Selection of power control parameters

Abstract

A solution for power control parameter selection is proposed. The solution comprises determining (300) a selection of radio access network parameters that have an impact on uplink power control from a set of given radio access network parameters, training a neural network (302) using the selection of radio access network parameters as input to determine the uplink power control parameters, and obtaining (304) a set of initial uplink power control parameters as output using the trained neural network with the selection of radio access network parameters as input.

Description

Technical Field

[0001] The exemplary and non-limiting embodiments of the present invention generally relate to wireless communication systems. The exemplary and non-limiting embodiments of the present invention particularly relate to apparatus and methods in wireless communication networks. Background Art

[0002] In wireless telecommunication systems, the control of transmission power is an important factor. Wireless transmissions of mobile terminals and network elements often cause interference to other transmissions of the network. Furthermore, in battery-operated mobile terminals, power consumption is an important factor. However, the transmission power should not be too low, as the transmission quality of the wireless connection may be degraded. Therefore, efficient power control mechanisms are constantly being developed. Summary of the invention

[0003] In order to provide a basic understanding of some aspects of the present invention, a simplified summary of the present invention is presented below. This summary is not an extensive overview of the present invention. It is not intended to identify the key / essential elements of the present invention, nor is it intended to define the scope of the present invention. Its sole purpose is to present some concepts of the present invention in a simplified form as a prelude to a more detailed description presented later.

[0004] According to one aspect of the present invention, the apparatus of claims 1 and 6 is provided.

[0005] According to one aspect of the present invention, the system of claim 8 is provided.

[0006] According to one aspect of the present invention, the method of claims 10 and 15 is provided.

[0007] According to an aspect of the invention, a computer program comprising instructions of claims 17 and 18 is provided.

[0008] One or more examples of implementation are set forth in more detail in the drawings and the following description. Other features will be clear from the description and drawings and from the claims. The embodiments and / or examples and features described in this specification that do not fall within the scope of the independent claims (if any) are to be interpreted as examples useful for understanding the various embodiments of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Embodiments of the present invention are described below by way of example only with reference to the accompanying drawings, in which:

[0010] Figure 1 and Figure 2 An example of a simplified system architecture for a communication system is shown;

[0011] Figure 3A and Figure 3B is a flow chart illustrating an embodiment of the present invention;

[0012] Figure 3C and Figure 3D An example of a neural network is shown;

[0013] Figure 4 An example of determining selection of radio access network parameters is shown;

[0014] Figure 5 An example of training of a neural network is shown;

[0015] Figure 6 An example of providing initial uplink power control parameters is shown;

[0016] Figure 7 An example of fine-tuning uplink power control parameters is shown; and

[0017] Fig. 8A , Figure 8B and Figure 8C A simplified example of an apparatus to which some embodiments of the invention apply is shown. DETAILED DESCRIPTION

[0018] The following embodiments are merely examples. Although the specification may refer to "an", "one" or "some" embodiments in several places, this does not necessarily mean that each such reference refers to the same embodiment, nor does it necessarily mean that the feature is only applicable to a single embodiment. Individual features of different embodiments may also be combined to provide other embodiments. In addition, the words "comprising" and "including" should be understood as not limiting the described embodiments to only those features that have been mentioned, and such embodiments may also include features, structures, units, modules, etc. that are not specifically mentioned.

[0019] Some embodiments of the present invention are applicable to user terminals, communication devices, base stations, eNodeBs, gNodeBs, distributed implementations of base stations, network elements of communication systems, corresponding components and / or any communication system or any combination of different communication systems that support the required functionality.

[0020] The protocols, communication systems, servers and user equipment used, especially in wireless communications, are evolving rapidly. Such developments may require additional changes to the embodiments. Therefore, all words and expressions should be interpreted broadly and are intended to illustrate rather than limit the embodiments.

[0021] In the following, different exemplary embodiments will be described using a radio access architecture based on Advanced Long Term Evolution (Advanced LTE, LTE-A) or New Radio (NR, 5G) as an example of an access architecture to which the embodiments may be applied, but the embodiments are not limited to such an architecture. By appropriately adjusting the parameters and procedures, these embodiments may also be applied to other types of communication networks with appropriate means. Some examples of other options for suitable systems are Universal Mobile Telecommunications System (UMTS) Radio Access Network (UTRAN), Wireless Local Area Network (WLAN or WiFi), Worldwide Interoperability for Microwave Access (WiMAX), Personal Communications Service (PCS), Wideband Code Division Multiple Access (WCDMA), systems using Ultra-Wideband (UWB) technology, sensor networks, Mobile Ad Hoc Networks (MANETs), and Internet Protocol Multimedia Subsystem (IMS), or any combination thereof.

[0022] Figure 1 An example of a simplified system architecture is depicted, showing only some elements and functional entities, which are all logical units, the implementation of which may differ from what is shown. Figure 1 The connections shown are logical connections; the actual physical connections may be different. It is clear to those skilled in the art that the system typically includes in addition to Figure 1 Other functions and structures than those shown.

[0023] However, the embodiments are not limited to the systems given as examples, but a person skilled in the art may apply the solutions to other communication systems providing the necessary characteristics.

[0024] Figure 1 The example of FIG. 1 shows a portion of an exemplary radio access network.

[0025] Figure 1 Devices 100 and 102 are shown. Devices 100 and 102 are configured to be in wireless connection with node 104 on one or more communication channels. Node 104 is further connected to core network 106. In one example, node 104 can be an access node, such as an (e / g)NodeB service device in a cell. In one example, node 104 can be a non-3GPP access node. The physical link from the device to the (e / g)NodeB is called an uplink or a reverse link, and the physical link from the (e / g)NodeB to the device is called a downlink or a forward link. It should be understood that the (e / g)NodeB or its functions can be implemented by using any node, host, server or access point suitable for such purpose.

