Technical Field

[0001] The present disclosure relates to network optimization and, more specifically, cooperative network optimization via selection of Physical Resource Block (PRB) blanking actions.

Background

[0002] Recent advances in machine learning have led to a number of opportunities for network optimization. One avenue for machine learned network optimization is mitigation of Passive Intermodulation (PIM).

[0003] Specifically, PIM mainly impacts Frequency Division Duplex (FDD) systems in which the wireless devices transmit and receive at the same time but on different frequency ranges. A common source of PIM is when two downlink (DL) carriers mix and produce unwanted products in neighboring UL bands. These products may arise from non-linearities in passive radio components, such as coaxial connections, cables, filters, antennas, and bolts, or due to other external reflections. When the PIM frequency components fall into a neighboring receive band, the overall receiver effective noise figure increases, which degrades UL Key Performance Indicators (KPIs) such as throughput, retainability, accessibility, etc.

[0004] Most conventional methods for PIM mitigation have focused on techniques that cancel PIM in the radio. More recently, baseband mitigation approaches such as PIM Avoidance Scheduling have been proposed. The main idea of baseband mitigation is to modify the DL scheduler such that the two downlink carriers causing PIM do not transmit at high load simultaneously as shown below. Figure 1 illustrates an example of baseband mitigation that avoids simultaneous transmission at high load. The example illustrated in Figure 1 can be referred to as PIM Avoidance Scheduling (PIM-AS). These conventional PIM-AS techniques reduce the PIM product power in the UL, and improve the UL KPIs, but at the cost of potential reduction in DL throughput.

[0005] Existing Radio Frequency (RF) solutions for PIM cancellation in the uplink receiver bands typically require the purchase and installation of additional hardware, which is costly. Furthermore, as network operators utilize more bands and higher FDD bandwidths in New Radio (NR), PIM is becoming a more frequent and more severe problem. As such cost effective, software- based solutions that can mitigate PIM are greatly needed. PIM cancellation hardware solutions typically have significant limitations in the maximum order and type of PIM products that can be cancelled, as well as convergence time and cancellation depth limits. Thus, even when PIM cancelation hardware resources are available, there is still a significant need for PIM mitigation solutions based on avoiding and modifying the co-scheduling of multiple carriers.

[0006] Existing PIM-AS solutions lack coordination across multiple cells and dynamic approaches/methods of determining the ideal DL PRBs to avoid scheduling on based on real-time UL and DL KPIs across multiple cells. Namely, existing PIM-AS approaches generally fail to consider efficient ways of incorporating feedback and real-time DL delay to load balance DL blanking. Additionally, existing solutions fail to offer methods to easily prioritize particular cells in terms of the UL PIM reduction desired and the DL degradation permitted. Finally, existing approaches generally lack methods to track the evolution of PIM intensity over time, and therefore do not adapt the DL blanking decisions over time based on the environment, deployment, and/or network changes. [0007] Figure 2 illustrates an example of the effects of blanking different parts of a DL carrier. Specifically, Figure 2 illustrates graphs of simulations of the power spectrum of two DL carriers and PIM distortion generated by mixing these carriers for two different DL blanking scenarios. In Figure 2, half of DL carrier 2 is blanked in both scenarios. However, in scenario 1, the lower half is blanked (solid line), while in scenario 2 the upper half is blanked (dashed line). The PIM power falling in carrier 2 UL spectrum in scenario 1 is approximately 14 dB lower than in scenario 2, averaging across the entire carrier frequency range. Real network scenarios and configurations are typically more complicated, and the problem of finding an optimum or sufficient DL blanking solution increases as the number of bands and carriers increases. In some configurations, the effect of blanking PRBs in the same bandwidth on the low side as the high side of a DL carrier is the same or very similar. Depending on the carrier configuration and PIM source, there may be a few blanking actions that provide the same required PIM level reduction, with the same total DL blanking bandwidth, but different PRBs or carriers selected for blanking. It should be noted that a large variation in the PIM power spectrum profiles exists, even when fixed total blanking bandwidth, DL traffic level, PIM source, and carrier configuration are fixed.

[0008] Figure 3 illustrates simulations of PIM spectra in victim UL channels for two 10 MHz carriers. Specifically, the PIM spectrum in which no blanking occurs is illustrated alongside a spectrum for 50 randomly chosen frequency domain PRB blanking sequences. This simple use case illustrates that choosing a random blanking action typically results in much higher PIM levels versus optimum, or near-optimum, blanking options for a given bandwidth of DL spectrum blanked. The number of possible blanking sequences increases as the number of carriers and bands increase, and the problem of finding an optimum or adequate blanking sequence increases accordingly. [0009] As such, variations in traffic, PIM sources, and on-the-fly carrier configurations present an additional challenge for PIM mitigation solutions based on DL spectrum blanking. The network performance cost of blanking DL PBRs also varies with blanking bandwidth and/or sequences, DL traffic level, PIM sources, and carrier configurations. This presents yet another challenge for PIM mitigation solutions based on DL spectrum blanking.

Summary

[0010] In some embodiments, a method performed by a network node for multi-cell cooperative network optimization is proposed. The method includes dynamically selecting, based at least in part on data indicative of network performance of one or more cells associated with two or more carriers, one or more Physical Resource Block (PRB) blanking actions for PRBs of the two or more carriers from a set of PRB blanking actions. The method includes scheduling, in accordance with the one or more PRB blanking actions, data transmission to the cells of at least one of the two or more carriers.

[0011] In some embodiments, dynamically selecting the one or more PRB blanking actions for the two or more carriers includes processing the data indicative of the network performance of the one or more cells associated with the two or more carriers with a machine-learned network optimization model to select the one or more PRB blanking actions for the two or more carriers from the set of PRB blanking actions. [0012] In some embodiments, prior to dynamically selecting the one or more PRB blanking actions for the PRBs of the two or more carriers from the set of PRB blanking actions, the method includes dividing, based at least in part on a configured granularity parameter, PRBs of each of the two or more carriers into a plurality of possible blanking groups. The method includes generating, based on the plurality of possible blanking groups, the set of PRB blanking actions, wherein the set of PRB blanking actions represents a set of all possible combinations between each of the plurality of possible blanking groups.

[0013] In some embodiments, generating the set of PRB blanking actions further includes removing one or more PRB blanking actions from the set of PRB blanking actions based at least in part on predicted intermodulation distortion (IMD) in at least one UL carrier of the two or more carriers.

