Method and system for organizing resources in a wireless network

Abstract

The present invention relates to a method for organizing resources in a wireless network. Multiple Quality of Service (QoS) parameters and Quality of Experience (QoE) feedback values ​​are collected for multiple applications. Variations in the collected values ​​are observed for a given application. A QoS degradation parameter that is most correlated with the QoE feedback variation is selected (702). An exponential parametric model is adjusted (704) to approximate the relationship between the QoE and the QoS degradation parameter. Resource organization (706) in the network is based on the adjusted model. The present invention further relates to a corresponding device and a corresponding computer program.

Description

[Technical Field]

[0001] The present disclosure relates to the field of telecommunications.

[0002] The present disclosure relates more particularly to a method and system for organizing resources of a wireless network. The present disclosure further relates to a corresponding non-transitory storage medium. Priority is claimed from European Patent Application No. 23305899.9, filed June 6, 2023, the contents of which are incorporated herein by reference. [Background technology]

[0003] The successful integration of applications into private 5G network deployments in industrial environments presents significant challenges. A key issue encountered during this process is optimizing application deployment to ensure seamless integration with existing applications. One key challenge is maintaining high quality of service (QoS) for newly deployed applications without sacrificing the quality of experience (QoE) of already deployed applications. Current solutions are inadequate because they cannot effectively address this issue. Therefore, there is a pressing need for the development of robust and efficient systems and methods for improving QoE while deploying multiple applications in communication networks, especially in industrial settings. Summary of the Invention [Means for solving the problem]

[0004] This disclosure improves the situation.

[0005] A method for organizing resources of a wireless network is proposed, the method being executed by at least one processing unit, comprising: a) collecting values ​​of a plurality of Quality of Service (QoS) parameters and Quality of Experience (QoE) feedback, each collected value being identified as being associated with an application; b) observing the variation of collected values ​​identified as being associated with the same given application; c) selecting a QoS degradation parameter, which is the QoS parameter whose value has the observed variation of the given application that is most correlated with the observed variation of the QoE feedback of the given application; d) adjusting an exponential parametric model that approximates the relationship between the observed variation in the QoE feedback of a given application and the observed variation in the selected QoS parameters of the given application; e) organizing resources of the wireless network based on the adjusted exponential parametric model.

[0006] A processing unit may be understood interchangeably as a software entity, a hardware entity, or a combination of one or more hardware and software entities, providing versatility in deployment options. A processing unit may correspond to, for example, a virtual network function such as a network data analysis function (NWDAF), or a proprietary function that collects feedback and QoS parameter values ​​from applications, or a virtual function implemented in an edge cloud server.

[0007] "Selecting" a QoS degradation parameter may refer, for example, to identifying or determining an appropriate parameter from among available QoS parameters, which may include, for example, latency, packet loss rate, bandwidth or jitter, and / or other parameters that can be derived or calculated from those mentioned, such as round trip time, network utilization, or error rate.

[0008] "Tuning" an exponential parametric model can be understood as determining, fitting, or learning the values ​​of parameters within the mathematical form of the exponential parametric model without modifying the mathematical form itself.

[0009] This method enables efficient management of resources in wireless networks by considering both QoS parameters and QoE feedback from applications. Furthermore, by identifying the QoS parameters that are most correlated with observed variations in QoE feedback, the method focuses on factors that are most likely to affect user experience.

[0010] Although the choice of the exponential parametric model is not arbitrary, it shows a strong tendency to converge towards an optimized value of QoE sensitivity. By accurately capturing the relationship between the observed variation in QoE feedback and the selected QoS parameters, the model enables effective resource allocation and optimization.

[0011] In an embodiment of the method, b), c), and d) are repeated for multiple applications deployed over the wireless network to obtain multiple adjusted exponential parametric models having a one-to-one relationship with the multiple applications; e) includes organizing resources of the wireless network based on each of the adjusted exponential parametric models.

[0012] This ensures that resources are allocated and managed in direct alignment with the requirements and behavior of each application.

[0013] In an embodiment of the method, the wireless network is a non-public network.

[0014] This allows for increased control and security over public networks.

[0015] In an embodiment of the method, the QoE feedback value for an application is derived from multiple client applications of said application.

[0016] This provides a more thorough assessment of the user experience of a given application, further contributing to improved resource allocation decisions.

[0017] In an embodiment of the method, the value of the QoE feedback comprises a mean opinion score and / or a peak signal to interference and noise ratio.

