Wireless communications system, a radio network node, a machine learning UNT and methods therein for transmission of a downlink signal in a wireless communications network supporting beamforming

1. A method performed in a radio network node for transmission of a downlink signal in a wireless communications network supporting beamforming, wherein the method comprises:

estimating uplink channel information for a radio link from a wireless device to the radio network node, wherein the wireless device is operating within a radio cell sector comprised in the wireless communications network;

obtaining downlink beamforming information applicable for the radio cell sector and related to the estimated uplink channel information, wherein the downlink beamforming information is obtained from a machine learning unit and is determined from relationships, between beamforming information and uplink channel information, that are determined based on stored pairs of reported downlink beamforming information and measured uplink channel information; and

transmitting, towards the wireless device, a beamformed downlink signal using the obtained downlink beamforming information.

2. The method of claim 1 , comprising:

triggering the wireless device to report downlink beamforming information used in transmission of one or more downlink signals, by the radio network node, that are received by the wireless device;

obtaining the reported downlink beamforming information from the wireless device;

measuring uplink channel information based on one or more uplink signals transmitted using the radio link from the wireless device to the radio network node; and

storing, in association with the machine learning unit, one or more pairs of reported downlink beamforming information and measured uplink channel information wherein each pair comprises reported downlink beamforming information and corresponding measured uplink channel information for the radio link.

3. The method of claim 1 , wherein the triggering and the measuring are performed during a selected period of time.

4. The method of claim 1 , further comprising training the machine learning unit, via neural network methods based on the stored pairs of reported downlink beamforming information and measured uplink channel information, to determine the relationships between beamforming information and uplink channel information.

5. The method of claim 1 , wherein

the estimated uplink channel information is associated with a carrier frequency f 0 for the radio link, and

the obtained output downlink beamforming information is associated with a target frequency f 1 for the radio link.

6. A method for a machine learning unit for a wireless communications network that is configured to support transmission of downlink signals via beamforming, wherein the method comprises:

performing a neural network training method on stored pairs of reported downlink beamforming information and measured uplink channel information to determine relationships between downlink beamforming information and uplink channel information;

retrieving downlink beamforming information related to received estimated uplink channel information for a radio link between a wireless device and a radio network node, wherein the downlink beamforming information is retrieved based on the determined relationships and wherein the downlink beamforming information is applicable for a radio cell sector comprised in the wireless communications network; and

providing the obtained downlink beamforming information to a radio network node, thereby supporting the radio network node to transmit, towards the wireless device operating within the radio cell sector, a beamformed downlink signal using the downlink beamforming information.

7. The method of claim 6 , wherein:

the neural network training method is performed using a neural network that is specific for the radio cell sector or a plurality of radio cell sectors including the radio cell sector; and

the neural network comprises an input layer, an output layer and one or more hidden layers;

each of the layers comprises one or more artificial neurons linked to one or more other artificial neurons of the same layer or of another layer;

each of the one or more artificial neurons has an activation function, an input weighting coefficient, a bias, and an output weighting coefficient; and

the weighting coefficients and the bias are changeable during training of the neural network.

8. The method of claim 6 , wherein performing the neural network training method comprises:

providing a known pair to the neural network, wherein the known pair includes known uplink channel information and corresponding known downlink beamforming information; and

for one or more of the artificial neurons, adjusting one or more of the input weighting coefficients, biases, and output weighting coefficients until the known downlink beamforming information is output from the neural network responsive to the corresponding known uplink channel information as an input to the neural network.

9. The method of claim 6 , wherein performing the neural network training method comprises, for one or more of the artificial neurons, calculating one or more of the input weighting coefficients, biases, and output weighting coefficients by executing a back propagation algorithm or a genetic algorithm on stored pairs of reported downlink beamforming information and measured uplink channel information.

10. The method of claim 6 , wherein:

the estimated uplink channel information is associated with a carrier frequency f 0 for the radio link, and

the obtained output downlink beamforming information is associated with a target frequency f 1 for the radio link.

