Controlling Radiating Elements

Abstract

An apparatus, method, and computer program product are disclosed. The apparatus may include means for receiving a performance metric of an antenna array including a plurality of radiating elements, the performance metric being based on performance data associated with the antenna array, the antenna array having a radiation configuration represented by configuration parameters. The apparatus may also include means for updating the configuration parameters based on the received performance metric by estimating new configuration parameters for moving the performance metric toward a target value. The apparatus may also include means for reconfiguring the radiation configuration of the antenna array based on the updated configuration parameters, thereby changing the physical geometry of the antenna array.

Description

Technical Field

[0001] Example embodiments relate to methods, systems, and computer programs for controlling one or more radiating elements (eg, radiating elements of one or more antennas that may form part of an antenna array). Background Art

[0002] An antenna can include one or more radiating elements that are fed with the same radio frequency signal. An antenna array can include multiple such antennas with separate radio frequency chains. Antenna arrays are used for various purposes, such as beamforming in telecommunications networks. Radiating elements can also be used to receive radio frequency energy. Summary of the Invention

[0003] A first aspect provides an apparatus comprising: a component for receiving a performance metric of an antenna array comprising a plurality of radiating elements, the performance metric being based on performance data associated with the antenna array, the antenna array having a radiation configuration represented by configuration parameters; a component for updating the configuration parameters according to the received performance metric by estimating new configuration parameters for moving the performance metric towards a target value; and a component for reconfiguring the radiation configuration of the antenna array based on the updated configuration parameters, such that the physical geometry of the antenna array is changed.

[0004] The reconfiguration component is configured to change the physical geometry by moving one or more radiating elements based on the reconfigured configuration.

[0005] The updating component is further configured to iteratively update the configuration parameters according to the received performance metrics generated by the updated configuration parameters until a stopping condition is reached.

[0006] The stopping condition corresponds to a fixed number of update iterations.

[0007] The stopping condition corresponds to the performance measure reaching a target value. The stopping condition corresponds to a fixed number of update iterations in which the performance measure does not move towards the target value.

[0008] The update component includes a machine learning model configured to estimate updated configuration parameters that are likely to move the performance metric toward a target value based on stored training data. The machine learning model is trained using data derived from initial connection quality and subsequently updated based on results caused by the updated configuration parameters. The machine learning model is trained using the initial configuration parameters and subsequently updated based on results of the reconfiguration caused by the updated configuration parameters. The machine learning model is trained using digital peers.

[0009] The performance metric is received from a computation component configured to receive performance data from physical nodes of at least a portion of the communication network of which the antenna array forms a part.

[0010] The computing means is associated with a base station of at least part of the cellular communication network of which the antenna array forms part.

[0011] The performance metrics are received from a simulation component that simulates physical nodes of at least a portion of the communication network of which the antenna array forms a part.

[0012] The simulated physical node is associated with a base station of at least part of a cellular communication network of which the antenna array forms part.

[0013] The computing component or the simulation component comprises a part of the device.

[0014] The received performance metric represents a measured or simulated value of one or more of data throughput, call loss rate, outage probability, and energy consumption associated with the antenna array.

[0015] The performance metric represents a measured or simulated value for each of a different number of user equipments within a cell of the cellular communication system associated with the antenna array.

[0016] The configuration parameters include one or more of the following: for the j-th radiating element of the i-th antenna of the antenna array:

[0017] -Location

[0018] - azimuth angle φi,j∈[0,2π]

[0019] -elevation angle θi,j∈[-π / 2,π / 2]

[0020] - simulated phase shift βi,j∈[-π,π]

[0021] -Analog signal gain αi,j∈[αmin,αmax]

[0022] And the means for reconfiguring the radiation configuration of the antenna array is configured to update one or more of the configuration parameters to produce corresponding changes in one or more of the radiating elements of the antenna array.

[0023] The means for reconfiguring the radiation configuration of the antenna array is configured to perform the reconfiguration in substantially real time or near real time based on the last received value of the performance metric.

[0024] On the other hand, a method is provided, comprising: receiving a performance metric for an antenna array comprising a plurality of radiating elements, the antenna array having a radiation configuration represented by configuration parameters; updating the configuration parameters according to the received performance metric by estimating new configuration parameters to move the performance metric toward a target value; and reconfiguring the radiation configuration of the antenna array based on the updated configuration parameters, such that the physical geometry of the antenna array is changed.

[0025] Reconfiguring includes changing the physical geometry by moving one or more radiating elements based on the reconfigured configuration.

[0026] The method further includes iteratively updating the configuration parameters based on the received performance metrics generated by the updated configuration parameters until a stopping condition is reached.

[0027] The stopping condition corresponds to a fixed number of update iterations. The stopping condition corresponds to the performance measure reaching a target value. The stopping condition corresponds to a fixed number of update iterations in which the performance measure does not move toward the target value.

[0028] The updating includes using a machine learning model that is configured to estimate updated configuration parameters that are likely to move the performance metric toward the target value based on stored training data.

[0029] The method also includes: training the machine learning model using data derived from the initial connection quality and subsequently updated based on the results caused by the updated configuration parameters. The method also includes: training the machine learning model using the initial configuration parameters and subsequently updated based on the results of the reconfiguration caused by the updated configuration parameters.

[0030] Machine learning models are trained on digital peers.

