Policy based multi-rat spectrum sharing in o-ran networks using ai / ML techniques

Abstract

A Multi-Radio Access Technology (RAT) Spectrum Sharing Optimization (MR-SS- Optimization) module and an AIML-Training module are configured to collect, analyze, model and predict cell utilization, number of active UEs and other parameters to allocate spectrum sharing resources. Protocols across E2, E1, F1, 01 and other 0-RAN interfaces are enhanced to communicate various parameters.

Description

POLICY BASED MULTI-RAT SPECTRUM SHARING IN O-RAN NETWORKS USING AI / ML TECHNIQUES BACKGROUND OF THE DISCLOSURE

[0001] The present disclosure is related to Open Radio Access Network (O-RAN)wireless networks and relates more particularly to policy based multi-RAT spectrum sharing in O-RAN Networks. SUMMARY

[0002] Described are systems, methods, and computer program products for aMulti-Radio Access Technology (RAT) Spectrum Sharing Optimization (MR-SS- Optimization) module. In an implementation, a system comprises the Multi-Radio Access Technology (RAT) Spectrum Sharing Optimization (MR-SS-Optimization) module, and the system is configured to execute method MR-SS-Optimization.

[0003] In an implementation, a method for the MR-SS-Optimization comprises:collecting performance measures; and analyzing the performance measures to decide on spectrum sharing parameters for each RAT k in a cell g for a given frequency band for a given time interval. The method can further comprise: deciding a maximum number of resources that can be allocated to a RAT k in the cell g during the given time interval, wherein the maximum number of resources for the RAT k in the cell g is denoted as maxResourcesRAT(g,k; ) for time interval ; and deciding a minimum number of resources that can be allocated to the RAT k in the cell g during the given time interval, wherein the minimum amount of the resources for the RAT k for the cell g is denoted minResourcesRAT(g,k; ) for time interval , andwherein that (, ; ) ( , ; ) for each RAT k in the cell g.The method can also further comprise deciding the resources to be allocated for the RAT k in the cell g during the given time interval; and communicating the decision to a plurality of DUs for the RAT k, wherein allocated resources for the RAT k for the cell g for time interval is denoted as allocResourcesRAT(g,k; ).

[0004] In an implementation, the method can also further comprise computing cellutilization for the RAT k in cell g for time interval , denoted as cellUtilRAT(g, k; ), in a percentage as: (, ; ) = ( , ; )( , ; )100. The method can also comprise asresources g interval , usedResourcesRAT(g,k; ) and the weighted average of used resources for atime interval prior to for RAT k in cell g.; and computing weighted average cell utilization for RAT k in cell g, denoted as cellUtilRATwAvg(g,k; ), using the cell utilization for RAT k in cell g, cellUtilRAT(g,k; ), for time interval and weighted average of cell utilization for the time interval prior to .

[0005] In an implementation, the method can comprise collecting UE capabilitiesrelated statistics indicating supported RATs by each SA UE in the cell g from a CU-CP of corresponding base station for RAT k, BS(RAT k), to identify a number of RRC-Connected SA UEs which support a given set of the RATs in the cell g, denoted as numUEs (g, {List of supported RATs}, ) for SA UEs supporting a list of supported RATs, for a time interval in cell g.

[0006] In an implementation, the MR-SS-Optimization module is in operativecommunication with an MR-SS-Artificial Intelligence Machine Learning (MR-SS AIML) Training module, and the method further comprises: predicting, by the MR-SS-AIML- Training module, a cell utilization for each cell g for a next time interval , wherein a cell utilization for each cell g, cellUtil(g;T) input is received from a Distributed Unit (DU) or from a Centralized Unit – Control Plane (CU-CP) to the MR-SS-Optimization module for every time interval T, and the MR-SS-AIML-Training module uses the input to predict resource utilization for the next time interval . The MR-SS-AIML-Training module can be modeled as a Time Series and Long Short-Term Memory (LSTM) or Transformer based deep neural network model. The method can further comprise training an AIML model to predict future values of the number of active UEs supporting various combinations of RAT.

[0007] The MR-SS-Optimization module can be hosted at a near-real-time RadioIntelligent Controller (Near-RT-RIC) and the MR-SS-AIML-Training module can be hosted at a non-real-time RIC (Non-RT-RIC). The method can further comprise collecting, from a CU-CP at Near-RT RIC over the E2 interface, the performance measures for each cell periodically (time interval ); collecting, from a DU at Near-RT RIC over the E2 interface, the performance measures for each cell periodically (time interval ); communicating, by the MR-SS-Optimization module at the Near-RT-RIC, the spectrum sharing parameters to the MR-SS-AIML-Training module at the Non-RT-RIC; training and providing by the MR-SS- AIML module at the Non-RT-RIC , the AI / ML model to the MR-SS-Optimization module at the Non-RT-RIC; at the MR-SS-Optimization module at the Near-RT-RIC, using the AI / ML model to predict spectrum sharing parameter values for a next time interval ; at the MR- SS-Optimization module, using the predicted spectrum sharing parameter values to decide the resources to be allocated to each RAT k in cell g for the next time interval (allocResourcesRAT(g,k; ); and enhancement of E2 interface. The peformamce measures collected from the CU-CP at the Near-RT RIC over the E2 interface for each cell periodically (time interval ) comprise a number of connected users in a cell, (, ; ) for cell g, RAT k for time interval ; and a number of RRC-connected UEswith capabilities to support a different set of RAT combinations, numUEs(g, {List of supported RATs}; ) for cell g for time interval . The performance measures collected from the DU at the Near-RT RIC over the E2 interface for each cell periodically (timeinterval ) comprise: a number of active users in a cell, ( , ; ) , for each RAT k incell g for time interval ; a total number of resources that are usedRAT k in cell g for time interval (i.e. usedResourcesRAT(g,k; )); a total number of resources that are used in cell g for time interval (i.e. usedResources(g; )); and a total number of resources that were allocated for RAT k in cell g for time interval (allocResourcesRAT(g,k; )). The method can further comprise communicating, by the MR-SS-Optimization module at theNear-RT-RIC, cellUtil(g; ), usedResourcesRAT(g,k; ), ( , ; ), numUEs(g, {List ofsupported RATs}; ) and other relevant spectrum sharing parameters to the MR-SS-AIML- Training module at the Non-RT-RIC; and at the MR-SS-Optimization module at the Near- RT-RIC, using the AI / ML model to find the predicted values of the parameters( ; ), ( , { }; ),the next time interval .at a CU-CP and the MR-SS-AIML-Training module can be hosted at a CU-UP or another server. The method can be implemented in an EN-DC architecture comprising: a 4G LTE RAT k1 4G LTE and a 5G NR RAT k2; a 4G CU-CP (CU-CP(k1)); and a 4G vDU ( DU(k1); a 5G CU-CP (CU-CP(k2)); and a 5G vDU ( DU(k2); wherein the MR-SS-Optimization module is located at the DU(k1) and the MR-SS-AIML-Training module is located at an SMO. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] FIG. 1A is a block diagram of a system architecture.

[0010] FIG. 1B shows an example of a User Plane Stack.

[0011] FIG. 2 shows an example of a Control Plane Stack.

[0012] FIG. 3 shows an example of high-level NG-RAN including a gNB CU and DU.

[0013] FIG. 4 shows an example of a Separation of CU-CP (CU-Control Plane) and CU-UP (CU-User Plane) in a 5G gNB.

[0014] FIG. 5A shows an example of an O-RAN architecture.

[0015] FIG. 5B shows a logical flow example of an O-RAN architecture.

[0016] FIG. 6 illustrates an overall structure of an LSTM network.

[0017] FIG. 7A illustrates is an E-UTRAN architecture.

[0018] FIG. 7B illustrates a NG-RAN architecture.

[0019] FIG. 7C illustrates a Standalone SA architecture and logical flow.

[0020] FIG. 8 illustrates an EN-DC architecture.

[0021] FIG. 9 illustrates spectrum resources shared among three RATs.

[0022] FIG. 10 illustrates a Non-Standalone (NSA) architecture..

[0023] FIG. 11A illustrates an example of MR-SS-Optimization module computedweighted average cell utilization.

[0024] FIG. 11B illustrates an example of MR-SS-Optimization module computedweighted average cell utilization.

[0025] FIG. 12 illustrates a logical flow for an MR-SS-Optimization module and anMR-SS-AIML-Training module.

[0026] FIG. 13A illustrates a logical flow for an MR-SS-Optimization module and anMR-SS-AIML-Training module.

[0027] FIG. 13B illustrates time intervals for computing an exponentially weightedaverage value for a cell utilization.

[0028] FIG. 14 illustrates a logical flow for an MR-SS-Optimization module and anMR-SS-AIML-Training module.

[0029] FIG. 15 illustrates a logical flow for an MR-SS-Optimization module and a CU-UP.

[0030] FIG. 16 illustrates a logical flow for an MR-SS-Optimization module and anMR-SS-AIML-Training module.

[0031] FIG. 17 illustrates a logical flow for an MR-SS-Optimization module and anMR-SS-AIML-Training module.

[0032] FIG. 18 illustrates a logical flow for an MR-SS-Optimization module and anMR-SS-AIML-Training module.

[0033] FIG. 19 illustrates a logical flow for an MR-SS-Optimization module and anMR-SS-AIML-Training module.

[0034] FIG. 20 illustrates a logical flow for an MR-SS-Optimization module and anMR-SS-AIML-Training module. DETAILED DESCRIPTION OF THE DISCLOSURE

[0035] In the following section an overview of Next Generation Radio AccessNetwork (NG-RAN) architecture and 5G New Radio (NR) stacks is given.5G NR (New Radio) user and control plane functions with monolithic gNB (gNodeB) are shown in FIGS. 1A-1B and FIG.2. For the user plane (shown in FIG.1A, which is in accordance with 3GPP TS 38.300), PHY (physical), MAC (Medium Access Control), RLC (Radio Link Control), PDCP (Packet Data Convergence Protocol) and SDAP (Service Data Adaptation Protocol) sublayers originate in the UE 101 and are terminated in the gNB 102 on the network side.