[0014] In some embodiments, generating the set of PRB blanking actions further includes processing the data indicative of the network performance of the one or more cells with an action set reduction portion of the machine-learned network optimization model to remove one or more PRB blanking actions from the set of PRB blanking actions.

[0015] In some embodiments, the two or more carriers include a first carrier of the network node and a second carrier of a second network node. In some embodiments, scheduling the data transmission to the cells of the at least one of the two or more carriers includes scheduling, in accordance with the one or more PRB blanking actions, the data transmission to the cells of the first carrier of the network node, and providing, to the second network node, data indicative of the one or more PRB blanking actions. [0016] In some embodiments, dynamically selecting the one or more PRB blanking actions for the two or more carriers from a set of PRB blanking actions includes processing the data indicative of the network performance of the one or more cells associated with the two or more carriers with an action estimation portion of the machine-learned network optimization model to obtain a respective set of estimated action values for the set of PRB blanking actions. Dynamically selecting the one or more PRB blanking actions includes dynamically selecting, based at least in part on the set of estimated action values, the one or more PRB blanking actions for the two or more carriers from the set of PRB blanking actions.

[0017] In some embodiments, processing the data indicative of the network performance with the action estimation portion of the machine-learned network optimization model to obtain the respective set of estimated action values further includes processing the data indicative of the network performance of the one or more cells associated with the two or more carriers with a load estimation portion of the machine-learned network optimization model to apply a load balancing penalty to each estimated action value of the set of estimated action values.

[0018] In some embodiments, the method further includes obtaining updated performance data indicative of updated network performance of the one or more cells associated with the two or more carriers. The method further includes adjusting one or more parameters of at least a portion of the machine-learned network optimization model based at least in part on the updated performance data.

[0019] In some embodiments, the method further includes processing the updated performance data with the action set reduction portion of the machine-learned network optimization model to remove one or more additional PRB blanking actions from the set of PRB blanking actions.

[0020] In some embodiments, the one or more PRB blanking actions are for optimization of (a) link adaptation, (b) connection procedure, (c) Passive Intermodulation (PIM) avoidance, or (d) inter-cell interference coordination.

[0021] In some embodiments, the data indicative of the network performance of the one or more cells associated with the two or more carriers includes (a) uplink Key Performance Indicators (KPIs), (b) downlink KPIs, or (c) both (a) and (b). In some embodiments, the uplink KPIs include user satisfaction KPIs and wherein the downlink KPIs comprise user dissatisfaction KPIs, the user satisfaction and user dissatisfaction KPIs are associated with users located at a cell edge.

[0022] In some embodiments, the one or more cells comprise a Frequency Division Duplexing (FDD) cell, and the two or more carriers comprise an uplink carrier for the FDD cell on a first frequency and a downlink carrier for the FDD cell on a second frequency that is different than the first frequency.

[0023] In some embodiments, the one or more cells associated with the two or more carriers comprises two or more cells on two or more respective carriers.

[0024] In some embodiments, a network node is proposed for multi-cell cooperative network optimization is proposed. The network node is adapted to dynamically select, based at least in part on data indicative of network performance of one or more cells associated with two or more carriers, one or more PRB blanking actions for PRBs of the two or more carriers from a set of PRB blanking actions. The network node is adapted to schedule, in accordance with the one or more PRB blanking actions, data transmission to the cells of at least one of the two or more carriers.

[0025] In some embodiments, a network node is proposed for multi-cell cooperative network optimization is proposed. The network node includes one or more transmitters, one or more receivers, and processing circuitry. The processing circuitry is configured to cause the network node to dynamically select, based at least in part on data indicative of network performance of one or more cells associated with two or more carriers, one or more PRB blanking actions for PRBs of the two or more carriers from a set of PRB blanking actions. The processing circuitry is configured to cause the network node to schedule, in accordance with the one or more PRB blanking actions, data transmission to the cells of at least one of the two or more carriers.

Brief Description of the Drawings

[0026] The accompanying drawing figures incorporated in and forming a part of this specification illustrate several aspects of the disclosure, and together with the description serve to explain the principles of the disclosure.

[0027] Figure 1 illustrates an example of baseband mitigation that avoids simultaneous transmission at high load;

[0028] Figure 2 illustrates an example of the effects of blanking different parts of a Downlink (DL) carrier;

[0029] Figure 3 illustrates simulations of Passive Intermodulation (PIM) spectra in victim Uplink (UL) channels for two 10 MHz carriers;

[0030] Figure 4 illustrates one example of a cellular communications system according to some embodiments of the present disclosure;

[0031] Figure 5 is an overview block diagram for machine-learned network optimization according to some embodiments of the present disclosure

[0032] Figure 6 illustrates an overview data flow diagram for multi-cell cooperative network optimization using Physical Resource Block (PRB) blanking actions;

[0033] Figure 7 illustrates a flowchart for determination of a set of PRB blanking actions according to some embodiments of the present disclosure;

[0034] Figure 8 illustrates a flowchart for performing a value-based reduction of a set of PRB blanking actions according to some embodiments of the present disclosure; [0035] Figure 9 demonstrates the effect of the value-based action set reduction illustrated in Figure 8 on initial buffer build-up from value-based action set reduction according to some embodiments of the present disclosure;

[0036] Figure 10 illustrates a flowchart for applying a load balancing correction according to some embodiments of the present disclosure;

[0037] Figure 11 is a schematic block diagram of a radio access node according to some embodiments of the present disclosure;

[0038] Figure 12 is a schematic block diagram that illustrates a virtualized embodiment of the radio access node of Figure 11 according to some embodiments of the present disclosure;

[0039] Figure 13 is a schematic block diagram of the radio access node of Figure 11 according to some other embodiments of the present disclosure;

[0040] Figure 14 is a schematic block diagram of a User Equipment (UE) device according to some embodiments of the present disclosure; and

[0041] Figure 15 is a schematic block diagram of the UE of Figure 14 according to some other embodiments of the present disclosure.

Detailed Description

[0042] The embodiments set forth below represent information to enable those skilled in the art to practice the embodiments and illustrate the best mode of practicing the embodiments. Upon reading the following description in light of the accompanying drawing figures, those skilled in the art will understand the concepts of the disclosure and will recognize applications of these concepts not particularly addressed herein. It should be understood that these concepts and applications fall within the scope of the disclosure.

