Gaussian interference networks: Lattice alignment

A new achievable rate region for the 3-user discrete memoryless interference channel

“…The best known inner bound to the capacity region is given by the Han-Kobayashi achievable rate region [2]. Recently, there has been a lot of activity for the three user channel in special cases such as the additive white Gaussian noise model with and without fading [6]- [10], [13]. For the Gaussian 3-user interference channel it has been shown in [8]- [10] that random lattice codes can provide rates that are beyond what can be obtained using random code ensembles.…”

Should one always connect to the base station with the strongest signal?

“…Note that the achieved degrees of freedom for almost all interference channels (with real valued channel gains) is already known to be K/2. In this work, we compare the degrees of freedom achieved using a "weak" pairing and lattice alignment [14] versus a "strong" pairing and modified Han-Kobayashi scheme. Note that lattice alignment for "strong" pairing channels is still an open problem, and thus we focus on a Han-Kobayashi type coding scheme in that case.…”

“…For the case of weak-pairing, [14], [13] shows that one can achieve the following total DoF for such a channel:…”

“…As lattice alignment is better understood for strong-interference channels (corresponding to "weak"-pairing) [14] than a mixed-interference channel (corresponding to "strong"-pairing), it is natural to combine "weak"-pairing with lattice alignment and contrast its performance with a combination of "strong"-pairing with Han-Kobayashi coding.…”

“…Theorem 3.1 ( [14]): For a symmetric interference channel as defined by Figure 3, the following sum-rate can be achieved employing lattice encoding/decoding: 1)a + K)) .…”