Channel feature extraction via model-based neural networks
A method for wireless communication by a receiving device, includes receiving, from a transmitting device, a latent representation of a channel sequence for a wireless signal. A decoder applies a physical propagation channel model to the latent representation to reconstruct the channel sequence for the wireless signal.
1 . A method for wireless communication by a receiving device, comprising:
receiving, from a transmitting device, a latent representation z t of a channel sequence H t for communicating a wireless signal; and
applying, via a decoder, both a physical propagation channel model and a neural network to the latent representation to generate a reconstructed channel sequence Ĥ t for the wireless signal, wherein the physical propagation channel model is represented by a sum of multipath components, the latent representation comprises a vector having portions controlled by an encoder to follow physical parameters of the physical propagation channel model such that the decoder maps at least one-portion of the vector for the latent representation to a physical parameter of the physical propagation channel model based on a position within the vector of the at least one portion of the vector for the latent representation, the neural network estimates unknown parameters of the physical propagation channel model, and the decoder is trained independently of the encoder.
2 . The method of claim 1 , in which each of the multipath components has a delay and a complex magnitude.
3 . The method of claim 1 , in which each of the multipath components has a transmitter (Tx) array response and a receiver (Rx) array response.
4 . The method of claim 3 , in which the Tx array response and the Rx array response are represented as learnable functions of azimuth and zenith angles of departure and azimuth and zenith angles of arrival, respectively.
5 . The method of claim 1 , further comprising capturing additional effects on the wireless signal not captured by the physical propagation channel model via additional learnable neural network layers.
6 . The method of claim 5 , in which the additional effects includes a receiver (Rx) chain impulse response and a transmitter (Tx) chain impulse response.
7 . The method of claim 1 , in which the latent representation is sent to the receiving device as channel state feedback.
8 . The method of claim 1 , in which the reconstructed channel sequence comprises a continuous channel response across a plurality of frequency bands separated by a guard band.
9 . The method of claim 1 , in which the latent representation comprises one or more of a set of delays, complex magnitudes, angles of departure, and angles of arrival.
10 . The method of claim 1 , in which the latent representation comprises an encoded channel sequence, and the receiving device uses the decoder to generate the reconstructed channel sequence from the latent representation.
11 . A method for wireless communication by a transmitting device, comprising:
receiving, via an encoder, an input comprising a channel sequence for a wireless signal;
processing, via the encoder, the channel sequence to produce a latent representation of the channel sequence for the wireless signal, the latent representation comprising a vector controlled by the encoder to follow physical parameters of a physical propagation channel model such that at least one portion of the vector of the latent representation maps to a physical parameter of the physical propagation channel model based on a position within the vector of the at least one portion of the vector of the latent representation, the physical propagation channel model being represented by a sum of multipath components, and the encoder being trained independently of the physical propagation channel model; and
transmitting the latent representation to the receiving device.
12 . The method of claim 11 , in which each of the multipath components has a delay and a complex magnitude.
13 . The method of claim 12 , in which each of the multipath components has a transmitter (Tx) array response and a receiver (Rx) array response.
14 . The method of claim 13 , in which the Tx array response and the Rx array response are represented as learnable functions of azimuth and zenith angles of departure and azimuth and zenith angles of arrival, respectively.
15 . The method of claim 11 , in which the latent representation is transmitted to the receiving device as channel state feedback.
16 . The method of claim 11 , in which the encoder comprises a neural network.
17 . The method of claim 11 , in which the transmitting device jointly trains an encoder-decoder pair such that a cascade of the encoder and a second decoder of the encoder-decoder pair reconstructs the channel sequence input at the encoder.
18 . An apparatus for wireless communication by a receiving device, comprising:
at least one memory; and
at least one processor coupled to the at least one memory, the at least one processor being configured:
to receive, from a transmitting device, a latent representation z t of a channel sequence H t for communicating a wireless signal; and
to apply, via a decoder, both a physical propagation channel model and a neural network to the latent representation to generate a reconstructed channel sequence H t for the wireless signal, wherein the physical propagation channel model is represented by a sum of multipath components, the latent representation comprises a vector having portions controlled by an encoder to follow physical parameters of the physical propagation channel model such that the decoder maps at least one portion of the vector for the latent representation to a physical parameter of the physical propagation channel model based on a position within the vector of the at least one portion of the vector for the latent representation, the neural network estimates unknown parameters of the physical propagation channel model, and the decoder is trained independently of the encoder.
19 . The apparatus of claim 18 , in which each of the multipath components has a delay and a complex magnitude.
20 . The apparatus of claim 18 , in which each of the multipath components has a transmitter (Tx) array response and a receiver (Rx) array response.
21 . The apparatus of claim 20 , in which the at least one processor is further configured to represent the Tx array response and the Rx array response as learnable functions of azimuth and zenith angles of departure and azimuth and zenith angles of arrival, respectively.
22 . The apparatus of claim 18 , in which the at least one processor is further configured to capture additional effects on the wireless signal not captured by the physical propagation channel model via additional learnable neural network layers.
23 . The apparatus of claim 22 , in which the additional effects includes a receiver (Rx) chain impulse response and a transmitter (Tx) chain impulse response.
