First network node and methods performed thereby for handling a set of nodes connected via a segmented front-haul

Abstract

A performed by a first network node (111) in a communications network (100) comprising a set of nodes (120) connected via a segmented front-haul. The first network node (111) obtains (601) information indicating conditions under which one or more devices (130) are to be served. The first network node (111) then determines (602) parameters of a configuration of at least two processing actions comprising: i) one or more subsets of nodes (121, 122, 123) to serve the devices (130), ii) selecting a respective aggregating node (141, 142, 143) for each subset, iii) determining, for packets to be exchanged between the devices (130) and a central node (112), a route, iv) updating the subsets of nodes (121, 122, 123), and v) determining precoding weights for the nodes in the subsets. The determining (602) is performed by jointly optimizing the more parameters. The first network node (111) then initiates (603) sequential execution of the actions applying the configuration.

Description

[0001]FIRST NETWORK NODE AND METHODS PERFORMED THEREBY FOR HANDLING A SET OF NODES CONNECTED VIA A SEGMENTED FRONT-HAUL TECHNICAL FIELD The present disclosure relates generally to a first network node, and methods performed thereby, for handling a set of nodes connected via a segmented front-haul. The present disclosure also relates generally to a computer program and computer-readable storage medium, having stored thereon the computer program to carry out this method. BACKGROUND Wireless devices within a wireless communications network may be e.g., User Equipments (UEs), stations (STAs), mobile terminals, wireless terminals, terminals, and / or Mobile Stations (MS). Wireless devices may be enabled to communicate wirelessly in a cellular communications network or wireless communication network, sometimes also referred to as a cellular radio system, cellular system, or cellular network. The communication may be performed e.g., between two wireless devices, between a wireless device and a regular telephone and / or between a wireless device and a server via a Radio Access Network (RAN) and possibly one or more core networks, comprised within the wireless communications network. Wireless devices may further be referred to as mobile telephones, cellular telephones, laptops, or tablets with wireless capability, just to mention some further examples. The wireless devices in the present context may be, for example, portable, pocket-storable, hand-held, computer-comprised, or vehicle-mounted mobile devices, enabled to communicate voice and / or data, via the RAN, with another entity, such as another terminal or a server. The wireless communications network covers a geographical area which may be divided into cell areas, each cell area being served by a network node, which may be an access node such as a radio network node, radio node or a base station, e.g., a Radio Base Station (RBS), which sometimes may be referred to as e.g., gNB, evolved Node B (“eNB”), “eNodeB”, “NodeB”, “B node”, Transmission Point, or BTS (Base Transceiver Station), depending on the technology and terminology used. The base stations may be of different classes such as e.g., Wide Area Base Stations, Medium Range Base Stations, Local Area Base Stations, Home Base Stations, pico base stations, etc…, based on transmission power and thereby also cell size. A cell may be understood as the geographical area where radio coverage is provided by the base station or radio node at a base station site, or radio node site, respectively. One base station, situated on the base station site, may serve one or several cells. Further, each base station may support one or several communication technologies. The base stations communicate over the air interface operating on radio frequencies with the terminals within range of the base stations. The wireless communications network may also be a non-cellular system, comprising network nodes which may serve receiving nodes, such as wireless devices, with serving beams. In 3rd Generation Partnership Project (3GPP) Long Term Evolution (LTE), base stations, which may be referred to as eNodeBs or even eNBs, may be directly connected to one or more core networks. In the context of this disclosure, the expression Downlink (DL) may be used for the transmission path from the base station to the wireless device. The expression Uplink (UL) may be used for the transmission path in the opposite direction i.e., from the wireless device to the base station. NR The standardization organization 3rd Generation Partnership Project (3GPP) is currently in the process of specifying a New Radio Interface called New Radio (NR) or 5G-Universal Terrestrial Radio Access (UTRA), as well as a Fifth Generation (5G) Packet Core Network, which may be referred to as Next Generation (NG) Core Network, abbreviated as NG-CN, NGC or 5G CN. In the current concept, gNodeB (gNB) denotes a 5G-NR BS, where one NR BS may correspond to one or more transmission and / or reception points. One of the main goals of NR is to provide more capacity for operators to serve ever increasing traffic demands and variety of applications. Because of this, NR may be able to operate on high frequencies, such as frequencies over 6 GHz, until 60 or even 100 GHz. Operation in higher frequencies makes it possible to use smaller antenna elements, which enables antenna arrays with many antenna elements. Such antenna arrays facilitate beamforming, where multiple antenna elements may be used to form narrow beams and thereby compensate for the challenging propagation properties. D-MIMO Distributed Multiple-Input-Multiple-Output (MIMO), or only D-MIMO, may be understood as a technology aimed at next-generation mobile systems, e.g., 5G advanced and Sixth Generation (6G), in which a geographical area may be covered by spreading out Access Points (APs), instead of dividing that area into disjoint regions, e.g., cells. The APs may be connected to central processing units (CPUs), with baseband capabilities, through fronthaul links, and they may need to collaborate to serve users, such as user equipments (UEs), in the area. It is expected that D-MIMO may guarantee high connectivity, may reduce interference, and improve spectral efficiency, resulting in a better user experience when compared to the traditional Centralized MIMO (C-MIMO) architecture, where all the antennas of the base station (BS) may be co-located. Modular D-MIMO: Radio Stripes and Radio Weaves To make deployment of a large number of distributed Multiple-Input and Multiple-Output (MIMO) access points simple and cost efficient, various approaches have been proposed, such as Radio Stripes and RadioWeaves. A common feature may be understood to be to use a shared fronthaul (FH) together with a high degree of integration and miniaturization, see Figure 1 and Figure 2. Sometimes, the electronic circuit containing the digital signal processing (DSP), antenna panel, and external interfaces, for power supply and data, may be denoted by antenna processing unit (APU). In this document, it may be referred to as Access Point (AP). Note however that APs may not be visible physical boxes, they may in some cases be only the location of a small integrated circuit inside of a protective cable. Figure 1 depicts a schematic example illustration of an AP 1 top view in panel a) and side view in panel b), respectively. The AP 1 may comprise one or more antenna elements 2, which may be located in an antenna panel 3, and an external interface 4 on a printed circuit board (PCB) 5. In panel b), the external interface 4 is depicted as a separate component. The DSP 6 is depicted under the antenna panel 3. Figure 2 is a schematic diagram representing two different network deployment and architecture examples. Panel a) depicts a non-limiting example of the RadioWeaves planar approach to interconnect APs. Panel b) depicts a non-limiting example of the RadioStripes linear approach to interconnect APs. It may be appreciated that a Central Processing Unit (CPU) 21 may be connected, via one or more connections, to a plurality of APs 22. A CPU may be understood as a node, e.g., Distributed Unit with coordinating capabilities for connection control among APs in a centralized manner. In case of ad-hoc structure, a CPU may not have any controller role. Each of the APs 22 may be connected to another AP 22 via one or more FH segments 23. The plurality of APs 22 may be arranged in a dispersed, planar, form in the radioweave, or in a linear form in the radio stripe. Segmented fronthaul The fronthaul structure used by e.g., RadioStripes and RadioWeaves may be referred to as segmented fronthaul. Each AP may be connected to one or more neighboring APs via interfacing segments that may be used for transferring power, DL data packets and precoding weights, UL combining weights and symbol estimates, etc. The existence of the plurality of segments between the APs may be understood to then result in a segmented fronthaul. An important property of such FH structure may be understood to be that a given AP may be generally not directly connected to the CPU but may signal to and from it may need to pass multiple segments to reach their destinations. To transfer signals to and from multiple UEs and multiple APs, careful routing solutions may be required to utilize the data transfer capabilities of the segments as fully as possible. The segment capacities in terms of data packets per time unit may be estimated e.g., in Gigabits per second (Gbps) or in UE packets / radio slot. Herein, routing may be understood to denote the necessary data, user-plane and / or control-plane data, forwarding throughout fronthaul network, and not addressing any Internet Protocol (IP)- level routing. In D-MIMO deployments, subsets of the available set of APs may be typically assigned to serve individual UEs. The subsets may be typically selected according to link conditions between UE-AP pairs, e.g., average path gains. The serving APs may then receive data from the CPU and transmit to their associated UEs. If all APs are directly connected to the CPU, the connections may be assumed to not affect the resulting serving quality. However, in D-MIMO with segmented fronthaul, such as RadioStripes / RadioWeaves, the data may need to be routed to serving APs over segments with limited capacity or bandwidth. The routing task performed in this D-MIMO fronthaul architecture may differ from established routing methods in conventional network structures, where Internet Protocol (IP) routing may be established, for example, multi-Transmission and Reception Points (TRPs) served by a gNB node. In the legacy context, FH routing may be rather static and may need to establish a long-term addressing path for FH data transfer; it may typically vary with load fluctuation or specific scheduling choices. Routing solution determination latency may thus be incurred only once and may not be a major issue. In the D-MIMO segmented fronthaul context, the routing may take place via a more resource-constrained connection network and the routing task may fail for some UEs and APs if the amount of packets approaches or exceeds the capacities of a certain segment. It may be understood to not be feasible to design the FH network with a fixed pre-allocation of resources for all serving AP combinations. Therefore, the routing approach may be understood to need to be determined anew for each co-scheduled UE set, e.g., typically whenever the served UE selection may change, which may occur every slot or at least at a millisecond (ms)-level. The routing update rate in the current context may thus be high and the computational load may need to be considered. To improve the routing efficiency, two-level routing approaches may be anticipated, where each UE may have one serving Access Point (AP) as an aggregating AP (AAP) that may serve as a single delivery point from a CPU and may split / distribute the UEs’ data to other APs. Figure 3 is a schematic diagram which shows a D-MIMO deployment in which a UE, depicted as a black square, is served by 4 APs, depicted by 4 larger circles, of which one, represented as a solid black circle, is an AAP. For the rest of the deployment in Figure 3, the other APs are depicted with small empty circles. Each AP is indicated with a capital letter. Each FH segment is indicated with a numeral. Figure 3 illustrates an example of the routing tasks in this two-level routing approach in such deployment. Two-level routing may comprise the following physical layer data transfers: a first routing level, routing level 1 (L1), from CPU subgroup APs, illustrated in Figure 3 by dashed arrows. In Figure 3, the first routing level is depicted with a solid arrow, whereas the second routing level is depicted with a dashed arrow. It may be understood that the routing stages L1 / L2 have no relation to the Layer 1 / 2 signal processing or protocols in the conventional notation, which may use the same abbreviations. Figure 4 is a schematic diagram illustrating these routing tasks into the different levels, L1 and L2. L1 and L2 may be viewed as independent parallel tasks that may share resources. To decouple them, they may be pipelined over two routing time units, e.g., slots / groups. This may allow prioritizing resources that may reach the AAP first for each UE in a given radio Transmission Time Interval (TTI). The first level L1, connecting the AAP, may be understood to be most critical for the UE to be served in the first place, and may be more robust if the L1 paths (CPU AAP) are shorter on the average. The second level L2 may help acquiring higher Signal-to-interference- plus-noise ratio (SINR) for a UE by including more serving APs. L2 may be more successful if the L2 paths are shorter. The figure depicts a pipelined execution of L1 and L2 steps towards multiple UEs. The time indices denote the TTI in which data is being transferred at each level. For example, during one pipeline phase, the L1 routing step may be conveying TTI n data from the CPU to the AAPs while simultaneously the L2 routing step transfers TTI n-1 data, delivered to the AAPs during the previous pipeline phase, from the AAPs to other serving APs. In typical examples, the routing resources, segment bandwidth, used for L1 and L2 at the segments may be common and shared. The fact that some packet contents may be the same may or may not be utilized. Earlier studies have evaluated FH connectivity handling and suitable AP group selection in systems with serial instead of direct connectivity between APs, e.g., in the segmented fronthaul setup. In [1], it has been observed that optimizing the two-level AP selection using the traditional user-centric approach, a matched-decision framework, whereby APs may be assigned based on primarily UE-AP channel qualities, yields suboptimal performance and more advanced algorithms may be required to maximize the performance. In [2], the impact of serial connections has been studied using a simplified Markov model of routing failures and it has been found that failure mitigation is necessary and may be achieved by partially or fully duplicating resources or cross-connecting in relevant regions of the FH network. Accordingly, FH connectivity handling and suitable AP group selection in systems with serial instead of direct connectivity between APs, e.g., in the segmented fronthaul setup, is difficult to optimize and prone to radio performance and connectivity degradation. SUMMARY As part of the development of embodiments herein, one or more challenges with the existing technology in FH connectivity handling and suitable AP group selection in systems with serial instead of direct connectivity between APs, e.g., in the segmented fronthaul setup will first be identified and discussed. The multi-stage / iterative approach presented in [1] may not be viable for relatively rapid AP grouping updates. It may be noted that the grouping and routing approach may need to be redone for every new co-scheduled UE set, which may be updated at the ms- scale, or, at the very least, at 10 ms-scale, if the same grouping is kept over multiple radio slots. The cost of the duplication required for the approach described in [2] may be prohibitive in a practical deployment. In [3], the specific use case of the present document, two-level AP selection, that is, the two pipeline stages L1 and L2, choosing the AAP and then the routing, and routing, is explored. Optimizing and performing initial AP group selection and routing as a joint process may provide best overall performance, e.g., in terms of the radio conditions of the serving APs and the ability to ensure successful routing connections. However, the joint optimization is excessively complex for practical solutions, including Artificial Intelligence (AI)-based