阅读说明：本技术 自适应无线通信的学习与部署 (Learning and deployment of adaptive wireless communications ) 是由 T·J·奥谢 于 2018-05-03 设计创作，主要内容包括：本发明涉及用于训练和部署通过射频RF信道进行的机器学习通信的方法、系统和设备,包含编码在计算机存储媒体上的计算机程序。所述方法中的一种方法包含：确定第一信息；使用编码器机器学习网络处理所述第一信息并生成第一RF信号以通过通信信道传输；确定第二RF信号,所述第二RF信号表示通过经由所述通信信道的传输改变的所述第一RF信号；使用解码器机器学习网络处理所述第二RF信号并生成第二信息作为对所述第一信息的重构；计算所述第二信息与所述第一信息之间的距离量度；以及基于所述第二信息与所述第一信息之间的所述距离量度更新所述编码器机器学习网络或所述解码器机器学习网络中的至少一个。(The present invention relates to methods, systems, and apparatus, including computer programs encoded on computer storage media, for training and deploying machine learning communications over Radio Frequency (RF) channels. One of the methods comprises: determining first information; processing the first information using an encoder machine learning network and generating a first RF signal for transmission over a communication channel; determining a second RF signal representative of the first RF signal altered by transmission via the communication channel; processing the second RF signal using a decoder machine learning network and generating second information as a reconstruction of the first information; calculating a distance measure 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 distance metric between the second information and the first information.)

1. A method executed 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;

processing the first information using an encoder machine learning network and generating a first Radio Frequency (RF) signal for transmission over a communication channel;

determining a second RF signal representative of the first RF signal altered by transmission via the communication channel;

processing the second RF signal using a decoder machine learning network and generating second information as a reconstruction of the first information;

calculating a distance measure 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 distance metric between the second information and the first information.

2. The method of claim 1, wherein updating at least one of the encoder machine learning network or the decoder machine learning network based on the distance metric between the second information and the first information comprises:

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

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

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

3. The method of claim 1, wherein the distance metric 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.

4. 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 of 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.

5. 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 representing a state of the communication channel from a plurality of channel modes; and

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

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

wherein the autoencoder includes at least one channel modeling layer that represents an effect of the communication channel on a transmitted waveform.

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

8. 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), the DNN, CNN, or RNN comprising parametric multiplications, additions, and nonlinear characteristics.

9. The method of claim 1, further comprising:

processing the first RF signal to generate a first analog RF waveform, the first analog RF waveform being input into the communication channel;

receiving a second analog RF waveform as an output of the communication channel, the output representing the first analog RF waveform as changed by the communication channel; and

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

10. 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.

11. A method of transmitting and receiving information over a communication channel, the method comprising:

determining an encoder and a decoder, at least one of the encoder and the decoder configured to implement encoding or decoding based on at least one of an encoder machine learning network or a decoder machine learning network trained to encode or decode information over a communication channel;

determining first information;

processing the first information using the encoder and generating a first RF signal;

transmitting, by at least one transmitter, the first RF signal over the communication channel;

receiving, by at least one receiver, a second RF signal representative of the first RF signal altered by transmission via the communication channel; and

processing the second RF signal using the decoder and generating second information as a reconstruction of the first information.

12. The method of claim 11, further comprising:

determining feedback information indicative of at least one of (i) a distance measure between the second information and the first information or (ii) channel state information about the communication channel; and

updating at least one of the encoder or the decoder based on the feedback information.

13. The method of claim 12, wherein updating at least one of the encoder or the decoder based on the feedback information further comprises:

determining a channel mode representing a state of the communication channel from a plurality of channel modes based on the feedback information; and

updating at least one of the encoder or the decoder based on the channel mode of the communication channel.

14. The method of claim 11, wherein the encoder implements an encoding mapping based on results of training an encoder machine learning network and the decoder implements a decoding mapping based on results of training a decoder machine learning network,

wherein the encoder machine learning network and the decoder machine learning network are co-trained as an auto-encoder to learn communications over a communication channel.

15. The method of claim 11, further comprising:

processing the first RF signal to generate a first analog RF waveform;

transmitting the first analog RF waveform over the communication channel using one or more transmit antennas;

receiving, using one or more receive antennas, a second analog RF waveform representative of the first analog RF waveform changed over the communication channel; and

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

16. A system, comprising:

at least one processor; and

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

determining first information;

processing the first information using an encoder machine learning network and generating a first RF signal for transmission over a communication channel;

determining a second RF signal representative of the first RF signal altered by transmission via the communication channel;

processing the second RF signal using a decoder machine learning network and generating second information as a reconstruction of the first information;

calculating a distance measure 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 distance metric between the second information and the first information.