[0043] Radio Node: As used herein, a "radio node" is either a radio access node or a wireless communication device.

[0044] Radio Access Node: As used herein, a "radio access node" or "radio network node" or "radio access network node" or "RAN node" is any node in a Radio Access Network (RAN) of a cellular communications network that operates to wirelessly transmit and/or receive signals. Some examples of a radio access node include, but are not limited to, a base station (e.g., a New Radio (NR) base station (gNB) in a Third Generation Partnership Project (3GPP) Fifth Generation (5G) NR network or an enhanced or evolved Node B (eNB) in a 3GPP Long Term Evolution (LTE) network), a high-power or macro base station, a low-power base station (e.g., a micro base station, a pico base station, a home eNB, or the like), a relay node, a network node that implements part of the functionality of a base station (e.g., a gNB Distributed Unit (gNB-DU) or a gNB Central Unit (gNB-CU)) or a network node that implements part of the functionality of some other type of radio access node.

[0045] Core Network Node: As used herein, a "core network node" is any type of node in a core network or any node that implements a core network function. Some examples of a core network node include, e.g., a Mobility Management Entity (MME), a Packet Data Network Gateway (P-GW), a Service Capability Exposure Function (SCEF), a Home Subscriber Server (HSS), or the like. Some other examples of a core network node include a node implementing an Access and Mobility Function (AMF), a User Plane Function (UPF), a Session Management Function (SMF), an Authentication Server Function (AUSF), a Network Slice Selection Function (NSSF), a Network Exposure Function (NEF), a Network Function (NF) Repository Function (NRF), a Policy Control Function (PCF), a Unified Data Management (UDM), or the like.

[0046] Communication Device: As used herein, a "communication device" is any type of device that has access to an access network. Some examples of a communication device include, but are not limited to: mobile phone, smart phone, sensor device, meter, vehicle, household appliance, medical appliance, media player, camera, or any type of consumer electronic, for instance, but not limited to, a television, radio, lighting arrangement, tablet computer, laptop, or Personal Computer (PC). The communication device may be a portable, hand-held, computer-comprised, or vehiclemounted mobile device, enabled to communicate voice and/or data via a wireless or wireline connection.

[0047] Wireless Communication Device: One type of communication device is a wireless communication device, which may be any type of wireless device that has access to (i.e., is served by) a wireless network (e.g., a cellular network). Some examples of a wireless communication device include, but are not limited to: a User Equipment device (UE) in a 3GPP network, a Machine Type Communication (MTC) device, and an Internet of Things (loT) device. Such wireless communication devices may be, or may be integrated into, a mobile phone, smart phone, sensor device, meter, vehicle, household appliance, medical appliance, media player, camera, or any type of consumer electronic, for instance, but not limited to, a television, radio, lighting arrangement, tablet computer, laptop, or PC. The wireless communication device may be a portable, hand-held, computer-comprised, or vehicle-mounted mobile device, enabled to communicate voice and/or data via a wireless connection.

[0048] Network Node: As used herein, a "network node" is any node that is either part of the RAN or the core network of a cellular communications network/system.

[0049] Note that the description given herein focuses on a 3GPP cellular communications system and, as such, 3GPP terminology or terminology similar to 3GPP terminology is oftentimes used. However, the concepts disclosed herein are not limited to a 3GPP system.

[0050] Note that, in the description herein, reference may be made to the term "cell"; however, particularly with respect to 5G NR concepts, beams may be used instead of cells and, as such, it is important to note that the concepts described herein are equally applicable to both cells and beams.

[0051] Various embodiments of the present disclosure are primarily exemplified within the context of passive intermodulation (PIM) avoidance. However, it should be noted that these embodiments are not limited to PIM avoidance. Rather, the Physical Resource Block (PRB) blanking techniques of the present disclosure can be utilized in a variety of different network optimization scenarios (e.g., link adaptation, connection procedure, inter-cell interference coordination, etc.).

[0052] In the context of PIM avoidance, the present disclosure proposes a multi-cell adaptive blanking mechanism that selects PRBs to blank in the DL carriers to minimize the interference from PIM caused by DL carriers on the UL carriers. To do so, instead of always blanking a specific set of PRBs, the method is able to adapt to the carrier, load, and interference through feedback and/or a machine learning based approach. Specifically, the algorithm is able to learn the best blanking configuration by collecting feedback from the environment. The feedback, which is based on the reward of the RL algorithm, can be configured to prioritize UL or DL traffic or certain KPIs. Moreover, the proposed mechanism can configured to consider other parameters in the optimization, such as achieving delay fairness through load balancing. Finally, we also propose mechanisms to decrease the complexity of the algorithm and minimize the network impacts during action exploration. [0053] In the context of PIM avoidance using PRB blanking actions, this solution provides the following key advantages:

- Measuring KPIs / Feedback to adapt: The proposed solution selects a blanking configuration based on the information and feedback available from the environment. Therefore, it is able to adapt to changes in the UL and DL loads, the carrier configuration, PIM interference intensity, etc. For instance, if a new carrier is configured, the algorithm will be able to capture the effects of PIM and adjust its policy to blank the PRBs that cause the most interference.

- Improved PIM reduction by monitoring DL load/delay: the proposed solution has a load/delay monitoring portion that balances the DL blanking decision based on the load of the individual carrier. This enables the blanking to be more on carriers with lower load and reduce the uplink PIM interference more aggressively than without such a measure of carrier load.

- Load balancing: related to the above, the proposed solution has a load balancing portion that balances the blanking decisions among multiple carriers to enable a dynamic blanking decision across multiple carriers that will prioritize blanking carriers with lower load and adapt the decision in real-time.

- Intelligence: the proposed mechanism uses machine learning (ML) to learn the best blanking configuration for a given environment (e.g., load, carrier configuration). Methods to expedite the ML algorithm learning and to minimize the impact of the ML learning process (e.g., via action exploration) on the network are presented.

- Relative Prioritization: the proposed algorithm is flexible and can be configured with relative UL and DL KPIs priorities, and relative priorities across different carriers. For instance, when an aggressor cell impacts multiple victim UL carriers, a blanking policy that improves the PIM more on one victim carrier versus the other can be learned. This is particularly desirable for current services such as public safety that may be running on particular carriers, and future low latency 5G+ services like augmented reality, loT that may be running on particular carriers.