[0018] MoS is a subjective assessment of QoE, while PSNR is an objective assessment of QoE. Combining both is a dual approach that further contributes to improving resource allocation decisions.

[0019] In an embodiment of the method, e) comprises orchestrating virtual resources including network resources, storage resources, and / or processing resources.

[0020] This is particularly advantageous in wireless network environments where efficient utilization of available virtual resources is important to maintain reliable, high-quality connections.

[0021] In an embodiment of the method, e) comprises organizing radio resources.

[0022] This is particularly advantageous in wireless network environments where efficient utilization of available radio resources is important to maintain reliable, high quality connections.

[0023] In an embodiment of the method, c) is - for each QoS parameter whose values ​​are collected, determining a gradient of the collected values ​​of the QoE feedback with respect to the collected values ​​of said QoS parameter; and selecting the QoS parameter for which the highest gradient has been determined as the QoS degradation parameter.

[0024] This ensures that resources are organized based on the QoS parameters that have the greatest impact on QoE for a given application.

[0025] In an embodiment of the method, c) is running a principal component analysis scheme to determine the principal components that exhibit the greatest variance for the collected quality of experience feedback; selecting a QoS parameter whose collected value is closest to said principal component as a QoS degradation parameter.

[0026] The dimensionality reduction inherent in PCA reduces the complexity of QoS parameter selection.

[0027] In an embodiment of the method, c) is classifying the QoS parameters into classes with respect to the observed variation of QoE; and selecting the class for which the maximum variation in QoE feedback is observed as the QoS degradation parameter.

[0028] Classifying QoS parameters into classes further reduces the complexity of QoS parameter selection.

[0029] A system for organizing resources of a wireless network is further proposed, the system comprising: a) collecting values ​​of a plurality of Quality of Service (QoS) parameters and Quality of Experience (QoE) feedback, each collected value being identified as relevant to an application; b) observing variations in collected values ​​identified as being associated with the same given application; c) selecting a QoS degradation parameter, which is the QoS parameter whose value has the observed variation of the given application that is most correlated with the observed variation of the QoE feedback of the given application; d) calibrating an exponential parametric model that approximates the relationship between the observed variation in the QoE feedback of a given application and the observed variation in the selected QoS parameters of the given application; e) comprising at least one processing unit configured to organize resources of the wireless network based on the adjusted exponential parametric model.

[0030] It is further proposed a computer program comprising instructions which, when executed by a processing unit, cause the processing unit to carry out the above method.

[0031] Unless otherwise indicated, and as will be apparent from the following description, throughout the description herein, terms such as "computing," "calculating," "generating," and the like are understood to refer to the actions and / or processes of a computer or computing system, or similar electronic computing device, to manipulate and / or transform data represented as physical quantities, such as electronic quantities, in the registers and / or memory of the computing system into other data similarly represented as physical quantities in the memory, registers, or other such information storage, transmission, or display device of the computing system.

[0032] Embodiments of the present invention may include an apparatus for performing the operations herein. This apparatus may be specially constructed for the desired purposes, or it may comprise a general-purpose computer or a digital signal processor ("DSP") selectively activated or reconfigured by a computer program stored in the computer. Such a computer program may be stored on a computer-readable storage medium, including, but not limited to, a floppy disk, an optical disk, a CD-ROM, any type of disk including a magneto-optical disk, a read-only memory (ROM), a random access memory (RAM), an electrically programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a magnetic or optical card, or any other type of medium suitable for storing electronic instructions and capable of being coupled to a computer system bus.

[0033] The processes presented herein are not inherently related to any particular computer or other apparatus. Various general-purpose systems may be used with programs in accordance with the teachings herein, or it may prove convenient to construct a more specialized apparatus to perform the desired method. The desired structure for a variety of these systems will be apparent from the description below. Additionally, embodiments of the present invention are not described with reference to any particular programming language. It will be understood that a variety of programming languages ​​can be used to implement the teachings of the present invention as described herein.

[0034] Other features, details and advantages are set forth in the following detailed description and drawings. [Brief explanation of the drawings]

[0035] [Figure 1] FIG. 1 illustrates an exemplary private 5G deployment.

[0036] [Figure 2] FIG. 2 illustrates an exemplary integration of an application with the private 5G deployment of FIG.

[0037] [Figure 3] FIG. 1 illustrates an exemplary quality of service mapping in an acore network.