[0036] As shown in FIG. 1B, which is a block diagram illustrating the user planeprotocols stacks for a PDU session, in accordance with 3GPP TS 23.501, PDU layer 9010 corresponds to the PDU carried between the UE 101 and the data network (DN) 9011 over the PDU session. As shown in FIG.1B, UE 101 is connected to the 5G access network (AN) 902, which AN 902 is in turn connected via an N3 interface to the Intermediate UPF (I-UPF) 903a portion of the UPF 903, which I-UPF 903a is in turn connected via anN9 interface to the PDU session anchor 903b portion of the UPF 903, and which PDU session anchor 903b is connected to the DN 9011. CU-UP of AN 902 is connected to UPF 903b via a Backhaul (BH) path. The PDU session can correspond to IPv4, IPv6, or both types of IP packets, when the PDU session is of type IPv4, IPv6 or IPv4v6, respectively. GTP-U shown in FIG.1B supports tunnelling user plane data over N3 and N9 interfaces and provides encapsulation of end user PDUs for N3 and N9 interfaces.

[0037] For the control plane, shown in FIG. 2, which is in accordance with 3GPP TS38.300, RRC (Radio Resource Control), PDCP, RLC, MAC and PHY sublayers originate in the UE 101 and are terminated in the gNB 102 on the network side, and NAS (Non-Access Stratum) originate in the UE 101 and is terminated in the AMF (Access Mobility Function) 103 on the network side.

[0038] NG-Radio Access Network (NG-RAN) architecture from 3GPP TS 38.401 isshown in FIGS.3-4. As shown in FIG.3, the NG-RAN 301 consists of a set of gNBs 302 connected to the 5GC 303 through the NG interface. Each gNB comprises gNB-CU 304 and one or more gNB-DU 305 (see FIG.3). As shown in FIG.4 (which illustrates separation of CU-CP (CU-Control Plane) and CU-UP (CU-User Plane)), E1 is the interface between gNB- CU-CP (CU-Control Plane) 304a and gNB-CU-UP (CU-User Plane) 304b, F1-C is the interface between gNB-CU-CP 304a and gNB-DU 305, and F1-U is the interface between gNB-CU-UP 304b and gNB-DU 305. As shown in FIG.4, gNB 302 can consist of a gNB-CU-CP 304a, multiple gNB-CU-UPs (or gNB-CU-UP instances) 304b and multiple gNB-DUs (or gNB-DU instances) 305. One gNB-DU 305 is connected to one gNB-CU-CP 304a, and gNB-CU-UP 304b is connected to one gNB-CU-CP 304a.

[0039] Open Radio Access Network (O-RAN) is based on disaggregated componentswhich are connected through open and standardized interfaces based on 3GPP NG-RAN. An overview of O-RAN with disaggregated RAN CU (Centralized Unit), DU (Distributed Unit), and RU (Radio Unit), near-real-time Radio Intelligent Controller (Near-RT RIC) and non- real-time RIC (Non-RT RIC) is illustrated in FIG.5A.

[0040] As shown in FIG. 5A, the CU (shown split as O-CU-CP 801a and O-CU-UP801b) and the DU (shown as O-DU 802) are connected using the F1 interface (with F1-C for control plane and F1-U for user plane traffic) over a mid-haul (MH) path. One DU can host multiple cells (e.g., one DU can host 24 cells) and each cell can support many users. For example, one cell can support 800 Radio Resource Control (RRC)-connected users and out of these 800, there can be 250 Active users (i.e., users that have data to send at a given point of time).

[0041] A cell site can comprise multiple sectors, and each sector can supportmultiple cells. For example, one site can comprise three sectors and each sector can support eight cells (with each cell being on a different frequency band in a given sector). One CU-CP (CU-Control Plane) can support multiple DUs and thus multiple cells. For example, a CU-CP can support 500 cells and around 100,000 User Equipments (UEs). Each UE can support multiple Data Radio Bearers (DRBs) and there can be multiple instances ofCU-UP (CU-User Plane) to serve these DRBs. For example, each UE can support 4 DRBs, and 400,000 DRBs (corresponding to 100,000 UEs) can be served by five CU-UP instances (and one CU-CP instance).

[0042] The DU can be located in a private data center, or it can be located at a cell-site. The CU can also be in a private data center or even hosted on a public cloud system. The DU and CU, which are typically located at different physical locations, can be tens of kilometers apart. The CU communicates with a 5G core system, which can also be hosted in the same public cloud system (or can be hosted by a different cloud provider). A RU (Radio Unit) (shown as O-RU 803 in FIG.8) is located at a cell-site and communicates with the DU via a front-haul (FH) interface.

[0043] The E2 nodes (CU and DU) are connected to the near-real-time RIC 132 usingthe E2 interface. The E2 interface is used to send data (e.g., user and / or cell KPMs) from the RAN, and deploy control actions and policies to the RAN at near-real-time RIC 132. The applications or services at the near-real-time RIC 132 that deploys the control actions and policies to the RAN are called xApps. During the E2 setup procedures, the E2 node advertises the metrics it can expose, and an xApp in the near-RT RIC can send a subscription message specifying key performance metrics which are of interest. The near- real-time RIC 132 is connected to the non-real-time RIC 133 (which is shown as part of Service Management and Orchestration (SMO) Framework 805 in FIG.5A) using the A1 interface. The applications that are hosted at non-RT-RIC are called rApps. Also shown in FIG.8 are eNB 806 (which is shown as being connected to the near-real-time RIC 132 and the SMO Framework 805) and O-Cloud 804 (which is shown as being connected to the SMO Framework 805).

[0044] As in FIG. 5B, E2 node (which is DU or CU) and Near-RT-RIC establish E2session using E2 SETUP REQUEST and E2 SETUP RESPONSE. Near-RT-RIC can subscribe to certain parameters from the E2 node (on behalf the xApp running at Near-RT-RIC) using the RIC SUBSCRIPTION REQUEST and E2 node acknowledges this message by sending RIC SUBSCRIPTION RESPONSE to the Near-RT-RIC. As part of this, xApp running at the Near- RT-RIC also provides the event triggers to E2 node, e.g. it can ask E2 node to REPORTsubscribed parameters periodically to the xApp or to REPORT these subscribed parameters based on certain events to the xApp. E2 node communicates subscribed parameters to Near-RT-RIC (and the xApp) using RIC INDICATION as shown in FIG.5b. After analyzing received parameters from the E2 nodes (and based on network operator policies), Near- RT-RIC can send RIC CONTROL REQUEST to take an action at the E2 node (e.g. influence mobility decision). E2 node acknowledges this message by sending RIC CONTROL ACKNOWLEDGE to Near-RT-RIC while E2 node takes action as asked by the Near-RT-RIC.

[0045] Near-RT-Architecture is specified in the O-RAN Alliance specification O-RANNear-RT-Architecture 6.0, O-RAN E2 Service Model (E2SM) KPM 5.0, O-RAN E2 Application Protocol (E2AP) 5.0 and O-RAN A1 Interface: Application Protocol 4.02.

[0046] In this section, basic structures and characteristics of neural networks will bediscussed. Recurrent neural networks (RNNs) are a family of neural networks that are suited for handling sequential data. RNNs use a hidden state associated with each time-step and the output at each time step is computed using the input and the previous hidden state. RNN architectures suffer from some limitations. First, RNNs fail to store information for a long period of time and thus don’t handle the situations well in which a reference to certain information stored quite a long time ago is required to predict the current output. Second, there is no fine control to select which part of the context needs to be carried forward and or can be “forgotten”. Also, in RNNs, gradients in early layers are computed as the product of terms from later layers. With this, gradients in early layers can grow exponentially large (e.g. if the terms in later layers are large enough) or can exponentially decrease (e.g. if terms in later layers are small).

[0047] Long Short-Term Memory (LSTM) networks, which are an extension ofrecurrent neural networks (RNNs), have been introduced to handle situations where RNNs do not work well. The basic difference between the architectures of RNNs and LSTMs is that the hidden layer of LSTM is a gated unit or a gated cell, which consists of four layers that interact with one another in a way to produce the output of that cell along with the cell state. The output and the cell state are then passed onto the next hidden layer. Unlike RNNs which have only a single neural net layer of tanh (hyperbolic tangent), LSTMs comprisethree logistic sigmoid gates and one tanh layer. It should be noted that output of sigmoid activation function is in the range (0 to 1) for any real value as input, while the output of tanh is in the range (-1 to 1).

[0048] Gates are provided in LSTM in order to limit the information that is passedthrough the cell, i.e., the gates determine which part of the information will be needed by the next cell and which part is to be discarded. The output is usually in the range of 0 to 1, where “0” means “reject all”, and “1” means “include all”. Information is retained by the cells, and the memory manipulations are done by the gates. Three gates are provided in LSTM: 1) forget gate; 2) input gate; and 3) output gate.

[0049] FIG. 6 illustrates the overall structure of an LSTM network 1201. The overallLSTM network can be represented by the following expressions: f= (W . (h , x ) + b )i = (W . (h , x ) + b ))c = tanh (W . (h , x ) + b )c = f c + i ch = o tanh (c )where, frepresents the forget gate;i represents the input gate;o represents the output gate;is the sigmoid activation function; tanh: hyperbolic tangent activation function; h: output of the previous LSTM block (at time stamp t-1);x : input at the current time stamp;W : weight matrix corresponding to the input gate;W : weight matrix corresponding to the forget gate;W : weight matrix corresponding to the output gate;W : weight matrix corresponding to the cell state;b : bias corresponding to the input gate;b : bias corresponding to the forget gate;b : bias corresponding to the output gate;b : bias corresponding to the cell state;c : represents candidate for cell state (memory) at time stamp t;c : cell state (memory) at time stamp t;h : output of the LSTM block (at time stamp t); and: element wise multiplication.

