NF 928 sends 905 back the reward for the next state, and then the NF 924 updates 906 the state. At this point, the ML model is trained and may output the offload action after the training procedure.

[0109] Further, in the case a generative model (for translating the action into the configuration file) needs to be trained, the NF 924 initiates 907 a training procedure for training the generative model. Once the generative model is trained, the NF 924 sends 908 the action to the NF 926 to start the translation. The NF 928 generate 909 the configuration file by translating the action with the generative model. Next, the NF 928 sends 910 the generated configuration file to the NF 924. In some embodiments of the present application, in order to transmit the configuration file to the network device, the configuration file may be sent to the network device via the NF 920. Therefore, the NF 924 sends 911 the generated configuration file to the NF 920, and then the configuration file may be loaded into the network device for triggering the offloading procedure.

[0110] In some embodiments of the present application, the reward (for example, including the impacted QoS parameters, and the obtained energy reduction) is generated according to the ML model based on the state information of the plurality of cells, and then the NF 924 registers 912 the reward together with the offloading identification, the metadata, the rule, and the configuration file. In some embodiments of the present application, the NF 920 registers 913 the reward together with the offloading identification, the metadata, the rule, and the configuration file. With these embodiments, details related to the offloading procedure may be registered at one or more desired NFs for records.

[0111] Fig. 9 illustrates a whole offloading procedure that includes training the ML model 220 and the generative ML model. In some embodiments of the present application, these ML models may be trained in advance, and then the pre-trained ML models may be deployed in respective NFs and output the corresponding action and configuration file during operations of the communication system. Fig. 10 illustrates a schematic signaling chart 1000 for generating a configuration file based on a trained ML model in accordance with some embodiments of the present disclosure. As illustrated in Fig. 10, the NF 920 sends 1001 a discovery request to the NRF 922 to obtain the address of RRC agent. The NRF 922 responds 1002 back to the NF 920 with the address of the RRC agent. The NF 920 sends 1003 the model request to the NF 924 to obtain the action from ML model 220. The NF 924 sends 1004 the action and state, and then the NF 928 returns 1006 the next state and the reward to the NF 924. The NF 926 generates 1007 the configuration file and then sends 1008 the generated configuration file to the NF 924. The NF 924 sends 1009 the generated configuration file and the reward to the NF 920, and registers 1010 the reward together with the offloading identification, the metadata, the rule, and the configuration file. The NF 920 registers 1011 the reward together with the offloading identification, the metadata, the rule, and the configuration file. As illustrated in Fig. 10, the ML model and the generative ML model are trained in advance and then the pre-trained ML model is directly called to determine the offloading action. Further, the pre-trained generative ML model is directly used to translate the offloading action to the configuration file.

[0112] Fig. 11 illustrates a schematic flowchart of a method 1100 for generating configuration information for traffic offloading in a communication system based on ML in accordance with embodiments of the present disclosure. As illustrated in Fig. 11, at a block 1110, user information is collected for a plurality of terminal devices in a plurality of cells served by a network device in the communication system. At a block 1120, an action is determined according to an ML model, here the action is to be implemented by the network device based on the user information and state information of the plurality of cells, the action being used for offloading a group of terminal devices in the plurality of terminal devices from a first cell in the plurality of cells to a second cell in the plurality of cells. At a block 1130, configuration information is obtained for the network device based on the action. At a block 1140, the configuration information is transmitted to the network device. [0113] In some embodiments of the present application, energy consumption and Quality of Service (QoS) attributes associated with a third and a fourth cell are determined based on state information of the third and fourth cells. A first objective function is generated for training the ML model based on the energy consumption and QoS attributes. The ML model is trained based on the first objective function.

[0114] In some embodiments of the present application, the energy consumption and QoS attributes are determined according to a pre-trained ML model deployed at a first network function in the communication system. Here, the pre-trained ML model describes an association relationship between the state information of the third and fourth cells and estimations for the energy consumption and QoS attributes.

[0115] In some embodiments of the present application, in order to generate the first objective function , an energy saving metric is determined based on a change of the energy consumption attributes caused by offloading one or more terminal devices from the third cell to the fourth cell. A QoS reduction metric is generated based on a change of the QoS attributes caused by offloading the one or more terminal devices from the third cell to the fourth cell. The first objective function is created based on the energy saving metric and the QoS reduction metric.

[0116] In some embodiments of the present application, in order to obtain the configuration information, respective thresholds are generated for the first and second cells in the configuration information based on a type of the action.

[0117] In some embodiments of the present application, a threshold, in the respective thresholds, for a cell in the first and second cells comprises at least one of: a threshold for a number of terminal devices in the cell, or a threshold associated with a physical resource block for the cell.

[0118] In some embodiments of the present application, the type of the action comprises any of: an increment type indicating that the number of terminal devices in a cell of the first and second cells is to be increased by a first step size, a decrement type indicating that the number of terminal devices in a cell of the first and second cells is to be decreased by a second step size, or a nochange type indicating that the number of terminal devices in a cell of the first and second cell is unchanged.