[0026] A communication system typically includes more than one (e / g)NodeB, in which case the (e / g)NodeBs may also be configured to communicate with each other via wired or wireless links designed for this purpose. These links may be used for signaling purposes. (e / g)NodeB is a computing device configured to control the radio resources of the communication system to which it is coupled. NodeB may also be referred to as a base station, an access point, or any other type of interface device, including a relay station capable of operating in a wireless environment. (e / g)NodeB includes or is coupled to a transceiver. From the transceiver of the (e / g)NodeB, a connection to an antenna unit is provided, which establishes a bidirectional radio link with the device. The antenna unit may include multiple antennas or antenna elements. (e / g)NodeB is further connected to the core network 106 (CN or next generation core NGC). Depending on the deployed technology, the (e / g)NodeB is connected to a service and packet data network gateway (S-GW+P-GW) or a user plane function (UPF) for routing and forwarding user data packets and providing connectivity for the device to one or more external packet data networks; and the (e / g)NodeB is connected to a mobility management entity (MME) or an access mobility management function (AMF) for controlling access and mobility of the device.

[0027] Exemplary embodiments of an apparatus are a subscriber unit, user device, user equipment (UE), user terminal, terminal equipment, mobile station, mobile device, etc.

[0028] Device generally refers to a mobile or static device (e.g., a portable or non-portable computing device), including wireless mobile communication devices that operate with or without a Universal Subscriber Identity Module (USIM), including but not limited to the following types of devices: mobile phones, smart phones, personal digital assistants (PDAs), cell phones, devices using wireless modems (alarms or measuring devices, etc.), laptops and / or touch screen computers, tablets, game consoles, notebook computers, and multimedia devices. It should be understood that a device can also be an almost exclusively uplink-only device, an example of which is a camera or video camera that loads images or video clips to a network. A device can also be a device with the ability to operate in an Internet of Things (IoT) network, which is a scenario in which objects are provided with the ability to transmit data over a network without human-to-human or human-to-computer interaction, such as for use in smart grids and connected vehicles. The device can also utilize the cloud. In some applications, a device can include a user-portable device with a radio component (such as a watch, headphones, or glasses), and the calculation is performed in the cloud.

[0029] The device shows a type of device to which resources on the air interface are allocated and assigned, and thus any features of the device described herein may be implemented by a corresponding device, such as a relay node. An example of such a relay node is a layer 3 relay (self-backhaul relay) towards a base station. The device (or in some embodiments a layer 3 relay node) is configured to perform one or more of the user equipment functions.

[0030] Various techniques described in this article can also be applied to cyber-physical systems (CPS) (systems that coordinate computing elements that control physical entities). CPS can enable the implementation and utilization of a large number of interconnected information and communication technology (ICT) devices (sensors, actuators, processors, microcontrollers, etc.) embedded in physical objects in different locations. Mobile cyber-physical systems (in which the physical system in question has inherent mobility) are a subcategory of cyber-physical systems. Examples of mobile physical systems include mobile robots and electronic devices transported by humans or animals.

[0031] Furthermore, although the apparatus has been described as a single entity, different units, processors and / or memory units may be implemented (not all of which are shown in FIG. Figure 1 ).

[0032] 5G enables the use of multiple-input multiple-output (MIMO) antennas, more base stations or nodes than LTE (the so-called small cell concept), including macro sites that operate in cooperation with smaller base stations and adopt various radio technologies, depending on service requirements, use cases and / or available spectrum. 5G mobile communications support a wide range of use cases and related applications, including video streaming, augmented reality, different ways of sharing data and various forms of machine-type applications (such as (massive) machine-type communications (mMTC)), including vehicle safety, different sensors and real-time control. 5G is expected to have multiple radio interfaces, such as below 6 GHz or above 24 GHz, cmWave and mmWave, and can also be integrated with existing traditional radio access technologies (such as LTE). Integration with LTE can be implemented as a system, at least in the early stages, in which macro coverage is provided by LTE and 5G wireless interface access comes from small cells through aggregation to LTE. In other words, 5G plans to support both inter-RAT operability (such as LTE-5G) and inter-RI operability (inter-radio interface operability, such as below 6 GHz - cmWave, 6 or above 24 GHz - cmWave and mmWave). One of the concepts considered for use in 5G networks is network slicing, in which multiple independent and dedicated virtual subnets (network instances) can be created within the same infrastructure to run services with different requirements for latency, reliability, throughput and mobility.

[0033] The current architecture in LTE networks is completely distributed in the radio and completely centralized in the core network. Low-latency applications and services in 5G require content to be close to the radio, leading to local bursting and multi-access edge computing (MEC). 5G enables analysis and knowledge generation to be performed at the data source. This approach requires the use of resources that may not be continuously connected to the network, such as laptops, smartphones, tablets, and sensors. MEC provides a distributed computing environment for application and service hosting. It also has the ability to store and process content close to cellular subscribers to speed up response time. Edge computing covers a wide range of technologies such as wireless sensor networks, mobile data acquisition, mobile signature analysis, collaborative distributed peer-to-peer self-organizing networks and processing (which can also be classified as local cloud / fog computing and grid / mesh computing), dew computing, mobile edge computing, cloudlets, distributed data storage and retrieval, autonomous self-healing networks, remote cloud services, augmented and virtual reality, data caching, Internet of Things (massive connectivity and / or latency critical), critical communications (autonomous vehicles, traffic safety, real-time analysis, time-critical control, healthcare applications).

[0034] The communication system can also communicate with other networks 112, such as a public switched telephone network, or a VoIP network, or the Internet, or a private network, or utilize services provided by them. The communication network may also be able to support the use of cloud services, for example, at least a portion of the core network operations may be performed as a cloud service (this is in the Figure 1 114). The communication system may also include a central control entity or the like that provides facilities for networks of different operators to collaborate, for example in spectrum sharing.