[0050] The information that is no longer useful in the cell state is removed with theforget gate 1201a as shown in FIG.6. Two inputs, xt (input at the particular time t) and ht-1 (previous cell output from time t-1), are fed to the gate and multiplied with weight matrices followed by the addition of bias. The resultant is passed through sigmoid activation function 1202a, which produces a binary output. For a particular cell state, i) if the output is 0, the piece of information is forgotten, and ii) if output is 1, the information is retained for future use. As noted above in connection with FIG.6, the forget gate is represented by the expression: f= (W . (h , x ) + b )

[0051] The addition of useful information to the cell state is done by the input gate1201b as shown in FIG.6. First, the information is regulated using the sigmoid function1202b, which filters the values to be remembered using inputs h and x . Then, a vector,, is created using the tanh function 1203a which is used to compute the new cell state (attime t), , as given below. The input gate and its associated functionality can berepresented by the following expressions: i= (W . (h , x ) + b )c = tanh (W . (h , x ) + b )c = f c + i c

[0052] The task of extracting useful information from the current cell state to bepresented as output is done by the output gate 1201d, which is also shown in FIG.6. First, a vector is generated by applying the tanh function 1203b on the cell state . Then, the information is regulated using the sigmoid function 1202c and filtered for the values to beremembered using inputs h and x . Lastly, the values of the vector and the filtered(regulated) values are multiplied to be sent as an output of this cell and input to the next cell. The output gate 1201d and its associated functionality can be represented by the following expressions: o= (W . (h , x ) + b )h = o tanh (c )

[0053] Multi-RAT Spectrum Sharing (MR-SS) is described in this section. With MR-SS, multiple radio access technologies (RATs) can share the same spectrum. For example, 4G (LTE), 5G NR and 6G can share same spectrum (i.e. same frequency band) and the amount of resources assigned to these RATs can be specified statically or can change dynamically. It is also referred to as Dynamic Spectrum Sharing (DSS) for the case where resources allocated to each RAT can change dynamically.

[0054] An E-UTRAN 141 architecture is illustrated in FIG. 7A. The E-UTRANcomprises eNBs 116, providing the E-UTRAN U-plane (PDCP / RLC / MAC / PHY) and control plane (RRC) protocol terminations towards the UE. The eNBs 116 are interconnected with each other by the X2 interface. The eNBs 116 are also connected by the S1 interface to the EPC (Evolved Packet Core), more specifically to the MME (Mobility Management Entity) by the S1-MME interface and to the Serving Gateway (S-GW) by the S1-U interface. The S1 interface supports a many-to-many relation between MMEs / Serving Gateways and eNBs 116.

[0055] E-UTRAN also supports MR-DC via E-UTRA-NR Dual Connectivity (EN-DC), inwhich a UE is connected to one eNB that acts as a MN and one en-gNB 106 that acts as a SN. An EN-DC architecture is illustrated in FIG.7A. The eNB 116 is connected to the EPC 140via the S1 interface and to the en-gNB 106 via the X2 interface. The en-gNB 106 might also be connected to the EPC 140 via the S1-U interface and other en-gNBs 106 via the X2-U interface. In EN-DC, an en-gNB 106 comprises gNB-CU and gNB-DU(s).

[0056] As shown in FIG. 7B, in the NG-RAN architecture, an NG-RAN node is either:a gNB 302, providing NR user plane and control plane protocol terminations towards the UE; or a ng-eNB 306, providing E-UTRA user plane and control plane protocol terminations towards the UE. (3GPP TS 38.30017.3.0.)

[0057] As shown in FIG. 7B, the gNBs and ng-eNBs are interconnected with eachother by the Xn interface. The gNBs and ng-eNBs are also connected by the NG interfaces to the 5GC, more specifically to the AMF (Access and Mobility Management Function) by the NG-C interface and to the UPF (User Plane Function) by the NG-U interface.

[0058] The gNB 302 and ng-eNB 306 host functions such as functions for RadioResource Management: Radio Bearer Control, Radio Admission Control, Connection Mobility Control, Dynamic allocation of resources to UEs in both uplink and downlink (scheduling), connection setup and release; session Management; QoS Flow management and mapping to data radio bearers; and Dual Connectivity.

[0059] In an example, control information (e.g., scheduling information) can beprovided for broadcast and / or multicast operation. The UE can monitor different bundle sizes for the control channel depending on the maximum number of repetitions.

[0060] FIG. 7C shows an example with 4G and 5G networks in standalone (SA) mode.4G SA UEs communicate with 4G eNB 116 and EPC 140 (4G Evolved Packet Core) while 5G SA UEs communicate with 5G gNB 302 and 5GC (5G Core) 303. In the 5G SA architectures, 5G gNB 302 communicates with 5G Core (and not with 4G Evolved Packet Core 140). Similarly, in the 4G architecture, 4G eNB 116 communicates with 4G Evolved Packet Core (EPC) 140 and not with 5G Core.

[0061] With this architecture, 5G SA UEs can get access to all the services offered by5G SA networks (such as network slicing etc.). N26 interface exists between MME (Mobility Management Entity) of EPC 140 and AMF (Access and Mobility Management Function) of 5GC and it is utilized for inter-RAT mobility between 4G RAT and 5G RAT in SA architecture.

[0062] With Multi-RAT Spectrum Sharing in the SA architectures, 4G SA UEs and 5GSA UEs share the same spectrum resources (e.g. with frequency band) as shown in FIG.7.

[0063] FIG. 8 shows an example of the EN-DC (E-UTRA NR Dual Connectivity)architecture where eNB 116 acts as an MN (Master Node), EPC 140 acts as CN (Core Network), and en-gNB 106 acts as a SN (Secondary Node). With EN-DC (or more generally with Multi-RAT Dual Connectivity architectures), a UE connects with the MN and CN, and can communicate with SN via MN for control plane. For user plane, it can get data via MN or SN or both.

[0064] E-UTRA-NR Dual Connectivity (EN-DC) is a dominant form of the Non-Standalone (NSA) architecture. It is generally referred to as NSA architecture as described hering, though there can be other types of NSA architectures too. In FIG.8, en-gNB 106 represents a gNB which can connect with EPC 140 and eNB 116. X2 interface connects en- gNB 106 and eNB 116 in the EN-DC architecture, and the en-gNB 106 provides 5G NR air- interface functionality for the UE. Note that X2 AP (X2 Application Protocol) running on X2- C is specified in 3GPP TS 36.423 version 18.1.0.

[0065] In FIG. 8, 4G eNB 116 acts as the Master Node (or Master eNB) and it alsodecides which S1-U bearers are handled by the radio of 4G LTE or 5G NR. Here, 4G eNB instructs MME (of EPC 140) which further informs S-GW (Serving Gateway, of EPC 140) whether to establish S1-U bearer towards 4G LTE or 5G NR base station. Based on certain policies (e.g.5G NR radio quality falling below a pre-defined threshold), S1-U bearer towards 5G NR can be split at 5G en-gNB 106 (with data being communicated over 5G NR and 4G LTE RATs as shown in FIG.8) or a path switch can be triggered where S1-U directly goes to LTE eNB.

[0066] 4G eNB 116 (MN) acts as anchor node for the NSA UE for control planesignaling is also referred to as Master-eNB 116 (MeNB).

[0067] In FIG. 8, NSA UE (e.g. UEs supporting 4G and 5G capabilities for EN-DCarchitectures), can communicate data via EPC 140, 5G en-gNB 106 and 4G eNB 116 (e.g. EPC 140 en-gNB 106 eNB 116 NSA UE for DL traffic) using 4G part of spectrum resources and via EPC and 5G gNB 106 (e.g. EPC 140 en-gNB 106 NSA UE)using 5G part of spectrum resources.

[0068] Note that some UEs can still communicate using 4G SA mode in the NSAarchitecture shown in FIG.8. Specifically, 4G standalone (SA) UEs communicate data via 4G eNB 116 and EPC 140 (e.g. EPC 140 eNB 116 4G SA UE for DL, and 4G SA UE eNB 116 EPC 140 for UL).

[0069] With Multi-RAT Spectrum Sharing in the NSA architectures, same frequencyband is used for 4G and 5G RATs as shown in FIG.8.

[0070] Exemplary advantages of the disclosure

[0071] Issues for SA Architecture and Multi-RAT Spectrum Sharing

[0072] FIG. 9 shows an example where spectrum resources, corresponding tofrequency band b, are shared among three RATs (denoted as RAT k1, RAT k2 and RAT k3). It considers standalone (SA) UEs and separate core networks for these RATs are used. For example, SA UE(RAT k1) supports RAT k1 and communicates with base station for RAT k1, BS(RAT k1) and with core network for RAT k1, denoted as CN(RAT k1). Each UE communicates with the base station and the core network corresponding to its RAT for data as well as control plane (i.e. signaling) traffic but frequency band b (which was initially planned for RAT 1) is shared among RAT 1, RAT 2 and RAT 3.

[0073] For FIG. 9, RAT k2 can denote a more advanced technology than RAT k1 andRAT k3 can denote a more advanced technology than RAT k2. For example, RAT k1 can denote 4G (LTE), RAT k2 can denote 5G NR and RAT k3 can denote 6G technology. As another example, RAT k2 can denote a more advanced technology than RAT k1 but RAT k2and RAT k3 can denote same technology. For example, RAT k1 can denote 4G LTE, RAT k2 can denote 5G NR and RAT k3 can also denote 5G NR. In this case, BS(RAT k2) and BS(RAT k3) can be from different vendors but BS(RAT k1), BS(RAT k2) and BS(RAT k3) can be sharing same frequency band in a cell.

[0074] BS(RAT k1), BS(RAT k2) and BS(RAT k3) can be from same or differentvendors. As described earlier, each BS consists of DU, CU (with CU-CP, CU-UP) and RU. For example, BS(RAT k2) where RAT k2 is 5G NR, consists of gNB-DU, gNB-CU-CP, gNB-CU-UP and RU. As another example, BS(RAT k1) where RAT k1 is 4G eNB 116, consists of eNB-DU, eNB-CU-CP, eNB-CU-UP and RU. Note that the RU used for Multi-RAT spectrum sharing can be common for these base stations.

[0075] One commercial SA UE can support modem with capabilities for more thanone RATs in SA mode though only one may be active at a time. For example, one SA commercial UE can be supporting capabilities for RAT k1, RAT k2 and RAT k3 or RAT k1 and k2) but usually only one of these are used (for communication with BS and CN) at a given point of time. Network can handoff such a SA UE from one RAT (e.g. RAT k2) to another RAT (e.g. RAT k3) based on certain policies. An interface exists between core networks (e.g. between CN of RAT 2 and CN of RAT 3) which helps with handover between two different RATs in the SA mode.

[0076] Maximum allowed resources (such as resource blocks or resource elements)for RAT k for cell g is denoted as maxResourcesRAT(g,k;T) for time interval T. For example, maximum of 25% resources can be allowed in a cell for RAT k1, maximum of 35% for RAT k2 and maximum of 30% for RAT k3 (and remaining can be for overhead related to Multi- RAT spectrum sharing) over a time interval T.