approaches. Thus, a more viable approach may be to perform the AP grouping and routing path establishment tasks separately and sequentially. While the sequential execution of separately configured tasks may provide the required baseline functionality, a performance loss or resource under-utilization may be observed if heuristic configuration settings are used in varying UE load conditions. Some examples of grouping-related parameters and configuration aspects that may affect routing success in different user and traffic conditions may include AP group sizes, AAP selection choices, etc. Similarly, the routing task may be controlled via a number of controllable algorithm and utility function configuration parameters. The preferred settings for these configurations may typically depend on many scenario variables, such as the number of co-scheduled users, their service requirements, fronthaul segment constraints, AP processing capacity constraints, etc. that may all affect the preferred settings values in an interdependent manner that may be difficult to capture via traditional rules or look-up tables. It was further found that certain configuration parameters, e.g., the AP group size and routing maximum path length, affected multiple overall performance metrics via multiple mechanisms whose interactions were not easily predictable or separable. There is thus a need for an improved approach for jointly optimizing the parameters and settings used for AP grouping, AAP selection, routing, etc. that may capture the non- trivial impact of a combination of multiple parameter choices in different operating scenarios and may adapt the settings as the scenarios may change over time. According to the foregoing, it is an object of embodiments herein to improve the handling of a set of nodes connected via a segmented front-haul in a communications network. According to a first aspect of embodiments herein, the object is achieved by a method, performed by a first network node. The method is for handling a set of nodes connected via a segmented front-haul. The first network node operates in the communications network comprising the set of nodes. The first network node obtains first information indicating one or more conditions under which one or more devices are to be served by at least one of the set of nodes in the communications network. The first network node determines, based on the obtained first information, one or more parameters of a configuration of at least two processing actions to serve the one or more devices. The processing actions comprise i) determining, out of the set of nodes, one or more first subsets of nodes. The one or more first subsets of nodes are to serve the one or more devices. Each first subset of the one or more first subsets of nodes is determined to serve a respective second subset of devices of the one or more devices. The processing actions also comprise ii) selecting a respective aggregating node for each subset of the one or more first subsets of nodes. The processing actions further comprise iii) determining, for one or more packets to be exchanged between the one or more devices and a central node to which all the respective aggregating nodes are connected, a route. The central node provides a backhaul connection to a core network of the communications network. The route is between the central node and each of the nodes in the respective first subset of nodes. The processing actions additionally comprise iv) updating the one or more first subsets of nodes based on a measure of fulfilment of the one or more conditions. The processing actions also comprise v) determining precoding weights for the nodes in the determined or updated one or more first subsets of nodes. The determining of the one or more parameters is performed by jointly optimizing the one or more parameters. The first network node then initiates sequential execution of the at least two processing actions applying the determined configuration. According to a second aspect of embodiments herein, the object is achieved by the first network node, for handling the set of nodes configured to be connected via the segmented front-haul. The first network node is configured to operate in the communications network configured to comprise the set of nodes. The first network node is further configured to obtain the first information configured to indicate the one or more conditions under which the one or more devices are to be served by at least the one of the set of nodes in the communications network. The first network node is also configured to determine, based on the first information configured to be obtained, the one or more parameters of the configuration of the at least two processing actions to serve the one or more devices. The processing actions are configured to comprise: i) the determining, out of the set of nodes, the one or more first subsets of nodes to serve the one or more devices. Each first subset of the one or more first subsets of nodes is configured to be determined to serve the respective second subset of devices of the one or more devices. The processing actions are also configured to comprise: ii) the selecting the respective aggregating node for each subset of the one or more first subsets of nodes. The processing actions are further configured to comprise: iii) the determining, for the one or more packets to be exchanged between the one or more devices and the central node to which all the respective aggregating nodes are configured to be connected, the central node being configured to provide the backhaul connection to the core network of the communications network, the route. The route is between the central node and each of the nodes in the respective first subset of nodes. The processing actions are additionally configured to comprise: iv) the updating the one or more first subsets of nodes based on the measure of fulfilment of the one or more conditions. The processing actions are also configured to comprise: v) the determining the precoding weights for the nodes in the one or more first subsets of nodes configured to be determined or updated. The determining of the one or more parameters is configured to be performed by jointly optimizing the one or more parameters. The first network node is also configured to initiate the sequential execution of the at least two processing actions applying the configuration configured to be determined. According to a third aspect of embodiments herein, the object is achieved by a computer program, comprising instructions which, when executed on at least one processing circuitry, cause the at least one processing circuitry to carry out the method performed by the first network node. According to a fourth aspect of embodiments herein, the object is achieved by a computer-readable storage medium, having stored thereon the computer program, comprising instructions which, when executed on at least one processing circuitry, cause the at least one processing circuitry to carry out the method performed by the first network node. By obtaining the first information, the first network node may be enabled to use the first information as input to determine which configuration to use in order to optimally serve the one or more. By determining the one or more parameters of the configuration of the at least two processing actions to serve the one or more devices, wherein the determining may be performed by jointly optimizing the one or more parameters, the first network node may enable that the one or more devices may be served with increased system throughput and user throughput, as well as reduced scheduling latency, and / or reduced number of retransmissions or packet drops. Determining the one or more parameters of the configuration of the at least two processing actions to serve the one or more devices may be understood to be a preparation for the sequential execution and therefore may enable that such execution configuration may be jointly optimized, as opposed to performing the steps “unaware” of each other, providing the performance benefits. By then initiating the sequential execution of the of the at least two processing actions applying the determined configuration, the first network node may enable that the one or more devices may be served, e.g., to send the one or more packets between the one or more devices and the central node, with increased system throughput and user throughput, as well as reduced scheduling latency, and / or reduced number of retransmissions or packet drops, while simplifying the system design and reducing the cost and overall complexity of the optimization process. The sequential process may be understood to allow approaching the performance potential, e.g., the multiple example metrics mentioned, without putting undue hardware and processing speed requirements on the implementation. Embodiments herein may enable to jointly optimize the parameters and settings used to execute the steps that may be performed in dynamic signal transmission in the communications network, so that non-trivial impact of a combination of multiple parameter choices in different operating scenarios may be captured. Replacing joint execution with joint parameter optimization and sequential execution may yield significant complexity reduction at moderate performance impact, and large performance improvement over legacy, non-coordinated parameter settings. BRIEF DESCRIPTION OF THE DRAWINGS Examples of embodiments herein are described in more detail with reference to the accompanying drawings, and according to the following description. Figure 1 depicts a schematic example illustration of an antenna processing unit, top view in panel a) and side view in panel b). Figure 2 is a schematic diagram representing two different network deployment and architecture examples, RadioWeaves in panel a) and RadioStripes in panel b). Figure 3 is a schematic diagram representing a D-MIMO deployment. Figure 4 is a schematic diagram representing routing tasks into the different levels. Figure 5 is a schematic diagram illustrating a communications network, according to embodiments herein. Figure 6 is a flowchart depicting an example of a method performed by a first network node, according to embodiments herein. Figure 7 is a schematic representation depicting aspects of a non-limiting example of a method according to embodiments herein. Figure 8 is a schematic representation depicting aspects of a non-limiting example of a method according to embodiments herein. Figure 9 is a schematic representation depicting aspects of a non-limiting example of a method according to embodiments herein. Figure 10 is a schematic representation depicting aspects of a non-limiting example of a method according to embodiments herein. Figure 11 is a schematic representation depicting aspects of a non-limiting example of a method according to embodiments herein. Figure 12 is a schematic representation depicting aspects of a non-limiting example of a method according to embodiments herein. Figure 13 is a schematic block diagram illustrating a non-limiting example of a first network node, according to embodiments herein. DETAILED DESCRIPTION Certain aspects of the present disclosure and their embodiments may provide solutions to the challenge discussed in the Summary section or other challenges. There are, proposed herein, various embodiments which address one or more of the issues disclosed herein. As a general overview, embodiments herein may be understood to relate to joint configuration of AP grouping and routing in D-MIMO with segmented fronthaul. More particularly, embodiments herein may be understood to relate to coordinated optimization of a D-MIMO processing sequence where the following multiple steps may be executed sequentially: 1) Selection of initial serving AP groups, 2) AAP selection, 3) L1 and L2 routing, 4) Adjustment of serving AP groups, and 5) Precoding. Embodiments herein may relate to a method for jointly and incrementally determining preferred configurations for the constituent steps in the sequence to achieve joint optimization of radio and routing performance, based on multiple inputs. It may be noted that performing e.g., AP grouping and routing in a single joint stage, UE positions and traffic information in, optimized serving AP sets with routing paths and precoding weights out, may be preferable but this may not be practical, even when using a large AI model, due to the resulting large product solution space. Embodiments herein may therefore be understood to not perform full joint grouping and routing determination per co- scheduled UE set realization. However, by jointly determining the relevant configuration parameters, a high degree of joint optimization may be achieved. Jointly may be understood to mean that configuration of a certain step may be determined while being aware of configurations of other steps that may interact with and affect each other. For example, choosing the grouping and determining the routing, as a consequence of the first. This may also mean that by backtracking on this decision another grouping may be chosen, and then another routing determination may be made Some of the embodiments contemplated will now be described more fully hereinafter with reference to the accompanying drawings, in which examples are shown. In this section, the embodiments herein will be illustrated in more detail by a number of exemplary embodiments. Other embodiments, however, are contained within the scope of the subject matter disclosed herein. The disclosed subject matter should not be construed as limited to only the embodiments set forth herein; rather, these embodiments are provided by way of example to convey the scope of the subject matter to those skilled in the art. It should be noted that the exemplary embodiments herein are not mutually exclusive. Components from one embodiment may be tacitly assumed to be present in another embodiment and it will be obvious to a person skilled in the art how those components may be used in the other exemplary embodiments. Note that although terminology from LTE / 5G has been used in this disclosure to exemplify the embodiments herein, this should not be seen as limiting the scope of the embodiments herein to only the aforementioned system. Other wireless systems with similar features, may also benefit from exploiting the ideas covered within this disclosure. Figure 5 depicts a non-limiting example of a communications network 100, in some examples also referred to as a wireless communications system, cellular radio system, or cellular network, in which embodiments herein may be implemented. The communications network 100 may typically be a 5G system, 5G network, NR-U or Next Gen System or network, Licensed-Assisted Access (LAA), or MulteFire. In particular embodiments, the communications network 100 may be a D-MIMO network, e.g., a D-MIMO network with segmented fronthaul. The communications network 100 may support or be a younger system than a 5G system, such as, for example a 6G system. The communications network 100 may support other technologies, such as, for example Long-Term Evolution (LTE), LTE-Advanced / LTE-Advanced Pro, e.g., LTE Frequency Division Duplex (FDD), LTE Time Division Duplex (TDD), LTE Half-Duplex Frequency Division Duplex (HD-FDD), LTE operating in an unlicensed band, etc… Other examples of other technologies the communications network 100 may support may be Wideband Code Division Multiple Access (WCDMA), Universal Terrestrial Radio Access (UTRA) TDD, Global System for Mobile Communications (GSM) network, Enhanced Data Rates for GSM Evolution (EDGE) network, GSM EDGE Radio Access Network (GERAN) network, Ultra-Mobile Broadband (UMB), network comprising of any combination of Radio Access Technologies (RATs) such as e.g. Multi-Standard Radio (MSR) base stations, multi-RAT base stations etc., any 3rd Generation Partnership Project (3GPP) cellular network, WiFi networks, Worldwide Interoperability for Microwave Access (WiMax), IoT, Narrowband Internet of Things (NB-IoT), or any cellular network or system. Thus, although terminology from 5G / NR and LTE may be used in this disclosure to exemplify embodiments herein, this should not be seen as limiting the scope of the embodiments herein to only the aforementioned systems. It may be understood that the layout of the communications network 100 depicted in Figure 5 is a non-limiting example. As depicted in Figure 5, the communications network 100 comprises a first network node 111 and a central node 112. Any of the first network node 111 and the central node 112 may be understood as a computer system. In some examples, any of the first network node 111 and the central node 112 may be implemented as a standalone server in e.g., a host computer in the cloud. Any of the first network node 111 and the central node 112 may in some