[0054] Figure 4 illustrates one example of a cellular communications system 400 in which embodiments of the present disclosure may be implemented. In the embodiments described herein, the cellular communications system 400 is a 5G system (5GS) including a Next Generation RAN (NG-RAN) and a 5G Core (5GC) or an Evolved Packet System (EPS) including an Evolved Universal Terrestrial RAN (E-UTRAN) and an Evolved Packet Core (EPC). In this example, the RAN includes base stations 402-1 and 402-2, which in the 5GS include NR base stations (gNBs) and optionally next generation eNBs (ng-eNBs) (e.g., LTE RAN nodes connected to the 5GC) and in the EPS include eNBs, controlling corresponding (macro) cells 404-1 and 404-2. The base stations 402- 1 and 402-2 are generally referred to herein collectively as base stations 402 and individually as base station 402. Likewise, the (macro) cells 404-1 and 404-2 are generally referred to herein collectively as (macro) cells 404 and individually as (macro) cell 404. The RAN may also include a number of low power nodes 406-1 through 406-4 controlling corresponding small cells 408-1 through 408-4. The low power nodes 406-1 through 406-4 can be small base stations (such as pico or femto base stations) or RRHs, or the like. Notably, while not illustrated, one or more of the small cells 408-1 through 408-4 may alternatively be provided by the base stations 402. The low power nodes 406-1 through 406-4 are generally referred to herein collectively as low power nodes 406 and individually as low power node 406. Likewise, the small cells 408-1 through 408-4 are generally referred to herein collectively as small cells 408 and individually as small cell 408. The cellular communications system 400 also includes a core network 410, which in the 5G System (5GS) is referred to as the 5GC. The base stations 402 (and optionally the low power nodes 406) are connected to the core network 410. [0055] The base stations 402 and the low power nodes 406 provide service to wireless communication devices 412-1 through 412-5 in the corresponding cells 404 and 408. The wireless communication devices 412-1 through 412-5 are generally referred to herein collectively as wireless communication devices 412 and individually as wireless communication device 412. In the following description, the wireless communication devices 412 are oftentimes UEs, but the present disclosure is not limited thereto. [0056] Figure 5 is an overview block diagram for machine-learned network optimization according to some embodiments of the present disclosure. In some embodiments, the machine-learned network optimization model 504 processes performance data 502 to obtain network optimization action(s) 508 (e.g., PRB blanking actions, etc.). Specifically, the machine-learned network optimization model 504 processes the performance data 502 to select the network optimization action(s) 508 from a set of network optimization actions 103. In some embodiments, the performance data 502 can include UL KPIs 502A (e.g., user satisfaction KPIs, etc.) and/or DL KPIs 502B (e.g., user dissatisfaction KPIs, etc.).

[0057] Alternatively, in some embodiments, the action selector 506 of the network node 500 selects the selected network optimization action(s) 508. Specifically, in some embodiments, the machine-learned network optimization model processes the performance data 502 to obtain a set of estimated action values 505 for at least a portion of the set of network optimization actions 503. The action selector 506 of the network node 500 can select the selected network optimization action(s) 508 based on the estimated action values 505. Alternatively, in some embodiments, the action selector 506 selects the selected network optimization action(s) 508 based at least in part on the performance data 5O2.After selecting the network optimization action(s) 508, the network node 500 can schedule transmissions using the scheduler 510 on cells served by carriers 512 based on the selected network optimization action(s) 508. For example, the network node 500 may serve a first cell of the cells 512 with a first carrier. The network node 500 can schedule transmission(s) on the first cell of the cells 512 based on the selected network optimization action(s) 508. A second cell of the cells 512 may be served by a carrier of second node. The network node 500 can provide data indicative of the selected network optimization action(s) 508 to the second node (e.g., using the scheduler 510, etc.) so that the second node can schedule transmissions on the second cell of the cells 512.

[0058] The network node 500 can receive additional performance data indicative of network performance of the cells 512 after scheduling transmissions with the scheduler 510 based on the selected network optimization action(s) 508. Based on this data, the machine-learned network optimization model 504 can be trained by the trainer 514. For example, the trainer 514 may evaluate an objective function that evaluates the network optimization action(s) and the updated performance data. The trainer 514 may then adjust one or more parameters of the machine-learned network optimization model 504 based at least in part on the objective function. In such fashion, the machine-learned network optimization model 504 is trained to select the most optimal network optimization action(s) from the set of network optimization actions 503.

[0059] As an example, the trainer 514 may initialize a reward function as: where R_i is the reward for action i, R_(C_j ) is carrier Cj^th's reward, P i,CJ) is the percentage of blanked PRBs of action i in carrier C_j. This example can include 4 carriers described by the following vectors:

"DL_freq_vect": [1812.5, 1865.0, 2120.0, 2630.0],

"DL_BW_vect": [13.5, 18.0, 13.5, 13.5],

"UL_freq_vect": [1717.5, 1770.0, 1930.0, 2510.0], "UL_BW_vect": [13.5, 18.0, 13.5, 13.5].

Using the described reward function, initial rewards can be computed such that: (l,0,0,0)^R _C1 (0, 1,0,0) ^R _Cz (0,0, 1,0) ~^R _Cs (0,0,0, 1) —>R _C4 , and the reward of action (w, x, y, z) is estimated as w x y z

- Rc "i - Rc "i - Rc "i - w + x + y + z ¹ w + x + y + z ^z w + x + y + z ³ w + x + y + z Rc ⁴

[0060] It should be noted that, in some embodiments, the trainer 514 may be located or otherwise executed at a node separate from network node 500. For example, in the case of limited compute and storage resources in network node 500 (e.g., when serving as a gNB-DU, etc.), at least a portion of the operations illustrated in Figure 5 can be implemented in a distributed manner. In this context, to train the machine-learned network optimization model 504, the network node 500 would send the information about the traffic, chosen action, and collected reward to a central entity (e.g., a supervising node, etc.). A trainer 514 of this central entity would then use this information to train or update the machine-learned network optimization model 504, which would be then transferred to the associated gNB-DU for inference.

[0061] In some embodiments, the set of network optimization action(s) 503 may be reduced based on the updated performance data. As an example, the updated performance data may indicate that the network optimization action(s) 508 selected (e.g., by the machine-learned network optimization model 504, etc.) failed to increase network performance. In response, the selected action(s) 508, and/or action(s) similar to the selected action(s) 508, can be removed from the set of network optimization actions 503 (e.g., PRB blanking actions on PRBs of carriers different to those associated with action(s) 502, etc.). For example, the machine-learned network optimization model 504 may process the updated data to identify and remove the action(s) from the set of network optimization actions 503 (e.g., with an action set reduction portion of the model, etc.).