[0038] [Figure 4] FIG. 1 illustrates an exemplary quality of experience determination for a video streaming application.

[0039] [Figure 5] FIG. 1 illustrates an exemplary peak signal-to-noise ratio to mean opinion score mapping.

[0040] [Figure 6] FIG. 1 illustrates an exemplary private 5G network deployment.

[0041] [Figure 7] FIG. 1 illustrates an exemplary workflow of the quality of experience sensitivity learning and network resource organization process. DETAILED DESCRIPTION OF THE INVENTION

[0042] Figure 1 shows a typical deployment of an application over a private 5G network in a smart factory, where the application considered is a video streaming application from clients located on the factory floor to a server (108) located in the edge cloud. The edge server collects data from the clients (102) and provides additional functionality such as image recognition or artificial intelligence (AI).

[0043] A 5G mobile communication network or 5G system includes a 5G core network (5GCN) and a 5G new radio access network (5G-RAN). User equipment (UE) can connect to the 5GCN via the 5G new radio access network and further to a data network (DN) such as the Internet.

[0044] A key component of 5GCN is network functions separated by services to manage network behavior. These network functions include: An Access and Mobility Management Function (AMF) that acts as a single entry point for UE connectivity; each Session Management Function (SMF) selected by the AMF to manage a user session for the service requested by the UE; A User Plane Function (UPF) that carries IP data traffic (user plane) between the User Equipment (UE) and external networks; and Policy Control Function (PCF), which provides a policy control framework; Unified Data Management (UDM) function for managing network user data; a Network Data Analytics Function (NWDAF) that collects data from other 5G network functions in support of other 5G network functions, processes these data, and outputs analytics, such as statistical analysis or machine learning predictions; a Network Exposure Function (NEF) that exposes 3GPP® core network capabilities to third parties; etc.

[0045] The 5G network stack includes an open wireless architecture layer, a network layer, an open transport layer, and an application layer (APP), which can be mapped to the presentation and application layers of the OSI stack.

[0046] The application function (AF) 106 is the control plane of the application layer and provides application services.

[0047] The steps for integrating an application into a deployment are shown in Figures 2 and 3 and include the following steps:

[0048] The AF, which serves as the control plane for the application, requests a session with a specific application quality of service from the private 5G network (202). The request includes various parameters such as the UE address, AF identifier, flow description or external application identifier, QoS reference, QoS parameters, alternative service requirements, DNN, and S-NSSAI, among others. Additional details regarding the parameters that may be included in the request are provided in section 4.15.6.6 of TS 23.502. If the AF is a trusted network function, the request is sent directly to the network's Policy Control Function node (PCF), or to the Network Exposure Function (NEF) if the AF is not a trusted AF.

[0049] The PCF / NEF authorizes the request and maps it to internal 5G network QoS parameters such as 5G QoS Identifier (5QI), Allocation and Retention Priority (ARP), Guaranteed Flow Bit Rate (GFBR), Maximum Flow Bit Rate (MFBR), Maximum Packet Loss Rate, Delay-Critical Resource Type, Notification Control, Reflected QoS Attribute (RQA), Session AMBR, and UE-AMBR parameters (204). The QoS parameters depend on whether the QoS flow is Guaranteed Bit Rate (GBR) or non-GBR.

[0050] The Session Management Function (SMF) sets up a QoS flow between the user equipment (UE) and the network (206). The UPF inserts a QoS flow indicator (QFI) corresponding to the network QoS, and a mapping is performed between the QFI and the data radio bearer set up between the UE and the gNB.

[0051] The QoS application profile is applied to the application server and application client to shape traffic to the application QoS requested by the AF (208). This step may be applied after authorization of the request by the PCF / NEF.

[0052] A data path for the application session is established between the application server and the application client based on the requested application QoS (210).

[0053] By following these steps, seamless integration of applications into deployments in private 5G networks can be achieved while ensuring that the application's QoS is maintained throughout the session.

[0054] Figure 3 illustrates, in a specific example, the process of establishing and managing QoS flows between an application and a private 5G network using QFI insertion and DRB mapping to ensure that different types of traffic receive appropriate QoS treatment.

[0055] Three different types of protocol data units (PDUs) are shown: Internet PDUs, App4 PDUs, and IMS PDUs. Seven IP flows (302) are shown representing different types of input data traffic, including Internet PDUs, best effort, App1 video, App2 video and App3 streams, App4 PDUs, App4 streams, and IMS PDUs, voice and video. Each IP flow has its own QoS requirements, such as bandwidth, delay, and packet loss rate.