[0035] Edge cloud technology can be introduced into the radio access network (RAN) by utilizing network function virtualization (NFV) and software defined networking (SDN). Using edge cloud technology can mean that access node operations will be performed at least partially in a server, host or node that is operatively coupled to a remote radio head or base station including the radio portion. Node operations can also be distributed among multiple servers, nodes or hosts. The application of cloudRAN architecture enables RAN real-time functions to be performed at or near the remote antenna site (in the distributed unit DU108) and non-real-time functions to be performed in a centralized manner (in the central unit CU 110).

[0036] It should also be understood that the work distribution between core network operations and base station operations may be different from that of LTE, or even non-existent. Some other technological advances that may be used are big data and all-IP, which may change the way networks are built and managed. 5G (or New Radio NR) networks are designed to support multiple hierarchical structures, where MEC servers can be placed between the core and the base station or NodeB (gNB). It should be understood that MEC can also be applied to 4G networks.

[0037] 5G can also make use of satellite communications to enhance or supplement the coverage of 5G services, for example to provide backhaul. Possible use cases are to provide service continuity for machine-to-machine (M2M) or Internet of Things (IoT) devices or for vehicle passengers, mobile broadband (MBB), or to ensure service availability for critical communications and future railway / maritime / aeronautical communications. Satellite communications can make use of geostationary earth orbit (GEO) satellite systems as well as low earth orbit (LEO) satellite systems, in particular mega-constellations (systems in which hundreds of (nano) satellites are deployed). Each satellite in a mega-constellation can cover several satellite-enabled network entities creating ground cells. Ground cells can be created by ground relay nodes or by gNBs located on the ground or in satellites.

[0038] It is clear to a person skilled in the art that the depicted system is only an example of a part of a radio access system, and in practice the system may include multiple (e / g)NodeBs, a device may access multiple radio cells, and the system may also include other devices, such as physical layer relay nodes or other network elements. At least one (e / g)NodeB may be a home (e / g)NodeB. In addition, in a geographical area of ​​the radio communication system, a plurality of different kinds of radio cells as well as a plurality of radio cells may be provided. The radio cells may be macro cells (or umbrella cells), which are large cells with a diameter typically up to tens of kilometers, or smaller cells such as micro, femto or pico cells. Figure 1 The (e / g)NodeB can provide any kind of these cells. A cellular radio system can be implemented as a multi-layer network including several cells. Usually, in a multi-layer network, one access node provides one or more cells, and therefore multiple (e / g)NodeBs are required to provide this network structure.

[0039] In order to meet the needs of improving the deployment and performance of communication systems, the concept of "plug and play" (e / g) NodeB is introduced. Usually, in addition to the home (e / g) NodeB (H(e / g)nodeB), the network that can use "plug and play" (e / g) NodeB also includes the home NodeB gateway or HNB-GW ( Figure 1The HNB gateway (HNB-GW) usually installed in the operator's network can aggregate services from a large number of HNBs back to the core network.

[0040] Figure 2 An example of a communication system based on 5G network components is shown. A terminal device, user terminal or user equipment 200 communicates with a data network 112 via a 5G network 202. The user terminal 200 is connected to a radio access network RAN ​​node, such as an (e / g)NodeB 206, which provides the user terminal with a connection to the network 112 via one or more user plane functions UPF 208. The user terminal 200 is also connected to a core access and mobility management function AMF 210, which is the control plane core connector for the (radio) access network and can be seen as a 5G version of the mobility management entity MME in LTE from this perspective. The 5G network also includes a session management function SMF 212, which is responsible for subscriber sessions, such as session establishment, modification and release, and a policy control function PCF 214, which is configured to manage network behavior by providing policy rules to the control plane functions. The 5G network may also include a location management function LMF 216, which may be configured to determine the location of the terminal device 200 based on information received from the terminal device and / or the gNB 206.

[0041] As in many wireless communication systems, also in 5G based systems, there are methods for the network to control the transmit power of the user equipment. User equipment power control (uplink PC) is useful for several reasons. For example, by setting limits on the transmit (Tx) power that is not required for reliable reception, unnecessary interference to other users can be reduced. In addition, by controlling the Tx power, UE battery power can be saved.

[0042] The 5G specification describes a signaling mechanism for sending open-loop (OL) power control parameters from the base station to the user equipment. For example, equation (1) shows how to set the physical uplink shared channel (PUSCH) transmission power based on the signaling control parameters.

[0043] P PUSCH =min(P max ,P0+αPL+10log 10 (M2 μ )+δ TF ),

[0044] (Equation 1)

[0045] Where P PUSCH is the transmission power of PUSCH (in dBm), P maxis the maximum transmit power of the device (in dBm); P0 is the target receive power (in dB), α is the fractional path loss compensation factor, where a value of 1.0 means that the path loss is fully compensated, PL is the path loss from the user equipment to the base station (in dB), μ is a subcarrier spacing related parameter, M is the number of resource blocks scheduled for transmission, and δ TF It is a parameter related to the modulation and coding scheme.

[0046] In the above equation, the P0 and α terms can be considered the most important.

[0047] The 5G specification also supports some enhancements to the basic uplink power control (UL PC), such as a closed loop correction part and the possibility of dedicated UE-specific PC commands. The power control parameters are signaled using RRC signaling. These parameters are typically cell-specific, i.e., they are broadcast per cell and can therefore be different between different cells.

[0048] In 5G, there are several different types of services available to users. One of the services is Ultra Reliable Low Latency Communication (URLLC). URLLC requires low latency (as low as 1ms one-way) and high reliability. The basic requirement is that 99.999% of the packets are received correctly within a given time budget. Another service is enhanced mobile broadband (eMBB), which provides high data rates over a wide coverage area.

[0049] One objective of uplink open loop power control is to select power control parameters such that network performance can be maximized given specified key performance indicators (KPIs). It is well known that open loop power control and the PC algorithm and parameters used can have a significant impact on system performance. The relationship between PC parameters and network performance depends on the environment of the user equipment and the base station. It can be assumed that for all deployments, reasonable performance can be achieved using fixed PC parameters, but customizing the PC parameters for the current network configuration and requirements will yield higher performance.