24 . The apparatus of claim 18 , in which the at least one processor is further configured to receive, via the receiving device, the latent representation as channel state feedback.
25 . The apparatus of claim 18 , in which the reconstructed channel sequence comprises a continuous channel response across a plurality of frequency bands separated by a guard band.
26 . The apparatus of claim 18 , in which the latent representation comprises one or more of a set of delays, complex magnitudes, angles of departure, and angles of arrival.
27 . The apparatus of claim 18 , in which the latent representation comprises an encoded channel sequence, and the receiving device uses the decoder to generate the reconstructed channel sequence from the latent representation.
28 . A apparatus for wireless communication by a transmitting device, comprising:
at least one memory; and
at least one processor coupled to the at least one memory, the at least one processor being configured:
to receive, via an encoder, an input comprising a channel sequence for a wireless signal;
to process, via the encoder, the channel sequence to produce a latent representation of the channel sequence for the wireless signal, the latent representation comprising a vector controlled by the encoder to follow physical parameters of a physical propagation channel model such that at least one portion of the vector of the latent representation maps to a physical parameter of the physical propagation channel model based on a position within the vector of the at least one portion of the vector of the latent representation, the physical propagation channel model being represented by a sum of multipath components, and the encoder being trained independently of the physical propagation channel model; and
to transmit, the latent representation to the receiving device.
29 . The apparatus of claim 28 , in which each of the multipath components has a delay and a complex magnitude.
30 . The apparatus of claim 29 , in which each of the multipath components has a transmitter (Tx) array response and a receiver (Rx) array response.
31 . The apparatus of claim 30 , in which the at least one processor is further configured to represent the Tx array response and the Rx array response as learnable functions of azimuth and zenith angles of departure and azimuth and zenith angles of arrival, respectively.
32 . The apparatus of claim 28 , in which the at least one processor is further configured to transmit the latent representation to the receiving device as channel state feedback.
33 . The apparatus of claim 28 , in which the encoder comprises a neural network.
34 . The apparatus of claim 28 , in which the at least one processor is further configured to jointly train, via the transmitting device, an encoder-decoder pair such that a cascade of the encoder and a second decoder of the encoder-decoder pair reconstructs the channel sequence input at the encoder.
35 . An apparatus for wireless communication by a receiving device, comprising:
means for receiving, from a transmitting device, a latent representation z t of a channel sequence H t for communicating a wireless signal; and
means for applying, via a decoder, both a physical propagation channel model and a neural network to the latent representation to generate a reconstructed channel sequence Ĥ t for the wireless signal, wherein the physical propagation channel model is represented by a sum of multipath components, the latent representation comprises a vector having portions controlled by an encoder to follow physical parameters of the physical propagation channel model such that the decoder maps at least one portion of the vector for the latent representation to a physical parameter of the physical propagation channel model based on a position within the vector of the at least one portion of the vector for the latent representation, the neural network estimates unknown parameters of the physical propagation channel model, and the decoder is trained independently of the encoder.
36 . A method for wireless communication by a transmitting device, comprising:
means for receiving an input comprising a channel sequence for a wireless signal;
mean for processing, via an encoder, the channel sequence to produce a latent representation of the channel sequence for the wireless signal, the latent representation comprising a vector controlled by the encoder to follow physical parameters of a physical propagation channel model such that at least one portion of the vector of the latent representation maps to a physical parameter of the physical propagation channel model based on a position within the vector of the at least one portion of the vector of the latent representation, the physical propagation channel model being represented by a sum of multipath components, and the encoder being trained independently of the physical propagation channel model; and
means for transmitting the latent representation to the receiving device.
37 . A non-transitory computer readable medium having encoded thereon program code for wireless communication by a receiving device, the program code being executed by a processor and comprising:
program code to receive, from a transmitting device, a latent representation z t of a channel sequence H t for communicating a wireless signal; and
program code to apply, via a decoder, both a physical propagation channel model and a neural network to the latent representation to generate a reconstructed channel sequence Ĥ t for the wireless signal, wherein the physical propagation channel model is represented by a sum of multipath components, the latent representation comprises a vector having portions controlled by an encoder to follow physical parameters of the physical propagation channel model such that the decoder maps at least one portion of the vector for the latent representation to a physical parameter of the physical propagation channel model based on a position within the vector of the at least one portion of the vector for the latent representation, the neural network estimates unknown parameters of the physical propagation channel model, and the decoder is trained independently of the encoder.
38 . A non-transitory computer readable medium having encoded thereon program code for wireless communication by a transmitting device, the program code being executed by a processor and comprising:
program code to receive an input comprising a channel sequence for a wireless signal;
program code to process, via an encoder, the channel sequence to produce a latent representation of the channel sequence for the wireless signal, the latent representation comprising a vector controlled by the encoder to follow physical parameters of a physical propagation channel model such that at least one portion of the vector of the latent representation maps to a physical parameter of the physical propagation channel model based on a position within the vector of the at least one portion of the vector of the latent representation, the physical propagation channel model being represented by a sum of multipath components, and the encoder being trained independently of the physical propagation channel model; and
program code to transmit, the latent representation to the receiving device.