CPC H04L 25/03929 (2013.01) [H04L 25/023 (2013.01); H04L 25/066 (2013.01)] | 20 Claims |
1. A method for adaptive coding and modulation, the method comprising:
training, utilizing one or more processors, a set of recurrent neural networks (RNNs) based on a set of communication channels; and
transmitting a tth set of transmit symbols in a tth time interval by transmitting an (i, t)th transmit symbol in the tth set of transmit symbols by an ith transmitter in a set of transmitters where 1≤i≤N, N is a size of the set of transmitters, 1≤t≤T, and T is a maximum number of symbol transmissions, transmitting the (i, t)th transmit symbol comprising:
generating an (i, t)th set of mapped symbols by:
obtaining an (i, t)th hidden states set of an ith RNNs subset of the set of RNNs based on an (i, r)th set of mapped symbols where 0≤r≤T; and
applying the ith RNNs subset on one of an ith data vector and a set of channel state information (CSI) samples of an ith channels subset of the set of communication channels based on the (i, t)th hidden states set,
wherein:
each mapped symbol in the (i, t)th set of mapped symbols comprises a respective complex value; and
each mapped symbol in an (i, 0)th set of mapped symbols is zero;
generating the (i, t)th transmit symbol from the (i, t)th set of mapped symbols; and
transmitting the (i, t)th transmit symbol through the ith channels subset.
|