[0119] In some embodiments of the present application, at least one of the first and second step sizes is determined by adapting a default step size based on a second objective function that is generated by the ML model based on the state information of the plurality of cells.

[0120] In some embodiments of the present application, in order to generate the respective thresholds in the configuration information, the respective thresholds are determined according to a pre-trained ML model that is deployed at a second network function in the communication system. Here, the pre-trained ML model describes an association relationship between the type of an action and estimations for the respective thresholds.

[0121] In some embodiments of the present application, the action is determined in response to receiving an offloading request from a third network function in the communication system.

[0122] In some embodiments of the present application, a rule defining constraints for the offloading is obtained from the third network function, and the action is determined further based on the rule.

[0123] In some embodiments of the present application, in order to transmit the configuration information to the network device, the configuration information is sent to the network device via the third network function.

[0124] In some embodiments of the present application, the state information of the plurality of cells is obtained via an address of a radio resource control agent provided by a fourth network function in the communication system.

[0125] In some embodiments of the present application, a third objective function is generated according to the ML model based on the state information of the plurality of cells. The configuration information is registered along with the third objective function at one or more network functions in the communication system.

[0126] In some embodiments of the present application, the network device is a base station in the communication system, and the first and second cells are within any of: the same sector of the base station; or different neighboring sectors of the base station.

[0127] In some embodiments of the present application, the user information comprises any of: mission criticality, tasks of user, user contexts, user locations regarding applications running on the plurality of terminal devices, and terminal device information.

[0128] In some embodiments of the present application, the ML model is pre-trained and serves as a digital twin in the communication system.

[0129] In some embodiments of the present application, the method 1100 may be implemented in any computing device in the communication system. Fig. 12 illustrates an example of a communication system 1200 in accordance with some embodiments.

[0130] In the example, the communication system 1200 includes a telecommunication network 1202 that includes an access network 1204, such as a radio access network (RAN), and a core network 1206, which includes one or more core network nodes 1208. The access network 1204 includes one or more access network nodes, such as network nodes 1210a and 1210b (one or more of which may be generally referred to as network nodes 1210), or any other similar 3rd Generation Partnership Project (3 GPP) access node or non-3GPP access point. The network nodes 1210 facilitate direct or indirect connection of user equipment (UE), such as by connecting UEs 1212a, 1212b, 1212c, and 1212d (one or more of which may be generally referred to as UEs 1212) to the core network 1206 over one or more wireless connections.

[0131] Examples of wireless communication over a wireless connection include transmitting and/or receiving wireless signals using electromagnetic waves, radio waves, infrared waves, and/or other types of signals suitable for conveying information without the use of wires, cables, or other material conductors. Moreover, in different embodiments, the communication system 1200 may include any number of wired or wireless networks, network nodes, UEs, and/or any other components or systems that may facilitate or participate in the communication of data and/or signals whether via wired or wireless connections. The communication system 1200 may include and/or interface with any type of communication, telecommunication, data, cellular, radio network, and/or other similar type of system.

[0132] The UEs 1212 may be any of a wide variety of communication devices, including wireless devices arranged, configured, and/or operable to communicate wirelessly with the network nodes 1210 and other communication devices. Similarly, the network nodes 1210 are arranged, capable, configured, and/or operable to communicate directly or indirectly with the UEs 1212 and/or with other network nodes or equipment in the telecommunication network 1202 to enable and/or provide network access, such as wireless network access, and/or to perform other functions, such as administration in the telecommunication network 1202.

[0133] In the depicted example, the core network 1206 connects the network nodes 1210 to one or more hosts, such as host 1216. These connections may be direct or indirect via one or more intermediary networks or devices. In other examples, network nodes may be directly coupled to hosts. The core network 1206 includes one more core network nodes (e.g., core network node 1208) that are structured with hardware and software components. Features of these components may be substantially similar to those described with respect to the UEs, network nodes, and/or hosts, such that the descriptions thereof are generally applicable to the corresponding components of the core network node 1208. Example core network nodes include functions of one or more of a Mobile Switching Center (MSC), Mobility Management Entity (MME), Home Subscriber Server (HSS), Access and Mobility Management Function (AMF), Session Management Function (SMF), Authentication Server Function (AUSF), Subscription Identifier De-concealing function (SIDF), Unified Data Management (UDM), Security Edge Protection Proxy (SEPP), Network Exposure Function (NEF), and/or a User Plane Function (UPF). [0134] The host 1216 may be under the ownership or control of a service provider other than an operator or provider of the access network 1204 and/or the telecommunication network 1202, and may be operated by the service provider or on behalf of the service provider. The host 1216 may host a variety of applications to provide one or more service. Examples of such applications include live and pre-recorded audio/video content, data collection services such as retrieving and compiling data on various ambient conditions detected by a plurality of UEs, analytics functionality, social media, functions for controlling or otherwise interacting with remote devices, functions for an alarm and surveillance center, or any other such function performed by a server.