[0056] A set of service data flow / traffic flow templates (SDF / TFT) in the user plane function (UPF) enables the step of performing QoS flow identification insertion 304. Thus, each flow is assigned a QoS flow identifier (QFI) from 1 to 7.

[0057] For each PDU, a corresponding Service Data Application Protocol module (SDAP) is provided to the gNB. The SDAP enables the step of performing Data Radio Bearer (DRB) mapping (306). The SDAP can merge multiple QoS flows into a single DRB for efficient network utilization. Five data radio bearers are shown here.

[0058] Finally, for each PDU, a corresponding Service Data Adaptation Protocol and Traffic Flow Template module (SDAP+TFT) is provided to the UE to further optimize the QoS flow between the application and the network. The UE finally receives the traffic as seven IP flows (308).

[0059] The Quality of Experience (QoE) of an application is a key factor in ensuring user satisfaction. QoE is defined as the mean opinion score describing the quality of a received application flow, and the main term for this QoE metric is the Mean Opinion Score (MoS). MoS values ​​can range, for example, from 1 (poor) to 5 (very good). Possible standardization of MoS is known from Recommendation ITU-T P.800.1 "Mean Opinion Score (MOS) Terminology, 2003." Reference and alternative QoS profiles are selected by the AF to maximize the application client's QoE. In the case of video streaming, the video client sends MoS ranking feedback to the server based on the quality of the received video stream, described in terms of the Peak Signal-to-Noise Ratio (PSNR), calculated as the difference between the received video and the original video.

[0060] 4 shows an example of how QoE may be calculated for a video streaming application. An original video is encoded by a video encoder (402) and transmitted over a network (404) to a video decoder (406). The decoded video is compared to the reference original video by a full-reference QoE monitoring system (408). A QoE metric, such as the MoS described above, may then be determined as a function of the results of the comparison.

[0061] Figure 5 shows a table (502) that can be used to map MoS with PSNR. The QoE-QoS relationship is an important tool for optimizing QoE at the application layer. The AF can calculate statistics of MoS received from application clients and construct a QoE-QoS function to select a QoS profile requested from the network that maximizes the application's QoE. However, this learning relies only on the application's MoS and QoS, which can cause problems when multiple applications share resources in a private 5G system.

[0062] In the context of multiple application deployment in a factory, the selection of reference and alternative QoS parameters may affect the QoE of other applications deployed in the network. For example, if the requested QoS is too high, the 5G core network may reject the AF's request because this request may affect the QoS that can be provided to other AFs deployed in the network. On the other hand, if the requested QoS is too low, there is no guarantee that the requested application QoS is sufficient to optimize the QoE. That is, the PSNR may be low, resulting in a MoS ranging from "not good" (rating 3) to "poor" (rating 1). If the requested QoS is too high, the 5G core network may reject the AF's request because this request may affect the QoE that can be provided to other AFs deployed in the network.

[0063] According to this disclosure, a technique for improving QoE in a network is proposed, specifically addressing the QoE degradation caused by high QoS demands from applications in a heavily loaded network. By learning the QoE sensitivity to specific QoS degradation parameters of each deployed application, the network can effectively organize network resources to optimize QoE. The proposed technique utilizes a simple and low-complexity learning model based on an exponential degradation model to approximate the QoE-QoS relationship.

[0064] Learning the sensitivity of QoE to QoS degradation is,e.g. collecting QoE feedback, such as Mean Opinion Score (MoS), from deployed application clients to assess application satisfaction; observing the variations of QoS parameters and identifying dominant QoS degradation parameters that significantly affect QoE; and learning the sensitivity of application satisfaction to QoS degradation by approximating the QoE-QoS relationship using an exponential parametric model.

[0065] Organizing network resources based on the learned QoE sensitivities includes: orchestrating specific virtual resources, including network, storage, and processing resources, to optimize the QoE of deployed applications; and / or and deriving system parameters for organizing radio resources and optimizing QoE for applications.

[0066] The advantages of the proposed technique lie in its ability to improve user satisfaction, improve resource allocation, employ low-complexity learning models, adapt to changing network conditions, improve efficiency, and offer broad applicability in the context of 5G deployment. Collectively, these advantages contribute to improved QoE and network performance in a variety of network environments and application scenarios.

[0067] The proposed technology fills gaps in current 5G technologies and contributes to the advancement of network optimization strategies.