examples be a distributed node or distributed server, with some of their respective functions being implemented locally, e.g., by a client manager, and some of its functions implemented in the cloud, by e.g., a server manager. Yet in other examples, any of the first network node 111 and the central node 112 may also be implemented as processing resources in a server farm. In a particular example, any of the first network node 111 and the central node 112 may be a distributed node, which some functions performed at any of nodes in the set of nodes 120 described below, and other functions performed at a central network node. In some non-limiting examples, any of the first network node 111 and the central node 112 may be a radio network node. As a radio network node, any of the first network node 111 and the central node 112 may be a transmission point such as a radio base station, for example a gNB, an eNB, or any other network node with similar features capable of serving a wireless device, such as a user equipment (UE) or a machine type communication device, in the communications network 100. In typical examples, any of the first network node 111 and the central node 112 may be a base station, such as a gNB. The first network node 111 may be understood as a network node having a capability to obtain and analyze radio information pertaining to the communications network 100. The central node 112 may be a CPU in the communications network 100. The central node 112 may be a node where one or more packets, e.g., in the downlink direction, may arrive, e.g., from a backhaul connection, for distribution to the set of nodes 120, via the fronthaul network of the communications network 100. In embodiments wherein the communications network 100 may be a D-MIMO network, the central node 112 may be a CPU of the D-MIMO network. In some embodiments, the first network node 111 and the central node 112 may be independent and separated nodes, as depicted in the non-limiting example of Figure 5. In another embodiment, the first network node 111 may be co-located or be the same node as the central node 112. In such an example, the first network node 111 and the central node 112 may communicate over a physical interface, e.g., an inter-chip / module interface on circuit boards or an inter-unit interface with plugged cables. In such an example, the radio and routing algorithms may be executed in logically separated Software (SW) units in the same Hardware (HW) / computational block and passed over an Application Programming Interface (API) or another SW interface. All the possible combinations are not depicted in Figure 5 to simplify the Figure. The communications network 100 may cover a geographical area, which in some embodiments may be divided into coverage areas, wherein each coverage area may be served by a radio network node, although, one radio network node may serve one or several coverage areas. As a radio network node, any of the first network node 111 and the central node 112 may be of different classes, such as, e.g., macro eNodeB, home eNodeB or pico base station, based on transmission power and thereby also coverage area size. In some examples, as a radio network node, any of the first network node 111 and the central node 112 may serve receiving nodes with serving beams. Any of the first network node 111 and the central node 112, e.g., as radio network node, may support one or several communication technologies, and its name may depend on the technology and terminology used. Any of the first network node 111 and the central node 112 may be directly connected to one or more core networks. The central node 112 provides a backhaul connection to a core network 115 of the communications network 100. The communications network 100 comprises a set of nodes 120. Any of the nodes in the set of nodes 120 may be understood to be an Access Point (AP). An AP may be understood as a hardware unit with transmission / reception capability comprising at least an antenna panel and radio circuitry, and a fronthaul connection via at least one segment. An AP may be also referred as an APU or Transmission Reception Point (TRP). Any reference herein to a “node” may be understood to refer to a node in the set of nodes 120, e.g., an AP. In typical embodiments, each of the nodes in the set of nodes 120 may be an AP or a TRP. The set of nodes 120 may be interconnected and configured in such a way that more than one panel may cooperate in coherent decoding of data from a given device, of the one or more devices comprised in the communications network and described below, and more than one panel may cooperate in coherent transmission of data to a device, of the one or more devices comprised in the communications network 100. The set of nodes 120 may comprise one or more first subsets of nodes 121, 122, 123. One or more devices 130 may be comprised in the communications network 100, whereof six devices are depicted in the non-limiting examples of Figure 5. Any of the one or more devices 130 comprised in the communications network 100 may be a wireless device or wireless communication device such as a 5G UE, or a UE, which may also be known as e.g., mobile terminal, wireless terminal and / or mobile station, a Customer Premises Equipment (CPE) a mobile telephone, cellular telephone, or laptop, e.g., with wireless capability, just to mention some further examples. Any of the one or more devices 130 comprised in the communications network 100 may be, for example, portable, pocket-storable, hand-held, computer-comprised, or a vehicle-mounted mobile device, enabled to communicate voice and / or data, via the RAN, with another entity, such as a server, a laptop, a Personal Digital Assistant (PDA), or a tablet, Machine-to-Machine (M2M) device, device equipped with a wireless interface, such as a headset, printer, goggles or a file storage device, modem, IoT device, sensor, or any other radio network unit capable of communicating over a radio link in a communications system. Any of the one or more devices 130 comprised in the communications network 100 may be enabled to communicate wirelessly in the communications network 100. The communication may be performed e.g., via a RAN, and possibly the one or more core networks, which may be comprised within the communications network 100. Each first subset of the one or more first subsets of nodes 121, 122, 123 may serve a respective second subset of devices 131, 132, 133 of the one or more devices 130. That is, one first subset of nodes 121, 122, 123 may be understood to be the nodes in the set of nodes 120 which may provide radio coverage to a respective subset of one or more devices of the one or more devices 130 comprised in the communications network 100. Hence, the set of nodes 120 may comprise a plurality of first subsets of nodes 121, 122, 123, one respective first subset of nodes 121, 122, 123, for every second subset of devices 131, 132, 133. In the non-limiting example of Figure 5, the one or more first subsets of nodes 121, 122, 123 comprise a first first subset of nodes 121 providing coverage to a first second subset of devices 131, a second first subset of nodes 122, providing coverage to a second second subset of devices 132, and a third first subset of nodes 123, providing coverage to a third second subset of devices 133. The communications network 100 may also comprise a respective aggregating node 141, 142, 143 for each subset of the one or more first subsets of nodes 121, 122, 123. An aggregating node, e.g., AAP, may be understood as a node which in the downlink, may receive the data packets for a respective device from a respective source node, e.g., the central node 112, and distribute them further to additional serving nodes, e.g., APs, of the set of nodes 120. This way, the amount of unique packets to be routed may be reduced compared to the number of individual respective source node-node pairs per device. In other words, an AAP may be understood as an aggregation point for a respective first subset of nodes 121, 122, 123 that may be serving a certain second subset of devices 131, 132, 133, at least receiving the data to be served and distributing it to the other serving nodes in the respective first subset of nodes 121, 122, 123. The same may apply in the uplink direction. In the non-limiting example of Figure 5, the first first second subset of nodes 121 comprises a first respective aggregating node 141, the second first subset of nodes 122 comprises a second respective aggregating node 142 and the third first subset of nodes 123 comprises a third respective aggregating node 143. Whether a node of the set of nodes 120 may be comprised in one of the first subsets of nodes 121 or in another of the first subset of nodes 122, 123 may vary from time to time and from device to device. Individual nodes of the set of nodes 120, as shown in Figure 5, may be comprised in more than one first subset of nodes 121. The first network node 111 may be configured to communicate within the communications network 100 with the central node 112 over a first link 151, e.g., a wired link or a radio link. The central node 112 may be configured to communicate within the communications network 100 with at least one of the nodes in the set of nodes 120 over a respective second link 152, e.g., a wired link or a radio link. Any of the nodes in the set of nodes 120 may be configured to communicate within the communications network 100 with one or more of the other nodes in the set of nodes 120 over a respective third link 153, e.g., a radio link or a wired link, which may be understood to be a respective FH segment. In general, the usage of “first”, “second”, and / or “third” herein may be understood to be an arbitrary way to denote different elements or entities, and may be understood to not confer a cumulative or chronological character to the nouns they modify. Several embodiments are comprised herein. It should be noted that the examples herein are not mutually exclusive. Components from one embodiment may be tacitly assumed to be present in another embodiment and it will be obvious to a person skilled in the art how those components may be used in the other exemplary embodiments. Embodiments of a method performed by the first network node 111, will now be described with reference to the flowchart depicted in Figure 6. The method may be understood to be for handling the set of nodes 120 connected via a segmented front-haul. The first network node 111 operates in the communications network 100 comprising the set of nodes 120. In some embodiments, the communications network 100 may be a D-MIMO network. Several embodiments are comprised herein. In some embodiments all the actions may be performed. In some embodiments, two or more actions may be performed. It should be noted that the examples herein are not mutually exclusive. One or more embodiments may be combined, where applicable. All possible combinations are not described to simplify the description. Components from one embodiment may be tacitly assumed to be present in another embodiment and it will be obvious to a person skilled in the art how those components may be used in the other exemplary embodiments. A non-limiting example of the method performed by the first network node 111 is depicted in Figure 6. Some actions may be performed in a different order than that shown in Figure 6. In Figure 6, actions which may be optional in some examples are depicted with dashed boxes. Action 601 Embodiments herein may be understood to be applied in a communications network such as the communications network 100, e.g., a distributed MIMO network, with the segmented fronthaul. In such a network, fronthaul / backhaul / transport may be provided to the central node 112 or central unit, e.g., a CPU, that may connect to a segmented interconnected fronthaul network that may connect to the set of nodes 120, e.g., individual APs. The segments may implement wired or wireless connections. In either case, the capacity or effective bandwidth of the segments may be determined or estimated. The first network node 111 may be co-localized or be the same as the CPU. The nodes in the set of nodes 120, e.g., APs, may not be reachable directly by the CPU, but via one or, typically, more segments. There may be multiple options to connect the nodes in the set of nodes 120 to the central node 112, and the segmented option may be used over the coverage area to connect them. Once deployed, the nodes in the set of nodes 120 and the FH network may be tightly integrated. Embodiments herein may be understood to aim to solve the task of serving the one or more devices 130 by the set of nodes 120, such as, for example, to distribute respective one or more packets, e.g., DL data packets, from a source node, e.g., the central node 112, or the CPU, to multiple serving nodes of the set of nodes 120, e.g., APs, per device of the one or more devices 130, where the packets in single or multi streams may need to be distributed to all serving nodes of the set of nodes 120, that may subsequently perform precoded data transmission to the respective device. The one or more packets may include encoded binary data, modulated symbol indices, or, in some embodiments, uncoded binary data packets. In this Action 601, the first network node 111 obtains first information indicating one or more conditions under which the one or more devices 130 are to be served by at least one of the set of nodes 120 in the communications network 100. The one or more conditions may be understood to be e.g., states or one or more scenario parameters of the communications network 100. In some embodiments, the one or more conditions comprise one or more of the following. According to a first option, the one or more conditions may comprise radio related conditions, such as, for example, those related to channels or channel ranks between the one or more devices 130 and particular nodes of the one or more set of nodes 120, that is, UE-AP channels or channel ranks. According to a second option, the one or more conditions may comprise one or more indicators of traffic of the devices of the one or more devices 130. The one or more indicators may be, for example, UE traffic Key Performance Indicators (KPIs), such as latency-optimized, Throughput (TP)-optimized, other application information, etc. According to a third option, the one or more conditions may comprise a first load of a transport network of the communications network 100. A transport network may comprise backhaul and fronthaul links. Hence, the load of a transport network may comprise backhaul and fronthaul data. According to a fourth option, the one or more conditions may comprise a second load of a radio access network of the communications network 100 or network load, such as, for example, the number of (nrof) co-scheduled devices of the one or more devices 130, etc. According to a fifth option, the one or more conditions may comprise one or more routing constraints. The one or more routing constraints may comprise, for example, an AP adjacency matrix or aggregated metrics describing it, segment capacities and availability. The adjacency matrix of the set of nodes 120 may describe the interconnectivity among the set of nodes 120. That is, for each node, its connected neighbors and connection segment capacities. The matrix may be constant over time for a given network deployment. Alternatively, it may variable and updated over time, e.g., when the nodes may be mobile and wireless FH connections may be used. Examples of wired segment capacities as of 2021 may be on the order of several Gbps or 5-10 UE packets per radio slot. The first information may be available as fixed system design information or extracted from the current system configuration. According to a sixth option, the one or more conditions may comprise one or more processing constraints of the set of nodes 120, such as, for example, precoding and combining computation budget, nrof devices of the one or more devices 130 that may be handled per node of the set of nodes 120, e.g., per AP, etc.), the second load, nrof the one or more devices 130, etc. According to a seventh option, the one or more conditions may comprise operating frequency of a bandwidth used by the communications network 100. For example, group size of each of the first subsets of nodes 121, 122, 123, that is, spatial diversity, may be more important at higher frequencies, e.g., mmWaves. According to an eighth option, the one or more conditions may comprise information on mobility of the one or more devices 130, e.g., UE mobility information. Obtaining may be understood as determining, deriving, calculating, retrieving, e.g., from a memory storage, and / or receiving, e.g., from any of the one or more devices 130, any of the nodes in the set of nodes 120, and / or another node comprised in the communications