Processing of communications signals using machine learning
One or more processors control processing of radio frequency (RF) signals using a machine-learning network. The one or more processors receive as input, to a radio communications apparatus, a first representation of an RF signal, which is processed using one or more radio stages, providing a second representation of the RF signal. Observations about, and metrics of, the second representation of the RF signal are obtained. Past observations and metrics are accessed from storage. Using the observations, metrics and past observations and metrics, parameters of a machine-learning network, which implements policies to process RF signals, are adjusted by controlling the radio stages. In response to the adjustments, actions performed by one or more controllers of the radio stages are updated. A representation of a subsequent input RF signal is processed using the radio stages that are controlled based on actions including the updated one or more actions.
1. A method performed by one or more processors to control processing of radio frequency (RF) signals using a machine-learning network, the method comprising:
receiving a first RF signal input to a radio communications apparatus, wherein the radio communications apparatus includes a plurality of radio stages and a plurality of controllers, each controller of the plurality of controllers configured to control a particular radio stage of the plurality of radio stages, the plurality of radio stages including a radio antenna and the plurality of controllers including an antenna controller for controlling the radio antenna;
processing the first RF signal using at least the radio antenna of the plurality of radio stages of the radio communications apparatus;
in response to the processing, generating, at an output of the radio antenna, a second RF signal for transmission in a communications system;
obtaining one or more observations about the second RF signal;
determining one or more metrics of the second RF signal;
adjusting, using at least one of the one or more observations or the one or more metrics, parameters of a machine-learning network, the parameters corresponding to one or more policies to process RF signals by controlling one or more radio stages of the plurality of radio stages, the one or more radio stages including the radio antenna;
in response to adjusting the parameters of the machine-learning network, updating one or more actions performed by one or more controllers corresponding to the one or more radio stages, including updating one or more actions performed by the antenna controller to select at least one antenna parameter from one of azimuth, elevation, impedance matching, antenna port, or another antenna operating characteristic; and
processing a subsequent RF signal input to the radio communications apparatus using at least the radio antenna controlled by the antenna controller based at least on the updated one or more actions.
2. The method of claim 1 , wherein the plurality of radio stages includes one or more programmable gain amplifiers or attenuators and the plurality of controllers includes a gain controller for controlling the one or more programmable gain amplifiers or attenuators,
wherein controlling the one or more radio stages of the plurality of radio stages by the machine-learning network includes controlling the gain controller, and
wherein updating the one or more actions performed by the one or more controllers includes controlling, using the gain controller, the one or more programmable gain amplifiers or attenuators to apply a different amount of gain to RF signals.
3. The method of claim 1 , wherein updating the one or more actions performed by the one or more controllers includes updating a selection of time-frequency-spatial spectrum resources to allocate to users in the communications system.
4. The method of claim 3 , wherein the one or more policies includes a policy corresponding to at least one of an optimal allocation of spectrum to users, an optimal allocation of time to users, an optimal allocation of frequency to users, or an optimal allocation of spatial slots to users.
5. The method of claim 1 , wherein the one or more policies includes a policy to detect presence of channel interference or distortion in communications channels in the communications system.
6. The method of claim 1 , wherein adjusting the parameters corresponding to the one or more policies comprises:
adjusting the parameters corresponding to the one or more policies to effect updates in processing performed by at least one radio stage to achieve at least one target objective;
following adjusting the parameters, executing the one or more adjusted policies;
in response to executing the one or more adjusted policies, determining one or more actions to be performed by a controller associated with the at least one radio stage to achieve the at least one target objective; and
providing the determined one or more actions to the controller.
7. The method of claim 6 , wherein the at least one target objective includes one of maximizing user experience, maximizing throughput, minimizing latency, minimizing overhead, minimizing detection time for particular RF signals, identifying particular types of bursts for wideband extraction, maximizing resource utilization for multi-user capacity, robust signal detection, identification or reception of a radio signal or interference, mitigation of interference, low bit error rate, low power usage, low bandwidth requirement, low processing complexity, or optimal resource allocation.
8. The method of claim 1 , wherein the machine-learning network includes one of an artificial neural network (ANN), a deep dense neural network (DenseNN) or a convolutional neural network (ConvNN) comprising a series of parametric multiplications, additions, and non-linearities, and
wherein adjusting parameters of the machine-learning network includes updating at least one of:
a connectivity in one or more layers of the ANN, or a weight of a connection in the one or more layers of the ANN, or
adjusting parameters of the machine-learning network using reinforcement learning, Deep Q-Learning, Double Q-Learning, policy gradients, or an actor critic method.