[0031] The performance metric is received from a computation component configured to receive performance data from physical nodes of at least a portion of the communication network of which the antenna array forms a part.

[0032] The computing means is associated with a base station of at least part of the cellular communication network of which the antenna array forms part.

[0033] The performance metrics are received from a simulation component that simulates a physical node of at least a portion of a communication network of which the antenna array forms a part.

[0034] The simulated physical node is associated with a base station of at least part of a cellular communication network of which the antenna array forms part.

[0035] The computing component or the simulation component comprises a part of the device.

[0036] The received performance metric represents a measured or simulated value of one or more of data throughput, call loss rate, outage probability, and energy consumption associated with the antenna array.

[0037] The performance metric represents a measured or simulated value for each of a different number of user equipments within a cell of the cellular communication system associated with the antenna array.

[0038] The configuration parameters include one or more of the following: for the j-th radiating element of the i-th antenna of the antenna array:

[0039] -Location

[0040] - azimuth angle φi,j∈[0,2π]

[0041] -elevation angle θi,j∈[-π / 2,π / 2]

[0042] - simulated phase shift βi,j∈[-π,π]

[0043] -Analog signal gain αi,j∈[αmin,αmax]

[0044] And reconfiguring the radiation configuration of the antenna array includes updating one or more of the configuration parameters to produce corresponding changes in one or more of the radiating elements of the antenna array.

[0045] Reconfiguring the radiation configuration of the antenna array includes performing the reconfiguration in substantially real time or near real time based on a last received value of the performance metric.

[0046] On the other hand, an apparatus is provided, comprising: at least one memory directly connected to at least one processor, the at least one memory comprising computer program code; and at least one processor, the at least one processor, together with the at least one memory and the computer program code, being arranged to perform the following method: receiving a performance metric for an antenna array comprising a plurality of radiating elements, the antenna array having a radiation configuration represented by configuration parameters; updating the configuration parameters according to the received performance metric by estimating new configuration parameters, so as to move the performance metric towards a target value; and reconfiguring the radiation configuration of the antenna array based on the updated configuration parameters, such that the physical geometry of the antenna array is changed.

[0047] The at least one processor, the at least one memory and the computer program code are further arranged to perform the method defined in association with the above aspects.

[0048] On the other hand, a computer program product is provided, comprising a set of instructions, which, when executed on an apparatus, are configured to cause the apparatus to perform the following method: receiving a performance metric for an antenna array comprising a plurality of radiating elements, the antenna array having a radiation configuration represented by configuration parameters; updating the configuration parameters according to the received performance metric by estimating new configuration parameters to move the performance metric toward a target value; and reconfiguring the radiation configuration of the antenna array based on the updated configuration parameters, such that the physical geometry of the antenna array is changed.

[0049] On the other hand, a non-transitory computer-readable medium is provided, comprising program instructions stored on the non-transitory computer-readable medium for performing a method, comprising: receiving a performance metric for an antenna array comprising a plurality of radiating elements, the antenna array having a radiation configuration represented by configuration parameters; updating the configuration parameters according to the received performance metric by estimating new configuration parameters to move the performance metric toward a target value; and reconfiguring the radiation configuration of the antenna array based on the updated configuration parameters, such that the physical geometry of the antenna array is changed.

[0050] On the other hand, a device is provided, comprising: at least one processor; and at least one memory, the at least one memory comprising computer program code, which, when executed by the at least one processor, causes the device to: receive a performance metric for an antenna array comprising a plurality of radiating elements, the antenna array having a radiation configuration represented by configuration parameters; update the configuration parameters based on the received performance metric by estimating new configuration parameters to move the performance metric toward a target value; and reconfigure the radiation configuration of the antenna array based on the updated configuration parameters, such that the physical geometry of the antenna array is changed.

[0051] The above references to "means" may refer to any one of, or any combination of, hardware, software, electrical or electronic circuitry configured to perform the recited function. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Example embodiments will now be described, by way of non-limiting examples, with reference to the accompanying drawings, in which:

[0053] Figure 1 is a schematic diagram of the antenna array.

[0054] Figure 2 is a block diagram of a system including an antenna array and a controller according to an example embodiment;

[0055] Figure 3 is based on Figure 2 A partial schematic diagram of another antenna array controllable by the controller;

[0056] Figure 4is based on Figure 2 A partial schematic diagram of another antenna array controllable by the controller;

[0057] Figure 5 is based on Figure 2 Partial schematic diagram of another antenna array controllable by the controller in Figure 2 ;

[0058] Figure 6 is a block diagram of another system including an antenna array and a controller according to an example embodiment;

[0059] Figure 7 is a block diagram illustrating a physical control system and a digital peer system used in accordance with an example embodiment;

[0060] Figure 8 is a flowchart illustrating operations that may be performed according to example embodiments;

[0061] Figure 9 is another flow chart illustrating operations that may be performed according to example embodiments;

[0062] Figure 10 is another flow chart illustrating operations that may be performed according to example embodiments;

[0063] Figure 11 is another flow chart illustrating operations that may be performed according to example embodiments;

[0064] Figure 12 is another flow chart illustrating operations that may be performed according to example embodiments;

[0065] Figure 13 According to an example embodiment Figure 2 or Figure 6 a schematic diagram of the components of; and

[0066] Figure 14 Non-volatile media according to some embodiments is shown. DETAILED DESCRIPTION

[0067] Example embodiments relate to controlling one or more radiating elements, such as one or more radio frequency antennas. Example embodiments may relate to apparatus, methods, and / or computer programs for controlling one or more antenna radiating elements (hereinafter referred to as "radiating elements") of one or more antennas of an antenna array; an antenna array is a physical device comprising multiple distinct antennas. The antennas of an array can effectively work together as a single antenna to transmit or receive radio frequency waves. Antenna arrays are used in various fields, such as in cellular communications.