[0062] Figure 6 illustrates an overview data flow diagram for multi-cell cooperative network optimization using PRB blanking actions. Specifically, Figure 6 illustrates a multi-cell cooperative PIM Avoidance mechanism for network optimization that dynamically blanks specific carriers) to reduce/avoid intermodulation interference in the uplink (e.g., prevents Physical Downlink Shared Channel (PDSCH) transmission on specific downlink PRBs). The network optimization method illustrated in Figure 6 is capable of adapting the Downlink (DL) blanking based on feedback obtained from the network Key Performance Indicators (KPIs) and load state. Namely, this mechanism adapts the ideal blanking action in real-time, and therefore provides an improved blanking configuration that changes with network and PIM conditions. In some embodiments, in a multi-cell environment, the proposed solution uses machine learning to learn the DL blanking configuration that provides the best tradeoff between Uplink (UL) and DL KPIs in real time, and across the KPIs of multiple carriers. In summary, the illustrated method is able to improve and adapt the PDSCH blanking actions by using feedback from the network and environment, such as the load, carrier configuration, PIM level, etc.

[0063] It should be noted that figure 6 is performed by a network node, which may be, e.g., a RAN node such as a base station 402 or a network node (e.g., a gNB-DU or gNB-CU) that performs part of the functionality of the base station 402. Further, in some implementations, the steps of Figure 6 may be distributed across two or more network nodes.

[0064] Specifically, a network node 500 (e.g., base station 402 of Figure 4, network node 1100 of Figures 11 and 13, etc.)), in some embodiments, determines a set of actions (step 604) (e.g., PRB blanking actions). To determine the set of PRB blanking actions, the network node 500 divides PRBs of each of two or more carriers into a plurality of possible blanking groups (step 604A). The network node 500 then generates the set of PRB blanking actions based on the plurality of possible blanking groups. The set of PRB blanking actions represents a set of all possible combinations between each of the plurality of possible blanking groups.

[0065] In some embodiments, after determining the set of PRB blanking actions, the network node 500 may reduce the size of the set of actions (e.g., a value-based action set reduction, etc.). For example, after generating the step of PRB blanking actions at step 604B, the network node 500 may remove one or more PRB blanking actions from the set of PRB blanking actions based at least in part on predicted intermodulation distortion (IMD) in the two or more carriers (step 605A). For example, the network node 500 may employ statistical operation(s) / model(s) (e.g., interpolation, A/B testing, feedback control, model predictive control, etc.) to identify particular carriers of importance (e.g., a set of carriers that introduces more interference than others, etc.), and remove PRBs of carriers that are relatively less important.

[0066] Additionally, or alternatively, in some embodiments the network node 500 may process data indicative of network performance of one or more cells associated with two or more carriers with an action set reduction portion of a machine-learned network optimization model to remove the one or more PRB blanking actions from the set of PRB blanking actions (step 605B). For example, the network node 500 may process the data with the action set reduction portion to determine carriers of particular importance, etc. The machine-learned network optimization model will be discussed in greater detail with regards to Figure 5.

[0067] Additionally, or alternatively, in some implementations, the network node 500 may process the data indicative of the network performance of the one or more cells with an action estimation portion of a machine-learned network optimization model to remove the one or more PRB blanking actions from the set of PRB blanking actions. For example, the network node 500 may process the data with the action set reduction portion to obtain a respective set of estimated action values for the set of PRB blanking actions, and may remove PRB blanking actions from the set that have an associated estimated action value that is under a threshold value.

[0068] After determining the set of PRB blanking actions, the network node 500 dynamically selects one or more PRB blanking actions for PRBs of the two or more carriers from the set of PRB blanking actions based at least in part on the data indicative of network performance of the one or more cells associated with the two or more carriers (step 608). In some embodiments, the data includes uplink KPIs (e.g., user satisfaction KPIs, etc.) and/or downlink KPIs (e.g., user dissatisfaction KPIs, etc.). In some embodiments, the uplink KPIs and/or downlink KPIs are associated with users located at a cell edge.

[0069] In some embodiments, the two or more carriers include an uplink carrier on a first frequency and a downlink carrier on a second frequency different than the first frequency. In some embodiments, the one or more cells includes a cell served by the uplink carrier and the downlink carrier. Alternatively, in some embodiments, one or more cells associated with the two or more carriers includes two or more cells on two or more respective carriers.

[0070] In some embodiments, to dynamically select the action(s), the network node 500 can process the data indicative of the network performance of the one or more cells associated with the two or more carriers with the action estimation portion of the machine-learned network optimization model to obtain a respective set of estimated action values for the set of PRB blanking actions (step 608A). Alternatively, in some embodiments, rather than processing the data with the machine-learned network optimization model to obtain the estimated action values, the network node 500 may determine the respective set of estimated action values using statistical operations / models. For example, the network node 500 may determine, based at least in part on the data indicative of the network performance of the one or more cells associated with the two or more carriers, a respective set of estimated action values for the set of PRB blanking actions using one or more statistical operations. Then, the network node can dynamically select the one or more PRB blanking actions for the two or more carriers from a set of PRB blanking actions based at least in part on the set of estimated action values.

[0071] Next, in some embodiments, the network node 500 can dynamically select the one or more PRB blanking actions for the two or more carriers from the set of PRB blanking actions based at least in part on the set of estimated action values (step 608D).

[0072] In some embodiments, the prior to dynamically selecting the PRB blanking action(s) at step 608D, the network node 500 considers load balancing for the PRB blanking actions. Specifically, the network node 500 can process the data indicative of the network performance with a load estimation portion of the machine-learned network optimization model to apply a load balancing penalty to each estimated action value of the set of estimated action values. In such fashion, load balancing can be taken into account when selecting PRB blanking action(s) at step 608D. Additionally, or alternatively, in some embodiments the network node 500 may determine and apply the load balancing penalties to the estimated action values based on the data indicative of the performance (e.g., using one or more statistical operations / models, etc.).