[0050] In addition, simulations show that even small deviations in PC parameters can cause URLLC service interruptions. This shows that PC parameters are extremely sensitive to URLLC delay and reliability. Therefore, PC parameter values ​​that may cause URLLC service interruptions or are similar to such PC parameter values ​​should be avoided.

[0051] Figure 3A An embodiment is shown in a flowchart of FIG. The flowchart shows an example of the operation of an apparatus or entity configured to train a neural network.

[0052] In step 300, the apparatus is configured to determine a selection of a radio access network parameter having an impact on uplink power control from a given set of radio access network parameters.

[0053] In step 302, the apparatus is configured to train a neural network using as input a selection of radio access network parameters, more than one radio access network configuration, one or more types of radio access network traffic, and varying radio access network conditions to determine uplink power control parameters.

[0054] In step 304, the apparatus is configured to use the trained neural network with the selection of the radio access network parameters as input to obtain a set of initial uplink power control parameters as output.

[0055] In one embodiment, the uplink power control parameters include a target received power P0 and a fractional path loss compensation factor α.

[0056] In one embodiment, the apparatus is configured to transmit the initial set of uplink power control parameters for further processing.

[0057] Figure 3B An embodiment is shown in the flowchart of FIG. The flowchart shows an example of the operation of an apparatus such as a base station or gNB.

[0058] In step 310, the apparatus is configured to receive an initial set of uplink power control parameters trained by a neural network.

[0059] In step 312, the apparatus is configured to utilize the set of initial uplink power control parameters in an operating radio access network and select parameters from the set of initial uplink power control parameters that give the best performance as uplink power control parameters to be applied to the radio access network.

[0060] In one embodiment, when utilizing the initial set of uplink power control parameters in operating the radio access network, the network element may be configured to apply each uplink power control parameter in the initial set of uplink power control parameters for a predetermined amount of time.

[0061] In one embodiment, when utilizing the initial set of uplink power control parameters in an operating radio access network, the network element may be configured to monitor the performance of the network and, if performance degrades, reject the current uplink power control parameters and apply utilization of parameters different from the initial set of uplink power control parameters.

[0062] Figure 3C An embodiment of a neural network 320 having one hidden layer is shown, and Figure 3D An embodiment of a compute node is shown.

[0063] Deep learning (also called deep structured learning or layer-wise learning) is part of a broader family of machine learning methods based on layers used in artificial neural networks.

[0064] Artificial neural networks (ANNs) include a set of rules designed to perform tasks such as regression, classification, clustering, and pattern recognition. ANNs achieve this through a learning process, where various examples of input data are shown along with the desired output. Through this, ANNs learn to identify the appropriate output for any input within the various training data. Learning with labels is called supervised learning, while learning without labels is called unsupervised learning. Deep learning typically requires a large amount of input data.

[0065] A deep neural network (DNN) 320 is an artificial neural network that includes a given number of hidden layers 324 between an input layer 322 and an output layer 326. Training of the neural network allows it to find the correct mathematical operations to transform the input into an appropriate output, even when the relationship is highly nonlinear and / or complex.

[0066] In this example, each hidden layer 324 includes nodes 328, 330, 332, 334, 336 in which calculations are performed. Figure 3D As shown, each node 336 combines input data 322 with a set of coefficients or weights 340 that amplify or attenuate the input 322, thereby assigning importance to the input 322 relative to the task the algorithm is trying to learn. The input-weight products are summed 342, and the sum is passed through an activation function 344 to determine whether and to what extent the signal should be further passed through the network 320 to affect the final result, such as classification behavior. In this process, the neural network learns to identify the correlation between certain relevant features and the best results.

[0067] In the case of classification, the output of the deep learning network 332 can be considered as the probability of a particular outcome, such as in this case, the probability that a given PC parameter is optimal. In this case, the number of hidden layers 324 can vary in proportion to the number of input data 322 used. However, when the number of input data 322 is higher, the accuracy of the result 326 is more reliable. On the other hand, when there are fewer layers 324, the calculation may take less time and thereby reduce latency. However, this is highly dependent on the specific neural network architecture and / or computing resources.

[0068] The initial weights 340 of the model can be set in various alternative ways. During the training phase, the initial weights 340 are adapted to improve the accuracy of the process based on errors in analytical decisions. Training the model is essentially a trial and error activity. In principle, each node 328, 330, 332, 4334, 336 of the neural network 320 makes a decision (input * weight), which is then compared with the collected data to find differences with the collected data. In other words, it determines the error, based on which the weights 340 can be adjusted. Therefore, the training of the model can be regarded as a correction feedback loop.

[0069] Typically, neural network models are trained using a stochastic gradient descent optimization algorithm, where the gradient is calculated using a back-propagation algorithm. The gradient descent algorithm seeks to change weights 340 so that the next evaluation reduces the error, which means that the optimization algorithm is navigating down the gradient (or slope) of the error. Any other suitable optimization algorithm that can provide sufficiently accurate weights 340 may also be used. Thus, the trained parameters of neural network 320 may include weights 340.

[0070] In the context of an optimization algorithm, the function used to evaluate a candidate solution (i.e., a set of weights) is called an objective function. Typically, for neural networks (where the goal is to minimize error), the objective function is often referred to as a cost function or a loss function. When adjusting weights 340, any suitable method can be used as a loss function, some examples being mean square error (MSE), maximum likelihood (MLE), and cross entropy.

[0071] As for the activation function 344 of the node 336, it defines the output 326 of the node 336 given an input or a set of inputs 322. The node 336 calculates the weighted sum of the inputs, possibly with a bias, and then makes a decision of "activation" or "inactivation" based on a decision threshold as a binary activation or using an activation function 334 that gives a nonlinear decision function. Any suitable activation function 344 can be used, such as a sigmoid, a rectified linear unit (ReLU), a normalized exponential function (softmax), sotfplus, tanh, etc. In deep learning, the activation function 344 is usually set at the layer level and applied to all neurons in the layer. The output 326 is then used as the input of the next node, and so on, until the desired solution to the original problem is found.

