| dc.description.abstract | We consider a wireless network consisting of multiple source nodes, a set of relays
and a destination node. Suppose the sources transmit their messages simultaneously
to the relays and the destination aims to decode all the messages. At the physical layer,
a conventional approach would be for the relay to decode the individual message
one at a time while treating rest of the messages as interference. Compute-and-forward
is a novel strategy which attempts to turn the situation around by treating
the interference as a constructive phenomenon. In compute-and-forward, each relay
attempts to directly compute a combination of the transmitted messages and then
forwards it to the destination. Upon receiving the combinations of messages from the
relays, the destination can recover all the messages by solving the received equations.
When identical lattice codes are employed at the sources, error correction to integer
combination of messages is a viable option by exploiting the algebraic structure of
lattice codes. Therefore, compute-and-forward with lattice codes enables the relay
to manage interference and perform error correction concurrently. It is shown that
compute-and-forward exhibits substantial improvement in the achievable rate compared
with other state-of-the-art schemes for medium to high signal-to-noise ratio
regime.
Despite several results that show the excellent performance of compute-and-forward,
there are still important challenges to overcome before we can utilize compute-and-
forward in practice. Some important challenges include the assumptions of \perfect
timing synchronization "and \quasi-static fading", since these assumptions rarely
hold in realistic wireless channels. So far, there are no conclusive answers to whether
compute-and-forward can still provide substantial gains even when these assumptions
are removed. When lattice codewords are misaligned and mixed up, decoding integer
combination of messages is not straightforward since the linearity of lattice codes is
generally not invariant to time shift. When channel exhibits time selectivity, it brings
challenges to compute-and-forward since the linearity of lattice codes does not suit
the time varying nature of the channel. Another challenge comes from the emerging
technologies for future 5G communication, e.g., autonomous driving and virtual
reality, where low-latency communication with high reliability is necessary. In this
regard, powerful short channel codes with reasonable encoding/decoding complexity
are indispensable. Although there are fruitful results on designing short channel
codes for point-to-point communication, studies on short code design specifically for
compute-and-forward are rarely found.
The objective of this dissertation is threefold. First, we study compute-and-forward
with timing-asynchronous users. Second, we consider the problem of compute-and-
forward over block-fading channels. Finally, the problem of compute-and-forward
for low-latency communication is studied. Throughout the dissertation, the research
methods and proposed remedies will center around the design of lattice codes in order
to facilitate the use of compute-and-forward in the presence of these challenges. | en |