[0073] Specifically, in some embodiments, the load balancing can be taken into account while selecting the PRB blanking actions. As an example, the network node 500 may select the PRB blanking actions such that, at each time t, the network node 500 selects the action A _t that maximizes:

Value(A t) = E[R _UL (Aj, t)] + a x R _DL (Aj,t) + B(A _i; t) where E[R _UL (A _h t)] is the expected UL reward for action A _ir R _DL (A t) is the expected DL reward of action A _if and is a factor that balances the load/blanking between the carriers. The factor that balances load/blanking between carriers and B(Ai, t) can be calculated such that:

[0074] In some embodiments, the expected UL reward E[R _UL A _b t)] may be or otherwise indicate a 1 - Z ' PIM, a Signal to Interference Plus Noise Ratio (SINR), and/or a I+N<threshold (e.g., a percentage of PRBs with interference plus noise below a threshold).

[0075] In some embodiments, the expected DL reward R _OL (A _; , t) of action A _t may depend on the buffer state, and can be more or less prioritized using a such that: therefore penalizing for DL traffic that is not in buffer and not transmitted.

[0076] After selecting the PRB blanking action(s), the network node 500 schedules data transmission to the cells of at least one of the two or more carriers (step 610) in accordance with the one or more PRB blanking actions. For example, the transmission can be scheduled such that the PRB(s) associated with the PRB blanking actions are blanked.

[0077] In some embodiments, the two or more carriers include a first carrier of the network node 500, and a second carrier of a second network node. To schedule the transmission, the network node 500 can schedule the data transmission to the cell(s) of the first carrier of the network node 500 in accordance with the one or more PRB blanking actions (step 610A). Next, the network node 500 can provide data indicative of the one or more PRB blanking actions to the second network node. In such fashion, the network node 500 can schedule its transmission(s) on its own respective cells, while also providing scheduling data / instructions to connected cells in a cooperative fashion, therefore further facilitating multi-cell cooperative network optimization.

[0078] In some embodiments, after scheduling the transmission and/or providing the action(s) at step 610, the network node 500 can collect feedback related to the PRB blanking action(s) and train / update the model based on the feedback (step 612). Specifically, in some implementations the network node 500 obtain updated performance data indicative of updated network performance of the one or more cells associated with the two or more carriers (step 612A). Next, the network node 500 can adjust one or more parameters of at least a portion of the machine-learned network optimization model based at least in part on the updated performance data (step 612B). In such fashion, the machine-learned network optimization model can be trained (e.g., via reinforcement learning, etc.) or otherwise updated to select the most optimal PRB blanking actions.

[0079] In some embodiments, the network node 500 may process the updated performance data with the action set reduction portion of the machine-learned network optimization model to remove one or more additional PRB blanking actions from the set of PRB blanking actions (step 614). For example, the action set reduction portion may identify, based on the updated performance data, that a certain type of PRB blanking actions (e.g., actions associated with a certain carrier, etc.) are relatively ineffective, and then remove these actions from the set of PRB blanking actions.

[0080] Figure 7 illustrates a flowchart for determination of a set of PRB blanking actions according to some embodiments of the present disclosure. Specifically, in one embodiment of the present disclosure, to determine the set of PRB blanking actions in step 604 of Figure 6, the network node 500 can read the DL carriers and the granularity of the blanking actions. For a given blanking action granularity, corresponding to the number of consecutive PRBs to blank, the PRBs in each carrier can be divided into a set of possible blanking PRB groups (step 700). Then, the set of PRB blanking actions can be generated as the set of all possible combinations between the PRB groups of the carriers (step 702). After generating the set of PRB blanking actions, the set is returned (step 704).

[0081] Figure 8 illustrates a flowchart for performing a value-based reduction of a set of PRB blanking actions according to some embodiments of the present disclosure. Specifically, in some embodiments, to perform the value-based action space reduction 605 of Figure 6, the network node 500 can estimate that a set of carriers in an N carrier configuration introduce more interference than others and limit or otherwise focus the exploration to this set of carriers, therefore reducing the set of PRB blanking actions. For example, as described in Figure 6, an initial set of actions is first determined (e.g., as described with regards to step 604 of Figure 5, etc.). Next, UL KPI are collected and extrapolated to all of the actions in the initial phase. Based on the UL KPI extrapolated to the actions, values can be determined for each action in the set of actions (step 800). The N actions with the highest values may be selected for inclusion in the reduced set of actions (step 802). This reduced set of actions is then returned (804). In other words, the reduction can, in some embodiments, be performed by initially performing a limited number of blanking actions across the different carriers, and thereafter employing statistical operations (e.g., interpolation, etc.) and/or learning mechanisms (e.g., an action set reduction portion of the model, etc.) to decrease the set of actions and focus on only actions will provide a good UL PIM reduction. Another benefit of such an approach is that it reduces gNb/eNb buffer build-up during the initial phases of the algorithm. Additionally, or alternatively, to decrease the complexity of the algorithm and minimize the compute and resources required, another embodiment of the present invention would only consider certain actions based domain expertise and mathematical computations of potential IMD hits in the UL from the DL carriers.

[0082] It should be noted that in some embodiments, the initial belief (e.g., the set of actions, etc.) can be updated as actions are iteratively selected and evaluated. More specifically, the initial N actions selected would change as the values associated with the actions are explored (e.g., as described with regards to step 614 of Figure 6).

[0083] Figure 9 demonstrates the effect of the value-based action set reduction illustrated in Figure 8 on initial buffer build-up from value-based action set reduction. Specifically, Figure 9A is a graph depicting the initial buffer build-up in TTIs across a number of iterations without reduction of the set of PRB blanking actions. Conversely, Figure 9B is a graph depicting the initial build-up in TTIs across a number of iterations with value-based reduction of the set of PRB blanking actions. As depicted, the valuebased reduction of the set of PRB blanking actions significantly reduces the initial buffer build-up, therefore leading to significant gains in efficiency.

[0084] Figure 10 illustrates a flowchart for applying a load balancing correction according to some embodiments of the present disclosure. Specifically, in some embodiments, to perform the load balancing correction 608C of Figure 6, the network node 500 can estimate UL and/or DL KPIs (e.g., from the data indicative of the performance, etc.) (step 1000). The network node 500 then applies an estimated load balancing penalty to the value (e.g., a value function) of each PRB blanking action (step 1002). After applying the estimated load balancing penalty, the network node 500 returns the updated values for the set of PRB blanking actions (1004). This correction can be a function of the real-time delay and is different for different carriers/cells. With such an approach UL PIM can be improved without incurring excessive delay in the DL carriers - and by load balancing this delay. In another implementation, the penalty factor may be simply used to discourage DL blanking on carriers that are deemed more important to the operator.