[0072] Figure 4 An example implementation of step 300 is shown, in which a selection of a radio access network parameter is determined.

[0073] First, a given set of radio access network parameters 400 is selected as a starting point. These network parameters may be assumed to be network parameters that may affect the power control parameters.

[0074] In one embodiment, the set of given radio access network parameters comprises two or more of base station locations, estimated path loss index, cell specific load, inter-site distance, average number of users per cell, network centre frequency and average packet size.

[0075] These radio access network parameters 400 are fed to a simulator 402 where the radio access network parameters and the power control parameters 404 are simulated at different values ​​of the selected parameters 400 to see which of them affects the power control parameters. The simulation is automatically monitored 406 and the power control parameters are evaluated with predefined key performance indicators (KPIs). Finally, if it is determined 408 that changing a particular network parameter affects the power control parameter selection, that network parameter is selected 410 as an input parameter to predict the optimal power control parameter. Those network parameters whose changed values ​​have no effect on the power control parameters are discarded 412.

[0076] Figure 5 An example implementation of step 302 is shown in which a neural network is trained to determine uplink power control parameters.

[0077] In the following, it is assumed that the uplink power control parameters include the target received power P0 and the fractional path loss compensation factor α, for example. However, those skilled in the art may also consider other parameters.

[0078] In one embodiment, in order to train the neural network-based power control parameter prediction algorithm, training data can be generated by simulating the network with eMBB services. Several different network configurations can be used, and each of them is evaluated using all combinations of P0 and α. Both parameters are discrete parameters. In this way, the same network with different power control parameters can be evaluated, and thus the optimal power control parameters for each network type can be determined.

[0079] In one embodiment, the optimal PC parameters are determined by calculating the following quantities for each simulation run:

[0080] S=xR mean +(1-x)R edge

[0081] (Equation 2)

[0082] Where R mean is the normalized average data rate, R edgeis the normalized data rate at the cell edge, and x is a design parameter with a value of 0.7, which is chosen to prioritize the average bit rate over the cell edge bit rate. Both data rates are normalized in the interval [0,1] to ensure that they have equal weight in the quality metric. In one embodiment, for a given network configuration, the PC parameters that give the highest value of S are considered optimal.

[0083] The above metric S is an example of a possible metric. Any other metric for defining optimal network operation may also be selected. The metric may be selected based on the type of service under which the network operates. For example, in a URLLC scenario, latency should be reflected in the equation. An example of a quality metric for a network with a mix of eMBB and URLLC services may be expressed as

[0084]

[0085] Where L URLLC is the delay of the URLLC traffic in seconds, and the parameter y can be selected based on the priority of eMBB and URLLC. This metric will allow the neural network to learn such PC optimizations where the risk of excessive delay is small while still achieving good eMBB performance.

[0086] In one embodiment, in order to separate different network configurations and allow the neural network to properly learn predictions of optimal PC parameters, the following features are used as input data to the neural network:

[0087] - Inter-site distance (ISD), and

[0088] - Number of users / cell (load).

[0089] The use of ISD can be used as an input because uplink PC operation is heavily dependent on the distance between the user equipment and the base station. The balance between users at the cell edge and at the cell center is critical to the overall performance. For example, with a smaller value of α, less compensation is given for the increased path loss, which results in less received power from cell edge users, but on the other hand, cell edge users also cause less interference to neighboring base stations. This balance depends on the ISD, which affects both the path loss of a user and how much interference the user generates to neighboring cells.

[0090] The impact of load on PC parameters comes from a variety of network effects. For example, load affects the scheduling and number of resource blocks allocated to the UE, which affects the available power per resource block. In addition, the modulation and coding scheme (MCS) is affected, and of course the interference situation of the network is also affected.

[0091] As is typical in neural network based solutions, the inputs may also be normalized to have unit variance before being fed to the neural network. It may be noted that quite a few parameters may be fed to the neural network, but in this simplified example only two parameters are considered. Other possible input parameters include path loss exponent, base station location and cell specific load.

[0092] The above parameters define the input data vector, while the best PC parameter combination represents the label, i.e. the output of the neural network. It can be noted that each (P0, α) pair can be encoded as a numerical class indicator, and thus the neural network basically outputs the probability of optimality of different PC parameter pairs. The one with the highest probability is usually selected as the classification result.

[0093] The training itself can be performed by collecting simulated data into a data storage device 500. Each data sample represents an input-output pair, where the input is a selected data vector and the output is an optimal PC parameter pair. The collected data is split 502 into separate parts, one of which is used for training 504 and one of which is used to verify 506 the performance of the neural network. The training loop 508 involves first feeding a batch of input samples through the neural network, after which the predicted classes are mapped to their corresponding PC parameter values ​​510, and the loss relative to the optimal PC parameter values ​​is calculated 512. For example, the value of the loss can be determined using the log-likelihood loss. The parameters of the neural network are then updated based on the calculated loss using stochastic gradient descent or any variant thereof. Once the training is completed 514, for example, after a predefined number of epochs / iterations has been reached, the final performance is determined by measuring 516 the prediction accuracy on a separate validation set.

[0094] When the neural network is trained, it can be used to give initial uplink power control parameters, as shown in step 304. Figure 6 An example of providing initial uplink power parameters is shown.

[0095] The selection of radio access network parameters 600 is taken as input to a trained neural network 602 which is configured to determine 604 initial uplink power control parameters, in this example P0 and α.

[0096] As described above, typically, a neural network includes an input layer, multiple hidden layers, and an output layer. In one embodiment, a four-layer neural network is used to determine the initial uplink power control parameters. However, other numbers of layers may also be used. In this example, there is therefore an input layer, two hidden layers, and an output layer.

[0097] In one embodiment, the neural network includes a total of four layers, where the hidden layer utilizes a rectified linear unit (ReLU) as an activation function, and the output layer is fed through a sigmoid function to map the output to a class probability. In the above embodiment, there are two input features that are mapped to predictions in a given number of classes, each class representing a specific combination of P0 and α.