network 100. The first information regarding the one or more devices 130, e.g., co-scheduled devices of the one or more devices 130, may for example be obtained from a scheduler in the central node 112 for the distributed MIMO network, or a relevant network segment. The scheduler, and / or associated algorithms such as link adaptation and channel estimation with respect to multiple nodes in the communications network 100 may also provide channel quality estimates, instantaneous or average, with respect to all or a large number of nodes in the communications network 100 for the one or more devices 130. The first information may be available from the radio scheduler or the core network 115. By obtaining the first information in this Action 601, the first network node 111 may then be enabled to use the first information as input to performing the next Action 602, wherein the first network node 111 may be enabled to determine which configuration to use in order to optimally serve the one or more devices 130, as will be described in Action 602. Action 602 In this Action 602, the first network node 111 determines, based on the obtained first information, one or more parameters of a configuration of at least two processing actions to serve the one or more devices 130. The processing actions comprise the following four actions, not necessarily in the order presented. A first processing action is determining, out of the set of nodes 120, the one or more first subsets of nodes 121, 122, 123 to serve the one or more devices 130. Each first subset of the one or more first subsets of nodes 121, 122, 123 is determined to serve a respective second subset of devices 131, 132, 133 of the one or more devices 130. In other words, the first processing action may be to perform the initial AP grouping (^^^). The first processing action may comprise determining the group sizes of the first subsets of nodes 121, 122, 123 per device of the one or more devices 130 or per device cluster, e.g., determining AP group sizes per UE or UE cluster, common or individual values per UE / cluster. The first action may also or alternatively comprise determining constraints on overlap of the first subsets of nodes 121, 122, 123, e.g., of the AP groups. At least two of the devices within at least one of the respective second subset of devices 131, 132, 133 may be co-scheduled. A second processing action is selecting a respective aggregating node 141, 142, 143 for each subset of the one or more first subsets of nodes 121, 122, 123. In other words, the second processing action may be to perform the initial AAP selection (^^^^). The second processing action may comprise determining a selection criterion for the respective aggregating node 141, 142, 143, e.g., an AAP selection criterion, and determining the constraints on the number of AAPs in the system. A third processing action is determining, for one or more packets to be exchanged between the one or more devices 130 and the central node 112 to which all the respective aggregating nodes 141, 142, 143 are connected, a route between the central node 112 and each of the nodes in the respective first subset of nodes 121, 122, 123. The central node 112 provides a backhaul connection to the core network 115 of the communications network 100. In other words, the third processing action may be to perform routing (^^). The third processing action may comprise determining: a) a routing strategy, such as connect all possible nodes of the set of nodes 120, connect a minimum set of nodes of the set of nodes 120, etc…., b) a maximum routing path length, c) a number of routing path options considered, etc. To be exchanged may comprise downlink communications, uplink communications, or both. The route may comprise a first route between the central node 112 and the selected respective aggregating node 141, 142, 143 and a second route between the selected respective aggregating node 141, 142, 143 and each of the nodes in the respective first subset of nodes 121, 122, 123. The first route may correspond to an L1 level routing and the second route may correspond to a L2 level routing. A fourth processing action is updating the one or more first subsets of nodes 121, 122, 123 based on a measure of fulfilment of the one or more conditions. In other words, the fourth processing action may be to perform a serving AP group adjustment. The fourth processing action may comprise determining an adjustment strategy, for example, remove unconnected nodes of the set or nodes 120, redo initial grouping and / or routing, … A fifth processing action is determining precoding weights for the nodes in the determined or updated one or more first subsets of nodes 121, 122, 123. In other words, the fifth processing action may be to perform the precoding weight computation, wherein (^^) may be understood as a matrix of , where ^ ∈ |^^| . The fifth processing action may comprise determining a weight structure, such as Matched Filter (MF), Zero-Forcing (ZF), Linear Minimum Mean Square Error (LMMSE), … Determining may be understood as calculating, deciding or deriving. The determining of the one or more parameters in this Action 602 is performed by jointly optimizing the one or more parameters. Jointly optimizing may be understood as that the determination of the one or more parameters may be performed so that all of the one or more parameters being determined may be optimized, as the determination of the one or more parameters as a whole, where the choice of one parameters may influence the other, as opposed to determining each parameter in isolation, considering the sequence of operations comprising two or more steps. The one or more parameters, e.g., two or more, may pertain to different steps of the sequence. The one or more parameters of the configuration may be understood as configurations, parameters, settings, or similar, and may be based on determining applying or selecting decision strategies. In some embodiments, joint optimization may be based on an improvement of one or more metrics selected from: system throughput, connection failure rate of the one or more devices 130, signal to interference plus noise ratio (SINR), transmission delay, a transmission diversity or transmission reliability, quality of service (QoS), and quality of experience (QoE). The system throughput may comprise, for example, an average UE throughput, a first indicator of whether the system or UE throughput may be above a first threshold or a first improvement, etc. The connection failure rate of the one or more devices 130 may comprise a UE connection failure rate, e.g., an AAP routing failure rate, a second indicator of whether the connection failure rate may be below a second threshold or a second improvement, etc. The SINR may comprise UE SINR values or distribution, a third indicator of whether the SINR man be above a third threshold or third improvement, etc. The transmission delay may comprise UE serving delay / latency, a fourth indicator of whether the transmission delay may be below a fourth threshold or fourth improvement, etc. The QoS metric may comprise a fifth indicator of whether UE QoS may be fulfilled. The QoE metric may comprise a sixth indicator of whether QoE may be fulfilled. The one or more metrics may be D-MIMO performance metrics. The determining in this Action 602 of the one or more parameters may be performed using artificial intelligence (AI) methods. The AI methods may comprise performing the method in Actions 601, 602 iteratively during a training phase, during which a model to predict the one or more parameters may be trained with a first set of data gathered during a first set of iterations, and an execution phase, during which the trained model may be used over one or more second iterations. Accordingly, the determination in this Action 602 may be implemented using AI / Machine Learning (ML) models. In some embodiments, the model(s) may be trained and / or gradually improved using Reinforcement-Learning (RL). One or more RL agents may be trained by receiving reward metrics based on a resulting performance of the communications network 100, e.g., D-MIMO performance, e.g., UE SINR, system throughput, failure rates, etc., and use these or their combination to improve the policy for determining the above configurations and settings for given input values. In further particular embodiments, the artificial intelligence methods may comprise using a plurality of RL agents, and wherein the determining in this Action 602 of the one or more parameters of the at least two processing actions may be performed by a collaborative optimization among the plurality of RL agents. The overall RL approach may be partitioned into multiple agents, where each agent may deal with a specific processing action or task, e.g., the group size or the selection of the group itself, and the two or more agents may work collaboratively to solve these tasks, exchanging their actions so that one agent may make a decision, e.g., about the group size, while knowing the decision, e.g., selection of group, of the other agent. The method may optionally further determine optimized specific first subset of nodes 121, 122, 123, e.g., AP group, and respective aggregating node 141, 142, 143, e.g., AAP selections, e.g., UE-specific group sizes, UE-specific actual group selections, or actual AAP selections for selected groups, specific routing solutions, etc., optionally using as additional input initial, non-optimized AP group selections. There may be several RL algorithms that may be utilized to perform the determining in this Action 602. In some embodiments, the first network node 111 may use a Soft Actor Critic (SAC)-Discrete algorithm, which may be a suitable algorithm to this problem for the following reasons. One reason is that it may allow for handling both continuous and discrete actions, since the problem to be solved may be understood to have both discrete, e.g., AAP grouping, and continuous actions, e.g., determining beam weights. Another reason is that it may consider the optimization of a temperature parameter that may be understood to weigh the importance between the entropy of un-explored a_t / s_t (Action / State) and the reward. In addition, the design of this algorithm may be extended in embodiments herein with a value, e.g., Q-Network, and a policy network that may incorporate 1) action branch, 2) per task value personalization, and other complexity reduction approaches. A non-limiting example of a basic discrete Soft-Actor-Critic algorithm that may be used in embodiments herein is shown below as Algorithm 1. Such algorithm may comprise the following steps. Steps (1-4) may take care of initialization of main and target Q, or value, networks and discrete policy network, which may also include continuous quantized values, and initialize the replay buffer. The main network may be understood as the model which may learn from the current set of experiences but may potentially forget previous actions that may yield higher rewards. The target network may be understood as the model that may remember previous actions, from previous experiences, that may have provided good results. The two may be used together to avoid suffering catastrophic forgetting which may happen if only the main network were to be used without the target network. The reason why both a main and a target network may be used may be understood to be to optimize the stability of the system, that is, allowing the first network 111 more time to evaluate more actions, instead of allowing it to update all the time, via considering the quick update of main models, where the target network may be updated every controlled period. This may be understood to be to overcome catastrophic forgetting. Steps (6-9) may be understood to focus on sampling actions from policy, states from the environment, and storing them in the replay buffer. Steps (10-14) may be understood to focus on updating Value (Q) and policy network weights, in addition to updating a temperature parameter and equalizing the target network by the local network every pre-decided time. That is, periodically the main network may be copied to the target network, making them equal. The temperature parameter may be understood to be responsible for determining the relative importance between the entropy term and the reward term in the policy maximization function, the entropy term may be understood to guide the policy to try out actions that may not have been tested before and its entropy with respect to state may be understood to be high. Before describing the pseudocode of Algorithm 1, the reward, states, and actions vectors may be defined. In the definition below any of the any of the one or more devices 130 may be referred to as a UE. This may be understood to be a non-limiting example, and any reference to UE may be understood to equally apply to a device. It may be noted that transmission reliability may be based on the Key Performance Indicator (KPI) chosen and threshold. If reliable communication is desired, the latency may be kept below a certain threshold and Signal to Interference Noise Ratio (SINR) above certain threshold, for most of transmission. Transmission diversity may be controlled by, e.g., adding a condition of having more than one node, e.g., AP, such that if special fading occurs because of one AP, another AP may be able to compensate. Definitions of reward and states: The states may be defined as: States: The actions may be defined as: Wherein the terms may be described as follows, wherein “m” may be understood to represent which UE, t may be understood to represent time, M may be understood to represent all UEs: ^^may be understood as the radio related conditions, e.g., radio-related, UE-AP channels or channel ranks; ^^^may be understood as the one or more indicators of traffic of the devices of the one or more devices 130, e.g., UE traffic KPIs, for example, latency-optimized, TP-optimized, other application info…; ^^may be understood as the second load of a radio access network of the communications network 100, e.g., the NW load for example, nrof co-scheduled UEs, etc.; ^^^may be understood as the one or more routing constraints, e.g., routing constraints, e.g., AP adjacency matrix or aggregated metrics describing it, segment capacities and availability; ^^^may be understood as the one or more processing constraints of the set of nodes 120, e.g., AP processing constraints, such as precoding and combining computation budget, nrof UEs that may be handled per AP, etc., NW load, nrof UEs, etc. ^^^, ^^,^^may be understood as the system throughput metric, e.g., throughput, of transferring sensing data, per node, and the corresponding weight; ^^^, ^^,^^may be understood as the connection failure rate of the one or more devices 130 metric, e.g., UE connection failure rate, e.g., AAP routing failure rate, e.g., the second indicator of whether the connection failure rate may be below the second threshold or a second improvement, and the corresponding weight; ^^^^^, ^^,^^^^^may be understood as the SINR metric, e.g., SINR per node and ^^,^^^^^may be understood asthe corresponding weight of the SINR; ^^, ^^,^may be understood as the transmission delay metric, e.g., UE serving delay / latency, the fourth indicator of whether the transmission delay may be below the fourth threshold or the fourth improvement, and the corresponding weight. ^^may be understood as the QoS and / or QoE metric, e.g., an indicator of whether UE QoS / QoE may be fulfilled; may be understood as the action space of the four processing actions ‘n’ (AP, AAP, R, {w,m}) at time ‘t’. System Actions: The set of processing actions may be represented as ^^,^∈ Accordingly, an example of action space representation may be, 1) ^^,^= ^^^,^, which may be understood to mean which AP may need to serve UE m, which may be the same for all UEs with the same cell, 2) ^^^^,^^^= ^^,^= ^{^,^}may be understood to represent the action of index of precoder ‘w’ for UE ‘3’, given specific AP, AAP, and route (via ^^^, ^^^^, ^^). It may be noted that ‘^’ in the sub-script of ^^may represent 1,2,3,4, for ease of noting However, ‘^’ in {^^, ^^^, ^^^, ^^^^^, ^{^,^}, ^^, ^^, ^^^, ^^, ^^^, ^^^} may represent device, e.g., UE ‘^’. It may be also noted that not all actions may be noted per ‘m’, except the because AP, AAP, and R actions may be understood to be for the whole UEs within a targeted cell. One example of such values that each of the action vector element may be represented by may be as follows: 1. ^^^∈ {1,2,3}, may be understood to mean that Access Points 1,2,3 may serve such device or group of devices; 2. ^^^^∈ {1} may be understood to mean that Access-Point 1 may be understood as the aggregation AP for such device; 3. ^^∈ {1,3,5}, may be understood to mean that the route of fronthaul may be through AP1→ AP3 → AP5 to device, and vice-versa. The routing component may be defined as an ordered path via the multiple segments, each connecting a specified pair of nodes of the set of nodes 120, e.g., an AP pair. Here the typical case where a pair of nodes of the set of nodes 120, e.g., an AP pair, has a single segment connecting them may be assumed. If there may be multiple segments, there may a straightforward extension. An equivalent representation may be to label the segments and provide a_R as the sequence of labels; 4. ^{^,^}∈ {11,9, 1}, may be understood to mean that the serving nodes of the set of nodes 120, e.g., APs, mentioned in ^^∈ {1,2,3}, may use one or more beams that may be precoded via indices {11,9,1}, respectively, wherein the sub-script “w” may be understood to represent the precoding index. In here, it may be assumed that there may exist numerated data beams which may have decided precoder weights, and may be tagged via indices. It may be noted that the Q-Function ( ^(^^)) , policy network's weights , and temperature cost ^(^)may be defined as in [4]. 