[0068] Example embodiments may involve adaptively controlling the radiating elements of one or more antennas. The adaptive control may be performed in real time or near real time to provide instant modifications to the current antenna array configuration based on, for example, current or very recent conditions (such as, network conditions, the number of user devices in one or more cells associated with the antenna array, etc.). These examples should not be considered as limiting the scope of the present disclosure and are given by way of example only. An antenna array configuration may, for example, refer to the characteristics of a single antenna and / or the characteristics of its single radiating elements in terms of its phase, gain, and position characteristics. Modifications to the antenna array configuration may be for the purpose of modifying the radiation or reception characteristics of one or more radiating elements. All or a subset of the radiating elements may be reconfigured in this manner.

[0069] refer to Figure 1 An antenna array 10 is shown schematically. Antenna array 10 includes a plurality of antennas 12-15. Each of antennas 12-15 has a respective RF input and output 16-19. Each of antennas 12-15 may include one or more radiating elements 20, which may be, for example, dipoles, which may be fed by the same RF signal via RF input and output 16-19. Each of antennas 12-15 may be associated with a respective RF processing chain.

[0070] Some characteristics of an antenna array are the spacing between the antennas and the overall size (relative to the wavelength), called the aperture. The size determines the array's directivity, which is its ability to focus radiated energy in certain directions. The number of antennas determines the amount of energy radiated / received.

[0071] refer to Figure 2 , the radiating elements 20 of each of the antennas 12-15 can generally share a single analog-to-digital converter (ADC) / digital-to-analog converter (DAC) 30. However, as indicated by the phase shifter and amplifier elements 22, the phase and amplitude (or gain) of the RF signal of each radiating element 20 can optionally be independently controlled in the analog domain. Phase and amplitude are example characteristics of a given radiating element 20.

[0072] In some embodiments, other characteristics of a given radiating element 20 may include one or more of the following: position, azimuth, and elevation. To this end, a mechanical positioning component may be associated with each radiating element 20. An exemplary mechanical positioning device may include a position actuator 24 ("PA"). Multiple such actuators 24, such as linear and / or rotary position actuators, may be used to change one or more of position, azimuth, and elevation.

[0073] Embodiments herein may provide a controller 26 for controlling the configuration of one or more radiating elements 20 by controlling one or more corresponding phase shifter and amplifier elements 22 and / or position actuators 24. Figure 2 , the controller 26 is shown as being connected to only one such phase shifter and amplifier element 22 and / or position actuator 24, but it will be appreciated that in practice the controller 26 may be connected to each device that it is to control.

[0074] In general, example embodiments may involve a controller 26 (or another processing device associated with the controller) receiving a performance metric L(Ω) for an antenna array 10 comprising a plurality of radiating elements 20. The performance metric L(Ω) may be based on performance data associated with the antenna array 10, the antenna array having a radiation configuration represented by configuration parameters. In some embodiments, the configuration parameters may correspond to one or more of the aforementioned characteristics of a given radiating element 20. For example, the jth radiating element of the i-th antenna of the antenna array may have the following configuration parameters:

[0075] -Location

[0076] -Azimuth φ i，j ∈[0, 2π]

[0077] - elevation angle θ i，j ∈[-π / 2,π / 2]

[0078] -Analog phase shift β i，j ∈[-π,π]

[0079] -Analog signal gain α i，j ∈[α min , α max ]The value of the configuration parameter may reflect the settings of the individual radiating elements 20 corresponding to the configuration.

[0080] Configuration parameters can be combined into a parameter vector:

[0081] ωi,j=[x T ,φi,j,θi,j,βi,j,ai,j] T ∈R 7 .

[0082] The antenna array 10 may have M antennas, where the i-th antenna has N i radiating elements. The set of all configuration parameters of the antenna array 10 can be expressed as

[0083] Ω={ωi, j: i=1,..., M: j=1,..., Ni}

[0084] The embodiment may further include: updating the configuration parameter Ω according to the received performance metric L(Ω), and To reconfigure the radiation configuration of the antenna array 10. Therefore, the updated configuration parameters The value of 1 reflects the updated settings to be applied to each radiating element 20. For example, this can be achieved by increasing or decreasing the phase shift αi,j, the signal gain αi,j, and then modifying the position xi,j, azimuth angle φi,j and elevation angle θi,j of each radiating element 20 via the position actuator 24.

[0085] The performance metric L(Ω) can be any metric. For example, the performance metric L(Ω) can be a value representing data throughput, such as cell aggregate throughput, 5th percentile throughput, median throughput, or geometric throughput. For example, the performance metric L(Ω) can be a value representing one or more of call loss rate, outage probability, energy consumption, etc. One or more of the above examples can be the basis for calculating the performance metric L(Ω), which is a value indicating the performance of one or more performance characteristics.