[0098] The neural network used to make the initial PC parameter predictions is shown in Table 1. It may be noted that the values ​​shown are only examples of possible structures of the neural network.

[0099]

[0100]

[0101] Table 1

[0102] When the example trained neural network solution was executed and compared with the selected PC value, this PC value can be considered a good default value for the test simulation data not used during training. The average results over all network configurations are shown in Table 2.

[0103] Average throughput Edge throughput Gain 1.6 times 2.0 times

[0104] Table 2

[0105] The gain of NN-based prediction is clear: on average, it can improve the total throughput of the cell by 60% while doubling the cell edge throughput improvement. The predicted PC parameters are beneficial for most network configurations. These simulation results show the advantages of the proposed approach. When studying the classification accuracy of the trained neural network, it correctly predicts the PC parameters 65-80% of the time for a randomly selected test dataset, demonstrating the high accuracy of selecting the best PC parameter pairs.

[0106] In another test case, irregular cells and more complex input parameter space were considered. The following four input features were considered:

[0107] - Number of users / cells (load)

[0108] - Group size

[0109] - Path loss coefficients A and B: PL = A + Blog_10(d), d = distance (km)

[0110] The packet size affects how many resource blocks are utilized in the transmission protocol. Increasing the number of resource blocks increases the power level requirements and therefore requires different power control parameters.

[0111] The importance of the path loss coefficient is that the path loss is compensated by the power control parameter and therefore directly affects the transmission power. Too low a transmission power indicates a low bit rate, especially for cell edge users. The path loss is different in different environments, e.g. in urban and rural areas. Therefore, the path loss coefficient provides information about the environment.

[0112] An example of a neural network architecture for this scenario is shown in Table 3.

[0113] layer Input size Output size activation Fully connected 4. 35 ReLU Fully connected 35 35 ReLU Fully connected 35 35 ReLU Fully connected 35 19 LogSoftmax

[0114] Table 3

[0115] As shown in Table 4, there is a gain in throughput for the randomly selected test set compared to the selected PC values ​​that are considered good defaults. The proposed neural network correctly predicted the PC parameters for about 76% of the trials.

[0116]

[0117] Table 4

[0118] It has been noted that the gains of the power control parameter prediction based on the neural network produced good results for the simulated network. However, in the actual network, the accuracy may not be that high. Therefore, it may be necessary to fine-tune the output of the neural network, i.e. the initial parameters. Figure 7 Examples of possible methods for fine-tuning parameters are shown.

[0119] The selection of radio access network parameters 600 is taken as input to a trained neural network 602, which is configured to produce as output an initial uplink power control parameter with probability 700. In this example, the parameters are the parameter pair P0 and α.

[0120] The unused power control parameter pair with the highest probability is selected 702. The probability is tested 704 against some preselected threshold. If the probability is less than the threshold, the process terminates 706. Otherwise, the parameter pair is used 708, and the network is operated 710, and the performance of the network is tracked. A performance metric is calculated 712, and compared 714 to a previously acquired metric (using other parameters). If the performance is higher, the current parameters are retained 716, and the process continues at step 702. If not, the previous parameters that gave better performance are used 718. The process can be repeated as long as there are PC parameter candidates with a probability above the threshold.

[0121] The process also includes a safeguard by using 718 previous parameters if the performance of the system suddenly degrades 720. This ensures robustness to erroneous PC parameter predictions.

[0122] Fig. 8A , Figure 8B and Figure 8C One embodiment is shown. The figures show simplified examples of apparatus to which embodiments of the invention may be applied.

[0123] It should be understood that the device is described here as an example to illustrate some embodiments. It is clear to those skilled in the art that the device may also include other functions and / or structures, and not all described functions and structures are required. Although the device has been described as one entity, different modules and memories may be implemented in one or more physical or logical entities.

[0124] Fig. 8A An apparatus configured to train a neural network is shown. 800 or a network element to which an embodiment of the present invention is applied. In some embodiments, the apparatus may be a network element or a part of a network element. In some embodiments, the apparatus may include several network entities connected to each other.

[0125] The apparatus 800 of this example includes control circuitry 800 configured to control at least a portion of the operation of the apparatus.

[0126] The apparatus may include a memory 802 for storing data. Furthermore, the memory may store software 804 executable by the control circuitry 800. The memory may be integrated in the control circuitry.

[0127] The apparatus also includes one or more interface circuit systems 806 configured to connect the apparatus to other devices and network elements of the radio access network. The interface 806 may provide a wired or wireless connection. The interface may provide access to Figure 8B or Figure 8C connection of network elements.

[0128] In one embodiment, the software 804 may include a computer program including program code means adapted to cause the control circuit system 800 of the apparatus to implement the above, for example, in conjunction with Figure 3A At least some of the embodiments described.

[0129] Figure 8B A network element device 810 to which an embodiment of the present invention is applied is shown. In some embodiments, the device may be a network element or a part of a network element. In some embodiments, the device may include several network entities connected to each other. In one embodiment, the device is a gNB or a part of a gNB.

[0130] The apparatus 810 of this example includes control circuitry 820 configured to control at least a portion of the operation of the apparatus.

[0131] The apparatus may include memory 822 for storing data. In addition, the memory may store software 824 executable by the control circuitry 820. The memory may be integrated in the control circuitry.

[0132] In one embodiment, the software 824 may include a computer program including program code means adapted to cause the control circuit system 820 of the apparatus to implement the above, for example, in conjunction with Figure 3B At least some of the embodiments described.

[0133] The apparatus also includes one or more interface circuit systems 826 configured to connect the apparatus to other devices and network elements of the radio access network. The interface 806 may provide a wired or wireless connection. For example, the interface may provide access to Fig. 8A connection between the device and other network elements of the radio access system.