9. The method of claim 1 , wherein the one or more observations include at least one of power spectrum estimates, cyclic features, time domain samples, current settings of the plurality of controllers, or historical information obtained by the radio communications apparatus from past experiences, and
wherein the one or more metrics include at least one of: a number of RF signals found, signal-to-noise ratio (SNR), symbol error distance, bit error rate (BER), frame error rate (FER), presence or type of electro-magnetic interference or distortion, information about signal types of the RF signals, information about users in the communications system, information describing a degree of goodness of the second RF signal, information about digital data corresponding to the second RF signal, spectral usage, or availability of spectral resources.
10. The method of claim 1 , wherein the communications system includes a cellular communications network, and wherein the radio communications apparatus includes one of a base station in the cellular communications network or a cellular phone in the cellular communications network.
11. One or more non-transitory machine-readable media storing instructions that, when executed by one or more processors, are configured to cause the one or more processors to perform operations to control processing of radio frequency (RF) signals using a machine-learning network, the operations comprising:
receiving a first RF signal input to a radio communications apparatus, wherein the radio communications apparatus includes a plurality of radio stages and a plurality of controllers, each controller of the plurality of controllers configured to control a particular radio stage of the plurality of radio stages, the plurality of radio stages including a radio antenna and the plurality of controllers including an antenna controller for controlling the radio antenna;
processing the first RF signal using at least the radio antenna of the plurality of radio stages of the radio communications apparatus;
in response to the processing, generating, at an output of the radio antenna, a second RF signal for transmission in a communications system;
obtaining one or more observations about the second RF signal;
determining one or more metrics of the second RF signal;
adjusting, using at least one of the one or more observations or the one or more metrics, parameters of a machine-learning network, the parameters corresponding to one or more policies to process RF signals by controlling one or more radio stages of the plurality of radio stages, the one or more radio stages including the radio antenna;
in response to adjusting the parameters of the machine-learning network, updating one or more actions performed by one or more controllers corresponding to the one or more radio stages, including updating one or more actions performed by the antenna controller to select at least one antenna parameter from one of azimuth, elevation, impedance matching, antenna port, or another antenna operating characteristic; and
processing a subsequent RF signal input to the radio communications apparatus using at least the radio antenna controlled by the antenna controller based at least on the updated one or more actions.
12. The one or more non-transitory machine-readable media of claim 11 , wherein the plurality of radio stages includes one or more programmable gain amplifiers or attenuators and the plurality of controllers includes a gain controller for controlling the one or more programmable gain amplifiers or attenuators,
wherein controlling the one or more radio stages of the plurality of radio stages by the machine-learning network includes controlling the gain controller, and
wherein updating the one or more actions performed by the one or more controllers includes controlling, using the gain controller, the one or more programmable gain amplifiers or attenuators to apply a different amount of gain to RF signals.
13. The one or more non-transitory machine-readable media of claim 11 , wherein updating the one or more actions performed by the one or more controllers includes updating a selection of time-frequency-spatial spectrum resources to allocate to users in the communications system.
14. The one or more non-transitory machine-readable media of claim 13 , wherein the one or more policies includes a policy corresponding to at least one of an optimal allocation of spectrum to users, an optimal allocation of time to users, an optimal allocation of frequency to users, or an optimal allocation of spatial slots to users.
15. The one or more non-transitory machine-readable media of claim 11 , wherein the one or more policies includes a policy to detect presence of channel interference or distortion in communications channels in the communications system.
16. The one or more non-transitory machine-readable media of claim 11 , wherein adjusting the parameters corresponding to the one or more policies comprises:
adjusting the parameters corresponding to the one or more policies to effect updates in processing performed by at least one radio stage to achieve at least one target objective;
following adjusting the parameters, executing the one or more adjusted policies;
in response to executing the one or more adjusted policies, determining one or more actions to be performed by a controller associated with the at least one radio stage to achieve the at least one target objective; and
providing the determined one or more actions to the controller.