[0135] As a whole, the communication system 1200 of Figure 12 enables connectivity between the UEs, network nodes, and hosts. In that sense, the communication system may be configured to operate according to predefined rules or procedures, such as specific standards that include, but are not limited to: Global System for Mobile Communications (GSM); Universal Mobile Telecommunications System (UMTS); Long Term Evolution (LTE), and/or other suitable 2G, 3G, 4G, 5G standards, or any applicable future generation standard (e.g., 6G); wireless local area network (WLAN) standards, such as the Institute of Electrical and Electronics Engineers (IEEE) 802.11 standards (WiFi); and/or any other appropriate wireless communication standard, such as the Worldwide Interoperability for Microwave Access (WiMax), Bluetooth, Z-Wave, Near Field Communication (NFC) ZigBee, LiFi, and/or any low-power wide-area network (LPWAN) standards such as LoRa and Sigfox.

[0136] In some examples, the telecommunication network 1202 is a cellular network that implements 3 GPP standardized features. Accordingly, the telecommunications network 1202 may support network slicing to provide different logical networks to different devices that are connected to the telecommunication network 1202. For example, the telecommunications network 1202 may provide Ultra Reliable Low Latency Communication (URLLC) services to some UEs, while providing Enhanced Mobile Broadband (eMBB) services to other UEs, and/or Massive Machine Type Communication (mMTC)/Massive loT services to yet further UEs.

[0137] In some examples, the UEs 1212 are configured to transmit and/or receive information without direct human interaction. For instance, a UE may be designed to transmit information to the access network 1204 on a predetermined schedule, when triggered by an internal or external event, or in response to requests from the access network 1204. Additionally, a UE may be configured for operating in single- or multi -RAT or multi-standard mode. For example, a UE may operate with any one or combination of Wi-Fi, NR (New Radio) and LTE, i.e. being configured for multi-radio dual connectivity (MR-DC), such as E-UTRAN (Evolved-UMTS Terrestrial Radio Access Network) New Radio - Dual Connectivity (EN-DC). [0138] In the example, the hub 1214 communicates with the access network 1204 to facilitate indirect communication between one or more UEs (e.g., UE 1212c and/or 1212d) and network nodes (e.g., network node 1210b). In some examples, the hub 1214 may be a controller, router, content source and analytics, or any of the other communication devices described herein regarding UEs. For example, the hub 1214 may be a broadband router enabling access to the core network 1206 for the UEs. As another example, the hub 1214 may be a controller that sends commands or instructions to one or more actuators in the UEs. Commands or instructions may be received from the UEs, network nodes 1210, or by executable code, script, process, or other instructions in the hub 1214. As another example, the hub 1214 may be a data collector that acts as temporary storage for UE data and, in some embodiments, may perform analysis or other processing of the data. As another example, the hub 1214 may be a content source. For example, for a UE that is a VR headset, display, loudspeaker or other media delivery device, the hub 1214 may retrieve VR assets, video, audio, or other media or data related to sensory information via a network node, which the hub 1214 then provides to the UE either directly, after performing local processing, and/or after adding additional local content. In still another example, the hub 1214 acts as a proxy server or orchestrator for the UEs, in particular in if one or more of the UEs are low energy loT devices.

[0139] The hub 1214 may have a constant/persistent or intermittent connection to the network node 1210b. The hub 1214 may also allow for a different communication scheme and/or schedule between the hub 1214 and UEs (e.g., UE 1212c and/or 1212d), and between the hub 1214 and the core network 1206. In other examples, the hub 1214 is connected to the core network 1206 and/or one or more UEs via a wired connection. Moreover, the hub 1214 may be configured to connect to an M2M service provider over the access network 1204 and/or to another UE over a direct connection. In some scenarios, UEs may establish a wireless connection with the network nodes 1210 while still connected via the hub 1214 via a wired or wireless connection. In some embodiments, the hub 1214 may be a dedicated hub - that is, a hub whose primary function is to route communications to/from the UEs from/to the network node 1210b. In other embodiments, the hub 1214 may be a non-dedicated hub - that is, a device which is capable of operating to route communications between the UEs and network node 1210b, but which is additionally capable of operating as a communication start and/or end point for certain data channels.

[0140] In embodiments of the present application, the network device may comprise any network node in the communication system. Figure 13 shows a network node 1300 in accordance with some embodiments. As used herein, network node refers to equipment capable, configured, arranged and/or operable to communicate directly or indirectly with a UE and/or with other network nodes or equipment, in a telecommunication network. Examples of network nodes include, but are not limited to, access points (APs) (e.g., radio access points), base stations (BSs) (e.g., radio base stations, Node Bs, evolved Node Bs (eNBs) and NR NodeBs (gNBs)).