[0077] These maximum allowed resources can be assigned to different RATs in astatic way (e.g. via configuration) or can be changed in semi-static way (e.g. over interval lengths such as few hundred ms) or can change dynamically (e.g. every 50 ms or 20 ms or 5 ms or 1 ms). Note that sum of resources for these RATs corresponding to frequency band bin cell g can be less than the total resources offered by band b due to some overhead associated with Multi-RAT Spectrum Sharing.

[0078] For a given time interval T, if spectrum resources allocated to RAT (e.g. forRAT k2) are less than what UEs in that cell need from this RAT k2, this cell for RAT k2 can become overloaded and some of the SA UEs which want to use RAT k2 but have capabilities to support RAT k1 also, can be handed over to RAT k1 and thus may not be able to use features offered by the network supporting RAT k2. This can also lead to performance degradation for some UEs. Also, the new UEs which want to establish session via RAT k2 can not be allowed to do during this time interval and can be redirected to another RAT or their connection establishment process can be delayed. Thus, it is important to allocate right number of resources to each RAT (for every time interval) when multi-RAT spectrum sharing is being used.

[0079] Issues for NSA Architecture and Multi-RAT Spectrum Sharing:

[0080] FIG. 10 shows an example of a non-standalone (NSA) architecture with NSAUE supporting RAT k1, RAT k2 and RAT k3. In the example shown in FIG.10, DL data is sent from the core network to the Base Station (BS) for RAT k2 (i.e. BS(RAT k2)), where data splitting decision is taken and DL data is sent to NSA UE via RAT k1, RAT k2 and RAT k3. In this case, BS(RAT k1) is the anchor node and the control plane signaling between UE and core network (CN) is carried out using the NSA UE – BS(RAT k1) – CN(NSA) path. Here, CN(NSA) denotes the core network for RAT k1 which is enhanced to support NSA UEs which can support over-the-air communication using RAT k1, RAT k2 and RAT k3. Spectrum resources are shared among RAT k1, RAT k2 and RAT k3. NSA UE can also split UL data and send across all these three RATs (if needed). Base station for RAT k2, BS(RAT k2), can also split DL data and send across two RATs (e.g. RAT k1 and RAT k2) or even one RAT (e.g. RAT k3), instead of three RATs in the example above.

[0081] Taking spectrum sharing decisions for multi-RAT scenarios based on current(and past) information has limitations and it can lead to decisions where lot more resources are given to a RAT where the UEs for that RAT cannot really use those resourcesor less resources are given to a RAT where there are many UEs ready to use those resources and can’t get enough resources.

[0082] Exemplary advantages of disclosed system and methods

[0083] Consider the scenarios given in FIG. 9 and FIG. 10. Here frequency band b isinitially planned only for RAT k1. With spectrum sharing, other RATs (such as RAT k2 and RAT k3) are also allowed to use this frequency band b. As discussed above, it is done using SA (Standalone) mode in FIG.9 and using NSA (Non-standalone) mode in FIG.10.

[0084] Method IA

[0085] In an implimenation, a Multi-RAT Spectrum Sharing Optimization (MR-SS-Optimization) module 202 is used to decide the resource split (of frequency band b) among different RATs (e.g. for RATs k1, k2 and k3) for a cell over a given time interval. This time interval can be T sub-frames where each sub-frame is equal to 1 ms. For example, this time interval T can be 500 ms or 200 ms or 50 ms 10 ms or even 1 ms.

[0086] The MR-SS-Optimization module 202 collects, predicts and analyzes variousperformance measures (or state variables) to decide spectrum sharing related parameters for each RAT k in cell g (for a given frequency band) for a given time interval. These parameters are described below.

[0087] The MR-SS-Optimization module 202 decides the maximum number ofresources that can be allocated to RAT k in a cell during a given time interval. Maximum number of resources (such as resource blocks or resource elements) for RAT k in cell g is denoted as maxResourcesRAT(g,k; ) for time interval . This is used for internal computation at the MR-SS-Optimization module 202.

[0088] The MR-SS-Optimization module 202 decides the minimum number ofresources that can be allocated to RAT k in a cell during a given time interval. Minimumnumber of resources for RAT k for cell g is denoted as minResourcesRAT(g,k; T ) for timeinterval . Note that ( , ; ) ( , ; ) for RATk in cell g. This is

[0089] The MR-SS-Optimization module 202 decides the resources to be allocatedfor each RAT k in a cell during a given time interval and communicates to the DUs for each RAT k. This is shown as output of the MR-SS-Optimization module 202 in FIG.12 and FIG. 13A.

[0090] Allocated resources for RAT k for cell g is denoted as allocResourcesRAT(g,k;) for time interval . For example, 25 resources out of total of 100 resources (for a given frequency band) may be allowed in a cell for RAT k1, 35 resources for RAT k2 and 30 resources for RAT k3 (and remaining can be for overhead related to Multi-RAT spectrum sharing) over a time interval .

[0091] Note that ( , ; ) ( , ; )( , ; ) for RAT k in cell g for each time interval . Also, note that( , ; ) ( ; ), where totalResources(g; )the RATs) for a given frequency band during the time interval . Note that totalResources(g; ) denotes the total number of available resources for a given frequency band in cell g over the time interval which can be shared among different RATs. For example, it can be 100 resource blocks for 1 ms time interval or equal to 50 x 100 = 5000 resources for 50 ms time interval or 200 x 100 = 20000 resources for 200 ms time interval.

[0092] The MR-SS-Optimization module 202 collects (or computes) variousparameters for each cell g for each time interval such as for time intervals ,, … , when z = m 1. Value of ‘p’ is less than ‘m’. For example, ‘m’ can beinterval and ‘p’ can equal to 100000. These parameters are specified below:

[0093] Number of resources (such as resource blocks) that are used for each RAT kin cell g for time interval is denoted as usedResourcesRAT(g,k; ) and is communicatedfrom DU to the MR-SS-Optimization module 202.

[0094] The MR-SS-Optimization module 202 computes cell utilization for RAT k incell g for time interval , denoted as cellUtilRAT(g, k; ), in percentage as follows (whenallocResourcesRAT( , ; ) is greater than zero)( ) ( , ; ), ; =( , ; ) 100

[0095] Forallocatedresources for RAT k2 and 85% of allocated resources for RAT k3 can be utilized for a given time interval .

[0096] cellUtilRAT(g,k; ) can be computed in percentage (between 0 and 100) oras a fraction (between 0 and 1).

[0097] Note that ( , ; ) ( , ; )( , ; ) for each RAT k and for each time interval .

[0098] The MR-SS-Optimization module 202 computes weighted average value ofused resources for RAT k in cell g and it is denoted as usedResourcesRATwAvg(g,k; ). This is computed using exponentially weighted moving average and (, ; ) = ( , ; ) + (1 )(, ; ) is the weighted average value of used resources forcell k for the time interval ( , ), which is the time interval before over awindow of length W. Also, is the time interval just before the time interval .

[0099] Note that this weighted average can be computed over a recent time windowof W sub-frames. Alternatively, a higher value of W can be chosen during busy hours and a lower value can be chosen during lean hours (i.e. when traffic volume is below a pre- defined threshold).

[0100] The MR-SS-Optimization module 202 computes weighted average cellutilization, denoted as cellUtilRATwAvg(g,k; ), using the cell utilization for RAT k, cellUtilRAT(g,k; ), for time interval and weighted average of cell utilization for thetime interval prior to . This is computed using exponentially weighted moving averageand ( , ; ) = ( , ; ) + (1 )( , ; ). Here, is chosen between 0 and 1, and( , ; ) is the weighted average cell utilization for the time interval( , ), which is the time interval before for a window of length W. Note thatthis weighted average can be computed over a recent time window of W sub-frames. Also, is the time interval just before the time interval .

[0101] Two examples are given in FIG. 11A and 11B. Here,,( , ; ) ( , ; )( , ; ) for each RAT k and for each time interval and this alwaysholds good with the methods specified here.

[0102] For the example shown in FIG. 11A, usedResourcesRAT(g,k; ) happens tobe less than minResourcesRAT(g,k; ) and is much less than the allocResourcesRAT(g,k; ), and this is not desirable as it leads to wastage of resources. In FIG.11B, usedResourcesRAT(g,k; ) is closer to allocResourcesRAT(g,k; ) and this is a more desirable scenario. The methods given here work to select minResourcesRAT(g,k; ), maxResourcesRAT(g,k; ) and allocResourcesRAT(g,k; ) such that wastage of resources is minimized.

[0103] Total number of resources (such as resource blocks) that are used in cell gfor time interval is denoted as usedResources(g; ). This can be computed by summing usedResourcesRAT(g,k; ) for each RAT k which is sharing cell g for time interval .

[0104] As indicated earlier, the total number of resources in cell g during time isdenoted as totalResources(g; ). The MR-SS-Optimization module 202 computes cell utilization for cell g for time interval , denoted as cellUtil(g; ), in percentage as follows: (; ) = ( ; )100in percentage (between 0 and 100)or as a fraction (between 0 and 1).

[0106] The MR-SS-Optimization module 202 collects UE capabilities relatedstatistics indicating supported RATs by each SA UE in that cell g from CU-CP of corresponding base station BS(RAT k). With this, the MR-SS-Optimization module 202 can get to know the number of RRC-Connected SA UEs which support a given set of RATs in the cell g (as part of UE capabilities). This is indicated as numUEs(g, {List of supported RATs}, ) for SA UEs supporting the ‘List of supported RATs, for time interval in cell g. This is shown as input parameter to the MR-SS-Optimization module 202 in FIG.12.

[00107] For example, numUEs(g, {k1, k2}; ) denotes the number of SA UEssupporting RAT k1 and k2 during time interval T, and numUEs(g, {k1, k2, k3}, ) denotes the number of SA UEs supporting RAT k1, k2 and k3 in cell g during time interval .

[00108] Note that only one of these RATs may be active for a SA UE but some of theseUEs can have capabilities to support multiple RATs (though one is active at a time) and that is captured as part of this step. Also, note that the number of UEs with specific RAT capabilities (e.g. number of UEs with RAT k1 and k2) can change during the time interval and in that case, maximum value for number of UEs for each RAT combination during time interval is taken as the value of this parameter.

[00109] CU-CP of each RAT k communicates the number of RRC-connected users tothe MR-SS-Optimization module 202. This is denoted as ( , ; ) for cell g, RAT k fortime interval . This denotes the average number of RRC-connected users for cell g, RAT k for time interval . This is shown as input parameter to the MR-SS-Optimization module 202 in FIG.12.