1 Initialise ^^^: ^ → ℝ|^|, ^^^: ^ → ℝ|^|, ^^: ^ → Initialise main network [0,1]|^|2Initialise ^^^^: ^ → ℝ|^|, ^^^^: ^ → ℝ|^|Initialize target network 3^̅^ ← ^^ , ^̅^ ← ^^Equalize target and main network weights 4^ ← ∅Initialize an empty replay buffer 5 For each iteration do 6 For each environment step do 7 ^^~^^(^^|^^) Sample action from the policy 8 ^^^^~^(^^^^|^^, ^^) Sample transition from the environment 9 ^ ← ^ ∪ {(^^, ^^, ^(^^, ^^), ^^^^)} Store the transition in the replay buffer 10 For each gradient step do11^^← ^^− ^^∇^^^^(^^)for ^ ∈ {1,2}Update the Q-Functionparameters 12^ ← ^ − ^^^^^^^(^) Update policy weights 13^ ← ^ − ^^^^ ^(^)Update temperature 14^^^ ← ^^^ +(1 − ^)^^^ for ^ ∈ {1,2}Update target network weights 15Output ^^, ^^, ^Optimized parametersAlgorithm 1 Basic Discrete SAC In some embodiments, the AI methods may comprise using one or more RL agents using: a) an approach selected from: i) Discrete SAC, and ii) Contextual Bandit, and wherein the approach is Discrete SAC, b) one or more policy functions, and c) one or more value functions, such as (^(^^)) above. A policy function, such as above, may be understood to take as input the state and action and produce a value, typically known as Q. That value may be used to determine the best action for the given state. The one or more policy functions may comprise: a first policy deciding on the one or more first subsets of nodes 121, 122, 123 to serve the one or more devices 130, that is, serving APs, a second policy of deciding on the respective aggregating node 141, 142, 143, e.g., aggregated AP, a third policy function on deciding the route, e.g., the routing APs, and a fourth policy function on deciding on the precoding beam weights A value function, such as (^(^^)) above may be understood to take as input the state and measure feature rewards that may be obtained from being in that state. The one or more value functions may comprise: a first value function output of the one or more first subsets of nodes 121, 122, 123 to serve the one or more devices 130, that is, serving APs, a second value function of deciding on the respective aggregating node 141, 142, 143, e.g., aggregated AP, a third value function of the route, e.g., routing APs and a fourth value function of precoder weights. In some examples, the overall RL approach may be partitioned into multiple agents, where each may agent deal with a specific task, e.g., the group size or the selection of the group itself, and the two or more agents may work collaboratively to solve these tasks. To that end, the agents may exchange their actions so that one agent may make a decision, about the group size, while knowing the decision of the other agent, on the selection of group, and vice- versa. In the same manner, additional agents may be coupled together to collaboratively fine- tune their action-value mappings. As an example, an action of an RL agent may be to determine the number of APs to serve a UE -- 1,2,3 etc. The action may be formulated in many ways, for example it may be a discrete selection of a number, or it may be defined as increase in number of nodes, e.g., APs, decrease and retain the same, or it may also be a continuous space. Alternatively, instead of one agent per task, there may be one agent per configuration category: i) determining the one or more first subsets of nodes 121, 122, 123, that is, the initial grouping, ii) selecting of respective aggregating node141, 142, 143, e.g., AAP selection, iii) determining the route, that is, routing, iv) updating the one or more first subsets of nodes 121, 122, 123, that is, grouping update, and precoding. This way it may be possible to capture e.g., the relationship between group adjustment and precoding, using communication between agents. For the case of determining, out of the set of nodes 120, the one or more first subsets of nodes 121, 122, 123, e.g., AP group selection, the action space may be a continuous value that may identify the first subsets of nodes, e.g., AP group, that may yield the highest reward. An Actor-Critic based approach may be used to solve this problem since such approaches may be understood to deal better with continuous action spaces, as opposed to Deep Q Network (DQN) based techniques which may be understood to be better for discrete action spaces. The RL agents may operate at the first network node 111, which in some embodiments may be the same as the central node 112, a CPU, where the relevant input information may be understood to be available in a collected form. Alternatively, they may be distributed at one or more of the set of nodes 120, e.g., APs, for some tasks that may only require local optimization, e.g., L2 routing. In one example, Monte Carlo Tree Search (MCTS) may be used to optimize on these processing actions jointly. MCTS may be understood to organize the search space in a tree structure and may use a neural network to identify which parts of the tree (solution) may be more likely to yield the highest reward, thus steering the exploration phase in that direction. The RL method may be initialized with a predetermined robust configuration as the start point for automatic optimization. The RL method may be further designed to learn and account for repetitive change patterns in e.g., overall load of the network, e.g., Network (NW) load, number of active devices, e.g., UEs, in different parts of the NW, or UE traffic types, and pro-actively apply changes to the configuration policy without needing to await reward degradation due to an out- of-date policy. Figures 7-11 describe different aspects of embodiments wherein the selected approach may be SAC. Figure 7 schematically illustrates how policy network and value function map in relation to the four optimization variables: i) the one or more first subsets of nodes 121, 122, 123 to serve the one or more devices 130, that is, serving APs, ii) respective aggregating node 141, 142, 143, e.g., aggregated AP, iii) route, e.g., deciding routing APs, and iv) beam weights. Each of boxes 701-704 on the left side of Figure 7 represents a single output of the policy network, wherein m denotes the number of devices. Each of the boxes 708-708 on the right side of Figure 7 represent one output of the value function which may be fed to the policy network. From top to bottom and left to right, the first box 701 represents a single output of the policy deciding on serving APs, the second box 702 represents a single output of the policy deciding on Aggregating AP, the third box 703 represents a single output of the policy deciding on routing APs, the fourth box 704 represents a single output of the policy deciding on precoder weights, the fifth box 705 represents one output of the value function of serving APs, the sixth box 706 represents one output of the value function of Aggregating AP, the seventh box 707 represents one output of the value function of routing APs, and the eight box 708 represents one output of the value function of precoder weights. It may be noted that policy boxes affect also in Figure 8, 9, 10, and value-function boxes only affect in Figure 10. The map is depicted with different patterns in Figures 7-11, so that the first policy deciding on the one or more first subsets of nodes 121, 122, 123 to serve the one or more devices 130, that is, serving APs may map to the first value function output of the one or more first subsets of nodes 121, 122, 123 to serve the one or more devices 130, that is, serving APs. The second policy of deciding on the respective aggregating node 141, 142, 143, e.g., aggregated AP, may map to the second value function of deciding on the respective aggregating node 141, 142, 143, e.g., aggregated AP. The third policy function on deciding the route, e.g., the routing APs, may map to the third value function of the route, e.g., routing APs. The fourth policy function on deciding on the precoding beam weights may map to the fourth value function of precoder weights. Architectures The algorithm used by the first network node 111 to perform this Action 602 may be implemented using different architectures to enhance different aspects of the policy and value networks. In a first group of embodiments, a first architecture, Architecture-1 (A1), may be used, which may be referred to as “Basic unified value unified policy network”. Figure 8 schematically illustrates this architecture, which may comprise a unified value network, wherein the value function may output a single value to the policy network, and unified policy network, wherein the policy function may output one head, output neuron, of the policy network that may represent all four actions (AP, AAP, R, W), which may be achieved by combining all actions into a single large integer. For example, if there are 2 actions for simplicity, and each has 0 or 1 possibilities, then a= 0,1,2,3 may be provided as output, where a=0 may represent action 1 and 2 have 0 values, then a=1 may represent action 1 and 2 have 1, then a=2, may represent action 1 equal 0, action 2 equal 1, and finally a=3, may represent action 1 equal 1, action 2 equal 0. In this first group of embodiments, an action in each of the first, second third and fourth policies may be chosen at 801, that is, a first subset of nodes 121, out of the set of nodes 120, may be chosen, as well as a first respective aggregating node 141, a first route and the precoder weights for a first device 131 of the one or more devices 130. It may be understood that while an AP may serve multiple devices, the precoder ‘w’ may change per device. The chosen actions may be implemented on the environment at 802. This may result in a new state and a reward at 803, and hence in a single generalized value 804 at 805. At 806 the policy, that is, the action representation, may be output to both, the value network, to help in criticizing the value, and to the environment, a simulator or a real system, deciding on AP, AAP, route, precoder. At 807 the value network yields an output to the policy, where it may help the policy making more accurate policy with respect to the current state and reward. The replay buffer may sample the state and reward saved tuples and may provide it as input to value network at 808 and to the policy network as input at 809. Note that this unified value network may also include, the target / main and double network illustrated in basic discrete SAC in Algorithm 1. However, this architecture may be associated with, e.g., a large action space, and an inefficient mixed value function that may connect a single value for all actions. For example, if there are 4 actions, and each has a large range of values, then the policy network may have a very large action space per one head, that is, output neuron, which may cause difficulties in identifying the optimal values among all this very large number. In a second group of embodiments, a second architecture, Architecture-2 (A3), may be used, which may be referred to as “Unified value and branched policies”. Figure 9 schematically illustrates this architecture, which may comprise a unified value network 804 and a branched policy network 901. It may be noted that this unified value network 804 may also include, the target / main and double network as illustrated in basic discrete SAC in Algorithm 1. However, the policy network 901 may also have two parts, one may be related to common policy 902, and the other part may be related to personalized branches per action 903, and each branch may have an individual loss policy function. Hence, this branched and personalized architecture may reduce the action space complexity. In a follow-up group of examples, the training of the common policy 902 may be retrained per all branches, or per each branch. In this second group of embodiments, an action in each of the first, second third and fourth policies may be chosen at 904, as explained for Figure 8. The chosen actions may be implemented on the environment at 905. This may result in a new state and a reward at 906, and hence in a single generalized value 804 at 907. At 908 all actions outputs 701, 702, 703, 704 may be aggregated and fed to the value network to help in improving the value function. At 909 the replay buffer may sample the state and reward saved tuples and provide it as input to value network and the policy network. At 910 the value network may output values per policy given state and reward to the common policy network. In a third group of embodiments, a third architecture, Architecture-3 (A3), may be used, which may be referred to as “Branched value and branched policies”. Figure 10 schematically illustrates this architecture, which may comprise a branched value network and branched policy network 901, and is illustrated in Algorithm 2. The policy network may be understood to have two parts, one may be related to common policy 902, and the other part may be related to personalized branches per action 903, and each branch may have an individual loss policy function. Hence, this branched and personalized architecture may be understood to reduce the action space complexity. In a follow-up example, the training of the common policy may be retrained per all branches, or per each branch. Unlike the second group of embodiments depicted in Figure 9, the third group of embodiments depicted in Figure 10 may also consider a branched value network approach. This may be understood to bring a more advanced relation between the value and corresponding action / states. A unified value network may be understood to bring a single value per all actions, whereas, the branched values depicted in Figure 10, may be understood to bring personalized values 1001 per each action / state pair. Note that, in Algorithm 2, the sub-scription of the cost functions and Neural Network weights ^^^,^^ e.g., may be understood to express the relevant operation per n’s action, that is, all possible actions. In Algorithm 1, when the cost function on all Neural Network weights may be used to backpropagate, it may be a function of all actions at once, because the output of the policy network may be one head, however, in Algorithm 2, the cost function may impact the backpropagation in each individual action network, since it may have multiple output layers or output heads. In this third group of embodiments, an action in each of the first, second third and fourth policies may be chosen at 1002, as explained for Figure 8. The chosen actions may be implemented on the environment at 1003. This may result in a new state and a reward at 1004, and hence in bring personalized values 1001 at 1005. At 1006 the value network heads may yield four outputs, wherein each output may be understood to represent the value of state reward per action policy (AP, AAP, R, w). Those outputs may input to the common policy to impact its action selection process. At 1007, the replay buffer may sample the state and reward saved tuples and provide it as input to value network and the policy network. Initialise ^^^,^: ^ → ℝ|^|, ^^^,^: ^ → ℝ|^|, ^^^: ^ → Initialise main network, [0,1]|^|, ^ ∈ {1, ^} where ^ = |^^^, ^^^^, ^^, ^{^,^}| = 4 represents the four actions. Initialise ^^^^,^: ^ → ℝ|^|, ^^^^,^: ^ → ℝ|^|, ^ ∈ {1, ^}Initialize target network^̅^,^← ^^,^, ^̅^,^← ^^,^Equalize target and main network weights ^ ← ∅ Initialize an empty replay buffer For each iteration do For each environment step do ^^,^~^^^^^^,^^^^) Sample action from the policy ^^^^~^(^^^^|^^ , ^^,^)Sample transition from the environment ^← ^ ∪ ^^^^, ^^,^ , ^^^^, ^^,^^, ^^^^^^Store the transition inthe replay buffer For each gradient step do ^^,^← ^^,^− ^^^∇^^^,^^^^^^,^^ for ^ ∈ {1,2}Update the Q-Functionparameters ^^← ^^− ^^^^^^^^^^(^^) Update policy weights ^^← ^^− ^^^^^ ^^(^^) Update temperature ^^^,^← ^^^,^+ (1 − ^)^^^,^for ^ ∈ {1,2} Update target network weights Output ^^,^, ^^,^, ^^Optimized