[0086] By modifying one or more configuration parameters, the performance metric L(Ω) may change. Thus, embodiments may involve modifying one or more configuration parameters Ω to move the performance metric L(Ω) toward a target. The target may represent an optimal condition, such as solving an optimization problem:

[0087] arg max L(Ω) (1)

[0088] Ω

[0089] But some other goals can be used, usually in the direction of improving technical performance.

[0090] In some embodiments, reinforcement learning (RL) and / or supervised learning (SL) methods are proposed to adapt the position and / or other controllable characteristics of the radiating elements 20 to improve (if not maximize) the performance metric L(Ω). In this way, embodiments enable the geometry of the antenna array 10 to be dynamically changed so that it can be automatically improved or optimized for any radio environment without human intervention. The main benefits are improved performance and reduced capital and overhead expenses.

[0091] The performance metric L(Ω) itself can be calculated from measurements based on real data associated with the antenna array 10, or from simulations. The performance metric L(Ω) can also be obtained in an explicit mathematical form.

[0092] Figure 3There is shown schematically a portion of another antenna array 30 comprising an antenna having twelve radiating elements 32. As indicated by arrow 34 shown relative to one such radiating element 32, each radiating element can be moved by an associated position actuator to a position xi,j.

[0093] Figure 4 A portion of another antenna array is illustrated in perspective view and may include an antenna 36 comprising an antenna mast 37 and a plurality of radiating elements 38 extending from the mast 37. As shown, the mast 37 may be rotated by a position actuator, and each radiating element 38 may be modified in terms of its angle relative to the mast, in terms of the length of each arm, and in terms of rotation about its own axis.

[0094] Figure 5 A portion of another antenna array 50, specifically a planar antenna array, is shown in top plan view. The array includes four antenna columns 51-54, each including two blocks 55, 56 of four antennas or radiating elements 57. The vertical position of each block 55, 56 of each antenna column 51-54 can be controllably moved. In another embodiment, the horizontal position of each block 55, 56 can additionally or alternatively be controllable. The movement can be controlled by any of the methods described above.

[0095] A controllable antenna array according to example embodiments may be combined Figures 3 and 4 Modification of two or more features in a

[0096] An example of how the control is achieved will now be described.

[0097] In general, control may be performed using hardware, software, or a combination thereof.

[0098] Figure 6 is a block diagram illustrating the functional modules of a controller 60, which are shown in association with an antenna array 62 and a connected base station 64. Although this embodiment is not standard-specific, depending on the standard used, the base station 64 can be, for example, a cellular base station, commonly referred to as a Node B (NB), an enhanced Node B (eNB), or a next-generation Node B (gNB). The base station 64 can be associated with a cell of a cellular communication network, in which a plurality of user equipment 66 can be located. The controller 60 can include a learning algorithm module 68 and an antenna array controller 70.

[0099] The learning algorithm can be configured to generate antenna array configurations This is then applied to the antenna array 62 by the antenna array controller 70 .

[0100] The base station may be configured to measure the performance metric L For example, the throughput achieved by the user equipment 66 within its cell is then forwarded to the learning algorithm 68 which is in turn configured to generate updated antenna configurations, etc., thereby reconfiguring the characteristics of the antenna array 62.

[0101] Figure 8 8 is a flow chart illustrating processing operations of a general method of an example embodiment. A first operation 801 includes receiving a performance metric for an antenna array. The performance metric may be received from a remote node of a base station, such as base station 64. A second operation 802 includes updating configuration parameters of antenna array 62 based on the received performance metric. A third operation 803 includes reconfiguring the antenna's radiation configuration based on the updated configuration parameters.

[0102] Now about Figure 6 We present a high-level view of the concept in

[15] , describing a more specific but still general learning algorithm applicable to any performance metric L(Ω). Figure 9 A corresponding flow chart that can represent the processing operations is provided in FIG. The reference numbers do not necessarily indicate the order of processing.

[0103] In a first operation 901 , we set t=0 and select a set of initial parameters Ωt.

[0104] In another operation 902, we calculate the perturbation configuration For example, for i = 1, ..., M, j = 1, ..., Ni, a random vector εi, j can be drawn from some distribution p(ε). For example, p(ε) can be a multivariate normal distribution with zero mean and fixed covariance matrix σ2I, that is, p(ε) = N(0, σ 2 I). For each radiating element of the antenna array 62 we can also have a different distribution pi,j(ε).

[0105] In another operation 903, the antenna array 62 may be configured according to:

[0106]

[0107] This defines the parameters

[0108] In another operation 904, the performance metric L is calculated. For example, measurement.

[0109] In another operation 905, it is determined whether a stop condition is reached. If not, another operation 906 may include calculating a parameter update based on:

[0110]

[0111] For some learning rate η>0. For p(ε)=N(O,σ 2 I), which boils down to:

[0112]

[0113] In the same or a different operation, the value of t is updated to t+1, and the process may return to operation 902 .

[0114] The initial antenna array configuration Ω can be obtained from the simulation t In this case, the performance metric L(Ω) is not measured by the actual system, such as data received from antenna array 62, but is simulated on a computer. A good initialization helps speed up the learning process. However, this is only optional and not required to implement the embodiments herein.

[0115] In some embodiments, if the updated parameter vector ω is determined t+1 or Including infeasible values, for example if they correspond to physically impossible radiating element positions, the parameter vector can be projected back to a set of feasible parameter vectors. For example, the entries of the vector can be clipped to certain minimum and / or maximum values.