[0141] Base stations may be categorized based on the amount of coverage they provide (or, stated differently, their transmit power level) and so, depending on the provided amount of coverage, may be referred to as femto base stations, pico base stations, micro base stations, or macro base stations. A base station may be a relay node or a relay donor node controlling a relay. A network node may also include one or more (or all) parts of a distributed radio base station such as centralized digital units and/or remote radio units (RRUs), sometimes referred to as Remote Radio Heads (RRHs). Such remote radio units may or may not be integrated with an antenna as an antenna integrated radio. Parts of a distributed radio base station may also be referred to as nodes in a distributed antenna system (DAS).

[0142] Other examples of network nodes include multiple transmission point (multi-TRP) 5G access nodes, multi-standard radio (MSR) equipment such as MSR BSs, network controllers such as radio network controllers (RNCs) or base station controllers (BSCs), base transceiver stations (BTSs), transmission points, transmission nodes, multi-cell/multicast coordination entities (MCEs), Operation and Maintenance (O&M) nodes, Operations Support System (OSS) nodes, Self-Organizing Network (SON) nodes, positioning nodes (e.g., Evolved Serving Mobile Location Centers (E-SMLCs)), and/or Minimization of Drive Tests (MDTs).

[0143] The network node 1300 includes a processing circuitry 1302, a memory 1304, a communication interface 1306, and a power source 1308. The network node 1300 may be composed of multiple physically separate components (e.g., a NodeB component and a RNC component, or a BTS component and a BSC component, etc.), which may each have their own respective components. In certain scenarios in which the network node 1300 comprises multiple separate components (e.g., BTS and BSC components), one or more of the separate components may be shared among several network nodes. For example, a single RNC may control multiple NodeBs. In such a scenario, each unique NodeB and RNC pair, may in some instances be considered a single separate network node. In some embodiments, the network node 1300 may be configured to support multiple radio access technologies (RATs). In such embodiments, some components may be duplicated (e.g., separate memory 1304 for different RATs) and some components may be reused (e.g., a same antenna 1310 may be shared by different RATs). The network node 1300 may also include multiple sets of the various illustrated components for different wireless technologies integrated into network node 1300, for example GSM, WCDMA, LTE, NR, WiFi, Zigbee, Z-wave, LoRaWAN, Radio Frequency Identification (RFID) or Bluetooth wireless technologies. These wireless technologies may be integrated into the same or different chip or set of chips and other components within network node 1300.

[0144] The processing circuitry 1302 may comprise a combination of one or more of a microprocessor, controller, microcontroller, central processing unit, digital signal processor, application-specific integrated circuit, field programmable gate array, or any other suitable computing device, resource, or combination of hardware, software and/or encoded logic operable to provide, either alone or in conjunction with other network node 1300 components, such as the memory 1304, to provide network node 1300 functionality.

[0145] In some embodiments, the processing circuitry 1302 includes a system on a chip (SOC). In some embodiments, the processing circuitry 1302 includes one or more of radio frequency (RF) transceiver circuitry 1312 and baseband processing circuitry 1314. In some embodiments, the radio frequency (RF) transceiver circuitry 1312 and the baseband processing circuitry 1314 may be on separate chips (or sets of chips), boards, or units, such as radio units and digital units. In alternative embodiments, part or all of RF transceiver circuitry 1312 and baseband processing circuitry 1314 may be on the same chip or set of chips, boards, or units.

[0146] The memory 1304 may comprise any form of volatile or non-volatile computer- readable memory including, without limitation, persistent storage, solid-state memory, remotely mounted memory, magnetic media, optical media, random access memory (RAM), read-only memory (ROM), mass storage media (for example, a hard disk), removable storage media (for example, a flash drive, a Compact Disk (CD) or a Digital Video Disk (DVD)), and/or any other volatile or non-volatile, non-transitory device-readable and/or computer-executable memory devices that store information, data, and/or instructions that may be used by the processing circuitry 1302. The memory 1304 may store any suitable instructions, data, or information, including a computer program, software, an application including one or more of logic, rules, code, tables, and/or other instructions capable of being executed by the processing circuitry 1302 and utilized by the network node 1300. The memory 1304 may be used to store any calculations made by the processing circuitry 1302 and/or any data received via the communication interface 1306. In some embodiments, the processing circuitry 1302 and memory 1304 is integrated.