[0068] FIG. 6 shows an example private 5G network deployment.

[0069] Figure 6 shares similar structure and elements as Figure 1. However, while Figure 1 focuses on illustrating a single application with a single UE, AF, and AS, Figure 6 illustrates the deployment of multiple applications over a private 5G network, which will include multiple application clients (606a, 606b, 606c), AFs (602a, 602b, 602c), and ASs.

[0070] The following description provides an overview of one exemplary implementation of the proposed technology by a network, more specifically by a functional entity of a 5G-CN comprising at least an interface and a processing unit (604). Figure 7 depicts one exemplary workflow for the operations performed by the processing unit (604).

[0071] The following discussion highlights two important aspects that are performed based on specific triggers and do not necessarily occur simultaneously or strictly periodically: (1) network learning of Quality of Experience (QoE) sensitivity to QoS degradation, and (2) network performance organization of network resources based on learned QoE sensitivity.

[0072] To initiate the learning process and orchestration, the network leverages various triggers, such as application QoS requests received from the Application Function (AF) to the 5G Core Network (5G-CN). The learning process will now be described in more detail.

[0073] The learning process requires, among other things, collecting QoE feedback from application clients deployed over the network. There are several possible implementations for collecting QoE feedback, as outlined below.

[0074] The QoE feedback from the application client may include, for example, a Mean Opinion Score (MoS). In the case of a video transmission, this may be an average assessment of the quality of the video transmission performed by the application client for a particular video transmission. The average assessment may be calculated over a certain number of frames set up by the application layer and / or the network.

[0075] The QoE feedback from the application client may also include the Peak Signal-to-Interference-and-Noise Ratio (PSNR). In the case of video transmission, this PSNR may correspond to a measure of the maximum distortion of a sequence of images measured at the client and received during video streaming.

[0076] The QoE feedback may include, for example, the worst-case MoS and / or the lowest PSNR between the downlink video streaming calculated at one or more application clients and the uplink video streaming calculated at one or more application servers.

[0077] The learning process further requires monitoring and observing the variations of the QoS parameters.

[0078] By analyzing the QoE feedback provided as input and the observed variations in the QoS parameters, the network can identify the QoS parameter that has the greatest impact on the QoE variations, hereafter referred to as the "QoS degradation parameter" 702. There are several possible implementations for calculating this parameter, as outlined below.

[0079] One approach involves measuring a QoS variable and calculating the gradient of the QoE with respect to the QoS variable. The QoS degradation parameter is determined by selecting the QoS variable that maximizes the QoE gradient.

[0080] Another approach utilizes principal component analysis (PCA) to determine QoS degradation parameters. By applying PCA to the data, the network can identify the QoS variables that are closest to the principal components that exhibit the greatest variance in QoE.

[0081] A classification-based approach is also feasible for obtaining QoS degradation parameters. In this approach, the network classifies QoS variables based on their impact on QoE variation. By analyzing the QoS classes, the network identifies the classes that cause the highest variation in QoE. This approach is particularly efficient when a large amount of application client feedback is available.

[0082] Regardless of the specific implementation, the calculation of QoS degradation parameters is performed separately for each deployed application, allowing the network to adapt its optimization efforts to the unique characteristics and requirements of each application, ultimately improving the overall QoE.

[0083] The application sensitivity learning module learns 704 the relationship between the QoE variations of a particular application and the corresponding QoS degradation parameters. This relationship, also called the "application sensitivity relationship," is specifically approximated using an exponential parametric model. Various possible techniques are suitable for performing this approximation.

[0084] For illustrative purposes, an exemplary iterative approach is described herein in which the QoS-QoE relationship for a given application is learned over a learning period T.

[0085] The time T may be time-divided into K elementary periods. For each elementary period k, and for each application, it is possible to learn an instantaneous sensitivity parameter that represents the sensitivity of the QoE of the given application to QoS degradation within that particular elementary period.

[0086] The calculation of the instantaneous sensitivity parameter may be performed, for example, using the following relationship, where the QoS degradation is denoted as Qd and the QoE parameter is denoted as Qe:

[0087]

number

[0088] Once the instantaneous sensitivity parameters have been determined, a long-term estimate of the sensitivity parameters over a learning period may be calculated. This calculation may include a smoothing parameter α, which may be transmitted by the application or selected by the network based on, for example, deployment conditions. The calculation of the long-term estimate of the sensitivity parameters may be performed, for example, using the following relationship:

[0089]

number

[0090] A sensitivity parameter β may be stored for each different application deployed in the network. A mapping indicating the correspondence between the application class indicator transmitted by the application in the application request and the sensitivity parameter β may also be stored.