[0085] Figure 11 is a schematic block diagram of a radio access node 1100 according to some embodiments of the present disclosure. Optional features are represented by dashed boxes. The radio access node 1100 may be, for example, a base station 402 or 406 or a network node that implements all or part of the functionality of the base station 402 or gNB described herein. As illustrated, the radio access node 1100 includes a control system 1102 that includes one or more processors 1104 (e.g., Central Processing Units (CPUs), Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs), and/or the like), memory 1106, and a network interface 1108. The one or more processors 1104 are also referred to herein as processing circuitry. In addition, the radio access node 1100 may include one or more radio units 1110 that each includes one or more transmitters 1112 and one or more receivers 1114 coupled to one or more antennas 1116. The radio units 1110 may be referred to or be part of radio interface circuitry. In some embodiments, the radio unit(s) 1110 is external to the control system 1102 and connected to the control system 1102 via, e.g., a wired connection (e.g., an optical cable). However, in some other embodiments, the radio unit(s) 1110 and potentially the antenna(s) 1116 are integrated together with the control system 1102. The one or more processors 1104 operate to provide one or more functions of a radio access node 1100 as described herein. In some embodiments, the function(s) are implemented in software that is stored, e.g., in the memory 1106 and executed by the one or more processors 1104.

[0086] Figure 12 is a schematic block diagram that illustrates a virtualized embodiment of the radio access node 1100 according to some embodiments of the present disclosure. This discussion is equally applicable to other types of network nodes. Further, other types of network nodes may have similar virtualized architectures. Again, optional features are represented by dashed boxes.

[0087] As used herein, a "virtualized" radio access node is an implementation of the radio access node 1100 in which at least a portion of the functionality of the radio access node 1100 is implemented as a virtual component(s) (e.g., via a virtual machine(s) executing on a physical processing node(s) in a network(s)). As illustrated, in this example, the radio access node 1100 may include the control system 1102 and/or the one or more radio units 1110, as described above. The control system 1102 may be connected to the radio unit(s) 1110 via, for example, an optical cable or the like. The radio access node 1100 includes one or more processing nodes 1200 coupled to or included as part of a network(s) 1202. If present, the control system 1102 or the radio unit(s) are connected to the processing node(s) 1200 via the network 1202. Each processing node 1200 includes one or more processors 1204 (e.g., CPUs, ASICs, FPGAs, and/or the like), memory 1206, and a network interface 1208.

[0088] In this example, functions 1210 of the radio access node 1100 described herein are implemented at the one or more processing nodes 1200 or distributed across the one or more processing nodes 1200 and the control system 1102 and/or the radio unit(s) 1110 in any desired manner. In some particular embodiments, some or all of the functions 1210 of the radio access node 1100 described herein are implemented as virtual components executed by one or more virtual machines implemented in a virtual environment(s) hosted by the processing node(s) 1200. As will be appreciated by one of ordinary skill in the art, additional signaling or communication between the processing node(s) 1200 and the control system 1102 is used in order to carry out at least some of the desired functions 1210. Notably, in some embodiments, the control system 1102 may not be included, in which case the radio unit(s) 1110 communicate directly with the processing node(s) 1200 via an appropriate network interface(s). [0089] In some embodiments, a computer program including instructions which, when executed by at least one processor, causes the at least one processor to carry out the functionality of radio access node 1100 or a node (e.g., a processing node 1200) implementing one or more of the functions 1210 of the radio access node 1100 in a virtual environment according to any of the embodiments described herein is provided. In some embodiments, a carrier comprising the aforementioned computer program product is provided. The carrier is one of an electronic signal, an optical signal, a radio signal, or a computer readable storage medium (e.g., a non-transitory computer readable medium such as memory).

[0090] Figure 13 is a schematic block diagram of the radio access node 1100 according to some other embodiments of the present disclosure. The radio access node 1100 includes one or more modules 1300, each of which is implemented in software. The module(s) 1300 provide the functionality of the radio access node 1100 described herein. This discussion is equally applicable to the processing node 1200 of Figure 12 where the modules 1300 may be implemented at one of the processing nodes 1200 or distributed across multiple processing nodes 1200 and/or distributed across the processing node(s) 1200 and the control system 1102.

[0091] In some embodiments, the module(s) 1300 include a network optimization module 1302. The network optimization module 1302 can be, or otherwise include, a module that implements the network optimization methods of the present disclosure. Specifically, the network optimization module 1302 can include algorithm(s) that can be implemented in a distributed unit (gNB-DU) and optimize the network by controlling the blanking decision(s) of associated cells. For example, the

[0092] As a particular example of network optimization, the network optimization module 1302 can select PRBs to blank in the DL carriers of node(s) to minimize the interference from PIM in the UL carriers. A feedback mechanism can be used to learn the blanking that maximizes a predefined reward. This reward can include UL and/or DL KPIs according to the operators preference. Moreover, the reward can include both UL and DL KPIs and associate a weight to either to prioritize the corresponding traffic. [0093] Specifically, in some embodiments, the network optimization module 1302 can include an action set determination submodule 1302A to determine a set of actions (e.g., as described with regards to Figure 7. Generally, determination of the set of actions is performed at the initiation of the network optimization module 1302, and can be repeated when the configuration of the site is changed. The set of actions (e.g., set of PRB blanking actions) can refer to the different combinations of DL PRB transmission that are allowed to be considered as blanking alternatives to reduce PIM. This set of actions can be determined using the steps outlined in Figure 7.

[0094] In some embodiments, the network optimization module 1302 can include an action value evaluation submodule 1302B that evaluates the value of actions within the set of actions (e.g., as described with regards to Figure 8, etc.). Specifically, it is possible to decrease the complexity of the network optimization module 1302 by first reducing the set of PRB blanking actions. This is particularly useful to minimize the impact of exploration on the performance of the system. Thus, instead of following conventional learning mechanisms that would explore the whole set of PRB blanking actions, the submodule 1302B can be configured to predict the value of the actions and only explore in a subset of the set of PRB blanking actions.

[0095] In some embodiments, the network optimization module 1302 can include a load balancing correction submodule 1302C that applies load balancing for the PRB blanking actions based on real-time feedback of the DL KPI impacts, such as delay on the different carriers (e.g., as described with regards to Figure 10). This is achieved by adding a penalty to the value/reward of the different blanking actions.