[0147] The communication interface 1306 is used in wired or wireless communication of signaling and/or data between a network node, access network, and/or UE. As illustrated, the communication interface 1306 comprises port(s)/terminal(s) 1316 to send and receive data, for example to and from a network over a wired connection. The communication interface 1306 also includes radio front-end circuitry 1318 that may be coupled to, or in certain embodiments a part of, the antenna 1310. Radio front-end circuitry 1318 comprises filters 1320 and amplifiers 1322. The radio front-end circuitry 1318 may be connected to an antenna 1310 and processing circuitry 1302. The radio front-end circuitry may be configured to condition signals communicated between antenna 1310 and processing circuitry 1302. The radio front-end circuitry 1318 may receive digital data that is to be sent out to other network nodes or UEs via a wireless connection. The radio front-end circuitry 1318 may convert the digital data into a radio signal having the appropriate channel and bandwidth parameters using a combination of filters 1320 and/or amplifiers 1322. The radio signal may then be transmitted via the antenna 1310. Similarly, when receiving data, the antenna 1310 may collect radio signals which are then converted into digital data by the radio front-end circuitry 1318. The digital data may be passed to the processing circuitry 1302. In other embodiments, the communication interface may comprise different components and/or different combinations of components.

[0148] In certain alternative embodiments, the network node 1300 does not include separate radio front-end circuitry 1318, instead, the processing circuitry 1302 includes radio front-end circuitry and is connected to the antenna 1310. Similarly, in some embodiments, all or some of the RF transceiver circuitry 1312 is part of the communication interface 1306. In still other embodiments, the communication interface 1306 includes one or more ports or terminals 1316, the radio front-end circuitry 1318, and the RF transceiver circuitry 1312, as part of a radio unit (not shown), and the communication interface 1306 communicates with the baseband processing circuitry 1314, which is part of a digital unit (not shown).

[0149] The antenna 1310 may include one or more antennas, or antenna arrays, configured to send and/or receive wireless signals. The antenna 1310 may be coupled to the radio front-end circuitry 1318 and may be any type of antenna capable of transmitting and receiving data and/or signals wirelessly. In certain embodiments, the antenna 1310 is separate from the network node 1300 and connectable to the network node 1300 through an interface or port.

[0150] The antenna 1310, communication interface 1306, and/or the processing circuitry 1302 may be configured to perform any receiving operations and/or certain obtaining operations described herein as being performed by the network node. Any information, data and/or signals may be received from a UE, another network node and/or any other network equipment. Similarly, the antenna 1310, the communication interface 1306, and/or the processing circuitry 1302 may be configured to perform any transmitting operations described herein as being performed by the network node. Any information, data and/or signals may be transmitted to a UE, another network node and/or any other network equipment.

[0151] The power source 1308 provides power to the various components of network node 1300 in a form suitable for the respective components (e.g., at a voltage and current level needed for each respective component). The power source 1308 may further comprise, or be coupled to, power management circuitry to supply the components of the network node 1300 with power for performing the functionality described herein. For example, the network node 1300 may be connectable to an external power source (e.g., the power grid, an electricity outlet) via an input circuitry or interface such as an electrical cable, whereby the external power source supplies power to power circuitry of the power source 1308. As a further example, the power source 1308 may comprise a source of power in the form of a battery or battery pack which is connected to, or integrated in, power circuitry. The battery may provide backup power should the external power source fail.

[0152] Embodiments of the network node 1300 may include additional components beyond those shown in Figure 13 for providing certain aspects of the network node’s functionality, including any of the functionality described herein and/or any functionality necessary to support the subject matter described herein. For example, the network node 1300 may include user interface equipment to allow input of information into the network node 1300 and to allow output of information from the network node 1300. This may allow a user to perform diagnostic, maintenance, repair, and other administrative functions for the network node 1300.

[0153] In some example embodiments, an apparatus capable of performing any of the method 1100 may comprise means for performing the respective operations of the method 1100. The means may be implemented in any suitable form. For example, the means may be implemented in a circuitry or software module. In some embodiments, an apparatus is proposed for traffic offloading in a communication system. The apparatus comprises: means for collecting user information of a plurality of terminal devices in a plurality of cells served by a network device in the communication system; means for determining, according to an ML model, an action that is to be implemented by the network device based on the user information and state information of the plurality of cells, the action being used for offloading a group of terminal devices in the plurality of terminal devices from a first cell in the plurality of cells to a second cell in the plurality of cells; means for obtaining configuration information for the network device based on the action; and means for transmitting the configuration information to the network device. Further, the apparatus may comprise means for implementing other steps in the above method 1100, and details will be omitted hereinafter.

[0154] In some example embodiments, a computing device is provided for traffic offloading in a communication system. The computing device comprises at least one processor coupled to a computer-readable memory unit, the memory unit comprising instructions that when executed by the at least one processor implement a method for traffic offloading in a communication system. The method comprises: collecting user information of a plurality of terminal devices in a plurality of cells served by a network device in the communication system; determining, according to an ML model, an action that is to be implemented by the network device based on the user information and state information of the plurality of cells, the action being used for offloading a group of terminal devices in the plurality of terminal devices from a first cell in the plurality of cells to a second cell in the plurality of cells; obtaining configuration information for the network device based on the action; and transmitting the configuration information to the network device.

[0155] In some embodiments of the present application, the method further comprises: determining energy consumption and Quality of Service (QoS) attributes associated with a third and a fourth cell based on state information of the third and fourth cells; generating a first objective function for training the ML model based on the energy consumption and QoS attributes; and training the ML model based on the first objective function.