17. The system of claim 16, wherein updating at least one of the encoder machine learning network or the decoder machine learning network based on the distance metric between the second information and the first information comprises:

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

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

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

18. The system of claim 16, wherein the distance metric 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.

19. The system of claim 16, 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 of 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.

20. The system of claim 16, wherein updating at least one of the encoder machine learning network or the decoder machine learning network further comprises:

determining a channel mode representing a state of the communication channel from a plurality of channel modes; and

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

21. The system of claim 16, wherein the encoder machine learning network and the decoder machine learning network are co-trained as an auto-encoder to learn communications over a communication channel, and

wherein the autoencoder includes at least one channel modeling layer that represents an effect of the communication channel on a transmitted waveform.

22. The system of claim 21, wherein the at least one channel modeling layer represents at least one of: (i) additive Gaussian thermal noise in the communication channel; (ii) a delay spread caused by a time-varying effect of the communication channel; (iii) phase noise caused by transmission and reception over the communication channel; or (iv) a phase offset, a frequency offset, or a timing offset caused by transmission and reception over the communication channel.

23. The system of claim 16, 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), the DNN, CNN, or RNN comprising parametric multiplications, additions, and nonlinear characteristics.

24. The system of claim 16, wherein the operations further comprise:

processing the first RF signal to generate a first analog RF waveform, the first analog RF waveform being input into the communication channel;

receiving a second analog RF waveform as an output of the communication channel, the output representing the first analog RF waveform as changed by the communication channel; and

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

25. The system of claim 16, wherein the communication channel comprises at least one of a radio communication channel, an acoustic communication channel, or an optical communication channel.

26. A system, comprising:

at least one processor; and

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

determining an encoder and a decoder, at least one of the encoder and the decoder configured to implement encoding or decoding based on at least one of an encoder machine learning network or a decoder machine learning network trained to encode or decode information over a communication channel;

determining first information;

processing the first information using the encoder and generating a first RF signal;

transmitting, by at least one transmitter, the first RF signal over the communication channel;

receiving, by at least one receiver, a second RF signal representative of the first RF signal altered by transmission via the communication channel; and

processing the second RF signal using the decoder and generating second information as a reconstruction of the first information.

27. The system of claim 26, wherein the operations further comprise:

determining feedback information indicative of at least one of (i) a distance measure between the second information and the first information or (ii) channel state information about the communication channel; and

updating at least one of the encoder or the decoder based on the feedback information.

28. The system of claim 27, wherein updating at least one of the encoder or the decoder based on the feedback information further comprises:

determining a channel mode representing a state of the communication channel from a plurality of channel modes based on the feedback information; and

updating at least one of the encoder or the decoder according to the channel mode of the communication channel.

29. The system of claim 27, wherein the encoder implements an encoding mapping based on results of training an encoder machine learning network and the decoder implements a decoding mapping based on results of training a decoder machine learning network,

wherein the encoder machine learning network and the decoder machine learning network are co-trained as an auto-encoder to learn communications over a communication channel.

30. The system of claim 27, wherein the operations further comprise:

processing the first RF signal to generate a first analog RF waveform;

transmitting the first analog RF waveform over the communication channel using one or more transmit antennas;

receiving, using one or more receive antennas, a second analog RF waveform representative of the first analog RF waveform changed over the communication channel; and

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

Technical Field

The present disclosure relates to machine learning and deployment for adaptive wireless communications, and in particular to for Radio Frequency (RF) signals.

Background

Radio Frequency (RF) waveforms are ubiquitous in many systems for communication, storage, sensing, measurement and monitoring. RF waveforms are transmitted and received over various types of communication media, such as air, underwater, or through outer space. In some scenarios, the RF waveform transmits information modulated onto one or more carrier waveforms operating at RF frequencies. In other scenarios, the RF waveform itself is the information, such as the output of a sensor or probe. The information carried in the RF waveforms is typically processed, stored, and/or transported via other forms of communication, such as via an internal system bus in a computer or via a local or wide area network.

Disclosure of Invention

The subject matter described in this disclosure may be generally embodied in methods, devices, and systems for training and deploying machine learning networks to communicate over RF channels, and in particular encoding and decoding information for communication over RF channels.

In one aspect, a method performed by at least one processor to train at least one machine learning network to communicate over an RF channel, the method comprising: determining first information; processing the first information using an encoder machine learning network and generating a first RF signal for transmission over a communication channel; determining a second RF signal representative of the first RF signal altered by transmission via the communication channel; processing the second RF signal using a decoder machine learning network and generating second information as a reconstruction of the first information; calculating a distance measure 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 distance metric between the second information and the first information. Other implementations of this aspect include corresponding computer systems, apparatus, and computer programs recorded on one or more computer storage devices, each configured to cause at least one operatively connected processor to perform the actions of the methods.