[0116] Alternatively, the learning rate η can be adjusted so that i,j remains feasible.

[0117] In some embodiments, operations 903 and 904 may be run multiple times to calculate the performance metric for different perturbations for the same antenna configuration. In this case, the gradient calculated in operation 906 may be averaged over these values.

[0118] In operation 906 , the learning rate η may be calculated and adjusted by any stochastic gradient descent (SGD) algorithm, such as ADAM, RMSProp, momentum, to name a few examples.

[0119] The stopping criteria in operation 905 can take a variety of forms. For example, the process can stop after a fixed number of training iterations, or during a fixed number of iterations L Stop if there is no improvement, or Stop when the desired value is reached. In some embodiments, there may be no stopping operation 905, i.e., the training may proceed indefinitely.

[0120] The second term in Equation (3) is also called the policy gradient. Various methods exist in the (deep) learning literature to improve the convergence rate of this algorithm; see, for example, E.G. Med. Chem. Soc. R.S. Sutton, D. McAllester, S. Singh, and Y. Mansour, “Policy gradient methods for reinforcement learning with function approximation,” in Proceedings of the 12th International Conference on Neural Information Processing Systems, NIPS, pp. 1057–1063, Cambridge, MA, USA, 1999. MIT Press.

[0121] The learning algorithm for the differential performance metric L(Ω) will now be described.

[0122] In some cases, the gradient can be calculated explicitly This may occur when L(Ω) is calculated analytically / numerically or by simulation.

[0123] refer to Figure 10 The processing steps will now be described by way of example. Note that other numerical optimization algorithms can be used in this case.

[0124] In the first operation 1001, we set t=0 and select a set of initial parameters Ω t .

[0125] In another operation 1002, we calculate the t An antenna array 62 is configured.

[0126] In another operation 1003, we calculate the performance metric L(Ω t ).

[0127] In another operation 1004, it is determined whether a stop condition is reached. If not, another operation 1005 may include calculating a parameter update based on:

[0128]

[0129] For learning rate η>0.

[0130] In the same or a different operation, the value of t is updated to t+1, and the process may return to operation 1002 .

[0131] Similar to some other embodiments, if the updated parameter vector ω is determined t+1 or Including infeasible values, for example if they correspond to physically impossible radiating element positions, the parameter vector may be projected back into a set of feasible parameter vectors. For example, the entries of the vector may be clipped to certain minimum and / or maximum values.

[0132] The learning rate η can be calculated and adjusted by any stochastic gradient descent (SGD) algorithm, to give some examples such as ADAM, RMSProp, momentum.

[0133] In some embodiments, operations 1002 and 1003 may be run multiple times to calculate the performance metric for different perturbations for the same antenna configuration. In this case, the gradient calculated in operation 1006 may be averaged over these values.

[0134] In some embodiments, the above initial configuration of the antenna array 62 can be obtained through simulation. That is, the above algorithm can be used to find the parameter Ω through simulation, which is then used as the initial parameter Ω° for the algorithm running on the real system. Figure 7 As schematically shown in Figure 7 The system shown includes e.g. Figure 6 The real system 80A is shown, along with a digital sibling system 80B, a term widely used to represent a realistic simulator of a system that is desired to be improved or optimized. Digital sibling system 80B can run the simulation faster than real time and significantly increase the speed of the training process. In other embodiments, initial parameters from another base station can be used, sometimes referred to as "transfer learning."

[0135] So far, for a given performance metric L(Ω t ) optimizes the antenna configuration parameters. However, there may be other parameters s∈R that should be explicitly considered P These parameters may include, for example, the number of user equipment 66 in the cell, the time of day, weather conditions, etc.

[0136] In certain examples, the goal is to find a mapping Ω = f(s) from these additional parameters to an optimal antenna array configuration to maximize the performance metric L(Ω, s).

[0137] In this case, we can assume that the function f has an adjustable parameter μ, i.e., f(s) = fμ(s), and we want to solve the optimization problem:

[0138] arg max L(fμ(s), s). (6)

[0139] μ

[0140] For the non-differentiable performance metric L(fμ(s), s), we can obtain the following algorithm. Figure 11A flowchart of the processing operation in this case is shown.

[0141] In the first operation 1101, let t = 0 and select a set of initial parameters μ t We also calculate Ω t =fμt(s).

[0142] In another operation 1102, we calculate the configuration of the perturbation This may involve obtaining the parameter s and calculating Ω t =fμt(s). For i=1,...,M,j=1,...,Ni, we can draw a random vector εi,j from some distribution p(ε). For example, p(ε) can be a vector with zero mean and covariance matrix σ 2 I fixed multivariate normal distribution, that is, p(ε)=N(o,σ 2 I). We may also have a different distribution pi,j(ε) for each radiating element of the antenna array.

[0143] In another operation 1103, we may configure the antenna array 62 according to the following:

[0144] This defines the parameters

[0145] In another operation 1104, the performance metric L is calculated. For example, measurement.

[0146] In another operation 1105, it is determined whether a stop condition is reached. If not, another operation 1106 may include calculating a parameter update based on:

[0147]

[0148] For learning rate η>0.

[0149] In the same or a different operation, the value of t is updated to t+1, and the process may return to operation 1102 .