[0156] In some embodiments of the present application, the energy consumption and QoS attributes are determined according to a pre-trained ML model deployed at a first network function in the communication system, the pre-trained ML model describing an association relationship between the state information of the third and fourth cells and estimations for the energy consumption and QoS attributes.

[0157] In some embodiments of the present application, generating the first objective function further comprises: determining an energy saving metric based on a change of the energy consumption attributes caused by offloading one or more terminal devices from the third cell to the fourth cell; determining a QoS reduction metric based on a change of the QoS attributes caused by offloading the one or more terminal devices from the third cell to the fourth cell; and creating the first objective function based on the energy saving metric and the QoS reduction metric.

[0158] In some embodiments of the present application, obtaining the configuration information comprises: generating respective thresholds for the first and second cells in the configuration information based on a type of the action.

[0159] In some embodiments of the present application, a threshold, in the respective thresholds, for a cell in the first and second cells comprises at least one of: a threshold for a number of terminal devices in the cell, or a threshold associated with a physical resource block for the cell.

[0160] In some embodiments of the present application, the type of the action comprises any of: an increment type indicating that the number of terminal devices in a cell of the first and second cells is to be increased by a first step size, a decrement type indicating that the number of terminal devices in a cell of the first and second cells is to be decreased by a second step size, or a no- change type indicating that the number of terminal devices in a cell of the first and second cell is unchanged.

[0161] In some embodiments of the present application, at least one of the first and second step sizes is determined by adapting a default step size based on a second objective function that is generated by the ML model based on the state information of the plurality of cells.

[0162] In some embodiments of the present application, generating the respective thresholds in the configuration information further comprises: determining the respective thresholds according to a pre-trained ML model that is deployed at a second network function in the communication system, the pre-trained ML model describing an association relationship between the type of an action and estimations for the respective thresholds.

[0163] In some embodiments of the present application, the action is determined in response to receiving an offloading request from a third network function in the communication system.

[0164] In some embodiments of the present application, the method further comprises: obtaining from the third network function a rule defining constraints for the offloading, and wherein the action is determined further based on the rule.

[0165] In some embodiments of the present application, transmitting the configuration information to the network device comprises: sending the configuration information to the network device via the third network function.

[0166] In some embodiments of the present application, the state information of the plurality of cells is obtained via an address of a radio resource control agent provided by a fourth network function in the communication system.

[0167] In some embodiments of the present application, the method further comprises: generating, according to the ML model, a third objective function based on the state information of the plurality of cells; and registering, at one or more network functions in the communication system, the configuration information along with the third objective function.

[0168] In some embodiments of the present application, the network device is a base station in the communication system, and the first and second cells are within any of: the same sector of the base station; or different neighboring sectors of the base station.

[0169] In some embodiments of the present application, the user information comprises any of: mission criticality, tasks of user, user contexts, user locations regarding applications running on the plurality of terminal devices, and terminal device information.

[0170] In embodiments of the present application, the ML model is pre-trained and serves as a digital twin in the communication system. [0171] In some embodiments, a computer readable storage medium having instructions stored thereon is provided. The instructions when executed by at least one processor can cause the at least one processor to carry out the functionality in accordance with any one of the embodiments described herein. In some embodiments, the computer readable medium may be a non-transitory computer readable storage medium. A computer readable storage medium may include but not limited to an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the computer readable storage medium would include an electrical connection having one or more wires, a portable computer diskette, a hard disk, a read-only memory (ROM), a random access memory (RAM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0172] In some embodiments, a computer program comprising instructions is provided. The instructions, when executed by at least one processor, cause the at least one processor to carry out the functionality in accordance with any one of the embodiments described herein. In one embodiment, a carrier containing the aforementioned computer program product is provided. The carrier is one of an electronic signal, an optical signal, a radio signal, or a computer-readable storage medium (e.g., a non-transitory computer-readable medium).

[0173] Although the computing devices described herein (e.g., network nodes) may include the illustrated combination of hardware components, other embodiments may comprise computing devices with different combinations of components. It is to be understood that these computing devices may comprise any suitable combination of hardware and/or software needed to perform the tasks, features, functions and methods disclosed herein. Determining, calculating, obtaining or similar operations described herein may be performed by processing circuitry, which may process information by, for example, converting the obtained information into other information, comparing the obtained information or converted information to information stored in the network node, and/or performing one or more operations based on the obtained information or converted information, and as a result of said processing making a determination. Moreover, while components are depicted as single boxes located within a larger box, or nested within multiple boxes, in practice, computing devices may comprise multiple different physical components that make up a single illustrated component, and functionality may be partitioned between separate components. For example, a communication interface may be configured to include any of the components described herein, and/or the functionality of the components may be partitioned between the processing circuitry and the communication interface. In another example, non- computationally intensive functions of any of such components may be implemented in software or firmware and computationally intensive functions may be implemented in hardware.