[00110] DU of each RAT k communicates the number of active users for cell g, RAT kfor each time interval , to the MR-SS-Optimization module 202. This is denoted as (, ; ) and it denotes the average number of active users for cell g, RAT k, for timeis shown as input parameter to the MR-SS-Optimization module 202 in FIG. 12.

[00111] The MR-SS-Optimization module 202 interacts with an AI / ML training module,called MR-SS-AIML-Training module, to predict parameters which are described below.

[0112] The MR-SS-AIML-Training module 204 predicts cell utilization for each cell gfor the next time interval . Note that Cell utilization for each cell g, cellUtil(g;T), is communicated from DU (or from CU-CP) to the MR-SS-Optimization module 202 for every time interval T and this is used to predict resource utilization for the next time interval .

[0113] As shown in FIG. 12, the MR-SS-Optimization module 202 keepscommunicating these values to the MR-SS-AIML-Training module 204 to train the AI / ML model to predict future values of cell utilization. It is modeled as a Time Series and LSTM or Transformer based deep neural network models are used for this purpose.

[0114] ( ; ) denotes predicted values of cell utilization (or load) forcell g for the time interval where can be the next time interval after . This is communicated from the MR-SS-AIML-Training module 204 to the MR-SS-Optimization module 202 as shown in FIG.12. In an alternate architecture shown in FIG.13A, the AI / ML model is downloaded from the MR-SS-AIML-Training module 204 to the MR-SS-Optimization module 202 and ( ; ) is inferred at the MR-SS-Optimizationmodule 202 using this model.

[0115] The MR-SS-AIML-Training module 204 predicts the cell load for each RAT kin cell g during a future time interval . As shown in FIG.12, the MR-SS-Optimization module 202 keeps communicating usedResourcesRAT(g,k; ) for each RAT k in cell g for each time interval to the MR-SS-AIML-Training module 204 to train the AI / ML model to predict future values of cell load for that RAT k. It is modeled as a Time Series and LSTM or Transformer based deep neural network models are used for this purpose.

[0116] ( , ; ) denotes predicted values of cell load forRAT k in cell g for the time interval . This is communicated from the MR-SS-AIML- Training module 204 to the MR-SS-Optimization module 202 as shown in FIG.12. In an alternate architecture shown in FIG.13A, the AI / ML model is downloaded from the MR-SS- AIML-Training module 204 to the MR-SS-Optimization module 202 and (, ; ) is inferred at the MR-SS-Optimization module 202using this AI / ML model.

[0117] The MR-SS-AIML-Training module 204 predicts the number of UEs withcapabilities to support different types of RATs during a future time interval Tm. As specified earlier, number of (observed) UEs with different RAT capabilities is communicated from CU-CP to the MR-SS-Optimization module 202 and is denoted as numUEs(g, {List of supported RATs}; ) for cell g for time interval .

[0118] As shown in FIG. 12, the MR-SS-Optimization module 202 keepscommunicating these values to the MR-SS-AIML-Training module 204 to train AI / ML model to predict future values of number of UEs supporting various combinations of RATs (e.g. number of UEs supporting RAT k1 only, RAT k1 and k2, RAT k1, k2 and k3, RAT k1 and k3 etc.). This is modeled as a Time Series and LSTM or Transformer based deep neural network models are used to predict future values of these parameters. Contextual information (such as data for different types of UEs which are being sold in the market by UE vendors) if available is also used.

[0119] ( , { }; ) denotes the predictedvalues of number of UEs with different RAT capabilities for cell g for the time interval . This is communicated from the MR-SS-AIML-Training module 204 to the MR-SS- Optimization module 202 as shown in FIG.12. In an alternate architecture shown in FIG. 13A, the AI / ML model is downloaded from the MR-SS-AIML-Training module 204 to theMR-SS-Optimization module 202 and ( , { }; ) isinferred at the MR-SS-Optimization module 202 using this model.

[0120] The MR-SS-AIML-Training module 204 predicts the number of active users inthe cell for each RAT k for a future time interval . As discussed earlier, the number of active users for every time (for each cell and for each RAT) are communicated from DU to the MR-SS-Optimization module 202 which further communicates these to the MR-SS- AIML-Training module 204 and this is used to predict number of active users for future time intervals. LSTMs or Transformer based deep neural network models are used to predict the number of active users for each RAT k.

[0121] ( , ; ) denotes the number of predicted active users during thetime interval from the MR-SS-AIML-Training module 204 to theMR-SS- 202 as shown in FIG.12. In an alternate architecture shown in FIG.13A, the AI / ML model is downloaded from the MR-SS-AIML-Training module 204 tothe MR-SS-Optimization module 202 and ( , ; ) is inferred at the MR-SS-Optimization module 202 using

[0122] The MR-SS-Optimization module 202 uses the measured (or observed) orcomputed parameters listed below for some of the prior time intervals and predicted parameters for the next time interval , to find near-optimal values of allocResourcesRAT(g,k; ) for each RAT k in cell g for the time interval .

[0123] In this specific case, considers the time intervals before the time interval. If z= m -1, denotes the time interval just before the time interval . As shown inFIG. 13B, the time intervals , … , for a window of length W (with z = m-1) areused for computing exponentially weighted average such as for used resources for RAT k in cell g at .

[0124] Parameters considered by the MR-SS-Optimization module 202 to computeallocResourcesRAT(g,k; ) include the following (with z = m-1): -usedResourcesRAT(g,k; )o This is equal to usedResourcesRAT(g,k; ) for z=m-1- usedResourcesRATwAvg(g,k; )o As discussed earlier, this is computed over a window of length W (i.e.time intervals time intervals , … , )- cellUtilRAT(g, k; )- cellUtilRATwAvg(g, k; )- cellUtil(g; )- numUEs(g, {List of supported RATs}, )- ( , ; )- ( , ; )- ( ; )- ( , ; )- ( , { }; )- ( , ; )- allocResourcesRAT(g,k; )- minResourcesRAT(g,k; )- maxResourcesRAT(g,k; )- Contextual information (if available) and other relevant parameters

[0125] The MR-SS-Optimization module 202 uses various policies to decide suitablevalues of minResourcesRAT(g,k; ) and maxResourcesRAT(g,k; ) and eventually suitable value of allocResourcesRAT(g,k; ) for RAT k in cell g for time interval .

[0126] An example policy that the MR-SS-Optimization module 202 uses to allocateresources for each RAT k in cell g for the next time interval is described here.

[0127] In this policy, the MR-SS-Optimization module 202 sets the value ofminResourcesRAT(g,k; ) for each RAT k for (the next) time interval in proportion to the weighted average value of used resources for RAT k in cell g, usedResourcesRATwAvg(g,k; ) for time interval . Here, , is a time interval just before .

[0128] In this example policy, the MR-SS-Optimization module 202 computesminResourcesRAT(g,k; ) as ( , ; ) = ( )( , ; ). Here, ( ) is a factor which depends onthe length of time

[0129] If length of time interval is less than or equal to a pre-specified threshold,indicated as thresIntervalLen, this scaling factor ( ) is chosen to be equal to 1.

[0130] If length of time interval is greater than the pre-specified threshold,thresIntervalLen, the scaling factor ( ) is chosen to be equal to (1 – ( )) where valueof ( ) increases as length of the time interval increases but is always less than orequal to 1.

[0131] For example, if length of time interval is 1 ms and threshold, thresIntervalLenis 2 ms, ( ) is taken equal to 1 for the time interval .

[0132] As another example, if length of time interval is 20 ms and threshold,thresIntervalLen is 2 ms, ( ) can be taken as 0.6 and thus ( ) to be equal to 0.4. Iflength of time interval Tm is 40 ms and threshold, thresIntervalLen is 2 ms, ( ) can betaken as 0.7 and thus ( ) to be 0.3.

[0133] With above policy, minResourcesRAT(g,k; ) for each RAT k in cell g fortime interval Tm are assigned such that( , ; ) ( )( ; ). Here ( ) is a pre-defined scaling factor. Its value can also dependon the length of the time interval and other parameters.

[0134] The MR-SS-Optimization module 202 sets the value ofmaxResourcesRAT(g,k; ) for each RAT k (in cell g for time interval ) in proportion to another factor which is denoted as predictFactor(g,k; ).

[0135] The MR-SS-Optimization module 202 computes predictFactor(g,k; ) foreach RAT k for the time interval Tm using the above computed minResourcesRAT(g,k; )and other predicted values such as ( , ; ),( , { }; ) , ( ; ) andg for the timeinterval is computed as follows: (, ; )= ( , ; ) +( , ; ) + ( , ; ) +( , { }; )

[0137] is the weight assigned to the usedResourcesRATwAvg term, is theweight assigned to the term, is the weight assigned to the term and is the weight assigned to the term. Each of these weights (i.e. , , and ) can take value 0 or higher. In one such policy, can be set to 1 but it can be set to other values too.

[0138] ( , ; ) for RAT k in cell g for time interval , isdefined as (, ; ) =,( , ; ) = maximum { ( , ; ) ( , ; ) , 0}

[0140] ( , {setof supported RATs including RAT k is defined as (, { }; ) =max { ( , { };( , { }; )),0}

[0141] As there can be multiple combinations of RATs for above computation of, a RAT combination is used which gives maximum value of (, { }; ).

[0142] For example, consider a scenario where three RATs (k1, k2, and k3) aresharing a frequency band in a cell. When computing values of for RAT k2, this method considers the number of UEs which support RAT combinations (k1,k2,k3), (k1,k2),(k2,k3) and computes ( , { }; ) forthese three RAT combinations.

[0143] In one example policy, it uses the RAT combination (comprising k2 for theabove case) which maximizes the values of (, { }; ) for the purpose ofcomputation of ( , ; ). Thus, it computes( , { 2}; ) to be the maximum of( , { : 1, 2, 3}; ),( , { : 1, 2}; ),( , { : 2, 3}; )

[0144] In another example policy, it uses average instead of maximum whilecomputing ( , { }; ). Thus, itcomputes ( , { 2}; ) to be theaverage of ( , { : 1, 2, 3}; ),( , { : 1, 2}; ),( , { : 2, 3}; )

[0145] These weights (i.e. , and ) can have static value or their values canchange dynamically too. If it is observed that the error in predicting goes above a threshold for some consecutive time intervals, value of can be lowered for some such time intervals. Similarly, if it is observed that the error in predicting goes above a threshold for some consecutive time intervals,value of can be lowered for some such time intervals. If it is observed that the error in predicting for a specific RAT combination goes above a threshold for some consecutive time intervals, value of can be lowered for some such time intervals. As the error in predicted parameter goes below a pre-specified threshold, value of the corresponding weight (i.e. , and ) are increased again and slowly restored to their original pre-configured values.