[0150] In some embodiments, operations 1103 and 1104 may be run multiple times to calculate the performance metric for different perturbations of the same antenna configuration. In this case, the gradient calculated in operation 1106 may be averaged over these values.

[0151] For the differentiable performance metric L(fμ(s), s), we propose the following method, which processes operations such as Figure 12 shown.

[0152] In the first operation 1201, we set t = 0 and select a set of initial parameters μ t We also calculate Ω t =fμt(s).

[0153] In another operation 1202, we calculate the perturbation configuration This may involve obtaining the parameter s and calculating Ω t =fμt(s).

[0154] In another operation 1203, we can calculate the t =fμt(s) to configure the antenna array 62.

[0155] In another operation 1204, we can calculate the performance metric L(Ω t , s).

[0156] In another operation 1205, it is determined whether a stop condition is reached. If not, another operation 1206 may include calculating a parameter update based on:

[0157]

[0158] For learning rate η>0.

[0159] In the same or a different operation, the value of t is updated to t+1, and the process may return to operation 1202 .

[0160] In some embodiments, operations 1203 and 1204 may be run multiple times to calculate the performance metric for different perturbations for the same antenna configuration. In this case, the gradient calculated in operation 1206 may be averaged over these values.

[0161] In some embodiments, the mapping fμ(s) may be performed using a neural network (NN), and μ may represent its trainable parameters, such as its weights and / or biases.

[0162] In some embodiments, if the updated parameter vector ω is determined t+1 or Including infeasible values, for example if they correspond to physically impossible radiating element positions, the parameter vector can be projected back to a set of feasible parameter vectors. For example, the entries of the vector can be clipped to certain minimum and / or maximum values. Alternatively, the learning rate η can be adjusted so that ω t+1 Always stay feasible. Another approach is to "penalize" the learning system, perhaps by feeding back a predefined negative loss when a non-feasible setting is chosen, so that by minimizing the loss, the system "learns" to avoid such configurations.

[0163] The learning algorithm may employ a machine learning model that may be initialized using data derived from initial connection quality (e.g., initial signal quality) and then updated based on the results resulting from the updated configuration parameters. The machine learning model in other embodiments may be initialized using the initial configuration parameters.

[0164] Figure 14 An apparatus according to an embodiment is shown. The apparatus may provide Figure 2 The controller functions shown and / or Figure 6 . The device comprises at least one processor 420 and at least one memory 410 directly or closely connected to the processor. The memory 410 comprises at least one random access memory (RAM) 410b and at least one read-only memory (ROM) 410a. Computer program code (software) 415 is stored in the ROM 410a. The device can be connected to, for example, the TX path of the antenna array 62 and the RX path from the base station 64 to send configuration data and receive performance metrics along the respective paths. The device can be connected to a user interface UI for instructing the device and / or for outputting results. The at least one processor 420 having at least one memory 410 and computer program code 415 can be arranged to cause the device to at least perform operations according to Figures 9 to 12 One or more methods.

[0165] FIG15 illustrates a non-transient medium according to some embodiments. A non-transient medium is a computer-readable storage medium. It may be that the non-transient medium stores computer program code, when used by a computer such as Figure 14 When the processor 42 of the device executes, the device executes according to Figures 9 to 12 One or more methods.

[0166] Memory can be volatile or non-volatile. Examples include RAM, SRAM, flash memory, FPGA module memory, DCD, CD, USB stick, and Blu-ray disc.

[0167] Unless otherwise stated or clear from the context, stating that two entities are different means that they perform different functions. This does not necessarily mean that they are based on different hardware. That is, each entity described in this specification may be based on different hardware, or some or all entities may be based on the same hardware. This does not necessarily mean that they are based on different software. That is, each entity described in this specification may be based on different software, or some or all entities may be based on the same software. Each entity described in this specification may be embodied in the cloud.

[0168] As non-limiting examples, implementations of any of the above blocks, devices, systems, techniques, or methods include some implementations as hardware, software, firmware, dedicated circuits or logic, general purpose hardware or controllers or other computing devices, or a combination thereof. Some embodiments may be implemented in the cloud.

[0169] It should be understood that the above description is what is presently considered to be the preferred embodiment. However, it should be noted that the description of the preferred embodiment is given by way of example only and that various modifications may be made without departing from the scope of the appended claims.

Claims

1. A device for controlling radiation, comprising: means for receiving a performance metric of an antenna array (10) comprising a plurality of radiating elements (20), the performance metric being based on performance data associated with the antenna array (10), the antenna array (10) having a radiating configuration represented by a configuration parameter; means for updating said configuration parameters according to said received performance metric by estimating new configuration parameters for moving said performance metric towards a target value; means for reconfiguring the radiation configuration of the antenna array (10) based on the updated configuration parameters, whereby the physical geometry of the antenna array (10) is changed; wherein the means for updating comprises a machine learning model configured to estimate, based on stored training data, updated configuration parameters that are likely to move the performance metric toward the target value; wherein the machine learning model is trained using initial configuration parameters and is subsequently updated based on results of a reconfiguration caused by the updated configuration parameters; The configuration parameters include one or more of the following: for the j-th radiating element of the i-th antenna of the antenna array (10): -Location - Azimuth φ i，j ∈[0,2π] - Elevation angle θ i，j ∈[-π / 2, π / 2] And the means for reconfiguring the radiation configuration of the antenna array (10) is configured to update one or more of the configuration parameters to produce corresponding changes in one or more of the radiation elements (20) of the antenna array (10).