parameters Algorithm 2. A3 - Branched Value & Policies of Discrete SAC. In a fourth group of embodiments, a fourth architecture, Architecture-4 (A4), may be used, which may be referred to as “Unified value and independent policies”. Figure 11 schematically illustrates this architecture, which may comprise a unified value network 804 and independent policies 1101. This approach may comprise completely independent policies 1102, that is an independent actor network per each action, fed by single value function 804 for all actions. This approach may be understood to have similarities to the hierarchical RL algorithm. In this fourth group of embodiments, values of the four actions per state / reward in each of the first, second third and fourth policies may be chosen at 1103, as explained for Figure 8. The chosen actions may be implemented on the environment at 1104. This may result in a new state and a reward at 1105, and hence in a generalized value 1106 at 1107. At both 1108 and 1103 a common value function may be fed as input to the individual policy network per each action. At 1109 the replay buffer may sample the state and reward saved tuples and provide it as input to value network and the policy network. In a fifth group of embodiments, a fifth architecture, may comprise an independent value and a policy network, that is, a parallel agent per each action. The approaches followed by the groups of embodiments described above may differ in their complexity / size, which may be used as the highest-level deployment-dependent method choice criterion, based on available computing resources and decision time budget, as well as training availability. Next, specific policy options that may be enabled in the system may probably be viewed as system- or hyperparameters that may need experimentation, where domain knowledge may be used to identify reasonable starting points. In one group of examples, the combination of the different reward segments may be realized via: 1) Aggregation using fixed weights and 2) Optimization over reward’s components and corresponding weights. This may be done via Lagrangian weights optimization. Reward segments may be used to produce agents of different skill, e.g., beginner, intermediate, expert. This may be created via different policy functions which may have different arch and may be more complex to use, e.g., require more hardware. As such, by reward segmentation skill may be forfeited in the interest of lower compute footprint. In another example, the AI model may be non-RL and trained using traditional batch learning, using as a loss function e.g., estimated / simulated degradation of the above metrics compared to conventional radio-optimized AP grouping with unconstrained routing. Such trained model may also be used as a starting point for RL-based improvement in a physical deployment. In a different example, the first network node 111 may make use of Stacked Contextual Bandits (SCB) for the purpose of solving the same problem – the SCB approach may be understood to not employ any neural networks and may operate in a discrete space thus allowing for a less computationally expensive approach, where every bandit may make use of a context, or state space as mentioned in the previous approach, and the decisions / choices of the previous bandits to optimize its decision while exploring using the Upper Confidence Bound (UCB) algorithm. As illustrated in the schematic block diagram of Figure 12, which depicts an SCB overview, 4 bandits may be considered in this approach; Bandit 1 (B1), Bandit 2 (B2), Bandit 3 (B3) and Bandit 4 (B4). B1 may be tasked with determining, the one or more first subsets of nodes 121, 122, 123, e.g., selecting the AP group which may be going to serve a given device, e.g., UE, or a group of devices. Given that this may be understood to a bandit, the choices of the one or more first subsets of nodes 121, 122, 123 may be fixed. Despite the rigidness of this approach, it may allow to control the complexity of the implementation by enabling to adjust how many different options there may be, for example, the list of choices may be AP_GROUP1, AP_GROUP2, AP_GROUP3, that is, 3 choices, where every choice may have a specific set of APs involved. Stage 1: At 1201, the input to B1 may then be the state {s} as previously described and the UE or UE_GROUP. The output may then be one of the one or more first subsets of nodes 121, 122, 123, e.g., AP_GROUP. Devices, e.g., UEs, may be grouped in different ways, most predominantly by their location but also, e.g., on their capabilities, for example, type of traffic such as e.g., Ultra Reliable and Low Latency Communications (URLLC) vs Mobile broadband (MBB) vs Machine Type Communication (MTC). Stage 2: In stage 2, at 1202, the contextual bandit B2 may be tasked with choosing the respective aggregating node 141, 142, 143, e.g., AAP, and using as input the output of B1 and the state {s}. B2 may choose between specific a set of respective aggregating nodes 141, 142, 143, such as APPs, e.g., AAP1, APP2, etc. Stage 3: In stage 3, at 1203, contextual bandit B3 may use as input state {s}, the output of B1 and the output of B2. B3 may choose L1 & L2 routing as such and it may choose from a predefined set of routes ROUTE_1, ROUTE_2, … ROUTE_3. At this stage, the first network node 111 may observe the total reward of the process, which may be monotonically increasing – if that is not the case, the system may be allowed to perform another selection of respective aggregating nodes 141, 142, 143, e.g., AP groups, and respective aggregating nodes 141, 142, 143, e.g., AAPs, and compare the new reward with the previous – if new reward is better, the first network node 111 may accept the new selection. Otherwise, the first network node 111 may proceed with the previous choice. Stage 4: Finally, at 1204, in stage 4 precoding may be selected, motivated by the previous decisions of B1, B2 and B3 and state {s}. The exploitation / exploration problem in this case may be formalized as follows: For a set of discrete trials t = 1,2,3, ... ∈ T and set of bandits Bi ∈ B 1. Observe bandit bi,t and a set Ai,t of actions, or choices as mentioned previously, with their feature vectors xt,aifor ai∈ Ai,t. xt,aimay be understood to hold the association between the bandit (bi) and the action of bandit I for action a. 2. For every choice, the bandit may receive a reward rt,ai,t3. The algorithm may be tasked to find the maximum reward for every bandit given the previous choice of the bandit. This finite set of choices may allow the first network node 111 to control the complexity of the process, thus tuning it to the number of available resources, but also the requirement of the data that may need to be collected. Examples of expected behaviors of the trained system Sizes of the one or more first subsets of nodes 121, 122, 123, e.g., AP group sizes The average path lengths of the first route, L1, and the second route, L2, may be to some extent controlled by e.g., selecting the sizes of the serving one or more first subsets of nodes 121, 122, 123, e.g., the serving AP group sizes, where larger size may help L1 success but may degrade L2 success. In one example, one or more sizes of the one or more first subsets of nodes 121, 122, 123, e.g., group sizes, may be selected larger if connecting the first nodes, e.g., APs, e.g., avoiding connection failure, may be performance-critical or prioritized, when routing resources may be limited. Limited routing resources may correspond to a relatively large number of simultaneously scheduled UEs, combined with relatively limited segment capacities. In one example, one or more sizes of the one or more first subsets of nodes 121, 122, 123, e.g., group sizes, may be selected smaller if connecting additional nodes, e.g., APs or higher aggregate SINR for the device may be performance critical or prioritized, when routing resources may be limited. In one example, the baseline or minimum number of nodes, e.g., APs, may be predetermined, and the first network node 111 may increase the group size for one or more devices, as necessary. In one example, the size of the one or more first subsets of nodes 121, 122, 123, e.g., group size, may be increased if this may improve the probability of connecting the respective aggregating node 141, 142, 143, e.g., AAP. That is, reduce the risk of routing failure for the AAP. In one example, the size of the one or more first subsets of nodes 121, 122, 123, e.g., group size, may be increased if the routing constraints are not likely to reduce the actual connection ratio of additional APs. In one example, the size of the one or more first subsets of nodes 121, 122, 123, e.g., AP group sizes, may be adjusted for certain parts of the FH network due to routing bottlenecks. In an alternative example, when designing the system and selecting the characteristics of the respective aggregating node 141, 142, 143, e.g., AAP, the first network node 111 may need to consider the routing robustness due to the closeness to the central node 112, e.g., the CPU. If a particular node of the set of node 120, e.g., which may be referred to as a Gateway, is close to the CPU, the other traffic of the respective aggregating node 141, 142, 143 may need to pass through it, hence more congestion may be understood to occur there. Therefore, more communication and computation resources may need to be allocated to that gateway node, e.g., AP. Selection of the respective aggregating node 141, 142, 143, e.g., AAP selection There may exist different selection criteria for the respective aggregating node 141, 142, 143, such as closest to CPU, best channel to the device, evenly distributed, etc. They may provide different trade-offs between routing success for the device population and the link quality and radio system performance. In one example, the respective aggregating nodes 141, 142, 143 for one or more groups may be selected as the closest node to the CPU in the group if connecting the first nodes, e.g, APs, e.g., avoiding connection failure, may be performance-critical or prioritized, when routing resources may be limited. Limited routing resources may correspond to a relatively large number of simultaneously scheduled UEs, combined with relatively limited segment capacities. In one example, the respective aggregating nodes 141, 142, 143 for one or more groups may be assigned the node with the best SINR with respect to the device if connecting additional nodes or higher aggregate SINR for the device may be performance critical or prioritized, when routing resources may be limited. In one example, if the respective aggregating nodes 141, 142, 143 also perform weight computations, the above principles may be combined with e.g., per-node processing constraints. The assignment of the respective aggregating nodes 141, 142, 143 may e.g., be distributed over the available set of nodes, where each node may not serve as a respective aggregating node 141, 142, 143 for more than a predetermined maximum number of devices. By determining the one or more parameters of the configuration of the at least two processing actions to serve the one or more devices 130 in this Action 602, wherein the determining 602 may be performed by jointly optimizing the one or more parameters, the first network node 111 may enable that the one or more devices 130 may be served with increased system throughput and user throughput, as well as reduced scheduling latency. Moreover, the search space for the subsequent steps may be reduced, since the nodes, e.g., APs, may be known and therefore the next step may then be computed faster Action 603 In this Action 603, the first network node 111 may initiate sequential execution of the at least two processing actions applying the determined configuration. Initiating may be understood as triggering, enabling, facilitating or starting. Sequential execution may be understood to mean that the first network node 111 may initiate execution of one processing action, and then proceed to the next, which may be affected by the first one, and so on. To initiate sequential execution in this Action 603 may comprise implementing the processing actions on the communications network 100 using the determined configuration. That is, performing actions of the processing sequence according to the determined configuration parameters, e.g., by configuring the set of nodes 120 accordingly. In some embodiments this may be performed, for example, in the embodiments wherein the first network node 111 may be a different node than the central node 112, providing the output of Action 602 to the central node 112, e.g., via the respective second link 152. The output of action 603 may comprise: i) the one or more first subsets of nodes 121, 122, 123 to serve the one or more devices 130, wherein each first subset of the one or more first subsets of nodes 121, 122, 123 may be determined to serve the respective second subset of devices 131, 132, 133 of the one or more devices 130; this output may comprise, e.g., the initial AP grouping (^^^), e.g., AP group sizes per UE or UE cluster, common or individual values per UE / cluster, constraints on overlap of AP groups, etc..; ii) the respective aggregating node 141, 142, 143 for each subset of the one or more first subsets of nodes 121, 122, 123; this output may comprise, e.g., the AAP selection (^^^^), e.g., AAP selection criterion, constraints on the number of AAPs in the system; iii) the route between the central node 112 and each of the nodes in the respective first subset of nodes 121, 122, 123; this output may comprise, e.g., the routing (^^), e.g., routing strategy, connect all possible APs, connect a minimum set of APs,…, max routing path length, number of routing path options considered, etc.; iv) the updating the one or more first subsets of nodes 121, 122, 123; this output may comprise, e.g., the serving AP group adjustment, e.g., adjustment strategy, such as remove unconnected APs, redo initial grouping and / or routing, …; and v) the precoding weights for the nodes in the determined or updated one or more first subsets of nodes 121, 122, 123; this output may comprise, e.g., the precoding weight computation (^^) may be a matrix of , where ^ ∈ |^^|; this output may comprise, e.g., the weight structure, such as MF, ZF, LMMSE, … As stated earlier, the determining in Action 602 of the one or more parameters may be performed using AI methods. The AI methods may comprise performing the method in Actions 601, 602, 603 iteratively during a training phase, during which a model to predict the one or more parameters may be trained with a first set of data gathered during a first set of iterations, and an execution phase, during which the trained model may be used over one or more second iterations. By initiating the sequential execution of the of the at least two processing actions applying the determined configuration in this Action 603, the first network node 111 may enable that the one or more devices 130 may be served, e.g., to send the one or more packets between the one or more devices 130 and the central node 112, with increased system throughput and user throughput, as well as reduced scheduling latency, while simplifying the system design and reducing the cost and overall complexity of the optimization process. Action 604 In this Action 604, the first network node 111 may initiate sending the one or more packets between the one or more devices 130 and the central node 112 using the sequential execution of the at least two processing actions applying the determined configuration. Initiating may be understood as starting, triggering, enabling or facilitating. Sending may comprise, e.g., providing or outputting, or transmitting, e.g., via the first link 151 and then the respective second link 152 and one or more respective third links 153. The sending the one or more packets between the one or more devices 130 and the central node 112 may be in the downlink or in the uplink. The initiating in this Action 604 of the sending may comprise transmitting the respective one or more packets using the determined precoding weights. By initiating the one or more packets between the one or more devices 130 and the central node 112 in this Action 604, the first network node 111 may enable that the one or more devices 130 may be served to send the one or more packets between the one or more devices 130 and the central node 112, with increased system throughput and user throughput, as well as reduced scheduling latency, shorter round trip and less packet drop, while simplifying the system design and reducing the cost and overall complexity of the optimization process. As a summarized overview of the foregoing, embodiments herein may be understood to relate to a method in a D-MIMO network with a segmented fronthaul network for collaborative optimization of a