[0146] allocResourcesRAT(g,k; ) is computed as the maximum ofminResourcesRAT(g,k; ) and ( ; ) ( , ; ), and is given as( , ; ) = maximum { ( ; )( , ; ), ( , ; )},

[0147] ( ; ) is chosen such that the following constraints hold good.

[0148] [Constraint I]: ( ; ) is chosen to be less than or equal to 1 for each RAT kfor each time interval Tm. Value of ( ; ) can be changed dynamically.

[0149] [Constraint II]: ( ; ) for each RAT k is chosen such that the sum ofallocResourcesRAT(g,k, ) for all RATs that are sharing spectrum in cell g, is less than or equal to total resourcesthat frequency band (i.e. totalResources(g; ) for the time interval ) minus the overhead due to the Multi-RAT spectrum sharing. (, ; ); )( )

[0150] [Constraint III]: ( ; ) is chosen such that the following also holds goodfor each RAT k for each time interval in cell g: (, ; ) ( , ; )( , ; )

[0151] Value of ( ; ) can be chosen based on various policies but it is done in away such that the above three constraints (i.e. Constraint I, II and III) on ( ; ) holdgood.

[0152] In one such policy, the MR-SS-Optimization module 202 chooses ( ; )randomly for each RAT k for each time interval as long as above constraints (i.e. Constraint I, II and III) are satisfied.

[0153] In another policy, consider a scenario where each UE in cell g supports eitherRAT k1 only, or both the RATs k1 and k2. In this scenario, if the weighted average measure of cell utilization for RAT k2 in cell g, i.e. cellUtilRATwAvg(g,k2; ), is above a (pre- specified) threshold and if there are active UEs which are k1 though some ofthese UEs also have capability to support RAT k2 also, the MR-SS-Optimization module 202chooses ( 2; ) to be higher than ( 1; ). This starts giving additional resources forRAT k2 and some of the existing UEs which support RATs ‘k1 and k2’ but are using RAT k1 and some new UEs which also support RAR k2, can be redirected to RAT k2.

[0154] As another example policy, consider a scenario where each UE supportseither RAT k1 only, or both the RATs ‘k1 and k2’ or the RATs ‘k1, k2 and k3’. In this scenario, if the weighted average measure of cell utilization for RAT k3 in cell g, i.e. cellUtilRATwAvg(g,k3; ), is above a (pre-specified) threshold and if there are active UEs which are usingor k2 though some of these UEs have the capability to supportRAT k3 also, the MR-SS-Optimization module 202 chooses ( 3; ) to be higher than( 1; ) and ( 2; ). This starts giving additional resources for RAT k3 and some ofthe existing UEs which support RATs ‘k1, k2 and k3’ but are using RAT k1 or k2, and some new UEs which also support RAR k3, can be redirected to RAT k2.

[0155] As another example policy, ( , ; ) is used for thecomputation of ( , ; ) also but with different weights as givenbelow:

[00156] ( , ; ) = ( , ; ) +( , ; ) + ( , ; ) +( , { }; )

[0157] Weights used here (i.e. w , , and ) take smaller valuescompared to the weights used for the computation of ( , ; ), i.e. ,, and .

[0158] Method IB

[0159] FIG. 14 shows the scenario where the MR-SS-Optimization module 202 ishosted as part of the Near-RT-RIC (as an xApp) and the MR-SS-AIML-Training module 204 is hosted at Non-RT-RIC.

[0160] As in FIG. 15, the MR-SS-Optimization module 202 at the Near-RT RICcollects following information from CU-CP over the E2 interface for each cell periodically (e.g. for every time interval ): Number of connected users in a cell, ( , ; ) for cell g, RAT k for timeinterval and this is communicated over the E2 interface. Number of RRC-connected UEs with capabilities to support different set of RAT combinations, numUEs(g, {List of supported RATs}; ) for cell g for time interval . E2 interface is enhanced to communicate this information. Other system KPIs

[0161] As in FIG. 16, the MR-SS-Optimization module 202 at the Near-RT RICcollects following information from the DU for each cell periodically (e.g. for every time interval T): Number of active users in a cell, ( , ; ) , for each RAT k in cell g for timeinterval Total number of resources that are used for RAT k in cell g for time interval (i.e. usedResourcesRAT(g,k; )) and this is communicated over the E2 interfaceTotal number of resources that are used in cell g for time interval (i.e. usedResources(g; )) Total number of resources that were allocated for RAT k in cell g for time interval (allocResourcesRAT(g,k; )) Other system KPIs (such as related to throughput, block error rate, latency etc.)

[0162] The MR-SS-Optimization module 202 at the Near-RT-RIC communicatescellUtil(g; ), usedResourcesRAT(g,k; ), ( , ; ), numUEs(g, {List of supportedRATs}; ) and other relevant parameters to the MR-SS-AIML-Training module 204 at the Non-RT-RIC.

[0163] The MR-SS-AIML module at the Non-RT-RIC uses LSTM or other suitablemodels and provides the AI / ML model to the MR-SS-Optimization module 202 at the Non- RT-RIC.

[0164] The MR-SS-Optimization module 202 at the Near-RT-RIC uses this AI / MLmodel to find the values of ( ; ),( , { }; ), ( , ; ) andnext time interval .earlier, the MR-SS-Optimization module 202 uses measured,computed and predicted parameters to decide about the resources which should be allocated to each RAT k in cell g for time interval (i.e. allocResourcesRAT(g,k; )). In FIG.14, this is decided at the Near-RT-RIC and communicated to DUs (for each RAT k) as shown in FIG.16.

[0166] Method IC

[0167] FIG. 17 shows the case when the MR-SS-Optimization module 202 is locatedat the CU-CP and the MR-SS-AIML-Training module 204 is located at the CU-UP (or some other entity such as an OAM server). In this scenario, there is a common CU-CP for eachRAT k for which spectrum sharing decision needs to be taken in a given cell g. There can be separate or common vDUs (virtual DUs) for each RAT k.

[0168] As in FIG. 17, DU for each RAT k communicates the following parameters tothe MR-SS-Optimization module 202 at the CU-CP using the F1 Application layer Protocol (F1AP) running over the F1-C interface between DU and CU-CP: -Number of active users in a cell, ( , ; ) , for each RAT k in cell g fortime interval -Total number of resources that are used for RAT k in cell g for time interval(i.e. usedResourcesRAT(g,k; )) -Total number of resources that were allocated for RAT k in cell g for timeinterval T (allocResourcesRAT(g,k; )) -Other system KPIs (such as related to throughput, block error rate, latencyetc.)

[0169] E1 Application layer Protocol (E1AP) running between CU-CP and CU-UP isenhanced to communicate the following parameters from the MR-SS-Optimization module 202 (running at CU-CP) to the MR-SS-AIML-Training module 204 (for the case when this AI / ML training module is running at the CU-UP): -Number of connected users in a cell, ( , ; ) for cell g, RAT k for timeinterval T -Number of RRC-connected UEs with capabilities to support different set ofRAT combinations, numUEs(g, {List of supported RATs}; ) for cell g for time interval -Cell load (or used resources) for RAT k in cell g for time interval T (i.e.usedResourcesRAT(g,k; )), -Total number of resources that were allocated for RAT k in cell g for timeinterval (allocResourcesRAT(g,k; )) -Other system KPIs

[0170] The trained AI / ML model is downloaded from the entity where the MR-SS-AIML-Training module 204 is located (e.g. CU-UP in FIG.17) to the MR-SS-Optimization module 202 at the CU-CP.

[0171] The MR-SS-Optimization module 202 (at the CU-CP) uses this AI / ML modelto find the values of ( ; ),( , { }; ), ( , ; ) and( , ; ) for the next time interval .

[0172] The MR-SS-Optimization module 202 at the CU-CP uses measured, computedand predicted parameters to decide about the resources which should be allocated to each RAT k in cell g for time interval (i.e. allocResourcesRAT(g,k; )). As shown in FIG.17, allocResourcesRAT(g,k; )) is communicated to DUs (for each RAT k).

[0173] FIG. 18 shows an alternative deployment model where the AIML trainingmodule (i.e. MR-SS-AIML-Training module 204) is located as part of the SMO (Service, Management and Orchestration) and the MR-SS-Optimization module 202 is located at the CU-CP. The MR-SS-AIML-Training module 204 downloads the trained model from the SMO to the CU-CP. Various parameters to train the AI / ML model are communicated from CU-CP to SMO over the O1 interface. The MR-SS-Optimization module 202 analyzes various parameters at the CU-CP, decides the resources to be allocated to each RAT k and communicates this to the corresponding DUs.

[0174] FIG. 19 shows a deployment model for two RATs (k1 and k2) where there aredifferent CU-CPs for each of these RATs. CU-CP(k1) is CU-CP for RAT k1 and CU-CP(k2) is CU-CP for RAT k2. Also, there are different vDU instances for each RAT and these are denoted by DU(k1) and DU(k2) in FIG.19. RAT k1 and k2 are sharing spectrum resources provided by a frequency band in cell g.

[0175] In this case, DU(k1) provides relevant parameters to CU-CP(k1) and DU(k2)provides relevant parameters to CU-CP(k2). The CU-CP(k2) communicates parameters received from DU(k2) to CU-CP(k1) and the inter-CU interface (e.g. X2 between 4G and 5G for EN-DC architectures) is enhanced for this purpose. The MR-SS-Optimization module 202 located at CU-CP(k1) communicates these parameters (for RAT k1 and k2) to the MR- SS-AIML-Training module 204 at the SMO.

[0176] As before, the trained AI / ML model is downloaded from SMO to the CU-CP(k1). The MR-SS-Optimization module 202 analyzes various parameters at the CU-CP, decides the resources to be allocated to each RAT k and communicates these to the corresponding DUs. CU-CP(k1) communicates allocResourcesRAT(g,k1; ) to DU(k1) via F1-C and allocResourcesRAT(g,k2; ) to CU-CP(k2) via inter-vCU interface (e.g. X2 for the EN-DC architectures). CU-CP(k2) communicates allocResourcesRAT(g,k2; ) to DU(k2) via F1-C.