2. The apparatus of claim 1, wherein the reconfiguration component is configured to change the physical geometry by moving one or more radiating elements (20) based on the reconfigured configuration.

3. The apparatus according to claim 1 or claim 2, wherein the updating component is further configured to iteratively update the configuration parameters according to the received performance metrics generated by the updated configuration parameters until a stop condition is reached. The apparatus of claim 3 , wherein the stopping condition corresponds to a fixed number of update iterations. The apparatus of claim 3 , wherein the stopping condition corresponds to the performance metric reaching the target value. 6 . The apparatus of claim 3 , wherein the stopping condition corresponds to a fixed number of update iterations in which the performance metric does not move toward the target value.

7. The apparatus of claim 1, wherein the machine learning model is trained using data derived from initial connection quality and subsequently updated based on results caused by the updated configuration parameters.

8. An apparatus according to any one of claims 1, 2, 4-7, wherein the performance metric is received from a computing component configured to receive performance data from physical nodes of at least a part of a communication network of which the antenna array (10) forms a part.

9. An apparatus according to claim 8, wherein the computing means is associated with a base station of at least part of a cellular communication network of which the antenna array (10) forms a part.

10. An apparatus according to any one of claims 1, 2, 4-7 or 9, wherein the performance metric is received from a simulation component which simulates physical nodes of at least a part of a communication network of which the antenna array (10) forms a part.

11. An apparatus according to claim 10, wherein the simulated physical node is associated with a base station of at least part of a cellular communication network of which the antenna array (10) forms a part.

12. An apparatus according to any one of claims 1, 2, 4-7, 9 or 11, wherein the received performance metric represents a measured or simulated value of one or more of data throughput, call loss rate, outage probability and energy consumption associated with the antenna array (10).

13. The apparatus of claim 12, wherein the performance metric represents the measured or simulated value for each of a different number of user equipments within a cell of a cellular communication system associated with the antenna array (10).

14. An apparatus according to any one of claims 1, 2, 4-7, 9, 11 or 13, wherein the means for reconfiguring the radiation configuration of the antenna array (10) is configured to perform the reconfiguration in a substantially real-time manner based on the last received value of the performance metric.

15. A method for controlling radiation, comprising: Receiving (801) a performance metric for an antenna array (10) comprising a plurality of radiating elements (20), the antenna array (10) having a radiation configuration represented by a configuration parameter; updating (802) the configuration parameters according to the received performance metric by estimating new configuration parameters for moving the performance metric towards a target value; as well as reconfiguring (803) the radiation configuration of the antenna array (10) based on the updated configuration parameters, whereby the physical geometry of the antenna array (10) is changed; wherein the means for updating comprises a machine learning model configured to estimate, based on stored training data, updated configuration parameters that are likely to move the performance metric toward the target value; wherein the machine learning model is trained using initial configuration parameters and is subsequently updated based on results of a reconfiguration caused by the updated configuration parameters; The configuration parameters include one or more of the following: for the j-th radiating element of the i-th antenna of the antenna array (10): -Location - Azimuth φ i，j ∈[0,2π] - Elevation angle θ i，j ∈[-π / 2, π / 2] And the means for reconfiguring the radiation configuration of the antenna array (10) is configured to update one or more of the configuration parameters to produce corresponding changes in one or more of the radiation elements (20) of the antenna array (10).

16. The method of claim 15, wherein the reconfiguring comprises: The physical geometry is changed by moving one or more radiating elements (20) based on the reconfigured configuration.

17. The method according to claim 15 or claim 16, further comprising: The configuration parameters are iteratively updated according to the received performance metrics produced by the updated configuration parameters until a stopping condition is reached. The method of claim 17 , wherein the stopping condition corresponds to a fixed number of update iterations. The method of claim 17 , wherein the stopping condition corresponds to the performance metric reaching the target value.

20. The method of claim 17, wherein the stopping condition corresponds to a fixed number of update iterations in which the performance metric does not move toward the target value.

21. The method of claim 15, further comprising: The machine learning model is trained using data derived from initial connection quality and subsequently updated based on results caused by the updated configuration parameters.

22. A method according to any of claims 15, 16, 18-21, wherein the performance metric is received from a computing component configured to receive performance data from physical nodes of at least part of a communication network of which the antenna array (10) forms a part.

23. A method according to claim 22, wherein the computing means is associated with a base station of at least part of a cellular communication network of which the antenna array (10) forms a part.

24. A method according to any one of claims 15, 16, 18-21 or 23, wherein the performance metric is received from a simulation component which simulates physical nodes of at least part of a communication network of which the antenna array (10) forms a part.

25. A method according to claim 24, wherein the simulated physical node is associated with a base station of at least part of a cellular communication network of which the antenna array (10) forms a part.

26. The method of any one of claims 15, 16, 18-21, 23 or 25, wherein the received performance metric represents a measured or simulated value of one or more of data throughput, call loss rate, outage probability and energy consumption associated with the antenna array (10).

27. The method of claim 26, wherein the performance metric represents the measured or simulated value for each of a different number of user equipment within a cell of a cellular communication system associated with the antenna array (10).

28. The method according to any one of claims 15, 16, 18-21, 23, 25 or 27, wherein the reconfiguration of the radiation configuration of the antenna array (10) comprises: The reconfiguration is performed in substantially real time based on a last received value of the performance metric.