[0091] The network performs network resource organization based on learned QoE sensitivity with respect to QoS degradation.

[0092] The network then performs network resource organization based on the learned QoE sensitivities (706).

[0093] Network resource organization by the network may involve organizing specific virtual resources (e.g., network, storage, processing resources) to optimize the QoE of deployed applications. A network hypervisor may, for example, create virtual machines dedicated to running the applications and map the virtual resources to the QoS degradation parameter (Qd) determined for each application as described above. By dynamically reorganizing resources based on the learned sensitivity parameter β of a given application that exhibits particularly high sensitivity, the network aims to specifically improve the QoE of this application and more generally improve the overall QoE of all deployed applications.

[0094] Network resource organization by the network may also include radio resource organization and system parameter derivation to optimize application QoE. Access and Mobility Management Function (AMF) and Session Management Function (SMF) nodes manage radio and session parameters, respectively. The Network Data Analysis Function (NWDAF) may learn sensitivity parameters and establish a mapping between radio / session QoS parameters and the QoS degradation parameter (Qd). This mapping enables the AMF-SMF to identify application PDU sessions and apply PDU session / radio parameter reconfiguration to maximize QoE based on the sensitivity parameter β learned for each application.

Claims

1. 1. A method for organizing resources of a wireless network, the method being performed by at least one processing unit (604), comprising: a) collecting a plurality of Quality of Service, QoS, parameters and Quality of Experience, QoE feedback values, each collected value being identified as being relevant to an application; b) observing the variations of said collected values ​​identified as being associated with the same given application; c) selecting (702) a QoS degradation parameter, the QoS parameter whose value has an observed variation of the given application that most correlates with the observed variation of the QoE feedback of the given application; d) adjusting (704) an exponential parametric model to approximate the relationship between the observed variation of the QoE feedback of the given application and the observed variation of the selected QoS parameters of the given application; e) organizing resources of the wireless network based on the adjusted exponential parametric model (706).

2. b), c), and d) are repeated for a plurality of applications deployed over the wireless network to obtain a plurality of adjusted exponential parametric models having a one-to-one relationship with the plurality of applications; e) includes organizing resources of the wireless network based on each of the adjusted exponential parametric models. The method of claim 1.

3. The method of claim 1 , wherein the wireless network is a non-public network.

4. The method of claim 1 , wherein the value of QoE feedback for an application comes from multiple client applications of the application.

5. The method of claim 1 , wherein the value of QoE feedback comprises a mean opinion score and / or a peak signal to interference and noise ratio.

6. The method of claim 1 , wherein e) comprises organizing virtual resources including network resources, storage resources, and / or processing resources.

7. The method of claim 1 , wherein e) comprises organizing radio resources.

8. c) is determining, for each collected QoS parameter, a slope of the collected value of QoE feedback relative to the collected value of the QoS parameter; and selecting the QoS parameter for which the highest gradient is determined as the QoS degradation parameter.

9. c) is running a principal component analysis scheme to determine principal components that exhibit the greatest variance for the collected quality of experience feedback; and selecting the QoS parameter whose collected value is closest to the principal component as the QoS degradation parameter.

10. c) is classifying said QoS parameters into classes with respect to said observed variations in QoE; and selecting as the QoS degradation parameter the class for which the greatest variation in the QoE feedback is observed.

11. 1. A system for organizing resources of a wireless network, the system comprising: a) collecting a plurality of Quality of Service (QoS) parameters and Quality of Experience (QoE) feedback values, each collected value being identified as relevant to an application; b) observing variations in said collected values ​​identified as being associated with the same given application; c) selecting (702) a QoS degradation parameter, the QoS parameter whose value has an observed variation of the given application that most correlates with the observed variation of the QoE feedback of the given application; d) adjusting (704) an exponential parametric model to approximate a relationship between the observed variation in the QoE feedback of the given application and the observed variation in the selected QoS parameters of the given application; e) at least one processing unit (604) configured to organize (706) resources of said wireless network based on said adjusted exponential parametric model.

12. A non-transitory storage medium storing computer program instructions which, when executed by a processing unit, cause the processing unit to perform the method of any one of claims 1 to 10.

Images