[0177] Method ID

[0178] FIG. 20 shows an EN-DC scenario where RAT k1 is 4G LTE and RAT k2 is 5GNR. There are separate CU-CP and vDU instances for these. 4G is represented by DU(k1) and CU-CP(k1) while 5G uses DU(k2) and CU-CP(k2).

[0179] The MR-SS-Optimization module 202 is located at the DU(k1) and the MR-SS-AIML-Training module 204 is located at the SMO.

[0180] As in FIG. 20, DU for each RAT k communicates the following parameters tothe MR-SS-AIML-Training module 204 at the SMO using the O1 interface: -Number of active users in a cell, ( , ; ) , for each RAT k in cell g fortime interval -Cell load (or used resources) for RAT k in cell g for time interval (i.e.usedResourcesRAT(g,k; )), -Total number of resources that were allocated for RAT k in cell g for the timeinterval T (allocResourcesRAT(g,k; )) -Other system KPIs

[0181] Alternatively, DU(k2) can communicate above parameters (i.e. number ofactive users and cell load for RAT k in cell g during time interval ) to the MR-SS- Optimization module 202 at DU(k1) using the inter-DU interface (such as D2 interface which is defined as inter-vDU interface in the patent application with a preliminary serial number of 18 / 591,782 having a filing date of February 29, 2024. ) and DU(k1) can communicate these parameters for RAT k1 as well as RAT k2 to the MR-SS-AIML-Training module 204.

[0182] As shown in FIG. 20, CU-CP for each RAT k communicates the followingparameters to the MR-SS-AIML-Training module 204 at the SMO using the O1 interface: -Number of RRC-connected UEs with capabilities to support different set ofRAT combinations, numUEs(g, {List of supported RATs}; ) for cell g for time interval -Number of RRC-connected UEs with capabilities to support different set ofRAT combinations, numUEs(g, {List of supported RATs}; ) for cell g for time interval -Other system KPIs

[0183] Alternatively, CU(k2) can communicate above parameters (for RAT k2) toCU(k1) using the inter-CU interface (such as X2 interface for EN-DC architectures) and CU(k1) can communicate these parameters for RAT k1 as well as RAT k2 to the MR-SS- AIML-Training module 204.

[0184] As described earlier, the MR-SS-AIML-Training-Module trains the AI / MLmodel using LSTM, Transformer based DNN or some other techniques. This trained model is downloaded from SMO to the DU(k1) where the MR-SS-Optimization module 202 is also located.

[0185] The MR-SS-Optimization module 202 at DU(k1) uses this AI / ML model tofind the values of ( , { }; ), ( ; ),( , ; ) and ( , ; ) for each RAT k for the next timeinterval .

[0186] The MR-SS-Optimization module 202 at DU(k1) uses measured, computedand predicted parameters to decide about the resources which should be allocated to each RAT k in cell g for time interval (i.e. allocResourcesRAT(g,k; )). As shown in FIG.20, allocResourcesRAT(g,k2; ) is communicated to DU(k2) using the inter-DU interface (such as D2 interface) and allocResourcesRAT(g,k1; ) is used for DU(k1).

[0187] Reference is made to Third Generation Partnership Project (3GPP), O-RANAlliance and the Internet Engineering Task Force (IETF) and related standards bodies in accordance with embodiments of the present disclosure. The present disclosure employs abbreviations, terms and technology defined in accord with Third Generation Partnership Project (3GPP), O-RAN Alliance and / or Internet Engineering Task Force (IETF) technology standards and papers, including the following standards and definitions.3GPP, O-RAN and IETF technical specifications (TS), standards (including proposed standards), technical reports (TR), RFCs and other papers are incorporated by reference in their entirety hereby, define the related terms and architecture reference models that follow. 3GPP TS 23.501 V 18.1.02024-06-26 3GPP TS 38.300 V 18.1.04-03-2024 3GPP TS 38.401 V 18.1.02024-03-29 3GPP TS 36.423 V 18.1.02024-03-29 O-RAN Near-RT-Architecture 6.0, O-RAN.WG3.RICARCH-R003-v06.00, R003, June 2024 O-RAN E2 Service Model (E2SM) KPM 5.0, O-RAN.WG3.E2SM-KPM-R003-v05.00, R003, June 2024O-RAN E2 Application Protocol (E2AP) 5.0, O-RAN.WG3.E2AP-R003-v05.00, R003, February 2024 O-RAN A1 Interface: Application Protocol 4.02, O-RAN.WG2.A1AP v04.02, R003, June 2024

[0188] Abbreviations5GC: 5G Core Network 5G NR: 5G New Radio 5QI: 5G QoS Identifier ACK: Acknowledgement AI: Artificial Intelligence AI / ML (or AIML): Artificial Intelligence and Machine Learning AM: Acknowledged Mode APN: Access Point Name ARP: Allocation and Retention Priority BO: Buffer Occupancy BS: Base Station BSR: Buffer Status Report CNN: Convolution Neural Network CMS: Centralized Management System CN: Core Networks CP: Control Plane CSI: Channel State Information CU: Centralized Unit CU-CP: Centralized Unit – Control PlaneCU-UP: Centralized Unit – User Plane DC: Dual Connectivity DL: Downlink DDDS: DL Data Delivery Status DNN: Data Network Name DNN: Deep Neural Network DQN: Deep Q Network DRB: Data Radio Bearer DSS: Dynamic Spectrum Sharing DU: Distributed Unit eNB: evolved NodeB (4G LTE Base Station) en-gNB: EPC: Evolved Packet Core EN-DC: E-UTRA-NR Dual Connectivity E-UTRA: Evolved UMTS Terrestrial Radio Access EWMA: Exponentially Weighted Moving Average GBR: Guaranteed Bit Rate gNB: gNodeB (5G NR Base Station) GTP-U: GPRS Tunnelling Protocol – User Plane IP: Internet Protocol L1: Layer 1 L2: Layer 2 L3: Layer 3 L4S: Low Latency, Low Loss and Scalable ThroughputLC: Logical Channel LESS: Low Energy Scheduler Solution LSTM: Long Short-Term Memory LTE: Long Term Evolution MAC: Medium Access Control MDP: Markov Decision Process MeNB: Master eNodeB MIB: Master Information Block ML: Machine Learning MN: Master Node MR: Multi-RAT MR-DC: Multi-RAT Dual Connectivity MR-SS: Multi-RAT Spectrum Sharing NACK: Negative Acknowledgement NAS: Non-Access Stratum NG-RAN: Next Generation Radio Access Network NR: New Radio NR-U: New Radio – User Plane NSA: Non-Standalone NSI: Network Slice Instance NSSI: Network Slice Subnet Instance NWDAF: Network Data Analytics Function O-RAN: Open Radio Access Network OAM: Operations, Administration MaintenancePCI: Physical Cell Identity PDB: Packet Delay Budget PDCP: Packet Data Convergence Protocol PDU: Protocol Data Unit PER: Packet Error Rate PF: Proportional Fair PHY: Physical Layer PRB: Physical Resource Block QCI: QoS Class Identifier QFI: QoS Flow Identifier QoS : Quality of Service RAN: Radio Access Network RAT: Radio Access Technology RB: Resource Block RDI: Reflective QoS Flow to DRB Indication RL: Reinforcement Learning RLC: Radio Link Control RLC-AM: RLC Acknowledged Mode RLC-UM: RLC Unacknowledged Mode RNN: Recurrent Neural Networks RQI: Reflective QoS Indication RRC: Radio Resource Control RRM: Radio Resource Management RTP: Real-Time Transport ProtocolRTCP: Real-Time Transport Control Protocol RU: Radio Unit S1-U: S1 User Plane S1-C: S1 Control Plane SA: Standalone SCTP: Stream Control Transmission Protocol SD: Slice Differentiator SDAP: Service Data Adaptation Protocol SIB: System Information Block SLA: Service Level Agreement SN: Secondary Node S-NSSAI: Single Network Slice Selection Assistance SST: Slice / Service Type TB: Transport Block TCP: Transmission Control Protocol TEID: Tunnel Endpoint Identifier UE: User Equipment UP: User Plane UL: Uplink UM: Unacknowledged Mode UPF: User Plane Function vDU: Virtual DU X2-C: X2 Control planeX2-U: X2-User plane

Claims

Claims 1. A method for a system comprising a Multi-Radio Access Technology (RAT) Spectrum Sharing Optimization (MR-SS-Optimization) module, the method comprising: collecting performance measures; and analyzing the performance measures to decide on spectrum sharing parameters for each RAT k in a cell g for a given frequency band for a given time interval.

2. The method of claim 1, further comprising: deciding a maximum number of resources that can be allocated to a RAT k in the cell g during the given time interval, wherein the maximum number of resources for the RAT k in the cell g is denoted as maxResourcesRAT(g,k; ) for time interval ; and deciding a minimum number of resources that can be allocated to the RAT k in the cell g during the given time interval, wherein the minimum amount of the resources for the RAT k for the cell g is denoted minResourcesRAT(g,k; ) for time interval , and wherein that ( , ; ) ( , ; ) for eachRAT k in the cell g.

3. The method of claim 1, further comprising: deciding the resources to be allocated for the RAT k in the cell g during the given time interval; and communicating the decision to a plurality of DUs for the RAT k, wherein allocated resources for the RAT k for the cell g for time interval is denoted as allocResourcesRAT(g,k; ).

4. The method of claim 1, further comprising: computing cell utilization for the RAT k in cell g for time interval , denoted as cellUtilRAT(g, k; ), in a percentage as:( , ; ) = ( , ; )( , ; )100.

5. Thecomputing a weighted average value of used resources for the RAT k in cell g, denoted as usedResourcesRATwAvg(g,k; ) using the used resources for RAT k in cell gfor time interval , usedResourcesRAT(g,k; ) and the weighted average of used resourcesfor a time interval prior to for RAT k in cell g and computing weighted average cell utilization for RAT k in cell g, denoted as cellUtilRATwAvg(g,k; ), using the cell utilization for RAT k in cell g, cellUtilRAT(g,k; ), for time interval and weighted average of cell utilization for the time interval prior to .

6. The method of claim 1, further comprising: collecting UE capabilities related statistics indicating supported RATs by each SA UE in the cell g from a CU-CP of corresponding base station for RAT k, BS(RAT k), to identify a number of RRC-Connected SA UEs which support a given set of the RATs in the cell g, denoted as numUEs (g, {List of supported RATs}, ) for SA UEs supporting a list of supported RATs, for a time interval in cell g.