11. A radio network node configured for transmission of a downlink signal in a wireless communications network arranged to support beamforming, wherein the radio network node comprises:

a wireless transmitter/receiver configured for downlink/uplink communication in a radio cell sector of the wireless communication network; and

processing circuitry operably coupled to the wireless transmitter/receiver, whereby the processing circuitry and the wireless transmitter/receiver are configured to:

estimate uplink channel information for a radio link from a wireless device to the radio network node, wherein the wireless device is operating within a radio cell sector comprised in the wireless communications network;

obtain downlink beamforming information applicable for the radio cell sector and related to the estimated uplink channel information, wherein the downlink beamforming information is obtained from a machine learning unit and is determined from relationships, between beamforming information and uplink channel information, that are determined based on stored pairs of reported downlink beamforming information and measured uplink channel information; and

transmit, towards the wireless device, a beamformed downlink signal using the obtained downlink beamforming information.

12. The radio network node of claim 11 , wherein the processing circuitry and the wireless transmitter/receiver are further configured to:

trigger the wireless device to report downlink beamforming information used in transmission of one or more downlink signals, by the radio network node, that are received by the wireless device;

obtain the reported downlink beamforming information from the wireless device;

measure uplink channel information based on one or more uplink signals transmitted using the radio link from the wireless device to the radio network node; and

store, in association with the machine learning unit, one or more pairs of reported downlink beamforming information and measured uplink channel information wherein each pair comprises reported downlink beamforming information and corresponding measured uplink channel information for the radio link.

13. The radio network node of claim 11 , wherein the processing circuitry and the wireless transmitter/receiver are further configured to trigger the wireless device and measure the uplink channel information during a selected period of time.

14. The radio network node of claim 11 , wherein the processing circuitry and the wireless transmitter/receiver are further configured to train the machine learning unit, via neural network methods based on the stored pairs of reported downlink beamforming information and measured uplink channel information, to determine the relationships between beamforming information and uplink channel information.

15. The radio network node of claim 11 , wherein

the estimated uplink channel information is associated with a carrier frequency f 0 for the radio link, and

the obtained output downlink beamforming information is associated with a target frequency f 1 for the radio link.

16. A machine learning unit for a wireless communications network that is configured to support transmission of downlink signals via beamforming, wherein the machine learning unit comprises processing circuitry configured to:

perform a neural network training method on stored pairs of reported downlink beamforming information and measured uplink channel information to determine relationships between downlink beamforming information and uplink channel information;

retrieve downlink beamforming information related to received estimated uplink channel information for a radio link between a wireless device and a radio network node, wherein the downlink beamforming information is retrieved based on the determined relationships and wherein the downlink beamforming information is applicable for a radio cell sector comprised in the wireless communications network; and

provide the obtained downlink beamforming information to a radio network node, thereby supporting the radio network node to transmit, towards the wireless device operating within the radio cell sector, a beamformed downlink signal using the downlink beamforming information.

17. The machine learning unit of claim 16 , wherein:

the processing circuitry is configured to perform the neural network training method using a neural network that is specific for the radio cell sector or a plurality of radio cell sectors including the radio cell sector; and

the neural network comprises an input layer, an output layer and one or more hidden layers;

each of the layers comprises one or more artificial neurons linked to one or more other artificial neurons of the same layer or of another layer;

each of the one or more artificial neurons has an activation function, an input weighting coefficient, a bias, and an output weighting coefficient; and

the weighting coefficients and the bias are changeable during training of the neural network.

18. The machine learning unit of claim 16 , wherein the processing circuitry is configured to perform the neural network training method by:

providing a known pair to the neural network, wherein the known pair includes known uplink channel information and corresponding known downlink beamforming information; and

for one or more of the artificial neurons, adjusting one or more of the input weighting coefficients, biases, and output weighting coefficients until the known downlink beamforming information is output from the neural network responsive to the corresponding known uplink channel information as an input to the neural network.

19. The machine learning unit of claim 16 , wherein the processing circuitry is configured to perform the neural network training method by, for one or more of the artificial neurons, calculating one or more of the input weighting coefficients, biases, and output weighting coefficients by executing a back propagation algorithm or a genetic algorithm on stored pairs of reported downlink beamforming information and measured uplink channel information.

20. The machine learning unit of claim 16 , wherein:

the estimated uplink channel information is associated with a carrier frequency f 0 for the radio link, and

the obtained output downlink beamforming information is associated with a target frequency f 1 for the radio link.