[0174] In certain embodiments, some or all of the functionality described herein may be provided by processing circuitry executing instructions stored on in memory, which in certain embodiments may be a computer program product in the form of a non-transitory computer- readable storage medium. In alternative embodiments, some or all of the functionality may be provided by the processing circuitry without executing instructions stored on a separate or discrete device-readable storage medium, such as in a hard-wired manner. In any of those particular embodiments, whether executing instructions stored on a non-transitory computer-readable storage medium or not, the processing circuitry can be configured to perform the described functionality. The benefits provided by such functionality are not limited to the processing circuitry alone or to other components of the computing device, but are enjoyed by the computing device as a whole, and/or by end users and a wireless network generally.

[0175] Any appropriate steps, methods, features, functions, or benefits disclosed herein may be performed through one or more functional units or modules of one or more virtual apparatuses. Each virtual apparatus may comprise a number of these functional units. These functional units may be implemented via processing circuitry, which may include one or more microprocessor or microcontrollers, as well as other digital hardware, which may include DSPs, special-purpose digital logic, and the like. The processing circuitry may be configured to execute program code stored in memory, which may include one or several types of memory such as ROM, RAM, cache memory, flash memory devices, optical storage devices, etc. Program code stored in memory includes program instructions for executing one or more telecommunications and/or data communications protocols as well as instructions for carrying out one or more of the techniques described herein. In some embodiments, the processing circuitry may be used to cause the respective functional unit to perform corresponding functions according one or more embodiments of the present disclosure.

[0176] While processes in the figures may show a particular order of operations performed by certain embodiments of the present disclosure, it should be understood that such order is exemplary (e.g., alternative embodiments may perform the operations in a different order, combine certain operations, overlap certain operations, etc.).

[0177] Those skilled in the art will recognize improvements and modifications to the embodiments of the present disclosure. All such improvements and modifications are considered within the scope of the concepts disclosed herein.

Claims

WHAT IS CLAIMED IS:

1. A computer-implemented method for traffic offloading in a communication system, comprising: collecting user information of a plurality of terminal devices in a plurality of cells served by a network device in the communication system; determining, according to a machine learning model, an action that is to be implemented by the network device based on the user information and state information of the plurality of cells, the action being used for offloading a group of terminal devices in the plurality of terminal devices from a first cell in the plurality of cells to a second cell in the plurality of cells; obtaining configuration information for the network device based on the action; and transmitting the configuration information to the network device.

2. The method of claim 1, further comprising: determining energy consumption and Quality of Service (QoS) attributes associated with a third and a fourth cell based on state information of the third and fourth cells; generating a first objective function for training the machine learning model based on the energy consumption and QoS attributes; and training the machine learning model based on the first objective function.

3. The method of claim 2, wherein the energy consumption and QoS attributes are determined according to a pre-trained machine learning model deployed at a first network function in the communication system, the pre-trained machine learning model describing an association relationship between the state information of the third and fourth cells and estimations for the energy consumption and QoS attributes.

4. The method of claim 3, wherein generating the first objective function further comprises: determining an energy saving metric based on a change of the energy consumption attributes caused by offloading one or more terminal devices from the third cell to the fourth cell; determining a QoS reduction metric based on a change of the QoS attributes caused by offloading the one or more terminal devices from the third cell to the fourth cell; and creating the first objective function based on the energy saving metric and the QoS reduction metric.

5. The method of claim 1, wherein the user information comprises any of: mission criticality, tasks of user, user contexts, user locations regarding applications running on the plurality of terminal devices, and terminal device information.

6. The method of claim 1, wherein obtaining the configuration information comprises: generating respective thresholds for the first and second cells in the configuration information based on a type of the action.

7. The method of claim 6, wherein a threshold, in the respective thresholds, for a cell in the first and second cells comprises at least one of: a threshold for a number of terminal devices in the cell, or a threshold associated with a physical resource block for the cell.

8. The method of claim 6, wherein the type of the action comprises any of: an increment type indicating that the number of terminal devices in a cell of the first and second cells is to be increased by a first step size, a decrement type indicating that the number of terminal devices in a cell of the first and second cells is to be decreased by a second step size, or a no-change type indicating that the number of terminal devices in a cell of the first and second cell is unchanged.

9. The method of claim 8, wherein at least one of the first and second step sizes is determined by adapting a default step size based on a second objective function that is generated by the machine learning model based on the state information of the plurality of cells.

10. The method of claim 6, wherein generating the respective thresholds in the configuration information further comprises: determining the respective thresholds according to a pre-trained machine learning model that is deployed at a second network function in the communication system, the pre-trained machine learning model describing an association relationship between the type of an action and estimations for the respective thresholds.

11. The method of claim 1, wherein the action is determined in response to receiving an offloading request from a third network function in the communication system.

12. The method of claim 11, further comprising: obtaining from the third network function a rule defining constraints for the offloading, and wherein the action is determined further based on the rule.