processing sequence comprising two or more of initial grouping, AAP selection, routing, and grouping update. The method may comprise: a) obtaining one or more NW scenario parameters, b) jointly optimizing, based on the NW scenario parameters, configuration parameters for two or more actions in the processing sequence, c) performing steps of the processing sequence according to the determined configuration parameters. In some examples, the collaborative optimization may comprise joint optimization, sequential optimization. In some examples, the first network node 111 may use RL agents, e.g., Discrete Soft Actor Critic and Contextual Bandit, with single / multiple policy and / or value networks per task / tasks. Certain embodiments disclosed herein may provide one or more of the following technical advantage(s), which may be summarized as follows. Embodiments herein may enable to jointly optimize the parameters and settings used to execute the steps that may be performed in dynamic D-MIMO signal transmission, so that non-trivial impact of a combination of multiple parameter choices in different operating scenarios may be captured. Replacing joint execution with joint parameter optimization and sequential execution may yield significant complexity reduction at moderate performance impact, and large performance improvement over legacy, non-coordinated parameter settings. The RL-based approach may also adapt the settings as the scenarios may change over time. As a result, the communications network 100, e.g., a D-MIMO system, may be enabled to operate at increased system throughput and user throughput and reduced scheduling latency. Figure 13 depicts a non-limiting example of the arrangement that the first network node 111 may comprise. The first network node 111 may be understood to be for handling the set of nodes 120 configured to be connected via the segmented front-haul. The first network node 111 is configured to operate in the communications network 100 configured to comprise the set of nodes 120. Several embodiments are comprised herein. It should be noted that the examples herein are not mutually exclusive. One or more embodiments may be combined, where applicable. All possible combinations are not described to simplify the description. Components from one embodiment may be tacitly assumed to be present in another embodiment and it will be obvious to a person skilled in the art how those components may be used in the other exemplary embodiments. The detailed description of some of the following corresponds to the same references provided above, in relation to the actions described for the first network node 111, and will thus not be repeated here. For example, in some embodiments, the communications network 100 may be configured to be, e.g., a distributed MIMO network, with the segmented fronthaul, where fronthaul / backhaul / transport may be provided to the central node 112 or central unit, e.g., a CPU, that may connect to a segmented interconnected fronthaul network that may connect to the set of nodes 120, e.g., individual APs. The first network node 111 is configured to obtain the first information configured to indicate the one or more conditions under which the one or more devices 130 are to be served by at least one of the set of nodes 120 in the communications network 100. The first network node 111 is also configured to determine, based on the first information configured to be obtained, the one or more parameters of the configuration of the at least two processing actions to serve the one or more devices 130. The processing actions are configured to comprise: i) the determining, out of the set of nodes 120, the one or more first subsets of nodes 121, 122, 123 to serve the one or more devices 130. Each first subset of the one or more first subsets of nodes 121, 122, 123 is configured to be determined to serve the respective second subset of devices 131, 132, 133 of the one or more devices 130. The processing actions are also configured to comprise: ii) the selecting the respective aggregating node 141, 142, 143 for each subset of the one or more first subsets of nodes 121, 122, 123. The processing actions are further configured to comprise: iii) the determining, for the one or more packets to be exchanged between the one or more devices 130 and the central node 112 to which all the respective aggregating nodes 141, 142, 143 are configured to be connected, the central node 112 being configured to provide the backhaul connection to the core network 115 of the communications network 100, the route. The route is between the central node 112 and each of the nodes in the respective first subset of nodes 121, 122, 123. The processing actions are additionally configured to comprise: iv) the updating the one or more first subsets of nodes 121, 122, 123 based on the measure of fulfilment of the one or more conditions. The processing actions are also configured to comprise: v) the determining the precoding weights for the nodes in the one or more first subsets of nodes 121, 122, 123 configured to be determined or updated. The determining of the one or more parameters is configured to be performed by jointly optimizing the one or more parameters. The first network node 111 is also configured to initiate sequential execution of the at least two processing actions applying the configuration configured to be determined. In some embodiments, the one or more conditions may be configured to comprise the one or more of: i) the radio related conditions, ii) the one or more indicators of traffic of the devices of the one or more devices 130, iii) the first load of the transport network of the communications network 100, iv) the second load of the radio access network of the communications network 100, v) the one or more routing constraints, vi) the one or more processing constraints of the set of nodes 120, vii) the operating frequency of the bandwidth configured to be used by the communications network 100, and viii) the information on the mobility of the one or more devices 130. In some embodiments, the determining of the one or more parameters may be configured to be performed using artificial intelligence methods. The artificial intelligence methods may be configured to comprise performing the method iteratively during the training phase, during which the model to predict the one or more parameters may be configured to be trained with the first set of data configured to be gathered during the first set of iterations, and the execution phase, during which the trained model may be configured to be used over the one or more second iterations. In some embodiments, the artificial intelligence methods may be configured to comprise using the one or more reinforcement learning agents using: a) the approach configured to be selected from: i) Discrete Soft Actor Critic, and ii) Contextual Bandit, and wherein the approach is Discrete SAC, b) the one or more policy functions, and c) the one or more value functions. In some embodiments, the artificial intelligence methods may be configured to comprise using the plurality of reinforcement learning agents. The determining of the one or more parameters of the at least two processing actions may be configured to be performed by the collaborative optimization among the plurality of reinforcement learning agents. In some embodiments, the joint optimization may be configured to be based on the improvement of the one or more metrics configured to be selected from: the system throughput, the connection failure rate of the one or more devices 130, the signal to interference plus noise ratio, the transmission delay, the transmission diversity or transmission reliability, the quality of service, and the quality of experience. The first network node 111 may be configured to initiate the sending the one or more packets between the one or more devices 130 and the central node 112 using the sequential execution of the at least two processing actions applying the configuration configured to be determined. In some embodiments, the initiating of the sending may be configured to comprise transmitting the respective one or more packets using the precoding weights configured to be determined. In some embodiments, at least one of the following may apply: a) the route may be configured to comprise the first route between the central node 112 and the respective aggregating node 141, 142, 143 configured to be selected and the second route between the respective aggregating node 141, 142, 143 configured to be selected and each of the nodes in the respective first subset of nodes 121, 122, 123; b) at least two of the devices within at least one of the respective second subset of devices 131, 132, 133 may be configured to be co- scheduled; c) each of the nodes in the set of nodes 120 may be configured to be a TRP or an AP; d) the communications network 100 may be configured to be the D-MIMO network; and e) the central node 112 may be configured to be the CPU of the D-MIMO network. The embodiments herein in the first network node 111 may be implemented through one or more processors, such as a processing circuitry 1301 in the first network node 111 depicted in Figure 13, together with computer program code for performing the functions and actions of the embodiments herein. A processor, as used herein, may be understood to be a hardware component. The program code mentioned above may also be provided as a computer program product, for instance in the form of a data carrier carrying computer program code for performing the embodiments herein when being loaded into the first network node 111. One such carrier may be in the form of a CD ROM disc. It is however feasible with other data carriers such as a memory stick. The computer program code may furthermore be provided as pure program code on a server and downloaded to the first network node 111. The first network node 111 may further comprise a memory 1302 comprising one or more memory units. The memory 1302 is arranged to be used to store obtained information, store data, configurations, schedulings, and applications etc. to perform the methods herein when being executed in the first network node 111. In some embodiments, the first network node 111 may receive information from, e.g., any node in the set of nodes 120, the central node 112 and the one or more devices 130, and / or another structure in the communications network 100, through a receiving port 1303. In some embodiments, the receiving port 1303 may be, for example, connected to one or more antennas in first network node 111. In other embodiments, the first network node 111 may receive information from another structure in the computer system 100 through the receiving port 1303. Since the receiving port 1303 may be in communication with the processing circuitry 1301, the receiving port 1303 may then send the received information to the processing circuitry 1301. The receiving port 1303 may also be configured to receive other information. The processing circuitry 1301 in the first network node 111 may be further configured to transmit or send information to e.g., any node in the set of nodes 120, the central node 112 and the one or more devices 130, and / or another structure in the communications network 100, through a sending port 1304, which may be in communication with the processing circuitry 1301, and the memory 1302. Those skilled in the art will also appreciate that the units comprised within the first network node 111 described above as being configured to perform different actions, may refer to a combination of analog and digital circuits, and / or one or more processors configured with software and / or firmware, e.g., stored in memory, that, when executed by the one or more processors such as the processing circuitry 1301, perform as described above. One or more of these processors, as well as the other digital hardware, may be included in a single Application-Specific Integrated Circuit (ASIC), or several processors and various digital hardware may be distributed among several separate components, whether individually packaged or assembled into a System-on-a-Chip (SoC). Also, in some embodiments, the different units comprised within the first network node 111 described above as being configured to perform different actions described above may be implemented as one or more applications running on one or more processors such as the processing circuitry 1301. Thus, the methods according to the embodiments described herein for the first network node 111 may be respectively implemented by means of a computer program 1305 product, comprising instructions, i.e., software code portions, which, when executed on at least one processing circuitry 1301, cause the at least one processing circuitry 1301 to carry out the actions described herein, as performed by the first network node 111. The computer program 1305 product may be stored on a computer-readable storage medium 1306. The computer- readable storage medium 1306, having stored thereon the computer program 1305, may comprise instructions which, when executed on at least one processing circuitry 1301, cause the at least one processing circuitry 1301 to carry out the actions described herein, as performed by the first network node 111. In some embodiments, the computer-readable storage medium 1306 may be a non-transitory computer-readable storage medium, such as a CD ROM disc, or a memory stick. In other embodiments, the computer program 1305 product may be stored on a carrier containing the computer program 1305 just described, wherein the carrier is one of an electronic signal, optical signal, radio signal, or the computer-readable storage medium 1306, as described above. The first network node 111 may comprise a communication interface configured to facilitate, or an interface unit to facilitate, communications between the first network node 111 and any node in the set of nodes 120, the central node 112 and the one or more devices 130, and / or another structure in the communications network 100. The interface may, for example, include a transceiver configured to transmit and receive radio signals over an air interface in accordance with a suitable standard. In other embodiments, the first network node 111 may comprise a radio circuitry 1307, which may comprise e.g., the receiving port 1303 and the sending port 1304. The radio circuitry 1307 may be configured to set up and maintain at least a wireless connection with any node in the set of nodes 120, the central node 112 and the one or more devices 130, and / or another structure in the communications network 100. Circuitry may be understood herein as a hardware component. Hence, embodiments herein also relate to the first network node 111 operative to operate in the computer system 100. The first network node 111 may comprise the processing circuitry 1301 and the memory 1302, said memory 1302 containing instructions executable by said processing circuitry 1301, whereby the first network node 111 is further operative to perform the actions described herein in relation to the first network node 111, e.g., in Figure 6 and / or any of Figures 7-12. As used herein, the expression “at least one of:” followed by a list of alternatives separated by commas, and wherein the last alternative is preceded by the “and” term, may be understood to mean that only one of the list of alternatives may apply, more than one of the list of alternatives may apply or all of the list of alternatives may apply. This expression may be understood to be equivalent to the expression “at least one of:” followed by a list of alternatives separated by commas, and wherein the last alternative is preceded by the “or” term. When using the word "comprise" or “comprising” it shall be interpreted as non- limiting, i.e., meaning "consist at least of". A processor may be understood herein as a hardware component. The embodiments herein are not limited to the above described preferred embodiments. Various alternatives, modifications and equivalents may be used. Therefore, the above embodiments should not be taken as limiting the scope of the invention. REFERENCES [1] Matched-Decision AP Selection for User-Centric Cell-Free Massive MIMO Networks, M Freitas, D Souza, DB da Costa, G Borges, AM Cavalcante … - IEEE Transactions on Vehicular Technology, 2023 [2] Cell-Free Massive MIMO With Segmented Fronthaul: Reliability and Protection Aspects, ALP Fernandes, DD Souza, DB da Costa, AM Cavalcante, JCWA Costa, IEEE Wireless Communications Letters 11 (8), 1580-1584 [3] Distributed MIMO Precoding with Routing Constraints in Segmented Fronthaul, J Sadreddini, Ö Haliloglu, A Reial, submitted to IEEE ICC 2023 Workshop - Industrial Private 5G-and-beyond Wireless Network. [4] Petros Christodoulou, “Soft Actor-Critic for Discrete Action Settings”, https: / / arxiv.org / pdf / 1910.07207.pdf