17. The one or more non-transitory machine-readable media of claim 16 , wherein the at least one target objective includes one of maximizing user experience, maximizing throughput, minimizing latency, minimizing overhead, minimizing detection time for particular RF signals, identifying particular types of bursts for wideband extraction, maximizing resource utilization for multi-user capacity, robust signal detection, identification or reception of a radio signal or interference, mitigation of interference, low bit error rate, low power usage, low bandwidth requirement, low processing complexity, or optimal resource allocation.
18. The one or more non-transitory machine-readable media of claim 11 , wherein the machine-learning network includes one of an artificial neural network (ANN), a deep dense neural network (DenseNN) or a convolutional neural network (ConvNN) comprising a series of parametric multiplications, additions, and non-linearities, and
wherein adjusting parameters of the machine-learning network includes updating at least one of:
a connectivity in one or more layers of the ANN, or a weight of a connection in the one or more layers of the ANN, or
adjusting parameters of the machine-learning network using reinforcement learning, Deep Q-Learning, Double Q-Learning, policy gradients, or an actor critic method.
19. The one or more non-transitory machine-readable media of claim 11 , wherein the one or more observations include at least one of power spectrum estimates, cyclic features, time domain samples, current settings of the plurality of controllers, or historical information obtained by the radio communications apparatus from past experiences, and
wherein the one or more metrics include at least one of: a number of RF signals found, signal-to-noise ratio (SNR), symbol error distance, bit error rate (BER), frame error rate (FER), presence or type of electro-magnetic interference or distortion, information about signal types of the RF signals, information about users in the communications system, information describing a degree of goodness of the second RF signal, information about digital data corresponding to the second RF signal, spectral usage, or availability of spectral resources.
20. The one or more non-transitory machine-readable media of claim 11 , wherein the communications system includes a cellular communications network, and wherein the radio communications apparatus includes one of a base station in the cellular communications network or a cellular phone in the cellular communications network.
21. A communications system comprising:
one or more processors; and
one or more machine-readable media storing instructions that, when executed by the one or more processors, are configured to cause the one or more processors to perform operations to control processing of radio frequency (RF) signals using a machine-learning network, the operations comprising:
receiving a first RF signal input to a radio communications apparatus, wherein the radio communications apparatus includes a plurality of radio stages and a plurality of controllers, each controller of the plurality of controllers configured to control a particular radio stage of the plurality of radio stages, the plurality of radio stages including a radio antenna and the plurality of controllers including an antenna controller for controlling the radio antenna;
processing the first RF signal using at least the radio antenna of the plurality of radio stages of the radio communications apparatus;
in response to the processing, generating, at an output of the radio antenna, a second RF signal for transmission in the communications system;
obtaining one or more observations about the second RF signal;
determining one or more metrics of the second RF signal;
adjusting, using at least one of the one or more observations or the one or more metrics, parameters of a machine-learning network, the parameters corresponding to one or more policies to process RF signals by controlling one or more radio stages of the plurality of radio stages, the one or more radio stages including the radio antenna;
in response to adjusting the parameters of the machine-learning network, updating one or more actions performed by one or more controllers corresponding to the one or more radio stages, including updating one or more actions performed by the antenna controller to select at least one antenna parameter from one of azimuth, elevation, impedance matching, antenna port, or another antenna operating characteristic; and
processing a subsequent RF signal input to the radio communications apparatus using at least the radio antenna controlled by the antenna controller based at least on the updated one or more actions.
22. The communications system of claim 21 , wherein the plurality of radio stages includes one or more programmable gain amplifiers or attenuators and the plurality of controllers includes a gain controller for controlling the one or more programmable gain amplifiers or attenuators,
wherein controlling the one or more radio stages of the plurality of radio stages by the machine-learning network includes controlling the gain controller, and
wherein updating the one or more actions performed by the one or more controllers includes controlling, using the gain controller, the one or more programmable gain amplifiers or attenuators to apply a different amount of gain to RF signals.
23. The communications system of claim 21 , wherein updating the one or more actions performed by the one or more controllers includes updating a selection of time-frequency-spatial spectrum resources to allocate to users in the communications system.
24. The communications system of claim 23 , wherein the one or more policies includes a policy corresponding to at least one of an optimal allocation of spectrum to users, an optimal allocation of time to users, an optimal allocation of frequency to users, or an optimal allocation of spatial slots to users.
25. The communications system of claim 21 , wherein the communication system includes a cellular communications network, and wherein the radio communications apparatus includes one of a base station in the cellular communications network or a cellular phone in the cellular communications network.