13. The method of claim 11, wherein transmitting the configuration information to the network device comprises: sending the configuration information to the network device via the third network function.

14. The method of claim 1, wherein the state information of the plurality of cells is obtained via an address of a radio resource control agent provided by a fourth network function in the communication system.

15. The method of claim 1, further comprising: generating, according to the machine learning model, a third objective function based on the state information of the plurality of cells; and registering, at one or more network functions in the communication system, the configuration information along with the third objective function.

16. The method of claim 1, wherein the network device is a base station in the communication system, and the first and second cells are within any of: the same sector of the base station; or different neighboring sectors of the base station.

17. The method of claim 1, wherein the machine learning model is pre-trained and serves as a digital twin in the communication system.

18. A computing device, comprising at least one processor coupled to a computer-readable memory unit, the memory unit comprising instructions that when executed by the at least one processor implement a method for traffic offloading in a communication system, the method comprising: collecting user information of a plurality of terminal devices in a plurality of cells served by a network device in the communication system; determining, according to a machine learning model, an action that is to be implemented by the network device based on the user information and state information of the plurality of cells, the action being used for offloading a group of terminal devices in the plurality of terminal devices from a first cell in the plurality of cells to a second cell in the plurality of cells; obtaining configuration information for the network device based on the action; and transmitting the configuration information to the network device.

19. The device of claim 18, wherein the method further comprises: determining energy consumption and Quality of Service (QoS) attributes associated with a third and a fourth cell based on state information of the third and fourth cells; generating a first objective function for training the machine learning model based on the energy consumption and QoS attributes; and training the machine learning model based on the first objective function.

20. The device of claim 19, wherein the energy consumption and QoS attributes are determined according to a pre-trained machine learning model deployed at a first network function in the communication system, the pre-trained machine learning model describing an association relationship between the state information of the third and fourth cells and estimations for the energy consumption and QoS attributes.

21. The device of claim 20, wherein generating the first objective function further comprises: determining an energy saving metric based on a change of the energy consumption attributes caused by offloading one or more terminal devices from the third cell to the fourth cell; determining a QoS reduction metric based on a change of the QoS attributes caused by offloading the one or more terminal devices from the third cell to the fourth cell; and creating the first objective function based on the energy saving metric and the QoS reduction metric.

22. The device of claim 18, wherein the user information comprises any of: mission criticality, tasks of user, user contexts, user locations regarding applications running on the plurality of terminal devices, and terminal device information.

23. The device of claim 10, wherein obtaining the configuration information comprises: generating respective thresholds for the first and second cells in the configuration information based on a type of the action.

24. The device of claim 23, wherein a threshold, in the respective thresholds, for a cell in the first and second cells comprises at least one of: a threshold for a number of terminal devices in the cell, or a threshold associated with a physical resource block for the cell.

25. The device of claim 23, wherein the type of the action comprises any of: an increment type indicating that the number of terminal devices in a cell of the first and second cells is to be increased by a first step size, a decrement type indicating that the number of terminal devices in a cell of the first and second cells is to be decreased by a second step size, or a no-change type indicating that the number of terminal devices in a cell of the first and second cell is unchanged.

26. The device of claim 25, wherein at least one of the first and second step sizes is determined by adapting a default step size based on a second objective function that is generated by the machine learning model based on the state information of the plurality of cells.

27. The device of claim 23, wherein generating the respective thresholds in the configuration information further comprises: determining the respective thresholds according to a pre-trained machine learning model that is deployed at a second network function in the communication system, the pre-trained machine learning model describing an association relationship between the type of an action and estimations for the respective thresholds.

28. The device of claim 18, wherein the action is determined in response to receiving an offloading request from a third network function in the communication system.

29. The device of claim 28, wherein the method further comprises: obtaining from the third network function a rule defining constraints for the offloading, and wherein the action is determined further based on the rule.

30. The device of claim 28, wherein transmitting the configuration information to the network device comprises: sending the configuration information to the network device via the third network function.

31. The device of claim 18, wherein the state information of the plurality of cells is obtained via an address of a radio resource control agent provided by a fourth network function in the communication system.

32. The device of claim 18, wherein the method further comprises: generating, according to the machine learning model, a third objective function based on the state information of the plurality of cells; and registering, at one or more network functions in the communication system, the configuration information along with the third objective function.

33. The device of claim 18, wherein the network device is a base station in the communication system, and the first and second cells are within any of: the same sector of the base station; or different neighboring sectors of the base station.

34. The device of claim 18, wherein the machine learning model is pre-trained and serves as a digital twin in the communication system.

35. A non-transitory computer readable medium having instructions stored thereon, the instructions when executed by at least one processor cause the at least one processor to perform a method according to any of claims 1 to 17.

36. A computer program product comprising instructions, the instructions when executed by at least one processor cause the at least one processor to perform a method according to any of claims 1 to 17.