Claims

CLAIMS:

1. A method performed by a first network node (111), the method being for handling a set of nodes (120) connected via a segmented front-haul, the first network node (111) operating in a communications network (100) comprising the set of nodes (120), the method comprising: − obtaining (601) first information indicating one or more conditions under which one or more devices (130) are to be served by at least one of the set of nodes (120) in the communications network (100), − determining (602), based on the obtained first information, one or more parameters of a configuration of at least two processing actions to serve the one or more devices (130), the processing actions comprising: i. determining, out of the set of nodes (120), one or more first subsets of nodes (121, 122, 123) to serve the one or more devices (130), wherein each first subset of the one or more first subsets of nodes (121, 122, 123) is determined to serve a respective second subset of devices (131, 132, 133) of the one or more devices (130), ii. selecting a respective aggregating node (141, 142, 143) for each subset of the one or more first subsets of nodes (121, 122, 123), iii. determining, for one or more packets to be exchanged between the one or more devices (130) and a central node (112) to which all the respective aggregating nodes (141, 142, 143) are connected, the central node (112) providing a backhaul connection to a core network (115) of the communications network (100), a route between the central node (112) and each of the nodes in the respective first subset of nodes (121, 122, 123), iv. updating the one or more first subsets of nodes (121, 122, 123) based on a measure of fulfilment of the one or more conditions, and v. determining precoding weights for the nodes in the determined or updated one or more first subsets of nodes (121, 122, 123), wherein the determining (602) of the one or more parameters is performed by jointly optimizing the one or more parameters, and − initiating (603) sequential execution of the at least two processing actions applying the determined configuration.

2. The method according to claim 1, wherein the one or more conditions comprise one or more of:i. radio related conditions, ii. one or more indicators of traffic of the devices of the one or more devices (130), iii. a first load of a transport network of the communications network (100), iv. a second load of a radio access network of the communications network (100), v. one or more routing constraints, vi. one or more processing constraints of the set of nodes (120), vii. operating frequency of a bandwidth used by the communications network (100), and viii. information on mobility of the one or more devices (130).

3. The method according to any of claims 1-2, wherein the determining (602) of the one or more parameters is performed using artificial intelligence methods, and wherein the artificial intelligence methods comprise performing the method (601, 602, 603) iteratively during a training phase, during which a model to predict the one or more parameters is trained with a first set of data gathered during a first set of iterations, and an execution phase, during which the trained model is used over one or more second iterations.

4. The method according to claim 3, wherein the artificial intelligence methods comprise using one or more reinforcement learning agents using: a) an approach selected from: i. Discrete Soft Actor Critic, SAC, and ii. Contextual Bandit, and wherein the approach is Discrete SAC, b) one or more policy functions, and c) one or more value functions.

5. The method according to claim 4, wherein the artificial intelligence methods comprise using a plurality of reinforcement learning agents, and wherein the determining (602) of the one or more parameters of the at least two processing actions is performed by a collaborative optimization among the plurality of reinforcement learning agents.

6. The method according to any of claims 1-5, wherein joint optimization is based on an improvement of one or more metrics selected from: − system throughput,− connection failure rate of the one or more devices (130), − signal to interference plus noise ratio, − transmission delay, − a transmission diversity or transmission reliability, − quality of service, and − quality of experience.

7. The method according to any of claims 1-6, further comprising: − initiating (604) sending the one or more packets between the one or more devices (130) and the central node (112) using the sequential execution of the at least two processing actions applying the determined configuration.

8. The method according to claim 7, wherein the initiating (604) of the sending comprises transmitting the respective one or more packets using the determined precoding weights.

9. The method according to any of claims 1-8, wherein at least one of: a) the route comprises a first route between the central node (112) and the selected respective aggregating node (141, 142, 143) and a second route between the selected respective aggregating node (141, 142, 143) and each of the nodes in the respective first subset of nodes (121, 122, 123), b) at least two of the devices within at least one of the respective second subset of devices (131, 132, 133) are co-scheduled, c) each of the nodes in the set of nodes (120) is a Transmission Reception Point, TRP, d) the communications network (100) is a distributed Multiple Input Multiple Output, D-MIMO, network, and e) the central node (112) is a central processing unit, CPU, of the D-MIMO network.

10. A first network node (111), for handling a set of nodes (120) configured to be connected via a segmented front-haul, the first network node (111) being configured to operate in a communications network (100) configured to comprise the set of nodes (120), the first network node (111) being further configured to: − obtain first information configured to indicate one or more conditions under which one or more devices (130) are to be served by at least one of the set of nodes (120) in the communications network (100),− determine, based on the first information configured to be obtained, one or more parameters of a configuration of at least two processing actions to serve the one or more devices (130), the processing actions being configured to comprise: i. determining, out of the set of nodes (120), one or more first subsets of nodes (121, 122, 123) to serve the one or more devices (130), wherein each first subset of the one or more first subsets of nodes (121, 122, 123) is configured to be determined to serve a respective second subset of devices (131, 132, 133) of the one or more devices (130), ii. selecting a respective aggregating node (141, 142, 143) for each subset of the one or more first subsets of nodes (121, 122, 123), iii. determining, for one or more packets to be exchanged between the one or more devices (130) and a central node (112) to which all the respective aggregating nodes (141, 142, 143) are configured to be connected, the central node (112) being configured to provide a backhaul connection to a core network (115) of the communications network (100), a route between the central node (112) and each of the nodes in the respective first subset of nodes (121, 122, 123), iv. updating the one or more first subsets of nodes (121, 122, 123) based on a measure of fulfilment of the one or more conditions, and v. determining precoding weights for the nodes in the one or more first subsets of nodes (121, 122, 123) configured to be determined or updated, wherein the determining of the one or more parameters is configured to be performed by jointly optimizing the one or more parameters, and − initiate sequential execution of the at least two processing actions applying the configuration configured to be determined.

11. The first network node (111) according to claim 10, wherein the one or more conditions are configured to comprise one or more of: i. radio related conditions, ii. one or more indicators of traffic of the devices of the one or more devices (130), iii. a first load of a transport network of the communications network (100), iv. a second load of a radio access network of the communications network (100), v. one or more routing constraints,vi. one or more processing constraints of the set of nodes (120), vii. operating frequency of a bandwidth configured to be used by the communications network (100), and viii. information on mobility of the one or more devices (130).

12. The first network node (111) according to any of claims 10-11, wherein the determining of the one or more parameters is configured to be performed using artificial intelligence methods, and wherein the artificial intelligence methods are configured to comprise performing the method iteratively during a training phase, during which a model to predict the one or more parameters is configured to be trained with a first set of data configured to be gathered during a first set of iterations, and an execution phase, during which the trained model is configured to be used over one or more second iterations.

13. The first network node (111) according to claim 12, wherein the artificial intelligence methods are configured to comprise using one or more reinforcement learning agents using: a) an approach configured to be selected from: i. Discrete Soft Actor Critic, SAC, and ii. Contextual Bandit, and wherein the approach is Discrete SAC, b) one or more policy functions, and c) one or more value functions.

14. The first network node (111) according to claim 13, wherein the artificial intelligence methods are configured to comprise using a plurality of reinforcement learning agents, and wherein the determining of the one or more parameters of the at least two processing actions is configured to be performed by a collaborative optimization among the plurality of reinforcement learning agents.

15. The first network node (111) according to any of claims 10-14, wherein joint optimization is configured to be based on an improvement of one or more metrics configured to be selected from: − system throughput, − connection failure rate of the one or more devices (130), − signal to interference plus noise ratio, − transmission delay,− a transmission diversity or transmission reliability, − quality of service, and − quality of experience.

16. The first network node (111) according to any of claims 10-15, being further configured to: − initiate sending the one or more packets between the one or more devices (130) and the central node (112) using the sequential execution of the at least two processing actions applying the configuration configured to be determined.

17. The first network node (111) according to claim 16, wherein the initiating of the sending is configured to comprise transmitting the respective one or more packets using the precoding weights configured to be determined.

18. The first network node (111) according to any of claims 10-17, wherein at least one of: a) the route is configured to comprise a first route between the central node (112) and the respective aggregating node (141, 142, 143) configured to be selected and a second route between the respective aggregating node (141, 142, 143) configured to be selected and each of the nodes in the respective first subset of nodes (121, 122, 123), b) at least two of the devices within at least one of the respective second subset of devices (131, 132, 133) are configured to be co-scheduled, c) each of the nodes in the set of nodes (120) is configured to be a Transmission Reception Point, TRP, d) the communications network (100) is configured to be a distributed Multiple Input Multiple Output, D-MIMO, network, and e) the central node (112) is configured to be a central processing unit, CPU, of the D-MIMO network.

19. A computer program (1305), comprising instructions which, when executed on at least one processing circuitry (1301), cause the at least one processing circuitry (1301) to carry out the method according to any of claims 1-9.

20. A computer-readable storage medium (1306), having stored thereon a computer program (1305), comprising instructions which, when executed on at least one processing circuitry (1301), cause the at least one processing circuitry (1301) to carry out the method according to any of claims 1-9.

Images