[0134] In one embodiment, at least some of the functionality of the network element apparatus 810 may be shared between two physically separate devices to form an operational entity. Thus, the apparatus may be viewed as depicting an operational entity comprising one or more physically separate devices for performing at least some of the described processes. Thus, utilizing such a shared architecture Figure 8B The apparatus may include a remote control unit RCU 830, such as a host computer or a server computer, which is operably coupled (e.g., via a wireless or wired network) to a remote distributed unit RDU 832 located in a base station. In one embodiment, at least some of the above processes may be performed by the RCU 830. In one example, the execution of at least some of the above processes may be shared between the RDU 832 and the RCU 830.

[0135] In one embodiment, the RCU 830 can generate a virtual network through which the RCU 830 communicates with the RDU 832. In general, virtual networking can involve the process of combining hardware and software network resources and network functions into a single software-based management entity (virtual network). Network virtualization can involve platform virtualization, which is often combined with resource virtualization. Network virtualization can be classified as external virtual networking that combines many networks or network parts into a server computer or host computer (e.g., to an RCU). External network virtualization is intended to optimize network sharing. Another type is internal virtual networking, which provides network-like functionality for software containers on a single system. Virtual networking can also be used to test terminal devices.

[0136] In one embodiment, the virtual network can provide a flexible allocation of operations between the RDU and the RCU. In practice, any digital signal processing task can be performed in the RDU or the RCU, and the boundaries of transferring responsibilities between the RDU and the RCU can be selected according to the implementation.

[0137] There is no absolute time sequence for the steps and related functions described above and in the accompanying drawings, and some steps may be performed simultaneously or in an order other than the given order. Other functions may also be performed between steps or within steps. Certain steps may also be omitted or replaced with corresponding steps.

[0138] The device or controller capable of performing the above steps can be implemented as an electronic digital computer, a processing system or a circuit system, which may include a working memory (random access memory, RAM), a central processing unit (CPU) and a system clock. The CPU may include a set of registers, an arithmetic logic unit and a controller. The processing system, the controller or the circuit system is controlled by a sequence of program instructions transmitted from the RAM to the CPU. The controller may contain a plurality of microinstructions for basic operations. The implementation of the microinstructions may vary according to the CPU design. The program instructions may be encoded in a programming language, which may be a high-level programming language such as C, Java, etc., or a low-level programming language such as a machine language or an assembler. The electronic digital computer may also have an operating system, which may provide system services to computer programs written in program instructions.

[0139] As used in this application, the term "circuitry" refers to all of the following: (a) hardware circuit implementations only, such as implementations in analog and / or digital circuitry alone, and (b) combinations of circuitry and software (and / or firmware), such as, as applicable: (i) a combination of (multiple) processors, or (ii) portions of (multiple) processors / software (including (multiple) digital signal processors), software, and (multiple) memories that work together to cause the device to perform various functions, and (c) circuitry, such as (multiple) microprocessors or portions of (multiple) microprocessors, that requires software or firmware to perform operations, even if the software or firmware is not actually present.

[0140] This definition of "circuitry" applies to all uses of the term in this application. As a further example, as used in this application, the term "circuitry" would also cover an implementation of only a processor (or multiple processors) or a portion of a processor and its (or their) accompanying software and / or firmware. For example, the term "circuitry" would also cover a baseband integrated circuit or an application processor integrated circuit for a mobile phone, or a similar integrated circuit in a server, cellular network device, or other network device, if applicable to the particular element.

[0141] An embodiment provides a computer program embodied in a distribution medium. The computer program includes program instructions. When the program instructions are loaded into an electronic device, the device is configured to control the device to execute the above-mentioned embodiment.

[0142] A computer program may be in source code form, object code form, or some intermediate form, and may be stored in some carrier, which may be any entity or device capable of carrying the program. Such carriers include, for example, recording media, computer memory, read-only memory, and software distribution packages. Depending on the processing power required, a computer program may be executed in a single electronic digital computer or distributed among multiple computers.

[0143] The device may also be implemented as one or more integrated circuits, such as an application specific integrated circuit ASIC. Other hardware embodiments are also possible, such as circuits constructed from separate logic components. Mixtures of these different implementations are also possible. When selecting an implementation method, those skilled in the art will consider requirements set, such as the size and power consumption of the device, the necessary processing power, production costs, and production volume.

[0144] One embodiment provides an apparatus comprising a component for determining a selection of radio access network parameters that have an impact on uplink power control from a set of given radio access network parameters; a component for training a neural network using the selection of radio access network parameters, one or more radio access network configurations, one or more types of radio access network services, and changing radio access network conditions as inputs to determine the uplink power control parameters; and a component for obtaining a set of initial uplink power control parameters as output using the trained neural network with the selection of radio access network parameters as input.

[0145] An embodiment provides an apparatus comprising a component for receiving a set of initial uplink power control parameters provided by a trained neural network; and a component for utilizing the set of initial uplink power control parameters in an operating radio access network and selecting parameters that give the best performance from the received set of initial uplink power control parameters as uplink power control parameters to be applied to the radio access network.

[0146] It is clear to a person skilled in the art that, as technology advances, the inventive concept can be implemented in various ways.The invention and its embodiments are not limited to the examples described above but may vary within the scope of the claims.

Claims

1. A device comprising: at least one processor; as well as at least one memory including computer program code, The at least one memory and the computer program code are configured to, with the at least one processor, cause the apparatus to at least perform: determining a selection of a radio access network parameter having an impact on uplink power control from a given set of radio access network parameters; training a neural network to determine uplink power control parameters using the selection of the radio access network parameters as input; Taking the selection of the radio access network parameters as input, using the trained neural network to obtain a set of initial uplink power control parameters as output; Wherein, when training the neural network, a performance metric is selected that includes at least one of the following: a normalized average data rate of a user equipment, a normalized cell edge rate, and a delay; as well as Parameters are selected from the set of initial uplink power control parameters which give the highest value for the selected performance metric.

2. The apparatus of claim 1 , wherein the at least one memory and the computer program code are configured to, together with the at least one processor, cause the apparatus to further perform: At least one of the following is utilized as input in said training of said neural network: one or more radio access network configurations, varying radio access network conditions, distances between network sites, number of users per cell.