[0096] In some embodiments, the network optimization module 1302 can include a PRB blanking action selection submodule 1302D that selects PRB blanking action(s). The PRB blanking actions can be selected as previously described (e.g., as described with regards to steps 608 and 610 of Figure 6, etc.).

[0097] Figure 14 is a schematic block diagram of a wireless communication device 1400 according to some embodiments of the present disclosure. As illustrated, the wireless communication device 1400 includes one or more processors 1402 (e.g., CPUs, ASICs, FPGAs, and/or the like), memory 1404, and one or more transceivers 1406 each including one or more transmitters 1408 and one or more receivers 1410 coupled to one or more antennas 1412. The transceiver(s) 1406 includes radio-front end circuitry connected to the antenna(s) 1412 that is configured to condition signals communicated between the antenna(s) 1412 and the processor(s) 1402, as will be appreciated by on of ordinary skill in the art. The processors 1402 are also referred to herein as processing circuitry. The transceivers 1406 are also referred to herein as radio circuitry. In some embodiments, the functionality of the wireless communication device 1400 described above may be fully or partially implemented in software that is, e.g., stored in the memory 1404 and executed by the processor(s) 1402. Note that the wireless communication device 1400 may include additional components not illustrated in Figure 14 such as, e.g., one or more user interface components (e.g., an input/output interface including a display, buttons, a touch screen, a microphone, a speaker(s), and/or the like and/or any other components for allowing input of information into the wireless communication device 1400 and/or allowing output of information from the wireless communication device 1400), a power supply (e.g., a battery and associated power circuitry), etc.

[0098] In some embodiments, a computer program including instructions which, when executed by at least one processor, causes the at least one processor to carry out the functionality of the wireless communication device 1400 according to any of the embodiments described herein is provided. In some embodiments, a carrier comprising the aforementioned computer program product is provided. The carrier is one of an electronic signal, an optical signal, a radio signal, or a computer readable storage medium (e.g., a non-transitory computer readable medium such as memory).

[0099] Figure 15 is a schematic block diagram of the wireless communication device 1400 according to some other embodiments of the present disclosure. The wireless communication device 1400 includes one or more modules 1500, each of which is implemented in software. The module(s) 1500 provide the functionality of the wireless communication device 1400 described herein.

[0100] In the preferred embodiments described herein, the machine-learned network optimization model is used. However, while the embodiments disclosed herein utilized a machine-learned network optimization model for network optimization operations, the present disclosure is not limited thereto. Other schemes for network optimization can be used. For example, to dynamically select PRB blanking actions, rather than processing data indicative of network performance with the action estimation portion of the machine-learned network optimization model, statistical operation(s) I can be used to determine the estimated values for the set of PRB blanking actions based on the data indicative of network performance. These estimated values can then be used to select the PRB blanking action to be applied.

[0101] Any appropriate steps, methods, features, functions, or benefits disclosed herein may be performed through one or more functional units or modules of one or more virtual apparatuses. Each virtual apparatus may comprise a number of these functional units. These functional units may be implemented via processing circuitry, which may include one or more microprocessor or microcontrollers, as well as other digital hardware, which may include Digital Signal Processors (DSPs), special-purpose digital logic, and the like. The processing circuitry may be configured to execute program code stored in memory, which may include one or several types of memory such as Read Only Memory (ROM), Random Access Memory (RAM), cache memory, flash memory devices, optical storage devices, etc. Program code stored in memory includes program instructions for executing one or more telecommunications and/or data communications protocols as well as instructions for carrying out one or more of the techniques described herein. In some implementations, the processing circuitry may be used to cause the respective functional unit to perform corresponding functions according one or more embodiments of the present disclosure.

[0102] While processes in the figures may show a particular order of operations performed by certain embodiments of the present disclosure, it should be understood that such order is exemplary (e.g., alternative embodiments may perform the operations in a different order, combine certain operations, overlap certain operations, etc.).

[0103] At least some of the following abbreviations may be used in this disclosure. If there is an inconsistency between abbreviations, preference should be given to how it is used above. If listed multiple times below, the first listing should be preferred over any subsequent listing(s).

• 3GPP Third Generation Partnership Project

• 5G Fifth Generation

• 5GC Fifth Generation Core

• 5GS Fifth Generation System

• AF Application Function

• AMF Access and Mobility Function

• AN Access Network

• AP Access Point

• ASIC Application Specific Integrated Circuit

• AUSF Authentication Server Function

• CPU Central Processing Unit • DCI Downlink Control Information

• DL Downlink

• DN Data Network

• DSP Digital Signal Processor

• eNB Enhanced or Evolved Node B

• EPS Evolved Packet System

• E-UTRA Evolved Universal Terrestrial Radio Access

• FDD Frequency Division Duplexing

• FPGA Field Programmable Gate Array

• gNB New Radio Base Station

• gNB-CU New Radio Base Station Central Unit

• gNB-DU New Radio Base Station Distributed Unit

• HSS Home Subscriber Server

• IMD Intermodulation Distortion

• loT Internet of Things

• IP Internet Protocol

• KPI Key Performance Indicator

• LTE Long Term Evolution

• MAC Medium Access Control

• MME Mobility Management Entity

• MTC Machine Type Communication

• NEF Network Exposure Function

• NF Network Function

• NR New Radio

• NRF Network Function Repository Function

• NSSF Network Slice Selection Function

• OTT Over-the-Top

• PC Personal Computer

• PCF Policy Control Function

• PDSCH Physical Downlink Shared Channel

• P-GW Packet Data Network Gateway

• PIM Passive Intermodulation

• PRB Physical Resource Block • PRS Positioning Reference Signal

• QoS Quality of Service

• RAM Random Access Memory

• RAN Radio Access Network

• ROM Read Only Memory

• RP Reception Point

• RRH Remote Radio Head

• RTT Round Trip Time

• SCEF Service Capability Exposure Function

• SMF Session Management Function

• TCI Transmission Configuration Indicator

• TP Transmission Point

• TRP Transmission/Reception Point

• UDM Unified Data Management

• UE User Equipment

• UL Uplink

• UPF User Plane Function

[0104] Those skilled in the art will recognize improvements and modifications to the embodiments of the present disclosure. All such improvements and modifications are considered within the scope of the concepts disclosed herein.