Learning and deployment of adaptive wireless communications

• US 10,217,047 B2
• Filed: 05/03/2018
• Issued: 02/26/2019
• Est. Priority Date: 05/03/2017
• Status: Active Grant

• Current Assignee: Virginia Tech Intellectual Properties, Inc.
• Original Assignee: Virginia Tech Intellectual Properties, Inc.
• Inventors: O'Shea, Timothy James
• Primary Examiner(s): Chen, Alan
• Application Number:
  US15/970,324
• Publication Number:
  US 20180322388A1
• Time in Patent Office: 299 Days
• Field of Search: None
• CPC Class Codes: G06N 3/04 Architecture, e.g. interconnection topology
• CPC Class Codes: G06N 3/044 Recurrent networks, e.g. Hopfield networks
• CPC Class Codes: G06N 3/045 Combinations of networks
• CPC Class Codes: G06N 3/08 Learning methods
• CPC Class Codes: G06N 3/082 modifying the architecture, e.g. adding, deleting or silencing nodes or connections
• CPC Class Codes: H04L 1/0001 Systems modifying transmission characteristics according to link quality, e.g. power backoff adaptive data allocation for multicarrier modulation H04L5/0044; controlling transmission power for radio systems H04W52/04
• CPC Class Codes: H04L 1/0033 arrangements specific to the transmitter
• CPC Class Codes: H04L 1/0036 arrangements specific to the receiver

Abstract

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training and deploying machine-learned communication over radio frequency (RF) channels. One of the methods includes: determining first information; using an encoder machine-learning network to process the first information and generate a first RF signal for transmission through a communication channel; determining a second RF signal that represents the first RF signal having been altered by transmission through the communication channel; using a decoder machine-learning network to process the second RF signal and generate second information as a reconstruction of the first information; calculating a measure of distance between the second information and the first information; and updating at least one of the encoder machine-learning network or the decoder machine-learning network based on the measure of distance between the second information and the first information.

Claims

1. A method performed by at least one processor to train at least one machine-learning network to communicate over a communication channel, the method comprising:

determining first information;

using an encoder machine-learning network to process the first information and generate a first RF signal for transmission through a communication channel;

determining a second RF signal that represents the first RF signal having been altered by transmission through the communication channel;

using a decoder machine-learning network to process the second RF signal and generate second information as a reconstruction of the first information;

calculating a measure of distance between the second information and the first information; and

updating at least one of the encoder machine-learning network or the decoder machine-learning network based on the measure of distance between the second information and the first information, wherein the updating comprises;

determining an objective function comprising the measure of distance between the second information and the first information;

calculating a rate of change of the objective function relative to variations in at least one of the encoder machine-learning network or the decoder machine-learning network;

selecting, based on the calculated rate of change of the objective function, at least one of a first variation for the encoder machine-learning network or a second variation for the decoder machine-learning network; and

updating at least one of the encoder machine-learning network or the decoder machine-learning network based on the at least one of the selected first variation for the encoder machine-learning network or the selected second variation for the decoder machine-learning network.

2. The method of claim 1, wherein the measure of distance between the second information and the first information comprises at least one of (i) a cross-entropy between the second information and the first information, or (ii) a geometric distance metric between the second information and the first information.

3. The method of claim 1, wherein updating at least one of the encoder machine-learning network or the decoder machine-learning network comprises at least one of:

updating at least one encoding network weight or network connectivity in one or more layers of the encoder machine-learning network, or

updating at least one decoding network weight or network connectivity in one or more layers of the decoder machine-learning network.

4. The method of claim 1, wherein updating at least one of the encoder machine-learning network or the decoder machine-learning network further comprises:

determining a channel mode, from among a plurality of channel modes, that represents a state of the communication channel; and

updating at least one of the encoder machine-learning network or the decoder machine-learning network based on the channel mode of the communication channel.

5. The method of claim 1, wherein the encoder machine-learning network and the decoder machine-learning network are jointly trained as an auto-encoder to learn communication over a communication channel, and

wherein the auto-encoder comprises at least one channel-modeling layer representing effects of the communication channel on transmitted waveforms.