Drawings

Fig. 1 shows an example of a Radio Frequency (RF) system implementing a machine learning encoder and decoder to perform learned communications over one or more RF channels;

FIG. 2 illustrates an example of a network structure of a machine learning encoder and decoder network that may be implemented in an RF system to perform learned communications over an RF channel;

FIG. 3 illustrates an example of training an RF system implementing a machine-learned encoder and decoder network to learn encoding and decoding over an RF channel;

FIG. 4 is a flow diagram illustrating an example method of training an RF system implementing a machine-learned encoder and decoder network to learn encoding and decoding over an RF channel;

FIG. 5 illustrates an example of basis functions that may be learned by an encoder machine learning network and/or a decoder machine learning network and used to communicate over an RF channel;

fig. 6 illustrates an example of transmit and receive RF signals that may be learned by an encoder machine learning network and a decoder machine learning network for communication over an RF channel;

FIG. 7 illustrates an example of a system deploying encoders and decoders implementing communications that perform learned over real-world RF channels using encoding and/or decoding techniques based on the results of training an encoding and decoding machine learning network;

FIG. 8 is a flow diagram illustrating an example method of deploying an encoder and decoder that performs learned communications over real world RF channels using encoding and/or decoding techniques based on a training encoding and decoding machine learning network; and is

Fig. 9 is a diagram illustrating an example of a computing system that may be used to implement one or more components of a system that performs learned communications over an RF channel.

Implementations may include one or more of the following features. In the method, updating at least one of the encoder machine learning network or the decoder machine learning network based on the distance metric between the second information and the first information includes: determining an objective function comprising the distance measure between the second information and the first information; calculating a rate of change of the objective function relative to a change in at least one of the encoder machine learning network or the decoder machine learning network; selecting a first variation for at least one of the encoder machine learning network or a second variation for the decoder machine learning network based on the calculated rate of change of the objective function; 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 change of the encoder machine learning network or the selected second change of the decoder machine learning network. In the method, the distance metric between the second information and the first information includes 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. In the method, updating at least one of the encoder machine learning network or the decoder machine learning network includes at least one of: updating at least one encoding network weight or network connectivity in one or more layers in the encoder machine learning network, or updating at least one decoding network weight or network connectivity in one or more layers in the decoder machine learning network. In the method, updating at least one of the encoder machine learning network or the decoder machine learning network further includes: determining a channel mode representing a state of the communication channel from a plurality of channel modes; and updating at least one of the encoder machine learning network or the decoder machine learning network based on the channel pattern of the communication channel. In the method, the encoder machine learning network and the decoder machine learning network are co-trained as an auto-encoder to learn communications over a communication channel, and wherein the auto-encoder includes at least one channel modeling layer that represents an effect of the communication channel on a transmitted waveform. In the method, the at least one channel modeling layer represents at least one of: (i) additive Gaussian thermal noise in the communication channel; (ii) a delay spread caused by a time-varying effect of the communication channel; (iii) phase noise caused by transmission and reception over the communication channel; (iv) a phase offset, a frequency offset, or a timing offset caused by transmission and reception over the communication channel. In the method, at least one of the encoder machine learning network or the decoder machine learning network includes at least one of a deep Dense Neural Network (DNN), a Convolutional Neural Network (CNN), or a Recurrent Neural Network (RNN), the DNN, CNN, or RNN including parametric multiplications, additions, and nonlinear characteristics. The method further comprises: processing the first RF signal to generate a first analog RF waveform, the first analog RF waveform being input into the communication channel; receiving a second analog RF waveform as an output of the communication channel, the output representing the first analog RF waveform as changed by the communication channel; and processing the second analog RF waveform to generate the second RF signal. In the method, the communication channel includes at least one of a radio communication channel, an acoustic communication channel, or an optical communication channel. Embodiments of the described technology may comprise hardware, methods or processes, or computer software on a computer-accessible medium.

In another aspect, a method is performed by at least one processor to deploy a learning communication system over an RF channel. The method comprises the following steps: determining an encoder and a decoder, at least one of the encoder and the decoder configured to implement encoding or decoding based on at least one of an encoder machine learning network or a decoder machine learning network trained to encode or decode information over a communication channel; determining first information; processing the first information using the encoder and generating a first RF signal; transmitting, by at least one transmitter, the first RF signal over the communication channel; receiving, by at least one receiver, a second RF signal representative of the first RF signal altered by transmission via the communication channel; and processing the second RF signal using the decoder and generating second information as a reconstruction of the first information. Other implementations of this aspect include corresponding computer systems, apparatus, and computer programs recorded on one or more computer storage devices, each configured to cause at least one operatively connected processor to perform the actions of the methods.