7. The method of claim 1, wherein the MR-SS-Optimization module is in operative communication with an MR-SS-Artificial Intelligence Machine Learning (MR-SS AIML) Training module, and the method further comprises: predicting, by the MR-SS-AIML-Training module, a cell utilization for each cell g for a next time interval , wherein a cell utilization for each cell g, cellUtil(g;T) input is received from a Distributed Unit (DU) or from a Centralized Unit – Control Plane (CU-CP) to the MR- SS-Optimization module for every time interval T, and the MR-SS-AIML-Training module uses the input to predict resource utilization for the next time interval .

8. The method of claim 7, wherein the MR-SS-AIML-Training module is modeled as a TimeSeries and Long Short-Term Memory (LSTM) or Transformer based deep neural networkmodel.

9. The method of claim 7, the MR-SS-AIML-Training module is modeled as a Time Seriesand Long Short-Term Memory (LSTM) or Transformer based deep neural network model..

10. The method of claim 7, further comprising training an AIML model to predict future values of the number of active UEs supporting various combinations of RAT.

11. The method of claim 7, wherein the MR-SS-Optimization module is hosted at a near- real-time Radio Intelligent Controller (Near-RT-RIC) and the MR-SS-AIML-Training module is hosted at a non-real-time RIC (Non-RT-RIC).

12. The method of claim 11, further comprising: collecting, from a CU-CP at Near-RT RIC over the E2 interface, the performance measures for each cell periodically (time interval ): collecting, from a DU at Near-RT RIC over the E2 interface, the performance measures for each cell periodically (time interval ); communicating, by the MR-SS-Optimization module at the Near-RT-RIC, the spectrum sharing parameters to the MR-SS-AIML-Training module at the Non-RT-RIC; training and providing by the MR-SS-AIML module at the Non-RT-RIC , the AI / ML model to the MR-SS-Optimization module at the Non-RT-RIC; at the MR-SS-Optimization module at the Near-RT-RIC, using the AI / ML model to predict spectrum sharing parameter values for a next time interval ; at the MR-SS-Optimization module, using the predicted spectrum sharing parameter values to decide the resources to be allocated to each RAT k in cell g for the next time interval (allocResourcesRAT(g,k; ); and enhancement of E2 interface.

13. The method of claim 12, further comprising:wherein the peformamce measures collected from the CU-CP at the Near-RT RIC over the E2 interface for each cell periodically (time interval ) comprise: anumber of connected users in a cell, ( , ; ) for cell g, RAT kfor time interval ; and a number of RRC-connected UEs with capabilities to support a different set of RAT combinations, numUEs(g, {List of supported RATs}; ) for cell g for time interval ; and wherein the performance measures collected from the DU at the Near-RT RIC over the E2 interface for each cell periodically (time interval ) comprise: anumber of active users in a cell, ( , ; ) , for each RAT k incell g for time interval ;a total number of resources that are used for RAT k in cell g for time interval (i.e. usedResourcesRAT(g,k; )); a total number of resources that are used in cell g for time interval (i.e. usedResources(g; )); and a total number of resources that were allocated for RAT k in cell g for time interval (allocResourcesRAT(g,k; )).

14. The method of claim 13, further comprising: communicating, by the MR-SS-Optimization module at the Near-RT-RIC, cellUtil(g; ), usedResourcesRAT(g,k; ), ( , ; ), numUEs(g, {List of supported RATs}; )and other relevant spectrum sharing parameters to the MR-SS-AIML-Training module at the Non-RT-RIC; and at the MR-SS-Optimization module at the Near-RT-RIC, using the AI / ML model tofind the predicted values of the parameters ( ; ),( , { }; ), ( , ; ) and( , ; ) for the next time interval .laim 7, wherein the MR-SS-Optimization module hosted at a CU-CP and the MR-SS-AIML-Training module is hosted at a CU-UP or another server. 16.The method of claim 7 wherein the method is implemented in an EN-DC architecture comprising: a 4G LTE RAT k1 4G LTE and a 5G NR RAT k2; a 4G CU-CP (CU-CP(k1)); and a 4G vDU ( DU(k1); a 5G CU-CP (CU-CP(k2)); and a 5G vDU ( DU(k2); wherein the MR-SS-Optimization module is located at the DU(k1) and the MR-SS- AIML-Training module is located at an SMO.

17. A system comprising a Multi-Radio Access Technology (RAT) Spectrum Sharing Optimization (MR-SS-Optimization) module, the system being configured to: collect performance measures; and analyze the performance measures to decide on spectrum sharing parameters for each RAT k in a cell g for a given frequency band for a given time interval.

18. The system of claim 17, the system being further configured to: decide a maximum number of resources that can be allocated to a RAT k in the cell g during the given time interval, wherein the maximum amount of resources for the RAT k in the cell g is denoted as maxResourcesRAT(g,k; ) for time interval ; and decide a minimum number of resources that can be allocated to the RAT k in the cell g during the given time interval, wherein the minimum amount of the resources for the RAT k for the cell g is denoted minResourcesRAT(g,k; ) for time interval , wherein that ( , ; ) ( , ; ) for eachRAT k in the cell g.

19. The system of claim 17, the system being further configured to: decide the resources to be allocated for the RAT k in the cell g during the given time interval; and communicating the decision to a plurality of DUs for the RAT k, wherein allocated resources for the RAT k for the cell g is denoted as allocResourcesRAT(g,k; ).

20. The system of claim 17, the system being further configured to: compute c cell utilization for the RAT k in cell g for time interval , denoted as cellUtilRAT(g, k; ), in a percentage as: () ( , ; ), ; = 10021. The system ofcompute a weighted average value of used resources for the RAT k in cell g, denoted as usedResourcesRATwAvg(g,k; ) using the cell utilization for the RAT k, cellUtilRAT(g,k; ), for time interval and the weighted average of cell utilization for a time interval prior to .

22. The system of claim 17, the system being further configured to: collect UE capabilities related statistics indicating supported RATs by each SA UE in the cell g from a CU-CP of corresponding BS(RAT k) to identify a number of RRC-Connected SA UEs which support a given set of the RATs in the cell g, denoted as numUEs (g, {List of supported RATs}, ) for SA UEs supporting a list of supported RATs, for a time interval in cell g.

23. The system of claim 17, wherein the MR-SS-Optimization module is in operative communication with an MR-SS-Artificial Intelligence Machine Learning (MR-SS AIML) Training module, the system being further configured: predict, by the MR-SS-AIML-Training module, a cell utilization for each cell g for anext time interval , wherein a cell utilization for each cell g, cellUtil(g;T) input is received from a Distributed Unit (DU) or from a Centralized Unit – Control Plane (CU-CP) to the MR- SS-Optimization module for every time interval T, and the MR-SS-AIML-Training module uses the input to predict resource utilization for the next time interval .

24. The system of claim 17, wherein the system is further configured to: wherein the MR-SS-AIML-Training module is modeled as a Time Series and LongShort-Term Memory (LSTM) or Transformer based deep neural network model.

25. The system of claim 17, the system being further configured to: predict the cell load for the RAT k in cell g during a future time interval . ; predict the number of UEs with capabilities to support different types of RATs during a future time interval Tm.

26. The system of claim 17, the system being further configured to: train an AIML model to predict future values of number of UEs supporting various combinations of RAT.

27. The system of claim 17, wherein the MR-SS-Optimization module is hosted at a near- real-time Radio Intelligent Controller (Near-RT-RIC) and the MR-SS-AIML-Training module is hosted at a and non-real-time RIC (Non-RT-RIC).

28. The system of claim 27, the system being further configured to: collect, from a CU-CP at Near-RT RIC over the E2 interface, the performance measures for each cell periodically (time interval ); collect, from a DU at Near-RT RIC over the E2 interface, the performance measures for each cell periodically (time interval ); communicate, by the MR-SS-Optimization module at the Near-RT-RIC, the spectrum sharing parameters to the MR-SS-AIML-Training module at the Non-RT-RIC; andtrain and provide by the MR-SS-AIML module at the Non-RT-RIC , the AI / ML model to the MR-SS-Optimization module at the Non-RT-RIC; at the MR-SS-Optimization module at the Near-RT-RIC, using the AI / ML model to predict spectrum sharing parameter values for a next time interval ; and at the MR-SS-Optimization module, using the predicted spectrum sharing parameter values to decide the resources to be allocated to each RAT k in cell g for the next time interval (allocResourcesRAT(g,k; ).

29. The system of claim 28, the system being further configured so that: the peformamce measures collected from the CU-CP at the Near-RT RIC over the E2 interface for each cell periodically (time interval ) comprise: anumber of connected users in a cell, ( , ; ) for cell g, RAT k fortime interval ; and a number of RRC-connected UEs with capabilities to support a different set of RAT combinations, numUEs(g, {List of supported RATs}; ) for cell g for time interval ; and the performance measures collected from the DU at the Near-RT RIC over the E2 interface for each cell periodically (time interval ) comprise: anumber of active users in a cell, ( , ; ) , for each RAT k in cell g fortime interval a total number of resources that are used for RAT k in cell g for time interval (i.e. usedResourcesRAT(g,k; )); a total number of resources that are used in cell g for time interval (i.e. usedResources(g; )); and a total number of resources that were allocated for RAT k in cell g for timeinterval (allocResourcesRAT(g,k; )).

29. The system of claim 28, the system being further configured to:communicate, by the MR-SS-Optimization module at the Near-RT-RIC, cellUtil(g; ),usedResourcesRAT(g,k; ), ( , ; ), numUEs(g, {List of supported RATs}; ) andother relevant spectrum sharing parameters to the MR-SS-AIML-Training module at the Non-RT-RIC; and at the MR-SS-Optimization module at the Near-RT-RIC, using the AI / ML model tofind the predicted values of ( ; ),( , { }; ), ( , ; ) andthe MR-SS-AIML-Training module is hosted at a CU-UP or another server. 31.The system of claim 23 wherein the system is implemented in an EN-DC architecture comprising: a 4G LTE RAT k1 4G LTE and a 5G NR RAT k2; a 4G CU-CP (CU-CP(k1)); and a 4G vDU ( DU(k1); and a 5G CU-CP (CU-CP(k2)); and a 5G vDU ( DU(k2); wherein the MR-SS-Optimization module is located at the DU(k1) and the MR-SS- AIML-Training module is located at an SMO.

Images