6. The method of claim 5, wherein the at least one channel-modeling layer represents at least one of (i) additive Gaussian thermal noise in the communication channel, (ii) delay spread caused by time-varying effects of the communication channel, (iii) phase noise caused by transmission and reception over the communication channel, or (iv) offsets in phase, frequency, or timing caused by transmission and reception over the communication channel.

7. The method of claim 1, wherein at least one of the encoder machine-learning network or the decoder machine-learning network comprises at least one of a deep dense neural network (DNN), a convolutional neural network (CNN), or a recurrent neural network (RNN) comprising parametric multiplications, additions, and non-linearities.

8. The method of claim 1, further comprising:

processing the first RF signal to generate a first analog RF waveform that is input into the communication channel;

receiving a second analog RF waveform as an output of the communication channel that represents the first analog RF waveform having been altered by the communication channel; and

processing the second analog RF waveform to generate the second RF signal.

9. The method of claim 1, wherein the communication channel comprises at least one of a radio communication channel, an acoustic communication channel, or an optical communication channel.

10. A system comprising:

at least one processor; and

at least one computer memory coupled to the at least one processor having stored thereon instructions which, when executed by the at least one processor, cause the at least one processor to perform operations comprising;

determining first information;

using an encoder machine-learning network to process the first information and generate a first RF signal for transmission through a communication channel;

determining a second RF signal that represents the first RF signal having been altered by transmission through the communication channel;

using a decoder machine-learning network to process the second RF signal and generate second information as a reconstruction of the first information;

calculating a measure of distance between the second information and the first information; and

updating at least one of the encoder machine-learning network or the decoder machine-learning network based on the measure of distance between the second information and the first information, wherein the updating comprises;

determining an objective function comprising the measure of distance between the second information and the first information;

calculating a rate of change of the objective function relative to variations in at least one of the encoder machine-learning network or the decoder machine-learning network;

selecting, based on the calculated rate of change of the objective function, at least one of a first variation for the encoder machine-learning network or a second variation for the decoder machine-learning network; and

updating at least one of the encoder machine-learning network or the decoder machine-learning network based on the at least one of the selected first variation for the encoder machine-learning network or the selected second variation for the decoder machine-learning network.

11. The system of claim 10, wherein the measure of distance between the second information and the first information comprises at least one of (i) a cross-entropy between the second information and the first information, or (ii) a geometric distance metric between the second information and the first information.

12. The system of claim 10, wherein updating at least one of the encoder machine-learning network or the decoder machine-learning network comprises at least one of:

updating at least one encoding network weight or network connectivity in one or more layers of the encoder machine-learning network, or

updating at least one decoding network weight or network connectivity in one or more layers of the decoder machine-learning network.

13. The system of claim 10, wherein updating at least one of the encoder machine-learning network or the decoder machine-learning network further comprises:

determining a channel mode, from among a plurality of channel modes, that represents a state of the communication channel; and

updating at least one of the encoder machine-learning network or the decoder machine-learning network based on the channel mode of the communication channel.

14. The system of claim 10, wherein the encoder machine-learning network and the decoder machine-learning network are jointly trained as an auto-encoder to learn communication over a communication channel, and

wherein the auto-encoder comprises at least one channel-modeling layer representing effects of the communication channel on transmitted waveforms.

15. The system of claim 14, wherein the at least one channel-modeling layer represents at least one of (i) additive Gaussian thermal noise in the communication channel, (ii) delay spread caused by time-varying effects of the communication channel, (iii) phase noise caused by transmission and reception over the communication channel, or (iv) offsets in phase, frequency, or timing caused by transmission and reception over the communication channel.

16. The system of claim 10, wherein at least one of the encoder machine-learning network or the decoder machine-learning network comprises at least one of a deep dense neural network (DNN) a convolutional neural network (CNN), or a recurrent neural network (RNN) comprising parametric multiplications, additions, and non-linearities.

17. The system of claim 10, wherein the operations further comprise:

processing the first RF signal to generate a first analog RF waveform that is input into the communication channel;

receiving a second analog RF waveform as an output of the communication channel that represents the first analog RF waveform having been altered by the communication channel; and

processing the second analog RF waveform to generate the second RF signal.

18. The system of claim 10, wherein the communication channel comprises at least one of a radio communication channel, an acoustic communication channel, or an optical communication channel.