29. An apparatus for controlling radiation, comprising: at least one memory directly connected to at least one processor, the at least one memory comprising computer program code; and said at least one processor, said at least one processor together with said at least one memory and said computer program code being arranged to perform the following method: Receiving (801) a performance metric for an antenna array (10) comprising a plurality of radiating elements (20), the antenna array (10) having a radiation configuration represented by a configuration parameter; updating (802) the configuration parameters according to the received performance metric by estimating new configuration parameters for moving the performance metric towards a target value; as well as reconfiguring (803) the radiation configuration of the antenna array (10) based on the updated configuration parameters, whereby the physical geometry of the antenna array (10) is changed; wherein the means for updating comprises a machine learning model configured to estimate, based on stored training data, updated configuration parameters that are likely to move the performance metric toward the target value; wherein the machine learning model is trained using initial configuration parameters and is subsequently updated based on results of a reconfiguration caused by the updated configuration parameters; The configuration parameters include one or more of the following: for the j-th radiating element of the i-th antenna of the antenna array (10): -Location - Azimuth φ i，j ∈[0,2π] - Elevation angle θ i，j ∈[-π / 2, π / 2] And the means for reconfiguring the radiation configuration of the antenna array (10) is configured to update one or more of the configuration parameters to produce corresponding changes in one or more of the radiation elements (20) of the antenna array (10).

30. The apparatus of claim 29, wherein the at least one processor with the at least one memory and the computer program code are further arranged to perform the method of any one of claims 16 to 28.

31. A computer program product, comprising a set of instructions, which, when executed on a device, are configured to cause the device to perform the following method: Receiving (801) a performance metric for an antenna array (10) comprising a plurality of radiating elements (20), the antenna array (10) having a radiation configuration represented by a configuration parameter; updating (802) configuration parameters according to the received performance metric by estimating new configuration parameters for moving the performance metric towards a target value; as well as reconfiguring (803) the radiation configuration of the antenna array (10) based on the updated configuration parameters, whereby the physical geometry of the antenna array (10) is changed; wherein the means for updating comprises a machine learning model configured to estimate, based on stored training data, updated configuration parameters that are likely to move the performance metric toward the target value; wherein the machine learning model is trained using initial configuration parameters and is subsequently updated based on results of a reconfiguration caused by the updated configuration parameters; The configuration parameters include one or more of the following: for the j-th radiating element of the i-th antenna of the antenna array (10): -Location - Azimuth φ i，j ∈[0,2π] - Elevation angle θ i，j ∈[-π / 2, π / 2] And the means for reconfiguring the radiation configuration of the antenna array (10) is configured to update one or more of the configuration parameters to produce corresponding changes in one or more of the radiation elements (20) of the antenna array (10).

32. The computer program product of claim 31 , wherein the instructions, when executed on an apparatus, are configured to perform the method of any one of claims 16 to 28.

33. A non-transitory computer readable medium comprising program instructions stored on the non-transitory computer readable medium for performing a method comprising: Receiving (801) a performance metric for an antenna array (10) comprising a plurality of radiating elements (20), the antenna array (10) having a radiation configuration represented by a configuration parameter; updating (802) configuration parameters according to the received performance metric by estimating new configuration parameters for moving the performance metric towards a target value; as well as reconfiguring (803) the radiation configuration of the antenna array (10) based on the updated configuration parameters, whereby the physical geometry of the antenna array (10) is changed; wherein the means for updating comprises a machine learning model configured to estimate, based on stored training data, updated configuration parameters that are likely to move the performance metric toward the target value; wherein the machine learning model is trained using initial configuration parameters and is subsequently updated based on results of a reconfiguration caused by the updated configuration parameters; The configuration parameters include one or more of the following: for the j-th radiating element of the i-th antenna of the antenna array (10): -Location - Azimuth φ i，j ∈[0,2π] - Elevation angle θ i，j ∈[-π / 2, π / 2] And the means for reconfiguring the radiation configuration of the antenna array (10) is configured to update one or more of the configuration parameters to produce corresponding changes in one or more of the radiation elements (20) of the antenna array (10).

34. An apparatus for controlling radiation, comprising: at least one processor; and at least one memory, the at least one memory comprising computer program code which, when executed by the at least one processor, causes the apparatus to: receiving a performance metric for an antenna array including a plurality of radiating elements, the antenna array having a radiation configuration represented by a configuration parameter; updating the configuration parameters according to the received performance metric by estimating new configuration parameters for moving the performance metric towards a target value; as well as reconfiguring the radiation configuration of the antenna array based on the updated configuration parameters, whereby a physical geometry of the antenna array is changed; wherein the means for updating comprises a machine learning model configured to estimate, based on stored training data, updated configuration parameters that are likely to move the performance metric toward the target value; wherein the machine learning model is trained using initial configuration parameters and is subsequently updated based on results of a reconfiguration caused by the updated configuration parameters; The configuration parameters include one or more of the following: for the j-th radiating element of the i-th antenna of the antenna array (10): -Location - Azimuth φ i，j ∈[0,2π] - Elevation angle θ i，j ∈[-π / 2, π / 2] And the means for reconfiguring the radiation configuration of the antenna array (10) is configured to update one or more of the configuration parameters to produce corresponding changes in one or more of the radiation elements (20) of the antenna array (10).

Images