Method for free space optical communication utilizing patterned light and convolutional neural networks

1. An apparatus comprising:

an optical communications receiver receiving a turbulence-distorted, optical signal comprising a plurality of fundamental modes encoded via a combinatorial multiplexings dictionary, the received optical signal comprising a plurality of two-dimensional images, each two-dimensional image of the plurality of two-dimensional images respectively representing received fundamental modes of the plurality of fundamental modes,

said receiver comprising a neural network trained to assign to each two-dimensional image of the plurality of two-dimensional images at least one respective active fundamental mode of the plurality of fundamental modes and a corresponding accuracy probability based on the dictionary.

2. The apparatus according to claim 1 , further comprising:

an optical communications transmitter transmitting an original, optical signal to said optical communications receiver, the optical signal being converted by an environment into the turbulence-distorted, optical signal.

3. The apparatus according to claim 2 , wherein said receiver comprises a demultiplexer, said demultiplexer comprising said neural network,

wherein said optical communications transmitter comprises a laser, a mode multiplexer communicating with said laser, and a processor communicating with said mode multiplexer to generate the original, optical signal.

4. The apparatus according to claim 3 , wherein said mode multiplexer comprises at least one of a spatial and temporal phase modulator and a spatial and temporal amplitude modulator,

wherein said spatial and temporal phase modulator comprises one of a spatial light modulator, a diffractive waveplate, and a phase plate

wherein said spatial and temporal amplitude modulator comprises one of a spatial light modulator and a coded mask.

5. The apparatus according to claim 2 , wherein said optical communications transmitter comprises one of a free-space optical communications transmitter and an underwater optical communications transmitter.

6. The apparatus according to claim 1 , wherein said optical communications receiver comprises an imager communicating with said demultiplexer.

7. The apparatus according to claim 1 , wherein said neural network comprises at least one of a Recurrent Neural Network, a Convolutional Neural Network, a Feed Forward Neural Network, a Long Term Short Term Memory Neural Network, a Residual Neural Network, a Multilayer Perceptron, a Hopfield Neural Network, a Stacked Autoencoder, and a Deep Belief Network.

8. The apparatus according to claim 1 , wherein said neural network comprises a plurality of weights,

the apparatus further comprising:

a training set of turbulence-distorted, multiplexed fundamental modes, said plurality of weights being trained on said training set.

9. The apparatus according to claim 1 , wherein said optical communications receiver comprises one of a free-space optical communications receiver and an underwater optical communications receiver.

10. A method comprising:

selecting a plurality of fundamental modes, the selected plurality of fundamental modes comprising a number of the plurality of fundamental modes and a type of the plurality of fundamental modes;

generating a first training set of turbulence-distorted fundamental modes based on the selected plurality of fundamental modes;

training a neural network using the first training set;

providing an optical communications receiver comprising the trained neural network.

11. The method according to claim 10 , wherein the optical communications receiver comprises the trained neural network,

the method further comprising:

receiving at the optical communications receiver a turbulence-distorted, optical signal encoded using a combinatorial multiplexings dictionary, the received optical signal comprising a plurality of two-dimensional images representing the plurality of fundamental modes; and

assigning to the each two-dimensional image of the plurality of two-dimensional images at least one respective active fundamental mode of the plurality of fundamental modes and a corresponding accuracy probability based on the dictionary using a neural network.

12. The method according to claim 11 , wherein the neural network comprises a demultiplexer, the demultiplexer comprising the trained neural network,

wherein the plurality of two-dimensional images represents multiplexed fundamental modes of the plurality of fundamental modes.

13. The method according to claim 12 , wherein said multiplexed fundamental modes of the plurality of fundamental modes comprise at least one of wavelength-multiplexed fundamental modes of the plurality of fundamental modes and spatial multiplexed fundamental modes of the plurality of fundamental modes.

14. The method according claim 12 , wherein the optical communications receiver comprises an imager communicating with the demultiplexer,

wherein said receiving at the optical communications receiver a turbulence-distorted, optical signal comprises:

receiving the turbulence-distorted, optical signal using the imager,

generating the plurality of two-dimensional images using the imager; and

transmitting the each two-dimensional image of the plurality of two-dimensional images from the imager to the demultiplexer.

15. The method according to claim 11 , wherein the dictionary comprises a plurality of the combinatorial multiplexings of the selected plurality of fundamental modes,

wherein the first training set of turbulence-distorted, fundamental modes comprises a plurality of turbulence-distorted realizations for each combinatorial multiplexings of the plurality of combinatorial multiplexings;

wherein the method further comprises:

generating the combinatorial multiplexings dictionary based on the plurality of combinatorial multiplexings of the selected plurality of fundamental modes.

16. The method according to claim 11 , wherein said assigning to the each two-dimensional image of the plurality of two-dimensional images at least one respective active fundamental mode of the plurality of fundamental modes and a corresponding accuracy probability comprises one of:

outputting, using the neural network, a single label corresponding to the plurality of combinatorial multiplexings and a probability distribution over the plurality of combinatorial multiplexings; and

outputting, using the neural network, a plurality of labels respectively corresponding to the fundamental modes, and a plurality of probability distributions of the selected plurality of fundamental modes.

17. The method according to claim 11 , wherein the neural network comprises a plurality of weights, the plurality of weights being set by the neural network trained on the first training set,

the method further comprising:

generating a second training set of turbulence-distorted fundamental modes based on the selected plurality of fundamental modes, the second training set being at least partly different from the first training set; and

training the neural network using the second training set, thereby fine-tuning the plurality of weights.

18. The method according to claim 17 , wherein the original, optical signal comprises error-correcting code data,

the method further comprising:

adjusting at least one selected weight of the plurality of selected weights, using one of a test signal and the error-correcting code data.

19. The apparatus according to claim 10 , wherein said optical communications receiver comprises one of a free-space optical communications receiver and an underwater optical communications receiver.