3. The apparatus according to claim 1 or 2, wherein the uplink power control parameter comprises: Target received power P0, and fractional path loss compensation factor α.

4. The apparatus of any preceding claim, wherein the given set of radio access network parameters comprises: base station locations, estimated path loss exponent, cell specific load, inter-site distance, average number of users per cell, network center frequency, and average packet size.

5. A network element in a radio access network, comprising: at least one processor; as well as at least one memory including computer program code, The at least one memory and the computer program code are configured to, with the at least one processor, cause the network element to at least perform: receiving an initial set of uplink power control parameters provided by a trained neural network; utilizing the initial set of uplink power control parameters in operating the radio access network and selecting the parameters giving the best performance from the received set of initial uplink power control parameters as the uplink power control parameters to be applied to the radio access network, wherein when utilizing the initial set of uplink power control parameters in operating the radio access network, each uplink power control parameter in the initial set of uplink power control parameters is applied for a predetermined amount of time; Wherein, when utilizing the initial set of uplink power control parameters in an operating radio access network, monitoring the performance of the network, and if the performance deteriorates, rejecting the current uplink power control parameters and applying parameters different from the initial set of uplink power control parameters.

6. A system comprising an apparatus and at least one network element, wherein the apparatus comprises: at least one processor; as well as at least one memory including computer program code, The at least one memory and the computer program code are configured to, with the at least one processor, cause the apparatus to at least perform: determining a selection of a radio access network parameter having an impact on uplink power control from a given set of radio access network parameters; training a neural network to determine uplink power control parameters using the selection of the radio access network parameters as input; Taking the selection of the radio access network parameters as input, using the trained neural network to obtain a set of initial uplink power control parameters as output; Wherein, when training the neural network, a performance metric is selected including at least one of the following: a normalized average data rate of a user equipment, a normalized cell edge rate, and a delay; and selecting parameters from the set of initial uplink power control parameters that give a highest value for the selected performance metric; and said network element in a radio access network, comprising: at least one processor; and at least one memory including computer program code, The at least one memory and the computer program code are configured to, with the at least one processor, cause the network element to at least perform: receiving an initial set of uplink power control parameters provided by a trained neural network; utilizing the initial set of uplink power control parameters in operating the radio access network and selecting the parameters giving the best performance from the received set of initial uplink power control parameters as the uplink power control parameters to be applied to the radio access network, wherein when utilizing the initial set of uplink power control parameters in operating the radio access network, each uplink power control parameter in the initial set of uplink power control parameters is applied for a predetermined amount of time; Wherein, when utilizing the initial set of uplink power control parameters in an operating radio access network, monitoring the performance of the network, and if the performance deteriorates, rejecting the current uplink power control parameters and applying parameters different from the initial set of uplink power control parameters.

7. The system of claim 6, wherein the apparatus is configured to transmit the set of initial uplink power control parameters to at least one network element.

8. A method in a radio access network, comprising: determining a selection of a radio access network parameter having an impact on uplink power control from a given set of radio access network parameters; training a neural network to determine uplink power control parameters using as input the selection of radio access network parameters, one or more radio access network configurations, one or more types of radio access network services, and varying radio access network conditions; Taking the selection of the radio access network parameters as input, using the trained neural network to obtain a set of initial uplink power control parameters as output; Wherein, when training the neural network, a performance metric is selected that includes at least one of the following: a normalized average data rate of a user equipment, a normalized cell edge rate, and a delay; as well as Parameters are selected from the set of initial uplink power control parameters which give the highest value to the selected performance metric.

9. The method according to claim 8, further comprising: At least one of the following is also used as input in the training of the neural network: distance between network sites, number of users per cell.

10. The method according to claim 8 or 9, wherein the uplink power control parameters include: Target received power P0, and fractional path loss compensation factor α.

11. A method for a network element in a radio access network, comprising: receiving an initial set of uplink power control parameters provided by a trained neural network; utilizing the initial set of uplink power control parameters in operating the radio access network and selecting the parameters giving the best performance from the received set of initial uplink power control parameters as the uplink power control parameters to be applied to the radio access network, wherein when utilizing the initial set of uplink power control parameters in operating the radio access network, each uplink power control parameter in the initial set of uplink power control parameters is applied for a predetermined amount of time; Wherein, when utilizing the initial set of uplink power control parameters in an operating radio access network, monitoring the performance of the network, and if the performance deteriorates, rejecting the current uplink power control parameters and applying parameters different from the initial set of uplink power control parameters.

12. A computer program product comprising instructions for causing an apparatus to at least: determining a selection of a radio access network parameter having an impact on uplink power control from a given set of radio access network parameters; training a neural network to determine uplink power control parameters using as input the selection of radio access network parameters, one or more radio access network configurations, one or more types of radio access network services, and varying radio access network conditions; Taking the selection of the radio access network parameters as input, using the trained neural network to obtain a set of initial uplink power control parameters as output; Wherein, when training the neural network, a performance metric is selected that includes at least one of the following: a normalized average data rate of a user equipment, a normalized cell edge rate, and a delay; as well as Parameters are selected from the set of initial uplink power control parameters which give the highest value to the selected performance metric.

13. A computer program product comprising instructions for causing a network element in a radio access network to perform at least the following operations: receiving an initial set of uplink power control parameters provided by a trained neural network; utilizing the initial set of uplink power control parameters in operating the radio access network and selecting the parameters giving the best performance from the received set of initial uplink power control parameters as the uplink power control parameters to be applied to the radio access network, wherein when utilizing the initial set of uplink power control parameters in operating the radio access network, each uplink power control parameter in the initial set of uplink power control parameters is applied for a predetermined amount of time; in, When utilizing the initial set of uplink power control parameters in an operating radio access network, monitoring performance of the network and if the performance degrades, rejecting current uplink power control parameters and applying parameters different from the initial set of uplink power control parameters.

Images